GENETIC AND SOMATIC EFFECTS OF IONIZING RADIATION

Transcription

1 GENETIC AND SOMATIC EFFECTS OF IONIZING RADIATION United Nations Scientific Committee on the Effects of Atomic Radiation 1986 Report to the General Assembly, with annexes UNITED NATIONS New York, 1986

2 NOTE The report of the Committee withot its annexes appears as Official Records of the General Assembly, Forty-first Session, Spplement No. 16 (A/41/16). The designations employed and the presentation of material in this pblication do not imply the expression of any opinion whatsoever on the part of the Secretariat of the United Nations concerning the legal stats of any contry, territory, city or area. or of its athorities, or concerning the delimitation of its frontiers or bondaries. UNITED NATIONS PUBLICATION Sales No. E.86.IX.9 ISBN P

3 ANNEXB Dose-response relationships for radiation-indced cancer CONTENTS INTRODUCTION..... _..... I. DOSE-RESPONSE RELATIONSHIPS FOR RADIATION-INDUCED CAN- CER A. The independent variable.... B. The dependent variables in experimental work... C. The dependent variables in epidemiological stdies I. The risk expressed in absolte terms The risk expressed in relative terms.... D. Temporal relationships.... I. Latent period Relationships between dose and tmor-free lifetime lost as a reslt of indced cancers.... E. Smmary and conclsions.... II. ASSUMPTIONS AND LIMITATIONS OF EXISTING MODELS.... A. Basic phenomena and inflence of modifying factors B. The threshold dose for cancer indction.... C. Unicelllar verss mlticelllar origin of neoplastic growth.... D. Atonomy of target cells for tmor indction E. Nmber of cells at risk.... F. Poplation heterogeneity and ss- Paragraphs ceptibility to cancer indction G. Smmary and conclsions III. DOSE-RESPONSE MODELS OF RADIATION-INDUCED CANCER A. Introdction IV. B. Derivation and characteristics of dose-response models The linear model The linear-qadratic model The qadratic model... C. Smmary and conclsions.... DOSE-RESPONSE RELATIONSHIPS IN EXPERIMENTAL SYSTEMS... A. Phenomena observed at the celllar and sbcelllar level I. Mtations in germ cells Mtations in somatic cells Chromosome aberrarions Oncogenic transformation of cells in vitro B. Tmors in experimental animals.. I. Myeloid lekaemia Thymic lymphoma Other reticlar tmors Tmors of the thyroid Mammary tmors Lng cancers Lng adenomas Ovarian tmors Pititary gland tmors.... JO. Harderian gland tmors Skin tmors Bone tmors (internal irradiation)... Paragraphs I Liver _ 14. Combined tmor data and indirect inferences C. Smmary and conclsions V. DOSE-RESPONSE RELATIONSHIPS FOR TUMOURS IN MAN A. Lekaemia.... I. Direct stdies Indirect evidence

4 B. Cancer of the breast Radiotherapy for mastitis Pnemotherapy patients.... C. Tmors of the thyroid Direct stdies Indirect evidence.... Paragraphs D. Cancer of the respiratory tract The risk expressed in absolte terms The risk expressed in relative terms E. Bone sarcoma.... l. Direct stdies The model of Marshall and Greer.... F. Interspecies comparisons.... G. Smmary and conclsions.... VI. SUMMARY.... VIL RESEARCH NEEDS.... Tables.... References.... Paragraphs Page Introdction 1. It has long been recognized by UNSCEAR that radiation-indced malignant diseases are the most important late somatic effect in hman poplations exposed at high doses for which direct observations are available [U6, U7, U9-Ul2. U24]. For evalation of radiological risk or detriment [12] this importance derives from the fact that these diseases are often lethal and they are the only statistically verifiable case of radiation-indced life shortening at intermediate and low doses [B 17, B 18, JI. K39, K4, U24]. Radiation-indced cancer belongs to those radiobiological effects whose freqency of occrrence (bt not severity) is believed-as a rle-to correlate with dose. a The postlated probabilistic natre of the relationship between dose and freqency of malignancyb has led to the acceptance of the term "stochastic" for effects of sch type. 2. Assessment of risk. from environmental and occpational radiation sorces in the dose region from fractions of mgy to a few tens of mgy, wold be greatly facilitated by knowledge of the shapes of the dose-response relationships for radiation-indced cancers in hmans. This knowledge is not available at present and is not likely to be obtained by direct observation. Two featres of the dose-response relations are most important for evalation of the risk at low doses: the possible presence of a threshold dose below which the effects wold not occr, and the shape of the dose-response crve. 3. Lack of threshold for a given effect is sally assmed if the response for this effect, plotted against the independent variable (casal factor. or, specifically, dose). permits extrapolation by eye to the origin of the coordinate system. or when the calclated regres- ~For brevity, the term "dose" is sed here instead of the more correct "absorbed dose". hin this annex, the term "carcinogenesis" is sed to inclde carcinoma, lekaemia or any other form of malignancy (sarcoma, lymphoma, etc.). The word "malignancy" is sed when the reference is to all sch malignant conditions, while "cancer" or "malignant tmor" refer only to solid or focal malignancies. The term "tmor" is sed withot qalification when it is either clear from the context, or nimportant, whether a malignant or a benign tmor is intended. sion line intercepts the abscissa at vales that are not significantly different from zero. Conversely. a threshold is sally assmed when the fitted regression fnction crosses the abscissa at a vale significantly greater than zero. However, proving or disproving a threshold below the levels of direct observation may be impossible, de to statistical flctations of the spontaneos level and of the presmably indced response. Therefore, assmptions regarding a threshold are based essentially on theoretical considerations of the mechanisms of radiation interaction with the biological targets for initiation of neoplasia. spplemented by empirical observations to spport the hypothesis. Althogh absence of the threshold is often assmed, this has not been proved for any form of radiation-indced malignancy [U6] and mst be regarded as a working hypothesis. 4. In annex I of the 1977 UNSCEAR report [U6]. the available data concerning experimental radiation carcinogenesis in nmeros animal species were reviewed. The large variation of ssceptibility to cancer indction in different tisses was emphasized. Physical and biological factors modifying the freqency of indction were discssed in great detail and interactions of other agents (e.g., virses) with radiation were also reviewed. The extreme complication and nsatisfactory nderstanding of the pathogenesis of all forms of cancer, inclding those indced by radiation. were particlarly stressed. 5. Varios physical factors. sch as dose, dose rate, and qality of radiation, were also considered, and general patterns cold be recognized in cases where sch factors were systematically stdied in a given strain of animals, and for specific tmor types. Among the patterns identified, were a sparing effect of dose fractionation and protraction of low-let radiation (x or gamma rays) pon the freqency of indced cancer and the absence, or even reversal, of sch an effect for high-let radiation (netrons, alpha particles). 6. Crrent theories of cancer indction by radiation and some other agents (virses. chemicals) were also briefly reviewed in the 1977 UNSCEAR report [U6]. None of these was able to accommodate all the known facts and to allow development of a theory or of 166

5 comprehensive models of cancer indction by ionizing radiation. It was recognized, however, that known carcinogens have a common target within the ssceptible cells, which is most likely the nclear DNA or genome. 7. Since the 1977 report [U6], new information has been pblished on experimental indction of cancer by radiation. Some of it refers to observations at intermediate and low doses. Of particlar importance is the information dealing with high-let particles (mostly netrons). This is reviewed here when it appears relevant to models of radiation-indced malignancies. 8. In annex I of the 1977 report [U6], the response relationships as a fnction of single acte doses for varios forms of experimental radiation-indced cancer-both after whole-body and localized irradiation-were reviewed thoroghly. Tmors cold be broadly sbdivided into three categories: (a) (b) (c) Those showing an increasing incidence with increasing dose p to a maximm, with a decline following that maximm (most forms): Those displaying a negative conelation between incidence and dose, as observed in tmors with an nsally high spontaneos freqency: Those showing no clear rise with increasing dose p to several Gy. For occpational and environmental exposre of man, type (a) is the most relevant. Schematic examples of dose-response crves are shown in Figre I. 9. UNSCEAR [U6] also identified a large variability of the net incidence of varios tmor types at intermediate to high doses between different species and. within species. between inbred strains. It was also fond that in many cases a particlar tmor cold be indced by radiation in only one or two strains of a given species, an observation that mst raise qestions as to whether sch tmors may represent adeqate models of corresponding hman diseases. Similar dobts wold also apply to some observed forms of dose-response relationships. In some cases. doseresponse relationships differ from species to species. althogh in many cases consistent patterns have been fond. For these reasons, the increased incidence per nit dose of a given form of cancer cannot-as a rle-be extrapolated between species. 1. For dose-response relationships of category (a) some reglarities were pointed ot that appeared to conform to other radiobiological phenomena occrring in single cells (e.g., cell killing, indction of mtations and chromosome aberrations). These common featres were as follows: (a) (b) (c) The RBE vales for densely-ionizing radiation relative to x and gamma rays are higher than I and decrease as doses increase: With acte doses of high-let radiation the dose-response relationship is closer to linearity than for sparsely-ionizing radiation, for which pward concave crvilinearity is sally observed: The tmor yield often shows little dependence on dose protraction and fractionation for high LET radiation. while for x and gamma rays the yield sally declines. 11. Since pblication of the 1977 report [U6], additional information has appeared on tmor indction and life shortening in the low and intermediate dose region. It shows that after acte (high dose rate) high LET exposre, in some cases at intermediate and in most cases at high doses. the incidence of tmors per nit dose decreases with increasing dose. For low-let acte exposres. sch a decline is sally observed only at high doses (above several Gy). These and other observations, together with some notable exceptions. will be discssed in detail in chapter IV. ~5... c :::, t' 4... "' ' - i: 3 "' ~ 2 Cl z ::: w DOSE (arbitrary nits) 6 Crve 1: Downwards concave Crve 2: Linear Crve 3: Upwards concave Crve 4: Threshold type Crve 5: A high spontaneos tmor rate declining with dose Figre I. Typical dose-response crves for radiation-indced tmors. Crve 1: Downwards concave; Crve 2: Linear, Crve 3: Upwards concave; Crve 4: Threshold type; Crve 5: A high spontaneoa tmor rate decllnlng with dose. The shaded area repreaents the spontaneos (control) Incidence. 167

6 12. In annex G of the 1977 report [U6]. UNSCEAR presented a comprehensive review of epidemiological data on radiation-indced cancer in man. Absolte risk estimates of mortality per nit dose were examined in detail for malignant diseases of varios organs, and confidence limits were attached to these estimates, derived in most cases from irradiation with doses at or above 1 Gy. Also, an approximate life-time risk of mortality from cancer at all sites per nit dose was estimated from the ratio of incidence of all nonlekaemic malignancies to the incidence of lekaemias in several grops (atomic bomb srvivors. American radiologists, and patients treated for ankylosing spondyli tis and metropathia haemorrhagica ). However. the previosly available dosimetry at Hiroshima and Nagasaki has been qestioned, and the new dosimetric system (DS86) is expected to yield improved risk coefficients for atomic bomb srvivors. who are one of the most important sorces of information. 13. The merit of sch risk estimates lies in the fact that thev are derived directly from hman data and ths av~id interspecies extrapolations of dobtfl validity. The precision of the 1977 risk estimates was the best possible nder the conditions of exposre and follow-p then available, bt many limitations of the estimates were discssed extensively. 14. Uncertainties regarding the shape of the doseresponse relationships for radiation-indced malignancies, and the related risk estimates in man, derive mostly from the following conditions: (a) (b) The short dration of follow-p of irradiated poplations compared to the length of latent period of most tmors. Whereas 2 to 3 years may be sfficient for manifestation of lekaernias and bone sarcomas after short-term irradiation. it is not so for other cancers. At present, the time distribtion of their latent periods is not known in detail. and for some cancers it depends pon dose (or dose rate for chronic exposre), and also on age at irradiation. In addition. several. if not all, radiation-indced tmors tend to have an age distribtion similar to that of their spontaneos conterparts. This means that greater absolte incidence per nit dose will be observed in cohorts of older than of yonger ages. On the other hand. the expression of indced cancers will be trncated in cohorts irradiated at older ages, owing to their limited srvival. Ths. the shape of the dose-response crve may depend pon the length of follow-p and on the age strctre of the irradiated poplation. In most stdies, the mean follow-p period is sbstantially shorter than the time needed for fll tmor expression. In addition, a negative correlation between dose and latent period cold. if present. affect the shape of the dose-response relationship, becase it cold not permit comparable tmor expression after high and low dose; The sex and age composition of the poplation nder stdy. Since for certain tmors the age at irradiation and sex have a prononced effect pon the risk of later development of the malignancy, a given dose-response relationship (c) (d) (e) (f) (g) (h) may not be representative of poplations of different composition: There is a prononced geographical. socioeconomic and ethnic variation of the spontaneos incidence of cancers in most organs. (For a review of this point see [D 17, W9].) This sggests that epidemiological observations on radiationindced cancer cannot be applied indiscriminately to poplations of different ethnic. socio-economic or geographic characteristics. On the other hand. sch differences are not necessarily reflected in the vale of the absolte risk or the shape of the dose-response relationship. For instance. in spite of large differences in the age-specific incidence. the age-corrected excess risk of breast cancer and dose-response crves in Japanese atomic bomb srvivors are very similar to those in women irradiated in the United States for medical reasons; Deficiencies of tmor ascertainment in retrospective stdies from available records of incidence or specific mortality: Difficlties in the selection of sitable comparison grops for the calclation of the expected (control) incidence or mortality de to a given tmor; Presence of confonding variables and modifying factors (promoters. inhibitors) that. if correlated with dose, or per se. cold modify the shape of the dose-response relation: Qestionable accracy of dosimetric estimates. particlarly when these involve retrospective reconstrctions of complex sitations. Nmeros examples of sch ncertainties are given in chapter V. In this category belongs also the very narrow range of doses to which a poplation may have been exposed. as well as the nonniformity of dose distribtion in the target organs. The latter condition may distort the relationship when the mean organ dose is sed as the independent variable and indction is not a first-order process; Statistical ncertainties in the estimates of incidence or mortality at given dose levels. 15. Becase of sch limitations, in view of the presence of nmeros and often nknown biological variables affecting cancer incidence. and of the lack of nderstanding of the pathogenesis of cancer, UNSCEAR cationed against the indiscriminate se of risk estimates nder conditions other than those for which they had been derived. For example, direct application of the risk coefficients to doses in the range from I mgy to. I Gy involves a procedre of linear extrapolation, i.e.. an assmption that the incidence per nit dose does not vary with dose. Sch a procedre cold, however. lead to over- or nderestimates of risk. depending on the actal shape of the dose-response crve. It was generally conclded, in the 1977 report, that the real risk per nit dose of low LET radiation at low doses and/or dose rates wold be nlikely to be higher, bt cold be sbstantially lower. than the vales derived for the range of a few tens of mgy. The derivation was based essentially on 168

7 observations made above I Gy, bt some redction of the effect at low doses was already assmed (e.g., for lekaemia by a factor of abot 2). 16. The v.'ide confidence limits on the data from man allow varios mathematical fnctions to be fitted to the same epidemiological series [B 17. B2, B24, C29, M31, R21, SSO]. Conseqently. the probability of being able to discriminate between the statistical goodness of fit of varios alternatives. or to reject some of them. is too low. In addition, the extrapolation of a relationship beyond the region of direct observation is always qestionable when the nderlying mechanism is not well nderstood. 17. In view of all these difficlties UNSCEAR has followed another approach in preparing the present annex. It has reviewed evidence at the sbcelllar and celllar level. from which inferences cold be made as to the possible natre of the dose-response for cancer initiation by radiation. It has also examined how initiation of canceros clones. and their progression to clinical tmors. may affect the shape of the doseresponse relationship. Finally, it has reviewed pblished models of cancer indction and tested them for compatibility with epidemiological and experimental findings. It is hoped that this complex exercise may help to establish, with some confidence, the shape of dose-response relationships, and thereby limit the ncertainty in the extrapolation of the risk to low doses. 18. Ths. the objectives of this annex may be smmarized as follows: (a) (b) (c) (d) (e) To review the critical assmptions involved in the formlation of models linking radiationindced cancer to dose; To review and discss dose-response re la tionshi ps for effects at the celllar level that cold conceptally be linked with malignant transformation: To discss models of cancer indction by radiation from the standpoint of reslting doseresponse relationships for tmors of some organs and tisses; To review the effects of the mode of dose delivery (dose rate, fractionation) and qality of radiation pon the dose-response relationships; To identify possible general trends. and interspecies similarities. broght abot by changes in the above variables pon the dose-response relationships for varios types of cancer. UNSCEAR wishes to stress that in prsing this exercise it does not intend to give more weight or credit to one or another model of tmor indction, nor to depart from previosly established policies in risk estimation adopted within the Committee. This review is meant to be a prely scientific analysis of data aiming at an assessment of systematic errors in the risk estimates derived from existing epidemiological evidence when one or another model is assmed in interpreting sch evidence. I. DOSE-RESPONSE RELATIONSHIPS FOR RADIATION-INDUCED CANCER 19. Radiation-indced cancer, as a stochastic phenomenon. can be analysed in terms of probabilistic concepts sch as the distribtion fnction of the time from irradiation to the occrrence of cancer. Depending on the data, the event may be assmed to occr at the time of diagnosis for readily apparent tmors, or at the time of death for rapidly lethal ones. The definitions and procedres of estimation are similar, bt the two cases shold not be confsed becase this cold lead to serios errors. particlarly v.'ith cancers for which effective therapy is available. 2. In experimental work. the poplations nder stdy consist generally of inbred animals standardized for species, strain. sex and age. They are irradiated nder controlled conditions and followed for a specified time or p to death. Appropriately matched control grops are followed concrrently nder similar conditions. The time of death and, at least in some experiments. the case of death, can be ascertained for each animal. Sch data may ndergo sophisticated statistical treatments. 21. Easily diagnosed or rapidly lethal tmors are readily discovered. For sch '"manifest' neoplasms, established mathematical procedres can be applied to correct for competing risks. e.g., intercrrent mortality not related to the tmor incidence. Under these conditions. the time to the expression of the tmor is known only for some of the individals in the collective, while others die or disappear from observation de to nrelated cases before a tmor is observed. For these latter individals one knows only that the hypothetical time to the tmor wold have been longer than the observation time. i.e., it wold lie to the right on the time scale. Hence one speaks of "right censored" data. 22. If tmors are "occlt", in the sense that they are discovered only incidentally in animals killed or dying for other reasons, one speaks of "doble censored" data. In this case, one knows either that the time to the expression of the tmor is shorter than the observed time of death or that the hypothetical time wold be longer than the time of death, according to whether the dead animals carry a tmor, or not. Under these conditions the expression "doble censored data" is sed (meaning that the data are both "left and right censored"') and the methods for a competing-risk-corrected analysis are more complex (see paragraph 31). There are special difficlties for partly lethal tmors. bt a for-point grading of the tmors is sally practicable and sfficient for the analysis: it ranges from "definitively incidental" to "definitively manifest (e.g., lethal)" [Pl8. Pl9]. 23. In epidemiological work on hman poplations, the sitation is qite different. The series are, in most cases, retrospective: the final data on morbidity and mortality are freqently incomplete: and the composition of the grop is often heterogeneos with respect 169

8 to sex, age. socio-economic stats, health conditions and exposre to carcinogenic or promoting agents other than radiation. Also. the control poplation is seldom flly adeqate; follow-p to extinction is rarely achieved owing to the long life span of man; and dosimetry is freqently ncertain. The statistical treatment of sch data mst obviosly follow methods different from those applying to prospective experiments. A. THE INDEPENDENT VARIABLE 24. Dose-response crves are fnctional relationships between an independent variable, the radiation dose in a given organ or tisse. and a dependent variable represented by a sitable measre of the response. The specific energy, z, absorbed in a cell or in its critical strctres, is a random variable. The mean vale of z, i.e., the absorbed dose, D, is commonly sed as the qantity of reference. bt at eqal vales of absorbed dose the distribtion of the vales of specific energy can vary greatly, depending on the tisse volme for which the specific energy is determined and the vale of the absorbed dose. as well as the radiation qality (see 111.B.2). Frthermore. the same dose may be delivered at different dose rates. Jn the present context. the following terminology will be adopted for sparsely-ionizing radiation: low doses,<.2 Gy: intermediate doses,.2-2. Gy; and high doses. > 2. Gy. For densely-ionizing radiation (e.g., fast netrons) doses <.5 and >.5 Gy will be referred to as low or high. respectively. with intermediate doses falling between the figres qoted. Low dose rates for all radiations are<.5 mgy min- 1 : high dose rates are >.5 Gy min- 1 ; and intermediate dose rates fall between these limits. Other qantities will at times be sed as the independent variable, sch as the injected activity of a specified radionclide, or the timeintegrated concentration of alpha-energy of shortlived radon daghters ltimately to be released in air. With some oscillations. sch qantities are proportional to dose. B. THE DEPENDENT VARIABLES IN EXPERIMENT AL WORK 25. In experimental work on radiation carcinogenesis, varios expressions of the response may be adopted (see annex I in [U6]). The simplest. and most commonly sed. is the fraction of animals incrring a tmor after irradiation with a given dose (crde incidence). It has been stressed repeatedly [FI, GI7-G19. Hl5, M32, R9. S37, U2-U5, U2-U22] that sch way of expressing the response leads to erroneos reslts. The reason is the interference of competing risks and of the different dration of life between animals receiving different doses. Actally, animals receiving the highest doses tend to die earlier and ths have less chance of expressing the tmors that may be indced. 26. Corrections for differences in the distribtion of srvival times between control and irradiated animals may be made by approximate methods, as, for example, in stdies by Ullrich and Storer [U2-U5. U2, U21, U23-U26]. In this approach, the data are trncated at the time when the grop is extingished throgh natral death and the observed incidence in the treated grop is corrected by a factor eqal to the ratio of the mean lifetime for the control and the irradiated grops. This approach can provide approximate age corrections, bt it may be misleading when the freqency of tmor appearance varies considerably with time after exposre. 27. Rigoros corrections for age and intercrrent mortality may be made by following the response of irradiated and control individals throghot their life after irradiation or dring a pre-selected post-irradiation period, with appropriate methods of investigation, inclding carefl post-mortem pathology. The relevant parameter is then the age- or time-dependent rate of tmor appearance [Cl8, CI9. HIS. K8, S37] or a related cmlative qantity that can be more readily determined in the experiment. The basic qantities in this approach and their competing-risk-corrected estimates for manifest tmors are: (a) The tmor rate, r(t), as a fnction of age or time (t) after irradiation. It is the probability at time t per individal to develop a tmor per nit time. This qantity. r(t). is to be interpreted as a mean vale for the poplation nder stdy. Since, for tmors diagnosed dring the lifetime. the actal time of origin of the tmor is nknown, the time of its first observation is generally sed: the time of death is sed for rapidly developing, lethal rmors. In experimental work one derives r(t) as an average vale in a grop of animals at time t. If N animals are observed (i.e.. are at risk) over the interval t - A t/2 to t + LI t/2, and n tmors appear within the interval, the estimate of the tmor rate is f(t) = Lln/NAt. For incidentally observed tmors a direct estimate of the tmor rate is impossible: the tmor prevalence can, however, be estimated (see paragraph 31): (b) The cmlative tmor rate, R(t). Estimates of this qantity are less affected by statistical flctations and are therefore more readily derived. The qantity is defined as the integral of the tmor rate from the time of exposre (t = ) p to time t: I R(t) = f r(t) dt ( 1.1) R(t) is the nmber of tmors per animal p to time t nder the hypothetical condition that one cold keep the nmber of animals at risk constant in spite of intercrrent mortality and the occrrence of tmors. R(t) exceeds, therefore, not only the crde incidence, bt also the incidence I(t), corrected for competing risks (see paragraphs 29 and 3). A competing-risk-corrected estimate of the integral tmor rate is [A I, N3, S37]: R(t) =.? (n/n;lf t) At=.? n/ni ( 1.2) I I for all i with iat < t. where ni is the nmber of tmors appearing within the time interval (i - I )At to ia t, and Ni is the actal nmber of 17

9 individals still at risk at this time, i.e., individals still withot a tmor. The standard error for eqation ( 1.2) can be obtained by the relationship [S37] O"R(<) = [/ (J; n/nr) I ( 1.3) 28. If mltiple non-lethal tmors occr. estimates of the tmor rate or the integral tmor rate can be based also on all observed tmors [S37]. In this case all animals still at risk are inclded in N; (eqation 1.2), regardless of whether these animals had developed a tmor or not. With this modification. similar estimates are obtained. provided that the animals withot previos tmor had experienced the same tmor rate as the animals that had already incrred a tmor. This is so becase both the nmerator and the denominator in eqation ( 1.2) are increased. If, on the other hand, there are inherent variations of the tmor rate within a poplation, or if the occrrence of a tmor increases the probability of sbseqent tmors, the tmor rate estimated from all tmors will be larger than the rate estimated from first tmors only. It is mandatory, therefore, to specify whether the estimates of the integral tmor rate are based on the first or on all observed tmors. For partly lethal or rapidly developing lethal tmors the estimate can be based only on first tmors. 29. A freqently sed qantity, related to the cmlative tmor rate. is the actarial incidence. or incidence corrected for competing risks, l(t). It is the probability of an animal at risk p to time t to have incrred a tmor. In the absence of competing risks, the actarial incidence eqals the crde incidence (see paragraph 25). In the presence of competing risks. and for manifest tmors, a qantity can be obtained in terms of the prodct limit estimate [Kl]: I(t) = l -J7p - n/n;] ( 1.4) where the prodct extends over a nmber of time intervals (i) p to time t: n; is the nmber of animals with tmors appearing within the time interval t; t;; and N; is the nmber of animals withot tmors still at risk at time t;. The standard error of the prodct limit estimate is expressed by the so-called Greenwood formla: a-l(t) = [l - I(t)] l/ J; n/nr ti::::; l (1.5) I When N is very small the log-rank test is preferable (see paragraph 33). 3. If few individals incr the tmor. the actarial incidence and the integral tmor rate. based on first tmors only. are nearly eqal. At high freqencies. the actarial incidence can approach l, and the integral tmor rate may exceed I. The sm limit estimate (eqation 1.2) is largely eqivalent to the prodct limit estimate (eqation 1.4), i.e., the integral tmor rate can also be obtained from the prodct limit estimate by the relationship: R(t) = -In [l - I(t)] (l.6) Similarly, the actarial incidence can be obtained from the sm limit estimate by the relationship: I(t) = 1 - exp [-R(t)] (1.7) 31. For occlt tmors (see paragraph 22). which freqently occr in short-lived animals. the actarial incidence (which is then sally called prevalence) or the integral tmor rate are more difficlt to estimate. Theoretical analyses have shown that a combination of serial killing and srvival data is reqired for sch estimates in the case of tmors with nknown degree of lethality or life shortening [C36. M22, N4, R8]. If occlt tmors are definitely non-lethal, estimates can be obtained by serial sacrifices at specified times after irradiation. However. this approach reqires large nmbers of animals. As Hoel and Walbrg have pointed ot [H29]. the method of isotonic regression (see also [B84. K38]) may be sed to estimate the competing-risk-corrected incidence from srvival experiments. This provides a maximm likelihood soltion with the constraint of monotonicity of the incidence. The algorithm for isotonic regression is straightforward and has been tilized for the analysis of radiation carcinogenesis [C36]. At present. however, there are no methods to derive standard errors. 32. In most experimental stdies in which tmors are seen in varios organs of the same animals, it is sally assmed that sch tmors occr independently of each other. However, Storer has shown [S53] that this is not necessarilv so. In irradiated female BALB/c mice. 21 ot of th~ 66 pairs of tmors tested showed significant positive or negative correlations. Some of the negative associations were de to rapid lethality cased by one of the tmors, and this cold be corrected for by appropriate methods allowing for intercrrent mortality. Of the remaining 13 significant correlations. 6 involved tmors known to be endocrine-related. and 7 applied to tmors of other organs. Alterations in host factors were believed to be responsible for the observed associations. These possible complications shold be borne in mind in the analysis of dose-response relationships on the assmption of random, i.e.. independent. tmor occrrence. 33. The Jogrank test. the Breslow test. or the wider class of non-parametric generalized rank-sm tests are sitable for the comparison of tmor rates in two or more grops in the case of manifest tmors [K37]. Analogos tests do not exist for doble censored data from srvival experiments. i.e.. for tmors fond incidentally. With sch data. one mst se tests based on the assmption of the eqality of competing risks in the two grops nder comparison, or one reqires knowledge of the degree of difference of competing risks. For experiments with serial killing, sitable standard tests exist. 34. The qantities discssed in paragraphs are not based on specific models. In experiments where varios grops. exposed to different doses. are compared. estimates may be sed that are also nonparametric bt are based on models. Most freqently. the proportional hazards model is sed. This model is based on the assmption that the tmor rate or the integral tmor rate in non-irradiated animals is increased by a dose-dependent factor: r(t,d) = },(D) r (t) or R(t,D) = }.(D) Ro(t) (1.8) 171

10 where ro(t) and R (t) are the tmor rates and integral tmor rates for the non-irradiated animals, and r{t,d) and R(t,D) are the tmor rates for the individals exposed to dose D. By eqations (I.I) and ( 1.6) one cold express this model in terms of the actarial incidence. I{t). However, sch expression wold be more complicated. The reason is that tmor rates and cmlative tmor rates from independent cases are additive. while the incidence is additive only when its vale is small. For manifest tmors. there is a relatively straightforward algorithm for calclation of the proportional hazard coefficients. l(d), employing the method of partial likelihood [K37]. For incidentally observed tmors, one mst make se of more complex methods. reqiring compter algorithms for non-linear optimization, with the constraint of monotonicity [C36. K38]. The analysis can also be based on the model of accelerated failre times by se of the non-linear optimization methods. This model assmes that the competing-risk-corrected incidence. I(t.D). rises earlier in a way that can be described by a dose-dependent acceleration of the incidence Io(t) for the control grops: I(t,D) = lo [a(d)t] and R(t.D) = R [a(d)t] ( 1.9) A similar model is that of time shift [C36], which assmes that the tmor rates, the integral tmor rates, or the incidence may attain the same vales at earlier times, in a manner that can be described by a forward shift in time: l(t.d) = I [t+s(d)] and R(t.d) = R {t +s(d)] (1.1) With both the accelerated time or the time shift models, algorithms for non-linear optimization are reqired for either manifest or incidentally fond tmors. 35. As a frther step towards the derivation of coherent time and dose dependencies, parametric models have been sed. Particlarly important among these is the so-called Weibll model [K37], which postlates tmor rates. and integral tmor rates, increasing as a power of time: r(t) = c tp and R(t)=ctP+ 1 /(p+ l) (I.II) The coefficient c is assmed to depend on dose, while the exponent p may or may not be treated as a parameter that varies with dose. The Weibll model is a special case both of the proportional hazards model and of the accelerated time model. Another freqently sed model envisages a log-normal distribtion of the times to the tmor. i.e., of a competing-risk-corrected incidence that depends on time as a log-normal sm distribtion. Varios other models. for example the logistic model, have also been tilized. 36. The preceding paragraphs refer only to acte irradiation. In case of continos or fractionated long-term exposre. additional complexities are introdced. Under sch conditions. the dose increases with time and it may be difficlt to identify the relevant vale of the accmlated dose. The process of cancer indction is followed by a period of growth ntil the tmor becomes observable. The dose absorbed dring this period is not relevant to the appearance of the tmor. Corrections may therefore be applied by sbtracting from the total dose the portion received after the presmed onset of neoplastic growth. C. THE DEPENDENT VARIABLES IN EPIDEMIOLOGICAL STUDIES 37. Whereas experimental stdies se inbred animals that are niform as regards sex. age at exposre and other conditions. no comparable niformity is ever encontered in epidemiological hman stdies. Moreover. varios hman poplations are often sbject to a spectrm of inflences, of which onlv some are known or acconted for. In an ideal ca~e. a mltivariate analysis shold be sed to assess the relative importance of factors other than radiation. As this is seldom possible. Jess rigoros analyses mst freqently be accepted which, in addition to the basic qantities previosly discssed, se other, somewhat crder ones. In epidemiological investigations, the follow-p may start shortly after irradiation (prospective stdies) or at later times when some or all of the expected tmors may have occrred (retrospective stdies). Reliable data collection is more easily achieved in the first case. bt the majority of epidemiological stdies arc based on retrospective analyses. 38. It is not the objective of this annex to deal in detail with all factors and variables affecting the accracy of risk assessment of radiogenic cancers. However, for the nderstanding of the following text it is necessary to discss briefly the expression of the response in absolte and relative terms. The qestion of risk projection beyond the period of direct observation is intimately linked to the se of so-called absolte and relative risk projection models. These will not be discssed in detail in this annex. 39. As mentioned above. the occrrence of radiationindced neoplasms may depend on time after irradiation and on the absorbed dose in a variety of ways. In epidemiological investigations it is possible, ideally. to envisage two different sitations freqently referred to as the absolte or the relative risk model. (a) (b) The radiation-indced excess tmor rate-or the incidence rate, as more freqently determined in epidemiological stdies-after a latent period increases independently of the spontaneos incidence bt as a fnction of absorbed dose, i.e.. the spontaneos and indced rates are additive. Panel (a) of Figre II illstrates this case schematically; The excess tmor rate. or the net incidence rate. is proportional to the spontaneos age-specific incidence rate. i.e.. the dose reslts in a mltiplicative effect of the spontaneos tmor rate over the life span. This relative risk model. corresponding to the proportional hazards 111odel described in paragraph 34. is shown in panel (b) of Figre II [C29]. 172

11 Lifetime Expression, Comparison of Absolte and Relative Risk Models w z w z a. Absolte Risk x. Incidence after Irradiation r 'I r; 'I ;1,- ~ / Spen11,,..,, / Incidence b. Relative Rilk / x, + e 9 x. ~ / '// / I I I I I.,,,.- 9 Limited Expression Time, Relative Risk Model, Comparison of Two Age at Exposre Grops w z w z c. Irradiation at Yonger Age Radiogenic Exce / I I I I,,..- d. Irradiation at Older Age Radiogenic Excess x, X + 2 t AGE' 9 x, AGE' X + e 9 xe is ag-e at expowra, Q t1 1ht minim1r laten1 period. Figre II. Radiation-Indced cancer Incidence sperimposed on spontaneos cancer Incidence, by age. Relative and absolte risk models. [C29) 4. For most hman tmors. the spontaneos tmor rate is a steep fnction of age. bt qantitatively this dependence varies. For radiation-indced tmors with relatively short latent periods and fll expression within a short interval (lekaemia and bone sarcoma) one may distingish between absolte or relative risk models. A simple absolte model will apply when irradiation at any age is followed by a temporary increase of the tmor rate that does not depend on the magnitde of the spontaneos agespecific rate. A relative model will be preferable when the temporary increase of the tmor rate becomes larger with increasing age at exposre. The latter case is illstrated in panels (c) and (d) of Figre II. 41. The relative risk projection model implies that the absolte (attribtable) risk increases with age; on the other hand, the absolte risk model implies that the excess risk when related to the spontaneos incidence rate, decreases with age. Actally, epidemiological data often show an intermediate pattern. and a decision as to which. model is most appropriate for risk estimation is not always possible [C29, Tl2]. However. recent evalations of cancer incidences in atomic bomb srvivors lend rather strong spport to the relative model. This is so becase. for the same age cohort, excess deaths from cancer other than lekaemia increase with age at death in proportion to the agespecific death rate from these cancers in the poplation of Japan [K39, S59]. A similar conclsion was reached when the time pattern of appearance of second cancers, presmably radiation-indced, was stdied in women treated by radiation for carcinoma of the cervix [B93]. Constancy of the relative risk of lng cancer with time after irradiation was also noted in Swedish iron-ore miners [R41] and in United States of America and Czechoslovak ranim miners [15]. The two projection models have been sed for prediction of the nmbers of varios tmor types in a poplation having an age distribtion similar to the one in the United States, assming a series of dose-response relationships and applying the appropriate corrections for intercrrent mortality by the life-table techniqe [C29]. When considering the risk of radiation-indced cancer, whatever the model applied. it is being crrently assmed that the distribtion of the sensitivity in hman poplation is nimodal althogh the character of the distribtion is virtally nknown. The qestion of the possible exceptionally elevated ssceptibility of some individals is discssed in paragraphs

12 1. The risk expressed in absolte terms 42. The tmor response of an irradiated poplation may be characterized by the time average net incidence rate, i r,,,, which is defined as the nmber of additional tmors diagnosed per person-year (PY) at risk. This qantity shold reflect a net increase above the spontaneos incidence rate in sitably matched controls. It is calclated from the formla ita = X/P- J/Q ( 1.12) where X is the nmber of persons with a diagnosed tmor in the exposed grop: P is the nmber of person-years in this same grop, obtained by smming the nmber of years at risk for all individals: J is the nmber of persons with the same type of tmor in a control grop that is matched or corrected for sex, age and calendar years: and Q is the nmber of person-years in the control grop. 43. The period at risk for an irradiated sbject is the time. sally in years, from irradiation to cancer diagnosis, or death. or to loss from observation, or termination of the srvey. For cancers other than lekaemia and bone sarcoma, an assmed minimm latency is sally sbtracted from the time at risk. The vale of ita is given as "cancer cases/person-year at risk", or more generally as the annal probability of occrrence of a specified cancer in a given poplation. The dose, D, will commonly be an average of the different doses that the members of the poplation have received. 44. If a sitable control grop is not available. ageand sex-specific incidence rates of tmors in the general poplation may be sed. Under these conditions, however. rmor ascertainment in the grops may not be flly comparable, and possible selection by conditions that prompted irradiation, or by other co-variates, may reslt in gross errors. 45. In order to obtain the time average net incidence, ITA- the vale obtained from eqation (1.12) is mltiplied by the average time at risk, i.e., by the average period of observation or, for predictions beyond follow-p, by an assmed time for fll expression of the malignancy. For lekaemia and bone sarcoma, this time is known not to exceed significantly 3 years (see chapter V). For other radiation-indced tmors. the period of expression is nknown. In order to derive meaningfl projections, a correction for intercrrent!'1nality becomes necessary, i.e., average life expectancies have to be sed. When the distribtion of the poplation by age and sex is known. life tables can be sed to this prpose. as in [C29]. As the qantities ita and ITA can be affected by significant errors, de to several circmstances, they mst be sed with cation. First, as the latent period is not known precisely, the correction for it can only be, at best, an approximation. Secondly (and apart from the above correction) it is nrealistic to assme constancy of the incidence rate over time; if the observation periods for two collectives exposed to different doses do not coincide. then the observations cannot be strictly comparable. 46. Annal risk coefficients for radiation-indced cancer can be expressed in terms of the time average net incidence rate per nit dose, FTA, that is, the probability per nit time per nit dose per person that a tmor can be seen in the observation period. This risk may depend on sex, age at irradiation, genetic disposition. the organ exposed, and a variety of other factors. The qantity is obtained by dividing ita by the mean absorbed dose received in the exposed grop. Nmerical vales are commonly given in "cases 1-6 a- Gy-1". 47. In order to obtain the life-time risk coefficient (net incidence per nit dose), FTA, the qantity FTA is mltiplied by the time at risk, i.e.. by an assmed average time of fll expression of the malignancy. The considerations in paragraph 45 that refer to the qantity ITA apply also to FTA 48. When the risk coefficients FrA and FrA are calclated for grops exposed to different average doses, changes of these qantities with increasing dose may provide approximate information abot the shape of the dose-response crve. Ths, constancy of the coefficient indicates proportionality between response and dose within the range of doses stdied; a rise or a decrease with increasing dose may reslt from an pward or downward concavity of the dose-response crve. However. if the radiation dose is correlated with a variable that affects the response (e.g.. age) sprios effects may be seen and normalization procedres are called for. Frthermore, if the doseincidence relationship is not linear, expressing the observations in terms of a probability per nit dose distorts the data and introdces additional inaccracies if individal doses within the exposed poplation deviate sbstantially from the average dose to the poplation. For instance, if the dose relationship contains a dose-sqared component, the contribtion to the response by individals with high doses is greater than if linearity applies. 2. The risk expressed in relative terms 49. When the risk is expressed in relative terms, the response is related to the risk of spontaneos cancers (incidence or mortality) in an nirradiated control poplation. The response variables sed are the standardized mortality ratio (SMR), defined as the ratio of mortalities in the exposed grop to the mortality in a control grop mltiplied by 1; and/or the relative risk (RR). which is the ratio of risks observed over expected (each expressed, for instance, per 1-6 person and over a given follow-p time). The latter qantity may be applied both to the incidence and mortality indices. The proportional hazard model specifies that if ).i is the incidence rate of a disease, or the mortality rate for a specific case of death among sbjects in dose grop i, it can be expressed as),;= Aw (RR;), where ).io is the backgrond or spontaneos rate of mortality or incidence (i.e., the rate experienced by sbjects in the absence of exposre to radiation) while RR 1 is the relative risk associated with dose grop i. The starting point for investigation of a radiation dose-response is the fnction 174

13 RR; = l + yd;. where y represents the excess relative risk coefficient ((RR -1) per nit dose). Variation of y with dose provides information as to the form of the RR-dose relationship. 5. Varios statistical methods. inclding mltiple regression analysis. have been sed to obtain doseresponse relationships, and are presented and discssed in detail in nmeros pblications, e.g.. [G36, J8. K42, S6. W 17]. Carefl corrections for sex, age at irradiation. attained age at observation and other confonding variables (e.g., ethnicity, exposre to other carcinogens and promoters) are necessary when the shape of the dose-response crve is the object of the stdy. D. TEMPORAL RELATIONSHIPS 51. As pointed ot in preceding paragraphs, the tmor rate, the cmlative tmor rate and the actarial incidence depend on time after irradiation and on other factors, sch as age at exposre or absorbed dose. It has already been mentioned that the absorbed dose may change the time dependence in a variety of ways. It is necessary, therefore, to consider temporal relations in somewhat more detail. I. Latent period 52. The latent period is the time between irradiation (e.g.. single acte exposre) and manifestation of a tmor. It may be divided conceptally into tre latency (the time reqired from initiation to the beginning of nrestrained growth) and tmor growth (time ntil the neoplasm can be diagnosed). Reported data are sally latent periods, i.e., the sm of the above times. The length of tre latency may be obtained by sbtracting an estimated time of tmor growth from the observed latent periods. There are considerable differences between the latent periods of varios tmors. Two malignancies in man. lekaemia and bone sarcoma, appear to have relatively short latencies and show an pper limit of the latent periods. so that their fll distribtion in a poplation having a normal age strctre can be observed. For other tmors the distribtion of the latent periods is nknown. becase it extends to very long times; the distribtions are then sally trncated by the length of the observation period. Median times recorded so far are of the order of 2 to 3 years, bt they shold. increase with extension of the follow-p of the respective grops. 53. If the follow-p is shorter than tht; minimm latent period. radiation-indced tmors cannot be expected. For more extended stdies an inferred minimm latent period is freqently sbtracted from the total time at risk. as mentioned before. The minimm latent period is often identified with the time after irradiation at which the incidence rate becomes statistically significant at a given level of confidence above the control vales. Figre III shows. however. that, for statistical reasons. the minimm latent period so estimated mst vary with dose even Q,_...,- cc c:: c:: is :c ::,... C :; :.; "' "',_ ~ C QI 4J "',_ QI :::,... - "' Q - C: C: e...,e Q...J TIME AFTER SINGLE ACUTE IRRADIATION Figre Ill. Presmed effect ol the magnitde of dose pon estimate ol minimm latent period. Two doses (D 1 < D:z) prodce different absolte effects, which have similar time distribtions ol the latent periods. Shaded area: spontaneos (control) Incidence rate; Broken line: pper confidence limit ol the spontaneos Incidence rate. 175

14 nder the assmption of the proportional hazard model, i.e.. when the tmor rate is increased by a constant factor withot change of the nderlying temporal dependence. Similarly, shorter minimm latent periods are estimated if the nmber of individals in the observed poplation is larger, becase predetermined levels of statistical significance are reached earlier in a larger sample. Other statistical procedres have been developed to analyse the dependence of the latent period on dose [L32]. 54. For some cancers. the minimm observable latend periods may change with age at irradiation. as sggested. for example. by stdies of hman breast carcinoma [826, Ml 7, Sl6. TI2. T23] and lng cancer incidence in Japanese atomic bomb srvivors [K39]. These radiation-indced tmors appear with high freqency at ages when the spontaneos incidence is also relatively high. It follows, therefore, that latency is longer for cohorts irradiated in childhood and adolescence than for individals exposed in their twenties or thirties. If so. the follow-p to inclde a complete (i.e., life-time) expression of the effect mst vary with age at exposre for different cohorts. It is not known whether this observation can be applied in a general sense to all forms of radiation-indced tmors with long latent periods. 55. In cohorts of advanced age, the life expectancy may be shorter than the average latent periods, and only a fraction of indced neoplasms will be seen. The yonger the age at irradiation, the fller the expression of indced tmors. In the common case of neoplasms with a distribtion of latent periods extending significantly beyond the follow-p. the doseresponse relationship cannot be based on complete data. The estimate of ITA will then change with the dration of the observation period; however. for those cancers for which the relative risk model appears adeqate. the relative risk shold be a less ssceptible qantity in this respect. Analogos changes of the dose-response relationships have been shown in experimental work [S37]. Whenever the distribtion of latent periods varies with the absorbed dose. the form of the derived dose-response relationships will depend critically on the length of the observation period. At low doses, the minimm latent periods may exceed the average life span of the individals and it has been postlated that so-called practical thresholds may reslt [E3, M8, MI, R28]. This inverse relationship between latency and dose is most common in stdies of continos long-term irradiation. It was first described for bone sarcoma indced by long-lived isotopes [848. E3. E4, F8, M4, Rl8-R2]. The present evidence does not sbstantiate the existence of sch relations for other tmors commonly indced by radiation in man. for which sfficiently detailed data are available, namely cancer of the breast [K39, L22], of the thyroid [S38), and of the lng and stomach [K39. L32]: an inverse relationship between absorbed dose and latent period may exist for lekaemia [L24], bt has not been definitely proved. 56. All the above factors mst be borne in mind in any discssion of the accracy and reprodcibility of dose-response relationships. Except for tmors with short latencies. sch as lekaemia and osteosarcoma. there is generally a lack of data extending to follow-p periods long enogh for the dose-response relationship to be assessed with reasonable completeness. 2. Relationships between dose and tmor-free lifetime lost as a reslt of indced cancers 57. As mentioned above, varios malignancies may be characterized by different latent periods. For some malignancies and modes of irradiation the latent periods (mean, median) may vary with dose, the mean vales normally increasing with decreasing dose. A qantity that may provide a first approximation to a common denominator for assessment of the harm per indced neoplasm is the tmor-free lifetime lost [16]. The relationship between this qantity and dose is therefore of interest. Very few papers provide information on the sbject, bt this will be reviewed. when it is available. in relation to different types of tmor. E. SUMMARY AND CONCLUSIONS 58. Cancer, as a stochastic radiation effect, can be stdied for its dependence on dose and time after irradiation. Different degrees of precision are attainable in experimental and epidemiological stdies of radiation-indced tmors. The methods for the estimation of dose-response relationships reflect these differences. 59. The absorbed dose in the relevant tisse is the independent variable for dose-response stdies. The magnitde of the dose and its temporal pattern of delivery (dose rate. fractionation) may vary greatly. In the present context. low doses are taken to be those p to.2 Gy of low-let radiation. while those in the interval Gy are regarded as high doses. with the intermediate doses lying between these ranges. Low and high dose rates are taken to be below.5 mgy min- 1 and above.5 Gy min- 1 respectively. Intermediate rates are those in between these vales. For high-let radiation low and high doses are assmed to range p to.5 Gy and down to.5 Gy, respectively. Dose rates may be classified in the same way as for low-let radiation. In certain stdies, more readily measrable qantities that are proportional to absorbed dose-at least over some range-may be sed as the independent variable. 6. In experiments on radiation carcinogenesis, varios measres of the response may be sed. The crde incidence, ncorrected for mortality. is not infreqently applied, bt it is nsatisfactory becase it neglects differences in life span after different exposres: accordingly, it may be strongly affected by intercrrent mortality de to cases other than cancer. Competing risk corrected qantities sch as the tmor rate. r(t), the cmlative tmor rate. R(t), or the actarial incidence or prevalence, I(t,D). can be sed for more meaningfl expression of the response. These qantities are always dependent on the postirradiation time, and mst be reported accordingly. They are obtained from the observed time seqence of 176

15 tmor appearance and from the variable nmber of individals at risk dring the observation period. They may be calclated by appropriate algorithms that are different for manifest tmors. i.e., rapidly lethal or readily discovered tmors, and for incidentally observed tmors. i.e.. tmors that do not contribte to lethality and that are fond only at necropsy. Only a few experiments have so far been analysed according to the reqirements of the competing risk theory. and this circmstance has often redced the potentially sefl information from nmeros stdies. 61. In epidemiological srveys, the response is. in most cases, assessed retrospectively, bt less accracy can be achieved than in well-designed controlled experiments. The genetic composition, the distribtions of ages and sex. the forces of selection, and the followp periods are often different in irradiated and control grops and reqire appropriate adjstments. Frthermore. the doses to individal members of a cohort are sally sbject to sbstantial variation, making it necessary to grop exposed individals over broad dose intervals. Also, the less rigoros qantities sed in epidemiological stdies, ITA FTA, and their rates. are complex fnctions of time post-irradiation. The scoring of radiation-indced tmors depends therefore, in a complex way, on the age at irradiation. dration and completeness of follow-p, latent period, and life expectancy of the poplation. The dration of follow-p may especially inflence the magnitde of the response and the shape of the derived doseresponse relationship in terms of the absolte (attribtable) risk. If the tmors of interest show an increment of the relative risk per nit dose that is not strongly dependent on age at irradiation, the shape of the observed dose-response relationship is Jess ssceptible to dration of the follow-p. provided the minimm latent period has been exceeded. the nmber of tmors is sfficient for verification of the postlated models. and other relevant co-variables have been corrected for. 62. Hman tmors with relatively short latent periods. e.g., lekaemia and bone sarcoma, are not expected to appear after abot 3 years. For these tmors, the dose-response relationships in yong cohons can be based on complete manifestation of the effect. For other tmors. of longer and nknown latency, dose-response relationships mst be based on the incomplete information reslting from trncation of the response by the short follow-p. A possible dependence of latent periods on dose is of critical importance in determining the shape of the doseresponse crve. When mean latent periods decrease with increasing dose, the shape of the relation depends on the observation time and, in some cases. practical thresholds may apply. However. except for 2 26Raindced bone sarcoma. little information of this natre is available from data on man. 63. Projection of tmor incidence or incidence rate beyond the period of follow-p for tmors with Jong latent periods may be achieved by two alternative models: the absolte and the relative risk projection model. ln its simple version, the first is based on the assmption that the rates of spontaneos and of radiation-indced tmors are independent. so that the rate of radiation-indced tmors does not increase proportionally to the age-dependent rate of spontaneos incidence. The second model in its simplest form assmes that irradiation has a mltiplicative effect on the spontaneos tmor rate and that the rate of indced tmors is a dose-related mltiple of the spontaneos rate. Recent data from several epidemiological stdies have shown that, in many cases. the relative risk projection model provides a better fit to the data than the simple absolte risk model. For nmeros tmors. however, no firm conclsions can be drawn as yet. II. ASSUMPTIONS AND LIMITATIONS OF EXISTING MODELS A. BASIC PHENOMENA AND INFLUENCE OF MODIFYING FACTORS 64. Common featres of malignant cells are the nrestrained growth, the ability to infiltrate neighboring tisses. and often the capacity to give rise to metastases. ln a general sense. canceros cells transmit their malignant characteristics from one generation of cells to another. and therefore somatic mtation cold be one of the mechanisms involved as an nderlying factor [B71. C32, Fl4. K2, K29, M5, R31. S43]. On the other hand. epigenetic factors sch as dis-differentiation. expression of previosly sppressed genes, activation of oncogenes (see annex A to this report) or others [B71, K26, S43, R31, S55] may be envisaged, alternatively or in parallel. Continos growth throgh celllar divisions might reslt either from inactivation of operating or reglatory genes. or from the inability of malignant cells to receive or recognize divisionrestraining signals. whatever their natre might be. 65. Owing to poor nderstanding of the origin of cancer. qantitative predictions regarding the effects of ionizing radiation (dependence on dose. dose rate, qality of radiation) cannot be developed from basic principles. Formlation of simplified hypotheses nder the form of models is at present a practical step towards nderstanding basic isses. To this end. detailed knowledge of mechanisms of cancer initiation and promotion wold be valable. bt even this cannot be secred at present in a comprehensive form. Qantitative models of cancer indction have been reviewed recently by Whittemore [W8] and Parfenov [P2]. 66. Before reviewing selected models, it is necessary to discss the concept of mlti-stage development of cancer. Since the classical stdies of Berenblm et al. on chemically indced skin cancer in mice (discssed recently with regard to in vitro oncogenic transformation by Borek [B76]) this concept has gained wide acceptance. The sbject has been repeatedly reviewed [B71, C32, Dl8, Fl3. Fl6. F2. M5, R31, R42. S43]. It is commonly assmed that cancer indction is started by a phenomenon of initiation, occrring most likely in one cell, althogh participation or modifying inflences from neighboring cells cannot be exclded. Frther development into a canceros clone probably 177

16 reqires several stages. covered nder the concepts of promotion and progression. which reslt eventally in an overt malignancy. These concepts have been reviewed by the Committee in annex I of the 1977 report [U6]. 67. Celllar phenomena that take place dring development of a canceros clone are not well nderstood. Direct involvement of DNA in oncogenic cell transformation has been demonstrated [B 14, L33. L34]. Lesions affecting its strctre and fnction mst therefore be relevant. Somatic mtation(s)-sch as changes of the DNA base seqence: transposition. deletion, translocation and transfection (from one cell to another) of genetic material: enhanced instability of DNA, leading to facilitation of non-repairable DNA damage; indction of recombination between nclear and mitocondrial DNA; distrbances of DNA methylation: activation of oncogenes-may be involved. Bt some of these mechanisms are largely hypothetical [871, 883. F21. G33. R3J, S43, S45, S55. V2. V3]. Initiation is followed by a Jong latent period, dring which secondary inflences (systemic. exogenos) may inflence the final otcome of the process. Dring this period, the action of secondary factors may accelerate and stimlate (promote) or inhibit. or even reverse the process (anti-promoters, sppressors of malignant transformation). Argments have recently been presented which show rather convincingly that not all initiated cells, or their clones. will eventally give rise to diagnosable tmors [F22]. 68. Whatever the natre of initiation, be it radiationindced somatic mtation in the classic sense (indction coefficient of 1-s to 1-6 Gy- 1 of low-let radiation per locs at risk), or any other phenomenon similar to in vitro cell transformation (indction coefficient of 1-2 to 1-4 Gy- 1 ), the yield of recovered malignant tmors per irradiated cell in viva is lower by many orders of magnitde. The possibility of a symmetrical "tmor-specific" chromosomal translocation was advocated [F21] as a very rare event that wold remove part of this difficlty. However, recent stdies in vitro and in viva (chapter IV) sggest that initiation may not be a rare event when expressed per cell at risk. Therefore. if it is assmed that these processes operate in viva with comparable efficiency, even for a small fraction of the ssceptible cells - 4 to 1-6 ), a significant nmber of cells wold be transformed every day becase of backgrond irradiation alone [M31]. 69. The theory of seqential development of cancer may help to explain the discrepancy between data in,.;tro and in vivo, and the rarity of radiation-indced cancer. If a series of consective rare events is reqired for the initiated cell to become malignant. the final probability of emergence of the malignant clone will be the prodct of the probabilities of each event. Assming that only the first event (initiation) is radiation-indced, and that the sbseqent ones are rare phenomena, perhaps independent of irradiation. clinical cancer cold be many orders of magnitde less likely than initiation itself. Yet, clinical cancer cold show a relationship with dose similar to that of initiation, even thogh (owing to the mlti-stage natre of the process and possible systemic interferences) Jess reglarity wold be expected than that observed for stochastic radiation-indced phenomena in single cells nder controlled conditions. That this may actally be the case is sggested by the reprodcibility of dose-response relationships for some cancers in animals of selected strains. Also, spporting this contention is the fact that the absolte risk coefficient for cancer of the same organs derived from epidemiological observations of different poplations (see chapter V) are often similar. althogh there are exceptions to this rle (see. for example, Figre XXVIII). This may imply that the modifying inflence of secondary factors may not greatly change the final probability of indction and the reprodcibility of the dose-response relationships. It has been postlated that for some forms of malignancy, sch as lekaemias. radiation may be the casal factor. not for the first, bt for the last event in a series of rare phenomena [L24, K39]. In any case. initiation is a necessary step. althogh only a small fraction of initiated cells may develop into a viable clone and into a clinical tmor. 7. The principal aim of a model is to predict the shape of the incidence crves as a fnction of dose, dose rate, qality of radiation and possibly other factors. Ideally, a model for radiation-indced cancer cold provide some basis for the evalation of risk at low doses [D9. Fl2, Tl5] if it cold characterize the following processes: (a) (b) (c) (d) Probability of malignant transformation of a defined target cell. This will most likely vary with the organ, dose, qality of radiation. and temporal pattern of dose distribtion. In this respect, the model shold inclde theoretical concepts concerning the natre of the initiating event(s) and not be prely empirical. Mathematical formlations shold be compatible with general knowledge of radiation and cancer biology; Probability of interaction at the systemic level between transformed cells or between developing clones; Probability of cell killing with dose. as nonviable cells cannot give rise to canceros clones. This factor, as discssed in chapter Ill, depends certainly pon the tisses concerned. some of which are particlarly ssceptible, and pon dose, dose rate and qality of radiation and on other irradiation conditions (in vitro. in sit) [D7, GJ4, GI6, H8, HIO, M31. M52, R2, S26, S27. U6. U7, UI7]: Effect of nmeros host factors. discssed in annex I of the 1977 UNSCEAR report [U6]. Among promoting factors, cell division is perhaps a good example [B36, 885, M6. N6. P22]. Enhanced proliferation can be indced by varios mechanisms, inclding radiation-indced cell death. a phenomenon which is likely to prodce a threshold-type dose response. Other factors cold be the differentiation of initiated cells, immnological srveillance that sppresses development of malignant clones with antigenic characteristics foreign to the host-system. and hormonal inflences: 178

17 (e) Effect of exogenos promoting agents leading to more rapid appearance of tmors, to an increase in tmor yield. and perhaps also to changes in the shape of dose-response crves. as shown by several experiments [F21. L26] inclding those on in vitro oncogenic transformation [H22. L26]. 71. Immnological srveillance may be impaired by nmeros factors. inclding ionizing radiation. The degree of impairment wold certainly depend on dose. on whether the whole body or only part of it is irradiated, and then on what part and what proportion. It may be postlated that a redction of immnological capacity of the organism follows a threshold-type response to irradiation. The importance of the immnological sppression in the low and intermediate dose range remains to be established althogh recent evidence points to a limited role of the immne system. Stdies of Mole et al. [M52] on indction of acte myelogenos lekaemia in CBA/H mice, sggest that, at least in this tmor-host system. impairment of immnological srveillance is not very important. Similar conclsions were also reached by Nolibe et al. in respect to Jng cancers indced by 239P2 in rats [N9]. 72. Hormonal dependence is well known for some cancers [U6]. By partly or totally inactivating endocrine organs and creating a hormonal imbalance. radiation may accelerate or impair malignant growth in other organs (e.g.. breast, thyroid, ovary). Again, it is likely that this type of response wold show prononced dose thresholds. Hormonal inflences cold also be responsible for the association of varios tmors, as observed in experimental animals [S53]. 73. In smmary. the action of modifying factors might change the reslting dose-response relationships. It is a basic weakness of the models discssed in chapter III that mathematical fnctions relating cancer incidence to dose reflect only initiation and sterilization of the initiated cells. This is becase the precise effect of other modifying factors on tmor indction-as a fnction of dose-cannot yet be described with reasonable credibility. even if some attempts have been made [M4. P22]. Models, therefore, provide greatly simplified concepts. based on plasible assmptions and hypotheses. which may help in nderstanding the role of dose, dose rate and radiation qality. The relative importance of the parameters reflecting these qantities in a model can be dedced by comparing data with predictions. However, some of the parameters are the componded expression of several elementary phenomena that cannot be resolved at present. 74. From available models. risk coefficients for man in the range from abot I mgy p to abot.1 Gy can only be obtained by extrapolation from the recorded vales (in most cases above 1 Gy) to zero dose. Effects of the host factors discssed above (cell renewal, immnological srveillance, hormonal balance) and exogenos modifying factors pon dose-response relationships are ncertain, and only approximate evalations may be given. For the prposes of the present annex, the levels of acte doses of low-let radiation can be broadly separated as follows: (a) <.2 Gy. At this level, data for most tisses are not directly available, bt the damage to tisse fnctions is considered negligible. Therefore. dose-response relationships cold be very close to those postlated by models that disregard host factors; (b) (c).2-2. Gy. Damage to tisses, cell proliferation and inflence of other host factors wold not be expected to play a dominant role. Distortions in the dose-response relations shold be relatively minor. and wold not grossly bias the extrapolations; 2.-1 Gy. Damage to tisses. to immnological fnctions (for irradiation of the whole or most of the body) and to hormonal feed-back systems may play a significant role. Extrapolations to low doses mst therefore become more complex, particlarly from the pper end of this range. Great cation is needed to avoid extrapolating from data that are already strongly affected by killing of initiated cells. as this might reslt irt a sbstantial nder-estimate of the risk at low doses; (d) > 1 Gy. Severe damage to tisses and organs will dominate the response. Data in this dose region are most difficlt to evalate and so complex that extrapolation to low doses cold be meaningless. Indced cell repoplation may also become a complicating factor for irradiation times longer than a few days. A corresponding approximate classification of fast netron doses wold inclde the ranges: (a)<.5 Gy; (b).5-.2 Gy; (c).2-2. Gy: and (d) > 2. Gy. It shold be noted that the relation between these ranges of low- and high-let radiation may not be interpreted in terms of RBE. B. THE THRESHOLD DOSE FOR CANCER INDUCTION 75. It has already been mentioned that for many types of cancer the epidemiological and experimental information available does not sggest the presence of a threshold dose. If initiation is at least partly atonomos. i.e., if it affects a single or very few cells. present nderstanding of the mechanisms involved sggests that: (a) (b) The natre of radiation interaction with the celllar target for cancer initiation is stochastic, in the sense that each increment of dose carries a given probability of effective interaction; The error-free repair of the DNA, which is the most likely target involved, leaves some fraction of the damage nrepaired and the error-prone repair may prodce misrepaired seqences in the DNA strctre. As the prodct of two probabilities greater than zero cannot eqal zero, the presence of a threshold cannot be assmed. 179

18 76. A different sitation obtains when the lifetime lost per cancer case (instead of the incidence) is related to dose or dose rate. Few malignancies show a definite inverse relationship between dose rate and latent period (see III.B.2), particlarly at chronic exposre. In sch cases (e.g., bone sarcomas indced by 9Sr or long-lived radim isotopes) practical thresholds may exist below which no life shortening occrs and a very low probability of tmor appearance applies. C. UNICELLULAR VERSUS MULTICELLULAR ORIGIN OF NEOPLASTIC GROWTH 77. Under the simplifying assmption that doseresponse relationships for radiation-indced cancer reflect mainly the kinetics of initiation, they shold inclde both the dose-dependence of the relevant intracelllar phenomena and the interactions of individal (contigos) cells. if indeed the latter play any role in the process [G3]. Therefore. the qestion whether a malignant clone arises from transformation of a single cell or of several contigos cells is of critical importance for the formlation of models. Becase nderstanding of the biological process of carcinogenesis is limited, this qestion cannot be resolved neqivocally at present. However, most experimental and hman observations tend to spport-even if indirectly-the monocelllar origin of neoplasms. Frthermore, the alternative mlticelllar hypothesis allows the development of models reslting in sch a variety of dose-response relationships as to render the exercise of model-fitting totally meaningless. 78. All these problems were reviewed in depth by Fialkow [F3, F4, F6]. Nowell [N5] and others [B6. FI3, F14). It appears from their work that the monocelllar alternative is gaining increasing acceptance, except for some rare hereditary tmors in sbjects with inborn predispositions (e.g.. mltiple nero-fibromatosis [F5]) and other rare tmors. sch as condylomata acminata [FlO]. Three main lines of evidence spport the nicelllar hypothesis [F3. F6. N5]: (a) The isoenzyme pattern of glcose-6-phosphate dehydrogenase in canceros cells derived from tmors of heterozygos women, in accordance with the hypothesis that the same X-chromosome of the homologos pair is fnctioning in all cells of a given neoplasm. A similar pattern was observed recently in 3-methylcholantrene-indced fibrosarcomas in Pgk-P/Pgk-1 b mice carrying X-chromosome inactivation mosaicism for the phosphoglycerate kinase (PGK-1) gene [Tl7]: (b) The homogeneity of immnogloblins prodced by almost every case of myeloma; (c) The common cytogenetic markers freqently detected in the cells of a single tmor. This evidence is limited, indirect and compatible with the alternative hypothesis that selection of one clone from among several initially transformed cells might lead to an apparently monoclonal composition of the tmor. So far, however. this possibility has not been widely spported. D. AUTONOMY OF TARGET CELLS FOR TUMOUR INDUCTION 79. Conceptally, even if initiation is taking place in one cell, the mlti-stage natre of cancer development and possible inflences of tisse and systemic origin cold introdce random flctations into the expression of cancers in viva. For radiation-indced cancer. therefore, sch reglarity and reprodcibility of doseresponse relationships as that seen for stochastic radiobiological phenomena in cltred cells wold not be expected. Mch of the information from animal experiments shows nevertheless that dose-response relationships are reasonably reprodcible. Proportionality between initiation freqency and expression may therefore be assmed as a plasible hypothesis. 8. On the basis of microdosimetric considerations Rossi and Kellerer [Rl7], Rossi [Rl5, Rl6] and Rossi and Hall [R43] have pointed ot that the observed crvilinear dose-response relationships for netronindced mammary tmors in rats and other tmors in mice are inconsistent with atonomos response becase at low and intermediate doses there shold be a direct proportionality (linearity) between dose and tmor incidence. Therefore, if the dose response is in fact non-linear. this mst mean that either initiation or expression are modified by dose-dependent factors, implying a lack of "cell atonomy". If sch a conclsion shold generally apply, theories concerned with the response of single cells may not be relevant to dose-response relationships for cancer indction. 81. In the original experimental data of Shellabarger on Sprage-Dawley rats [S15), 14 months after an acte x-ray exposre, the incidence of all mammary tmors (adenomas pls carcinomas) per rat showed an approximately linear relationship with dose, p to.84 Gy, whereas.43 MeV netron irradiation, with doses from.1 to.64 Gy. reslted in a tmor incidence roghly proportional to the sqare root of the dose [Sl5]. Rossi and Kellerer arged [Rl7] that. since cell killing wold be negligible at sch low netron doses, the form of the observed relationship reqired the interplay of radiation effects in several cells. For mammary fibroadenomas, sch an interaction leads, apparently, to a redction of the tmor rate per nit dose even at netron doses of a few tens of mgy. 82. Earlier data by Shellabarger et al. [S 15) were later extended, with follow-p of the animals to abot 1 days [S37]. Tmors scored inclded fibroadenomas and adenocarcinomas. the latter malignant forms being 9-16% of the total. The data expressed as crde incidence or mortality-corrected indices, r(t,o), R(t.D), at varios post-irradiation times, essentially confirmed the previos findings. A very high RBE was seen at low netron doses for both tmor types (abot 1 at.1 mgy) and the dose-response for the combined tmors was concave downwards for netrons and approximately linear for x rays. There was also a significant shift of the age-specific r(t) cl've in the irradiated animals (acceleration). 18

19 83. When the net R(t) vales at 8 days (the latest follow-p time for all bt one dose grops) were considered separately for fibroadenomas and adenocarcinomas, the following conclsions cold be drawn: (a) (b) (c) For all tmors combined the downward concavity of the crve after irradiation with netrons reslts essentially from a low point for the fibroadenomas at.64 Gy. Sch an effect is not seen for the adenocarcinomas. where the relation is approximately linear, althogh the experimental scatter precldes accrate assessments in this case: At 8 days, the departre from linearity of R(t) for the fibroadenomas is mch less prononced than at 4 days [S 15] or than that for the incidence of adenocarcinoma [S37]; No accont was taken of cell killing in these experiments, althogh this may have not been negligible already at.64 Gy. 84. In view of their high spontaneos incidence (8% throghot the lifetime in females), the mammary tmors in Sprage-Dawley rats mst be regarded as an nsal model. This view is confirmed by the facts that only a minor part of these tmors is histologically malignant [Sl2, Sl3, S14, Sl5, Ul7. V4, V5] and that the very high sensitivity of the Sprage-Dawley rat to the indction of mammary tmors is not seen in other rat strains [B43. B92, C15. S7, Sl 1. VlO]. 85. In smmary. the shape of the dose-response relationship reported for netron- and x-ray-indced fibroadenomas at low doses in the mammary gland of Sprage-Dawley rats in one stdy is rather exceptional among all data on radiation-indced tmors. The low yield of adenomas per nit dose at.64 Gy, which has been interpreted to show an inhibitory inflence by the neighboring cells. may also be explained in terms of cell killing. Therefore, the observations discssed do not necessarily imply that malignant tmors arise from more than one initially transformed cell, or from interaction of several cells. 86. Other data on benign lng adenomas in RFM mice [U21, U22] may be discssed in this context. Pathogen-free animals were irradiated locally over the thorax with graded doses of fission netrons (.5 to 1.5 Gy) and the nmber of tmors in dissected lngs was conted 9 months after exposre. The effect was flly manifest at 6 months after.5 Gy, and intercrrent mortality was negligible. A peaked response, with a maximm arond.25 Gy, was seen following single netron doses. Splitting of the dose into two eqal fractions. given at 24-hor or 3-day intervals. reslted in no sparing effect of fractionation. Plotting the mean nmber of adenomas per animal at 9 months against dose gives a slightly pward concave crve in the range of O to.25 Gy. A qadratic model fits the data very well (P >.99) bt non-threshold linearity is not exclded (P >.8): also, a threshold of.5 Gy cannot be rejected, particlarly since the gamma-ray crve sggests sch a possibility. It is also possible that increasing doses of netrons p to.25 Gy may have an enhancing effect, interpreted either as acceleration or promotion. Sch an effect, however, is the opposite of that seen for mammary fibroadenomas in Sprage-Dawley rats. Reslts of other experiments with netron indction of varios tmors in BALB/c mice [U23] were also advocated as evidence that indction cannot be treated as a phenomenon related to atonomos cells. However, a correction for sterilization of initiated cells cold reslt in approximately linear dose-response crves p to.2 Gy. 87. There is yet another explanation for the above data, i.e., that single-cell initiation is modified by enhancing or inhibiting dose-related effects. The natre of these effects (direct or abscopal, celllar or hmoral) cannot be specified at present. Tmors reqiring hormonal stimlation for their development (as in the case of the ovary) often have threshold dose-responses. In essence, therefore, none of the data discssed are incompatible with the hypothesis that tmor initiation by radiation may take place in a single cell. which from this stand-point mst be considered relatively atonomos. 88. It has been pointed ot [G 15] that the simltaneos initiation of canceros growth in several interacting cells wold be eqivalent to a mlti-target event. If so, the sigmoid shape of the crve shold become more prononced with an increasing nmber of participating cells. There is, however. little data to spport sch a possibility and no known experimental system to test sch hypotheses. 89. It may be conclded that if the crve is not obviosly sigmoid or threshold-like pon visal inspection, the idea of a monocelllar origin of the tmor may be accepted. The argment of a mono- verss a plri-celllar theory, as well as the qestion of the degree of celllar atonomy in the origin of cancer, will not be finally resolved as long as the nderstanding of tmor pathogenesis remains nsatisfactory. As a rle, however. there appears to be no solid biological evidence against the monocelllar origin of neoplasms. The assmptions that malignancy is initiated in a single cell, and that there is a reasonable proportionality between tmor indction and expression. have been incorporated-explicitly or implicitly-into most models of cancer indction by radiation. This view is explicitly accepted here as a working hypothesis, even thogh it is recognized that different opinions have been expressed [B83, M52, R43]. E. NUMBER OF CELLS AT RISK 9. The nmber of cells at risk of cancer indction in any organ is not known. When discssing this sbject. most athors arge that the nmber of cells in which initiation takes place mst be proportional to the nmber of cells at risk in a given tisse. and that the latter is proportional to the fraction of tisse irradiated. Most models or theories of cancer indction by radiation have accepted this assmption. It does not follow from it that tmor incidence, spontaneos or radiation-indc ed, is proportional to the total nmber of cells in any organ. or in the body of varios animal 181

20 species [849, M9], becase there is no obvios correlation between body size in varios species and incidence of tmors. 91. Epidemiological data on lekaemia in man. after irradiation of varios fractions of the total active bone marrow. are broadly consistent with the notion that the probability of indction is proportional to the nmber of irradiated cells at risk, on the assmption that these are niformly distribted throghot the marrow. Absolte annal risk coefficients derived from varios stdies fall within a range which is narrower than the variation of the fraction of the irradiated tisse (annex G in the 1977 UNSCEAR report [U6]). The same conclsion may be derived from experiments on cancer indction by total and partial irradiation of the mammary tisse in rats [828. SIO. Sl2] and by varios areas of skin in mice [Hll. Hl2, HI3]. 92. An important and related qestion is the effect on tmor incidence of a non-niform tisse dose distribtion. The extreme case is that of the so-called hot particles, i.e., particles with very high concentrations of radionclides that are deposited in an organ sch as the respiratory tract or the liver. Under sch circmstances. the dose averaged over the entire organ wold be mch lower than that in the immediate vicinity of any sch particle. Several athors addressed this qestion by stdying carcinogenesis in the lng [I3. L35, S54]. In all cases, as expected from theoretical predictions [M9. MI OJ, hot particles were fond to be less effective than similar doses of the same radiation qality distribted in a more niform manner. F. POPULATION HETEROGENEITY AND SUSCEPTIBILITY TO CANCER INDUCTION 93. Bam [B 16] raised the qestion of whether linear extrapolation of the risk from the high dose and doserate region of low-let radiation to the low-dose domain is always conservative, in view of a possible difference in ssceptibility to cancer indction between grops in the general poplation. To exemplify the sitation, Bam [B 16] constrcted a model in which three poplation sb-grops, represented to 1. 1 and 89%, varied in ssceptibility (expressed as D vales) in the ratio of 1: I :, respectively. For each of the sb-grops he postlated a dose-response relationship of the form (2.1) 94. For the most sensitive sb-grop (l %) he assmed that cancer indction might be a single-target effect (n = l) with Do of.1 Gy; 1% wold exhibit a lower sensitivity (D = 1 Gy); and for the rest of the poplation D = oo. According to Bam. the composite response of the total poplation is dominated in the low-dose region by the highest sensitivity of the 1 % sb-grop. In the range from abot.1 Gy to a few Gy, the relationship wold be a crve. concave downwards. with response roghly proportional to D 5. The risk per nit dose above I Gy wold be less than below.1 Gy. 95. It is not easy to envisage why ssceptibility between grops shold be reflected only by D. and, if so. why by a factor as large as 1. Moreover, the assmption that the blk of the poplation shold be totally non-ssceptible to indction of malignant disease(s) lacks jstification in Bam's paper [B 16]. No epidemiological evidence spports this contention. even thogh it is known that incidence of radiationindced cancer in man is generally low. Examples of dose-response relationships advocated by Bam to spport his argment do not stand p in the face of criticism. The athor presented data for acte lekaemia at Nagasaki, and for all cancers. all lekaemias, stomach, lng, and breast cancer at Hiroshima. where incidences rise with a power of dose between.1 and.8. At present. dosimetric ncertainties preclde a more detailed discssion of atomic bomb srvivor data. bt preliminary evalations [K39, S46] of the dose-response relationships for lekaemia, breast cancer and total malignancies for the period p to sing new dose estimates [L27]. do not spport the above conclsions (this information is still tentative). Therefore, Bam's hypothesis is neither satisfactory on theoretical gronds nor spported by epidemiological observations. The sbject itself, however, is obviosly important and warrants some exploration. 96. Progress in cancer genetics has shed some light on possible mechanisms of differential ssceptibility to cancer development in individals with varios inherited traits [G 12. M35]. The relevance of these phenomena to radiation-indced cancer reqires attention. The relationship between genetic heterogeneity and incidence of spontaneos cancer, and cancer indced by radiation or other carcinogens, was discssed. in detail in annex I of the 1982 UNSCEAR report [U24]. It is again reviewed in annex A to this report. The homozygotes for some genes (ataxia telangiectasia. Fanconi's anaemia and several others) have a significantly increased natral incidence of some forms of cancer and an enhanced sensitivity to other celllar and sbcelllar radiation effects. Varios forms of impaired DNA repair are seen in most of these conditions. In view of their rarity. the affected sbjects do not pose a pblic health problem, bt the heterozygos carriers of these genes cold also have an increased risk of developing spontaneos cancer. Calclations have shown [V6] that a few percent of individals in the general poplation shold carry sch genes in a heterozygos state. If their capability for DNA repair shold be redced, this might reslt in an increased proneness to cancer development. Althogh it is not known at present whether these heterozygos sbjects may also be more prone to radiation-indced cancer, this possibility shold not be overlooked. 97. If risk coefficients at low doses were higher than those reported in most stdies at abot I Gy. then perhaps sch evidence cold be interpreted to sggest the presence of poplation grops particlarly ssceptible to cancer development. A higher effectiveness of low doses was claimed by Mancso et al. [M3. K 17. K32], who stdied the mortality of workers employed in the Hanford works, Richland, United States, dying 182

21 with cancer from for whom death certificates and data on personal radiation dosimetry were available. This information was reviewed by several athors [A 7. 14, G5. G25. H2. M3, R3. S2] who pointed to the low statistical power of the original data and to nmeros other difficlties in their interpretation. From carefl examination of all the evidence, it appears that these data are at present inadeqate to prove or disprove that the risk of radiation-indced cancer at low doses is higher than estimated by UNSCEAR in its 1977 report [U6]. This matter will, however, be kept nder review. G. SUMMARY AND CONCLUSIONS 98. The pathogenesis of tmors, both spontaneos and radiation-indced, is at present poorly nderstood. The probability of cancer indction cannot therefore be derived from general principles. The data available indicate that carcinogenesis is likely to be a complex, mlti-stage process. Initiation at the celllar level is followed by a long period of latency. dring which nmeros changes take place ntil the tmor becomes clinically manifest. The probability of some of these changes is very low, and dring the intermediate stages nmeros exogenos factors cold accelerate or inhibit the development of potentially canceros cell clones. 99. There is considerable variation between organs and tisses in the process of cancer development, presmably related to nmeros endogenos factors. Among these. the hormonal and the immnological factors are known to be of importance. at least for some forms of cancer. Significant cell killing in tisses may stimlate compensatory cell proliferation and therefore increase the probability of cancer development. The ways in which. and degrees to which, radiation cold modify sch mechanisms. particlarly in relation to dose, are not well nderstood. It appears likely. however. that dose thresholds sch as those operating for non-stochastic effects might apply. This wold imply that low or intermediate doses (below approximately 2. Gy of gamma rays and.3-.5 Gy of netrons) delivered actely might have little or no modifying effect. 1. Under the circmstances described, for predictive prposes the se of simplified models of tmor indction becomes one of the possible choices. even thogh the degree of simplification is sally sch as to exclde consideration of the above modifying factors. This important limitation of the models shold be kept in mind. 1 I. In view of the mlti-stage natre of the phenomena described. and of the small probability of occrrence of each step. the proportion of initially transformed cells giving rise to an overt tmor mst be extremely small. It may be postlated that at low doses and dose rates, where the host factors mentioned above are not expected to modify primary events appreciably. there is a broad similarity of relationships between dose and tmor incidence. on the one hand, and between dose and probability of cell initiation and srvival. on the other. Exogenos modifying factors (e.g., promoters and inhibitors) cold alter the slope, and perhaps the shape, of the dose-response crve in some instances; in man very little is known in this respect, apart from the effects of tobacco smoke on lng cancer indction. 12. Monocelllar and monoclonal hypotheses are sally assmed in models of radiation carcinogenesis. Most biological data spport the notion that cancer. as a rle, starts from a single cell. bt neqivocal. direct evidence to spport this point is lacking. and alternative interpretations are possible. The monocelllar hypothesis is critical in extrapolating the risk from high to low doses. The alternative hypothesis. that concomitant initiation or interaction of several cells is reqired, lacks sfficient experimental spport and makes extrapolation from intermediate to low doses difficlt or impossible at present. The postlate of monocelllar origin of cancer mst be treated. for the time being. as a working hypothesis. 13. The nmber of cells at risk of radiation-indced malignant transformation is sally assmed to be proportional to the fraction of an organ or tisse irradiated. The experimental evidence in favor of sch an assmption is very strong, bt. again. it mst be regarded simply as a plasible working hypothesis. Gross non-niformities of the tisse doses. sch as those reslting from deposition of the so-called hot particles, normally reslt in a lower tmor incidence than for comparable doses distribted niformly. 14. It has been proposed that sensitive sb-grops in the general poplation might invalidate any linear extrapolation of cancer risk to low doses of radiation, a procedre hitherto believed to provide pper vales of the risk. It is known that the very rare homozygos carriers of particlar genes (e.g.. ataxia telangiectasia. Fanconi's anaemia) are more prone to cancer and that their cells are particlarly sensitive to other effects of radiation, sch as cell killing. It is also tre that the healthy carriers of these same genes in the heterozygos form (a few percent of the general poplation) are at greater risk of developing spontaneos tmors and also have a slightly enhanced sensitivity to other radiobiological effects. Whether these same individals may also be at a higher risk of radiation-indced cancer cannot be proved at present and this qestion reqires frther stdy. The epidemiological evidence in favor of the notion that higher risk coefficients per nit dose of low-let radiation may apply at low, rather than at high. doses appears at present nconvincing. III. DOSE-RESPONSE MODELS OF RADIATION-INDUCED CANCER 15. Althogh in principle a large variety of mathematical expressions and models cold be fitted to experimental and epidemiological data on radiationindced tmors. for the prpose of this annex the Committee decided to limit the analysis only to those models that appear to be spported by general knowledge of celllar and sbcelllar radiobiology. 183

22 A. INTRODUCTION 16. When either the crde or the actarial incidence of varios radiation-indced experimental tmors is plotted against a wide range of doses. then, for low-let radiation. a peak incidence is normally seen at intermediate doses. followed by a decline at high doses. For high-let particles, the essentially linear initial slope gradally decreases with dose. In some cases, it even becomes negative. The blk of data is consistent with this generalization [U6]. Evidence in man is mch less certain becase only a few series cover a sfficiently wide range of doses. Nevertheless. the incidence per nit dose of bone sarcoma [M29, R 18], lng carcinoma [K22. S3] and cancer of the breast [S 16] decreases at very high doses. Ths. a model of radiation-indced cancer shold allow for two opposing trends: a rise of probability of indction with dose and then a decline. reslting presmably either from killing of irradiated potentially malignant cells [G 16. M52, R2]. or, in some cases. from increased mortality rate at high doses (15. 17]. Mathematically, this trend may be simlated by mltiplying the indction dose-response fnction by a term describing the probability of cell {or organism) srvival as a fnction of dose. 17. Conceptally, there are two possible approaches to the formlation of models. The first is based on radiobiological considerations, relating. by analogy. the radiation-indced cancer incidence and the inflence of varios parameters (dose, dose rate, fractionation) to other effects in single cells. Broad similarities have been observed in nmeros cases. The alternative approach is an empirical one, aimed at approximating the data {incidence verss dose) by simple mathematical fnctions. In practice, the two approaches are often applied in combination. 18. There is no reason why any model might be valid for all radiation- indced malignancies: actally. a more reasonable approach wold be to determine separately which model is applicable to which type of cancer, a conclsion spported by the blk of experimental data. There are examples of radiation-indced cancer with prononced thresholds {e.g., the chymic lymphoma of the C57BL mose, ovarian tmors in many mose strains). Indirect mechanisms (destrction of an overwhelming part of target cells in the thyms: killing of a large fraction of hormonally-active cells in the ovary) may explain sch sigmoid relationships. Similar data are exceptional in man (see chapter V). Ths, even if threshold-type relationships cannot absoltely be exclded, the most easily indced hman tmors, sch as those of breast, thyroid and lng (ndifferentiated small cell type), as well as lekaemia, do not indicate the existence of a threshold. Therefore. models sed for these malignancies do not postlate a threshold. In principle, the models discssed below have been formlated in terms of the absolte risk, bt they seem eqally applicable to a relative risk increment. B. DERIVATION AND CHARACTERISTICS OF DOSE-RESPONSE MODELS 1. The linear model 19. The simplest and oldest linear model was developed for x-ray-indced point mtations in Dro- sophila [G24, M46]. It was based on the postlate that each energy deposition in a celllar target carries a probability of changes in the genome (mtations) and that these may be either irreparable or the reslt of misrepair. Since radiation energy is transferred in discrete events. all effects shold ltimately be linear with dose at low doses when only few cells absorb some energy and their nmber is directly proportional to dose. The probabilities of an effect assmed to be an all-or-none phenomenon in relation to each dose increment shold be additive. and dose rate or fractionation wold not be expected to alter the magnitde of the final effect. 11. In the corse of time it became obvios that the linear model mst be limited to effects showing singlehit kinetics after low- or high-let irradiation. However. the postlated additivity of effects and the doserate independence does not always apply to single-hit effects [EIO, H35], showing that in sch cases repair processes mst also operate to make an otherwise linear relationship dose-rate sensitive. It appears that when a track passing throgh the sensitive volme is not 1% effective, then its effects may be sbject to repair. In other words. the presence of dose-protraction or dosefractionation effects may not be incompatible with single-hit kinetics For high-let radiation, the linearity of the initial slope, modified by phenomena of satration or cell inactivation (killing), has been widely accepted [B2. II, M66. U6, U7, U9, UIO]. The dose response is described by the following eqation: (3.1) If cells are sensitive to inactivation. the relationship will depart from linearity, even at low or intermediate doses. owing to cell killing. The lower the mean lethal dose (l/(j 1 ), the more downwards concave the relationship will be. Recent experiments on netron-indced tmors in animals (see chapter IV) sggest that this phenomenon-or others [S37. U23. U25, U26]-does in fact operate. casing departre from linearity. even at a few tens of milligray or a few tenths of a gray. This is explained by the fact that cell killing by high LET particles is very effective with exponential nonthreshold srvival crves. I 12. If eqation (3. I) reflects the kinetics of tmor incidence as a fnction of the high-let dose, extrapolation to low doses wold tend to nder-estimate the risk by a factor of e-p,d. Examples and limitations of this postlate will be discssed in chapter IV, in the light of available experimental data Alternatively, the data can normally be fitted over some range of doses by an eqation of the type: l(d) = a 1 \ID (3.2) Sch a prely descriptive model implies no linear slope at low doses, even for very radioresistant cells. This seems nlikely. however, on radiobiological gronds. In this latter model. as in any other model involving a downward concavity of the dose-response crve, dose fractionation wold prodce an increased 184

23 incidence of tmors if the intervals between fractions were sfficiently long for the effects of each fraction to be considered independent [R37]. By a similar mechanism. the effects at low dose rate cold also be enhanced, in comparison with the high dose rate An interesting variant of the linear model is that developed by Mayneord and Clarke [M9]. They discssed dose-response crves and their relationship with time. in regard to both dose rate and latency in the appearance of tmors. This is of importance becase irradiation time and latency may be comparable with the life span of an exposed individal and cold, therefore, affect the manifestation of the response. Their point of departre is that irradiation confers a probability of cancer appearance in the ftre [B48]. following a stochastic process. This probability may be estimated by integrating the response of single cells as a fnction of dose over the nmber of cells at risk. A linear relationship between the probability of malignant cell transformation and dose was assmed at low doses where cell killing is irrelevant. Becase mch experimental evidence from chemically- and radiation-indced tmors [B48, E3. E4, F8] pointed to an increase in latency with decreasing dose rate. this particlar featre was also incorporated into the model [M9, M JO] The reslt of the analysis was that in all cases. assming a linear relationship between each increment of dose and risk (the latter distribted in time). nonlinear (concave pwards) dose-response relationships were obtained. often with an apparent threshold. This is de to the inverse relation assmed between time of tmor appearance and the dose. According to this model. therefore. in most sitations a linear extrapolation of the risk from high to low doses and dose rates is likely to yield a conservative risk estimate. The dose-response relationships shold tend to linearity for: (a) long observation times. compared with the median latent period; (b) very heterogeneos poplations, i.e., poplations with a wide distribtion of the latent period after single irradiation; (c) little variation of the median tmor appearance time with the dose. In general. however. the theoretical analysis of Mayneord and Clarke [MI OJ does not spport an overall linear relationship between dose and cmlative tmor rate over finite time intervals in poplations having a standard age distribtion Inverse relationships between the mean latent period (from the start of exposre to tmor diagnosis) and dose (or dose rate) are seen mostly for irradiations of long dration. for instance in dogs and mice injected with 226Ra and 99Sr [M33, R28, R32]: in Chinese hamsters. in which the latent period of liver tmors decreases with increasing activity of injected 239P [B87]; in rats injected with thorotrast. which prodces tmors of liver and spleen [Wl5]; or rats exposed over periods of weeks or months to varios concentrations of radon reslting in lng carcinomas [C34]. Finally, a negative correlation of latency verss mean dose rate was seen in hman dial painters with 226Ra and 228Ra body brdens [E3, E4. Ml 4, PI I, Rl8]. It is interesting to note that the inverse doselatency relationship was mch less prononced for indction of bone sarcomas after single injections of short-lived 224 Ra in rats [M67]. For brief exposres sch as for atomic bomb srvivors. there is, at least in the yonger cohorts, a sggestion that the mean indction period for lekaemia is negatively correlated with dose [L32], althogh the reference in qestion seems to compare mostly radiogenic cases with mostly natrally occrring ones. Sch a relationship was not detected for breast. stomach and lng cancer in atomic bomb srvivors [BI7, L32. Ml7, K39. Tl2] and for breast and thyroid cancers reslting from medical irradiation [B26. S16, S38]. Moreo\'er. of all cases showing an inverse relation of the latent period against dose. an neqivocal departre from linearity was fond only for 9oSr-indced bone sarcomas. The sitation for hman osteosarcomas indced by 226 Ra and 228 Ra is not totally clear. and the dose-response relation for lekaemia in atomic bomb srvivors is still not available. In conclsion. the inverse relationship between latency and dose does not apply to many tmors. Ths, an pward concavity of the doseresponse relationship as a reslt of this factor alone cannot be viewed as a general phenomenon. 2. The linear-qadratic model 117. For the blk of experiments on single-cell systems, and for many end-points. the prevailing form of the dose-response crve after acte low-let irradiation is concave pward. the response rising with a power of dose of between l and 2. Broadly similar shapes were also observed for tmors indced by the same radiation. This form of the response is thoght to imply that for the occrrence of the final effect the interaction of two elementary effects, sally referred to as sblesions [K4], is reqired. Repair phenomena and dose-rate effects are sally acconted for by the interaction of these sblesions in time and space [L4, S39. S4], bt other interpretations were also proposed [G29. G32]. For effects showing crvilinear responses to low-let irradiation. linear dose-responses are freqently observed with densely-ionizing particles. {a) The microdosimetric approach 118. Microdosimetric considerations are important becase the distribtion of absorbed energy divided by the mass in the volmes believed to be the primary targets of radiation action (i.e., cells or cell nclei) is very inhomogeneos at low and intermediate doses [F7]. Ths the distribtion of specific energy, z (energy absorbed divided by the mass of these volmes). may be more relevant for the analysis of the effects than the organ or tisse mean dose, D Many experiments indicate that for a nmber of effects the probability of occrrence. E. after low-let radiation, rises with a power of dose close to 2. This may be acconted for by the theory of dal radiation action which assmes that the indction of two nonspecified sblesions is necessary for the prodction of the effect (lesion). In the approximation that was first developed in detail [K4], it was assmed that the probability of sblesion interaction does not depend 185

24 on their distance within a certain volme (the site). Conseqently. the probability of effect shold rise in proportion tq_ the mean of the sqare of the specific energy, i.e., z 2, in the site. It was shown also that z 2 is a fnction of the absorbed dose. D: z 2 = (D + D 1 (3.3) where the ( is a weighted average of the specific energy deposited by single events (energy depositions in the critical strctres by individal charged particles). Ths, the probability of a radiobiological effect, E. shold be E = K((D + D2) (3.4) The qantity ( has the dimension of dose and its vale denotes the dose at which the contribtion of the linear eqals that of the qadratic term. For x and gamma rays. and for sites of I µm diameter, the vale for ( falls within the range of.3-.8 Gy; for medimenergy netrons, it is as large as 16 Gy. For smaller sites. the vales of ( are considerably larger. bt the ratio of the vales for netrons and for sparsely ionizing radiation is less dependent on site size. 12. Consideration of the vales of ( for radiations with different vales of LET and their comparison with biological data. indicates that intracelllar distances at which sblesions can interact mst be of the order of.1 to 1 µm [K4]. High vales of ( for netrons and alpha particles (relative to x and gamma rays) accont for a blk of radiobiological observations. They show that the REE of high-let radiations varies as an inverse fnction of the sqare root of their dose and this can be nderstood on the basis of microdosimetric concepts. This type of dependence can hold down to doses where the response to both radiations becomes linear and the REE reaches a platea Sitable time fnctions can be inserted into the basic eqation (3.4) to describe the interaction probability dring the time interval between the formation of two sblesions [R 12). One of the conclsions is that the linear term for single events in eqation (3.4) shold be independent of dose rate. Therefore. any effect of time distribtion may only reslt in a modification of the qadratic term. This relationship between irradiation time, T, and mean time between indction of a sblesion and its repair. r, is reflected by a redction factor, G, that affects the qadratic term: G = 2(r/T) 2 (1/r e-tlr) (3.5) In the presence of recovery, the crves of RBE verss absorbed dose do not change in shape, bt are shifted towards higher doses by a factor depending on the dose distribtion in time and the recovery time. r. For a constant dose rate, if the exposre time is mch longer than the recovery time, the yield of elementary lesions becomes simply proportional (linear) to dose The theory of dal radiation action has been applied by several athors to the interpretation of biological reslts, for example to cell inactivation by ionizing radiation [B62, F7]. In this model. the slope of the srvival crve depends explicitly on radiation qality throgh the qantity ( [K4]. The theory has also been sed as a theoretical basis for the linearqadratic model for many effects. inclding mtations, chromosome aberrations and initiation of oncogenic transformation. However, experiments with ltrasoft x rays [C23. G9. G JO. G 11] and with spatially correlated ions [888, G3I. G32, K4I. R38] indicate a strong dependence of the interaction probability of the sblesions on distance, with a dominance of shortrange interactions (mch less than.1 µm) for denselyionizing radiation. The applicable version of the theory [K5] is a model of higher complexity that is Jess easily tested experimentally Alternative explanations were considered for the commonly observed RBE-dose and dose-response relationships for low- and high-let radiation. In agreement with earlier ideas of Haynes [H27]. Goodhead et al. [G29, G32] pt forward the notion that a nmber of celllar and sbcelllar radiation effects depend both on the prodction of the primary lesions and on a dose-dependent inactivation of repair mechanisms. Whereas the probability of the primary lesions is taken to increase linearly with dose, impairment of the repair system wold depend on absorbed dose and specific energy. It has been postlated that the degree of impairment of repair shold display a prononced dependence pon dose and/or dose rate of low-let radiation. For high-let particles there wold normally be no error-free repair, independently of dose. Therefore, basically linear dose-responses wold be expected for indction at the celllar level of each of the varios effects separately The above concepts have not been widely tested. A variety of factors have been reported that may inflence the expression of potential damage. and ths change linear dose-effect relationships to crvilinear ones, and vice versa [Fl8, H22, L26]. {b) Characterisrics of rhe linear-qadraric model 125. Upward concave dose-response crves that fit well the cmlative cancer incidence as a fnction of acte doses of low-let radiation are in themselves no proof that cancer is actally initiated in accordance with a linear-qadratic relationship. Crvilinearity cold well reslt from an inverse relationship between dose and latent period [MIO]. However. when followp data show a constant latent period with dose. it may be assmed, as a working hypothesis, that the pward concavity reflects a linear-qadratic kinetics of initiation A generalized linear-qadratic dose-response relationship for radiation-indced cancer wold be as follows: where I(D) is the incidence (in an ideal case corrected for intercrrent mortality): D is the dose; a is the spontaneos incidence: a I and a 2 are coefficients for the linear and dose-sqared terms of cancer initiation; 186

25 and S(D) is the probability of srvival of transformed cells and may be written as S(D) = e-<jj,d + /J,DJ (3.7) In the eqation (3.6), initiation may also be treated as a simplification of a more general expression, which incldes some repair at low and intermediate doses and dose rates [B95] To test compatibility of the model with experimental and epidemiological data, the parameters in eqation (3.6) mst be assigned some vale. This can only be done by: (a) reasonable analogy with celllar radiobiological effects: or (b) dedction of the vales from empirical stdies. As to the parameters for cell srvival. these are known for many cell lines (e.g., Barendsen [B5. B6, B9-B 12, B45]; Goodhead et al. [C23. G9. GlO]: Elkind [E2]: and others [Al7. B19. B27, Cl, C37, D3, D8, F2, Fl9, H2, H6, 16, K2. K24, L6, Ml6, M6, R39. S33, Tl8, Tl9, W6, W7]), bt are not necessarily applicable to specific tmors The nmber of nknown coefficients in eqations (3.6) and (3. 7) is too large to permit their calclation withot some constraints. One of these is that all parameters shold be positive or eqal to zero [C29]. However. this is often insfficient for reasonably stable estimates of fj I and {J 2, particlarly when a I and a 2 are estimated at the same time. Attempts were made, therefore [B7, B47], to derive independently the qotient a/a 2. which facilitates calclations by lowering the degrees of freedom Assming that the theory of dal radiation action [K4] applies to the initiation of cancer. there mst be a sbcelllar strctre in which the sblesions appear and an interaction distance below abot I µm where the final lesion leading to malignant transformation is prodced. Kellerer and Rossi [K4-K7. Rl2] did not specify the relevant strctre for cell transformation. bt the most likely target of radiation action appears to be the genome. In the light of this assmption, and of available evidence. it is not nexpected that models postlating a linear-qadratic relationship presppose a direct action of radiation on DNA or make reference to phenomena involving DNA (chromosome aberrations, mtations) for derivation of the relevant parameters. 13. Barendsen [B7] selected the vales for a I and a 2 on the basis of dose-response relationships for chromosomal aberrations in lymphocytes and the ronded vales of a/a 2 for x and gamma rays were set at.5 and 1. Gy, respectively. On a similar principle, Abrahamson [A3] and Brown [B47] assmed that. if hman radiation-indced cancers are de to mtations or chromosome aberrations in single cells, then information on the qotient a 1 /a 2 cold be derived from extensive data available in the literatre. However, the large variability of the qotient preclded the se of nmbers at their face vale (e.g.. mean or median). Brown showed also an inverse relationship between the DNA content per haploid genome in cells of varios species and the a/a 2 qotient, the vales extending over ranges of 13 and 1 4, respectively For mammalian cells, a/a 2 vales for mtations and chromosome aberrations, as reviewed by Brown [B47]. clstered arond I Gy (geometric mean, 1.27 Gy). The vales of the P if P 2 ratio for cell sterilization were generally mch higher. with a geometric mean of Gy. The difference is de mainly to higher rnles of the linear parameter for cell killing, in accordance with conclsions [B9. BI I] that at low doses there mst be a sbstantial component in cell killing that cannot be identified with the indction of chromosomal aberrations. Brown conclded that if the initial radiation lesion triggering a hman cancer is of mtational or cytogenetic natre. then risk estimates derived at 1-2 Gy of low-let'high-dose-rate exposre wold, as a maximm. over-estimate the cancer risk at low doses and dose rates by factors of abot 2 and abot 4, respectively Concerning cell srvival. it has been pointed ot repeatedly [B7. E2. F2. K2. M21. Sl7] that the low-let srvival crves can be described by varios models and, among others, by fnctions of the ''single-target" pls "mlti-target-single-hit" type S(D) = e-(dl1dol (1 -(I -e-dlndo) ] (3.8) where S(D) = SofS, 1D and D are mean inactivation doses for single and mltiple targets. respectively. and n is the target mltiplicity. Alternatively. the linear-qadratic form may be applied. as in eqation (3.6) Whether for each cell type there is an initial slope [PI in eqation (3. 7) and (e- 1 1 ) in eqation (3.8)] is still a matter of debate. bt this qestion is rather immaterial in the present context as both fnctions describe satisfactorily the experimental data for srviving fractions between 1. and.1 [B9. Bll. F2, M21]. Eqation (3.6) has been most commonly sed in relation to the linear-qadratic model of cancer indction (see [U6, annex G] and [B7. C29. M31]). For high-let radiation the contribtion of the qadratic term is negligible As a first approximation. for low and intermediate doses, particlarly of low-let radiation, the killing term S(D) may be neglected. This cold be jstified nder the assmption that repoplation of the target cells after irradiation has a promoting effect on the development of radiation-indced cancer roghly proportional to the degree of cell killing [P22]. If. in addition, some of the terms in eqation (3.6) and (3. 7) are omitted, a transition is possible from the linear-qadratic model to the pre linear or pre qadratic ones (Figre IV). As can be seen from eqation (3.6), the model does not assme the presence of a threshold There are many methodological and biological factors [BIO, Cl, D8. H6, K24. L6]. which make it difficlt to select appropriate srvival parameters. In particlar, cells irradiated in sit might exhibit higher srvival capacity than when irradiated in cltre or in sspension [G14. H8, HIO, M6] (however, for a contrary view, see [A 17]). By sing a range of srvival parameters sitable to describe the most and least 187

26 Genera 1 fonn linear I (D)=a +a 1 D w z w Cl - z w z Cell killing ~ attenates I(D) z I(D)=(a. +a. 1 D+a 2 D 2 )exp(-~ 1 D-~ 2 D 2 ) DOSE (D) DOSE (D) Qadratic I (D )=a. +a 2 D 2 Linear-qadratic I (D)=a. +a 1 D+a 2 D 2 w z w Cl - z w z w Cl z DOSE {D) Figre IV. Varios types of dose-response crves. (C29] DOSE (D) sensltlve cell lines, one may examine the degree to which cell killing may affect the shape of the tmor incidence data and. hence, the reliability of extrapolation from the intermediate to the low doses. This shold provide some indication of the degree of overor nder-estimation of the risk, within the range of vales selected for the parameters in qestion In an independent analysis. UNSCEAR selected two vales of the a/a 2 qotient applying to x and gamma rays:.5 and 2. Gy. For srvival. the bone marrow stem cell was selected as the most sensitive. Its srvival crve is described by /J 1 = Gy- 1 and /3 2 = Gy- 2 [B6]. For the least sensitive cell. a hypothetical line was assmed with srvival parameters /3 1 = 1-1 Gy- 1 and /3 2 = Gy- 2 [B6]. In addition, other srvival parameters for radiationindced hman lekaemia cells (Table l from [Ul 7]) were sed. as follows: /J 1 = Gy- 1, and /Ji= 1-1 Gy- 2 To normalize the calclated data. it was frther assmed that the lifetime cmlative incidence at 3. Gy may be 15, cases per 1 6 (5 1-6 Gy- 1 ). The reslts in terms of the yield of cases from 1-3 to 3 Gy for all combinations are plotted in Figre V From sch an analysis. the following tentative conclsions may be drawn. First. the sensitivity to cell killing has a more prononced effect on the shape of the dose-response relationships than the a 1 / a 2 qotient. (Similar conclsions were reached by Mole [M31].) Secondly, for the cells most sensitive to killing, the w z: ::g z DOSE (Gy) Figre V. Expected cmlative Incidences of a radiation Indced cancer according to a linear-qadratic model with a1/a2 qotients=.5 and 2 Gy and cell srvival fnctions Smax and Smtn as given In paragraph 136. Incidence normalized to 15, cases per 1s at 3 Gy. For fll explanation, see paragraphs 136 and

27 relationship is concave downward, with maxima at Gy. Since sch crves are not observed for hman cancers after low-let irradiation. it is possible that the assmed sensitivity is too high for in vivo irradiation. This wold be in accordance with the lower sensitivity of single cells irradiated in sit. which is de mostly to a wider sholder of the srvival crve [D7, G14, H8. HlO. M6. S26. S27]. Thirdly. for the cells least ssceptible to killing (the carcinomaros cells according to Barendsen [B6. B7]), the overestimate of the tmor yield per nit dose from 1-2 Gy to 1-1 mgy (relative to linear extrapolation) ranges from at aifa 2 =.5 Gy, to at aifa 2 = 2 Gy, respectively. The maximm over-estimation of the risk will reslt from totally neglecting the cell-killing effect. In sch a case. extrapolation from 2 and I Gy down to abot 1 mgy wold involve the following over-estimate: a/a 2 (Gy).5 2. O\ er-esrimare by a/actor of" from 2 Gy from l Gy The above differences for a if a 2 of.5-2 Gy cold not be easily-if at all-detected between estimates derived from I or 2 Gy and the exceptional estimates available for doses of a few tens of mgy per single dose. The lack of statistically significant difference is therefore compatible with both the linear or the linear-qadratic model Inspection of eqations (3.6) and (3. 7) indicates that cell killing may alter significantly the incidence of a given cancer throgh interaction with the spontaneos incidence, a At very high levels of the latter and with high sensitivity of the cells to sterilization, a dose-related decrease of the incidence wold be expected. This in fact was seen in experiments with animals, showing a negative slope of the doseresponse relationships for the so-called reticlm cell sarcoma [MI9] The slope of the dose-response crve in the complete linear-qadratic model will be described by a differential: d/dd i(d) = [(a 1 + 2a 2 D)- (/J 1 + 2P2D) (ao + aid + + a 2D2)] e-<p1d + P2D&J (3.9) As D approaches zero this eqation simplifies to: d/dd i(d) = a 1 + 2a2D - P 1 a - D(/J 1 a 1 + 2P2ao) (3.1) To examine the interaction with dose of a, a 1 and P, and its inflence on the crve at low dose, two hman cancers were selected: breast carcinoma, with a high spontaneos incidence and moderate sensitivity to cell killing of initiated cells; and lekaemia, with rather low vales of a and high sensitivity to sterilization. The notional parameters selected were: Breast cancer: a = a- 1 /1 1 = Gy- 1 a 1 = a- 1 Gy-' P2 = Gy- 2 a 2 = a- 1 Gy- 2 Lekaemia: a = a- 1 a 1 = I 1-4 a-1 Gy-I a 2 = I 1-4 a- 1 Gy- 2 (J 1 = Gy- 1 /12 = Gy-2 When the vales of d/dd i(d) were calclated at 1. 1 and 1 mgy, their ratios to the respective vales of a 1 were: Dose (Gy).1././ Lekaemia: Breast cancer: Ths, at very low doses interaction of cell killing. predominantly with the spontaneos incidence. wold introdce some degree of over-estimation of the risk, if this interaction shold be neglected in the extrapolation procedre. It appears nlikely. however. that this over-estimation might exceed another factor of 2, over that estimated in paragraph When extrapolation is made from intermediate or high doses of low-let radiation. delivered at low dose rates or in small fractions (of size mch below aifai). then, de to repair processes, the qadratic terms of eqation (3.6) shold be redced. Under these circmstances. dose-response relationships approaching linearity wold be expected in which the risk coefficient per nit dose is the same over the range of low and intermediate doses. For high-let radiation, the contribtion of the dose-sqared component for most effects in single cells is negligible p to a few tens of Gy. Therefore, the model may be treated as linear, althogh for some effects the slope of the line may not be dose-rate independent (see III.B. l). Sch observations cannot be explained solely on biophysical considerations It has been pointed ot [U17] that the linearqadratic model of cancer initiation is a simplistic concept and not a pathogenetic theory. The great complexity and the interaction of varios phenomena at the celllar level preclde its acceptance as a generalized theory of cancer initiation in all tisses and nder all circmstances. The dose dependence of the many modifying mechanisms operating at the tisse and systemic level is still too obscre to allow qantitative predictions over the entire range of doses. However. as discssed in chapter II, it may be postlated that the inflence of modifying factors might not be very important in the range of doses p to 1-2 Gy of low-let radiation (or their high-let eqivalent). Under these circmstances, appropriate extrapolation may be acceptable down from the dose region of 1-2 Gy of x or gamma radiation (or an eqivalent dose of netrons) to obtain risk estimates for the low-dose domain With all these reservations in mind, the linearqadratic model has been applied to the reslts of selected animal tmor experiments where dose. dose rate, dose fractionation, dose protraction and LET were systematically investigated (see chapter IV). In several cases [B7. P22, U2, U3, U25, Yl] the reslts were in basic agreement with the model's expectations. 189

28 3. The qadratic model 143. This model is based on the assmption that for the indction of a given effect two consective events are necessary. or rwo concrrent events separated in space in sch a way that their prodction by a single ionization track is extremely nlikely. Except for irradiation of flattened bone lining cells (see V.E.2.), this assmption is difficlt to jstify on microdosimetric gronds for low-let irradiation of normal cells in sit. In tmor radiobiology, this concept may be traced back to the theory of bone sarcoma indction by fj emitters [M26. M45, M47]. Proportionality to the sqare of the dose has been noted for varios cancers [Hl2. Hl3. M26. M52, Rl8]: this does not exclde the presence of a linear component, bt simply assmes that it may neither be demonstrated nor rled ot on statistical gronds. Therefore. if present, sch a component mst be relatively small Recently. Mole et al. [M52] have shown that dose-response relationships for radiation-indced acte lekaemia in CBA/H mice may be characterized by an approximately pre qadratic or linear initiation term for x rays or netrons, respectively (see IV.B.l). On the basis of microdosimetric considerations the apparent absence of a linear term in the dose-response eqation for x rays is difficlt to explain when the diameter of a ncles is of the order of few µm. The proposed hypothesis [M52] was that the interaction of two neighboring cells is necessary for the initiation of this cancer. Initiation by netrons is in direct proportion to dose and this was explained by the fact that a single recoil track cold cross the nclei of two neighboring cells. However, for netrons of maximm effectiveness (4 kev) the range of the secondary protons is less than most nclear diameters The experimental basis of these hypotheses is still debatable, and needs frther experimental testing. Moreover. the postlates are difficlt to reconcile with the reslts of transformation experiments of cells in vitro. irradiated at low density (see chapter IV) where it is nlikely that two neighboring cells may be crossed by a single recoil track or may interact dring oncogenic transformation by x rays. Explanations other than the microdosimetric ones mst be fond to accommodate all these facts When fitted to the experimental data obtained at intermediate or high doses. the qadratic model will predict effects at low doses to be mch less than implied by the linet1r or the linear-qadratic models within the sal range of a 1 and a 2 vales. The inflence of the time pattern of dose administration cannot be specified a priori for the qadratic model: if repair is fast and dose-rate-dependent. dose protraction or fractionation cold lead to a considerable redction of the effect. and zero effect cold not be exclded The qadratic model has been sed by the Committee on the Biological Effects of Ionizing Radiation (BEIR Committee) [C29] to obtain the lower bondary of likely linear-sqare dose responses at low doses. Three special models have recently been proposed which assme. or derive from experiments, the basic postlate of the dose-sqared concept. They refer to lekaemia in CBA mice [M52], skin cancer in rats [VI] and bone sarcoma in man [M4]. They are reviewed in chapters IV and V. respectively. C. SUMMARY AND CONCLUSIONS 148. Qantitative models of radiation-indced cancer mst allow for two facts: first, that tmor incidence rises with dose at low doses: and, second. that at high doses the effect sally declines (either in absolte terms or per nit dose). This shape of the crve may be modelled by the prodct of two terms: an initiation term increasing as a fnction of dose, and a cellsrvival term declining with it. The laner term can be neglected at low doses. bt this simplification may lead to misinterpretations as the dose increases. There is no reason why any one model shold be niversally applicable to all radiation-indced types of cancer. Models shold be tested in each instance by statistical criteria. In each given tmor system. they shold allow for the effect of the main radiobiological variables, sch as dose, dose rate. dose fractionation, radiation qality The linear model implies direct proportionality between dose and the probability of tmor indction. Absence of dose-rate and fractionation effects was thoght to be an inherent featre of the linear model, becase the biological lesion is assmed to reslt from a single-track effect. However, examples are known where this is not tre, and linear dose-response relationships show shallower slopes with decreasing dose rate. There are few examples of celllar or carcinogenic effects following linear kinetics after low LET irradiation, while the effects prodced by high LET particles often conform to these characteristics. However. there are examples where dose-rate effects for netrons are fond. in an opposite direction (enhancement) from those normally seen for low-let exposre. The linear model may be regarded as an pper bondary for the more general linear-qadratic model when the doses at which the linear eqals the dose-sqared components are high or very high. 15. The model of Mayneord and Clarke assmes linearity between dose and cancer initiation at the celllar level. It also assmes a log-normal distribtion of the latent periods for overt tmors and an inverse relationship of latency to dose. Soltion of this model for observation times comparable to hman life expectancy reslts in dose-response relationships which are pward concave. sometimes with apparent thresholds. The critical assmption leading to sch reslts is the shortening of the mean latent period with dose. If this were a general rle, then extrapolation of risk from high to low doses wold always lead to over-estimates. However, for some radiation-indced tmors in man (thyroid, stomach, breast. lng), the latent period is not obviosly dose-dependent; also, crvilinearity is not seen in several cases where there is an inverse relation of the latent period with dose. Ths, it is impossible to generalize the notion that jst for this reason crvilinear responses wold normally apply. 19

29 J 51. When corrected for cell killing, nmeros biological end-points indced by low-let radiation in single cells show an exponential dependence on dose with exponents of between I and 2. On the contrary, the same end-points for high-let radiation show essentially linear dose-responses. Conseqently, the RBE is an inverse fnction of dose. Low-LET radiation shows prononced inflences of dose rate and fractionation, in contrast with high-let. for which sch phenomena are seldom observed or are observed to a lesser extent. Most of these facts may be acconted for by microdosimetric considerations. They predict that, as a rle. linear-qadratic dose-responses shold be expected. where the relative contribtion of the two terms is comparable (for low-let). This prediction is in accordance with nmeros experimental data on single cells, and so are the effects of dose rate and dose fractionation that may be dedced from the kinetics of repair of the sblesions postlated in the linear-qadratic model Recent data on ltra-soft x rays, however, do not conform to the predictions of the theory of dal radiation action in its early approximate formlation. The disagreement cold be reconciled in the generalized version of the theory [B98], assming that some parameters of the biological effect previosly taken to be invariable are actally sbject to some variability: or that. in case of low-let radiation, the rate of repair depends on dose and on dose rate. Which of these two alternatives is likely to apply remains to be answered The linear-qadratic model may be characterized by the qotient of the indction constants (a/ai). which varies with the radiation qality and the specific biological effect. For high-let particles, this qotient is so high that the contribtion of the dose-sqared term may normally be neglected. The model then becomes linear. For low-let radiation (considering chromosomal exchanges. mtations and indction of some malignancies) the a/a 2 qotient is between.5 and 1.5 Gy. The over-estimation of the probability of effects at abot 1 mgy from single-dose data at 1-2 Gy (actely delivered) by linear (as opposed to linear-qadratic) extrapolation wold vary from 1.5 to 3. for an assmed reasonable set of parameters Similar over-estimations wold apply to the effects of low-let radiation delivered at low dose rates, irrespective of the total doses below the level when cell killing becomes important. This wold reslt from the disappearance of the qadratic component of the dose-response crves owing to repair of s blesions Introdcing a cell-killing term into a linearqadratic model leads, in general. to a lesser degree of risk over-estimation at low doses. The effect of sch a term may be analysed by applying a range of experimental cell-srvival parameters. Since the killing effect of a given dose might be less in vivo than in vitro. the over-estimation involved in the extrapolation to low doses and dose rates wold probably fall between that obtained from typical cell srvival parameters in vitro and that obtained by neglecting cell killing entirely For fission netrons and alpha particles, the a/a 2 qotient is of the order of several tens of Gy and the microdosimetric linear-qadratic model predicts a linear crve, relatively independent of dose rate. At doses where departre from linearity becomes sbstantial. a significant nder-estimation of the risk mst reslt from linear extrapolation to zero doses. Fractionation or protraction of dose may reslt in enhancement of the effect even at low and intermediate doses, contrary to what is seen at intermediate doses for sparsely-ionizing radiation, where the effect decreases pon fractionation and protraction. These observations, related to high-let particles, cannot be explained on biophysical gronds Dose-response crves after low-let irradiation have been reported for some tmors that sggest either the absence of. or a very small, linear term. In sch cases. the crve approaches the prely qadratic, indicating perhaps a very low probability of interaction of lesion in two targets crossed by a single ionizing track. or the presence of highly efficient dosedependent repair processes. Qadratic models have been proposed for lekaemia in CBA mice and cancer of the skin and bone. They predict a very prononced redction of the effect by lowering the dose rate or by dose fractionation. Sbstantial over-estimation of the risk wold derive in these cases from linear extrapolation to low doses and dose rates from high or intermediate single doses. IV. DOSE-RESPONSE RELATIONSHIPS IN EXPERIMENT AL SYSTEMS A. PHENOMENA OBSERVED AT THE CELLULAR AND SUBCELLULAR LEVEL 158. In recent years. there has been good progress in nderstanding the role of oncogenes, and of related phenomena at the macromoleclar level, in the casation and development of cancer. The newest data are reviewed in annex A to the present report. This progress. however, has left a considerable gap in the nderstanding of how these phenomena may be linked to mechanisms of radiation carcinogenesis. and what are the relevant moleclar targets [G34]. From qalitative analysis of the available information it appears that a rather wide spectrm of radiation interaction with genetic material shold be considered The present section, therefore. discsses doseresponse relationships for radiobiological effects, at the sbcelllar and celllar level. that might be relevant to the indction of malignancy. Some effects, sch as in vitro cell transformation, are probably closely related to in vivo carcinogenesis; the relevance of others is not yet well established, bt is not implasible. Among the latter effects are mtations, becase there is strong evidence of involvement of the genome in both the process of malignant transformation and of canceros growth (see II.A). Mtational events that cold be linked with carcinogenesis may inclde point mtations, chromosomal deletions and 191

30 translocations, as well as transposition of the genetic material [B7 I, B72, C32, F 14, S43, S56, Y3]. It has been postlated [F21] that tmor specific" symmetrical chromosomal translocations may be particlarly relevant for two reasons: first. there are cancers whose cells consistently show aberrations of this type (e.g., chronic myelogenos lekaemia, Brkitt's lymphoma); second, this chromosomal aberration is compatible with the cell's continos ability to divide, a pre-condition for a transformed cell to develop into a canceros clone. l. l\ltations in germ cells 16. The freqency of point mtations as related to dose in germ cells of male and female mice. and of other species, was reviewed in depth in annex H of the 1977 UNSCEAR repon [U6] and in annex I of the 1982 UNSCEAR report [U24]. The most recent findings are dealt with in annex A to this report. In smmary. there is an nqestionable effect of dose rate and fractionation pon specific locs mtations indced in germ cells of male and female mice by low LET radiation. Interpretations of these findings differ coilsiderably. Abrahamson and Wolff [A2, A3] postlated that mtational lesions are both one- and twotrack events, and tried to fit linear-qadratic eqations to the data. The attempt was criticized by Rssell [R24. R25. R26] who has long been advocating an alternative hypothesis, namely, that mtations themselves are single-track phenomena. and the apparent mltiple-track events reflect damage and satration of the repair processes at higher doses and dose rates. UNSCEAR reviewed both hypotheses in detail and, in the light of present evidence, accepted that of Rssell as a more likely explanation. 2. :\ltations in somatic cells (a) The Tradescantia system 161. The data obtained on plant cells of the species Tradescantia are very valable becase: (a) mtations can be scored down to 3 mgy of x rays (and correspondingly lower doses of netrons); (b) the amont of data on the effects of dose. dose rate and radiation qality is far better than for any other celllar system: (c) the effect is e, idently a somatic mtation (althogh its moleclar mechanism is not clear) and the nderlying lesion is most likely chromosomal; and (d) the form of the dose-response crve is similar to that for dicentrics and deletions in mammalian T lymphocytes Dose-response relationships for somatic mtations in Tradescantia, indced by acte.25-kvp x rays, are plotted on a log-log scale in Figre VI [N2, S2, UI9]. Below.1 Gy. the crve is linear, bt then its slope increases p to I Gy, where a dose-sqared component sets in. leading to an pward concavity of the crve. Above 1-2 Gy, the crve bends and declines, reflecting. perhaps. preferential killing of the mtated cells (the effect is scored in terms of freqency per srviving cell). The data points below a: ~... 'I... :z: w > 1..1 w.1 :z: "" a. w ~ c:, ::; Figre VI. Constant high dose rate, varying dose, x rays experimenta 1 points at high dose rate a,o I.. ~.,~, DOSE (Gy) 11 Dose-response crve for indced pink mtations In Tradescantia. [B99] I Gy are well fitted by a non-threshold regression eqation of the type I= a 1 D + a 2 2, where a 1 and a 2 arc constants, presmably for single- and two-track events. Mean vales of a 1 and a 2 dedced from three stdies [S2, N2, UI9] amonted to 5.1 (range ) 1-2 Gy- 1 and 6.5 ( ) 1-2 Gy-2, respectively [NI]. In terms of the linear-qadratic model, the a 1/a 2 qotient for x rays varied from.33 to.83 Gy. For gamma rays. 1 and 11 2 are Gy- 1 and Gy-2 respectively [NI]. Stdies with monoenergetic netrons ( Me V) prodced a linear dose-response p to.5-.1 Gy (S19, UI3] of the form I = a 1. At high and intermediate doses. the RBE of netrons verss x rays was inverselv proportional to the sqare root of the netron dose Redction of the netron dose rate had almost no effect pon the yield of mtations [N2, U 13]. When the x-ray dose rate was redced from abot I Gy min- 1 to 5-.5 mgy min- 1, the freqency of mtations was an increasingly linear fnction of the dose, approaching a linear response of the form I = a 1. When the dose rate of x or gamma rays was progressively lowered [M 18]. and a constant dose of.8 Gy delivered over increasing time intervals. lower mtation freqencies were observed approaching asymptotically the limiting vale of the respective a 1 for gamma rays. Fractionation of an x-ray dose of.6 Gy into two acte fractions of.3 Gy each [M 18] prodced no redction of the mtation yield p to a 15-minte fractionation interval, bt the yield declined with frther spacing of the fractions p to 24 hors. Splitting an x-ray dose of.5 Gy into two.25 Gy fractions given at 5 hors from each other did not redce the response any frther [U27) The inflence of fractionation of two x-ray doses (. 7 and 4 Gy) was stdied pon indction and repair of two types of somatic mtations (pink and colorless) in this system (Wl6]. The intervals between 192

31 the two fractions varied from.5 to 12 hors. Recovery for the two types of mtations differed. in that pink mtations showed recovery at both doses, whereas colorless events showed some recovery at lower doses, bt an increased effect at higher doses. Ths, two types of mtations indced by x rays in the same system may have different dose and time kinetics In smmary. the effects of dose, dose rate. fractionation and qality of radiation pon Tradescantia mtations provide. in general. a good example of the linear-qadratic model of radiation action. (b) Mammalian somatic cells 166. In recent years, significant progress has been made in the methodology of stdies of point mtations in somatic cells in cltre. Enzyme deficiencies are indced by mtagenic agents. and mtant cells are screened for resistance of developing clones to sbstances that are toxic to normal cells, e.g.. to 8-azaganine, 6-mercaptoprine or 6-thioganine in hypoxanthine-ganine-phosphoribosyltransferase (HG-PRT) deficient mtants. Biochemical mechanisms and methods of investigation were reviewed in detail in annex H of the 1977 UN SC EAR report [U 6] and in annex I of the 1982 UNSCEAR report [U24] and are discssed again in annex A to the present report In smmary, both linear and crvilinear (concave pwards) relationships have been observed for indction of J:{G-PRT deficiency in varios cell lines by sparsely-ionizing x or gamma rays, and the character of the dose-response crve is evidently related to the genome of the cell stdied and not only to the qality of radiation applied [Al 3, A 14, B42. CIO, C23. C24. E8, Kl6]. In cases of crvilinear dose-response relationships, after acte x or gamma irradiation there was also a prononced redction of the yield of mtations by dose fractionation. For densely-ionizing radiation (alpha particles. helim, boron and nitrogen ions). the dose-response relationship for indction of HG-RPT deficiency, both in hman and hamster cells, was essentially linear [C2, Tl8]. Indction of forward mtations in Chinese hamster cells by x rays reslted in crvilinear doseresponse relationships (concave pwards): moreover prononced dose-rate effects were observed [S25] Experiments on radiation-indced mtations reslting in HG-PRT deficiency show some correlation between the shape of the srvival crve for low LET radiation and the kinetics of indction of somatic mtations in the same cell lines. This observation has been generalized by Thacker et al. [TIO. T9, T8], Mnson and Goodhead [M36]. Leenhots and Chadwick [L3. C2] and Goodhead et al. [G29] on different radiations and cells in varios stages of the cell cycle. in the sense that there is a relationship of the type lns = -K.M where S is the cell srvival (SofSo), and M the freqency of indced mtations. This correlation is corroborated by other stdies (e.g.. [R5]). Thacker [T8] has shown that it applies to exponentially growing cells of several species (man, hamster and mose) with the same proportionality constant K. However. the mtation freqency per lethal lesion seems higher for high-let radiation and lower for ltrasoft x rays [G34] When Thacker et al. stdied the same relationship [Tl 9], by irradiating V-79-4 hamster cells grown to platea phase withot re-feeding (most cells in the G 1 phase), they fond a similar association of freqency of indced mtations with (log) srvival. This was tre both after high (1.7 Gy min- 1 ) and low dose rate (down to 3.4 mgy min- 1 ) gamma irradiation. There was considerable repair of sb-lethal and sbmtational damage with decreasing dose rate. However, when the cells were held for 5 hors after irradiation, before trypsinization and seeding. the potentially lethal damage was effectively repaired, whereas that leading to mtation (HGPRT locs) remained naffected. Goodhead et al. [G29] interpreted all this to show that the average nmber of mtagenic lesions in a cell is proportional to the average nmber of lethal lesions indced by a given radiation. However. reslts on repair of the potentially lethal damage show that phenomena involved in cell killing and mtation indction can be separated. 17. This conclsion remains to be confirmed. However, if the correlation were generally valid, a linear relationship between freqency of indced mtations and dose of low-let radiation wold be exceptional. becase exponential cell-srvival crves for x and gamma rays are also exceptional (Al7. B63, C9, C21-C23, F2, K21. N8. P7. Pl5. S33. W6. W7]. sigmoid cell srvival generally being the rle [B5. B9-BJ2, B19. B27. B45. CJ, C37, 3, 8, E2. Fl. F2. Fl9. H2. H6, K2. K24, L6, Ml6, M6, S33. Tl8, W6, W7]. In most cases stdied after netron and alpha-particle irradiation. the srvival of diploid mammalian cells is. with few exceptions [HI. C2]. a simple exponential fnction of dose [B5, B9-BI2. B29. B45. J6, K2. K3, R39, Tl8]. {a) 3. Chromosome aberrations Translocarions in germinal cells 171. As discssed in annex Hof the 1977 UNSCEAR report [U6]. the freqency of chromosomal translocations indced by x and gamma rays in spermatogonia, spermatids and oocytes shows a prononced dose-rate and dose-fractionation effect, and data can be fitted to a linear-qadratic eqation. On the other hand, netrons prodce essentially linear dose-reponse relationships withot any significant protraction or fractionation effect. These conclsions have been confirmed by more recent data reviewed in annex I of the 1982 UNSCEAR report [U24] and in annex A to the present report. {b) Somatic cells 172. Stdies of chromosomal aberrations in hman lymphocytes have been reviewed in the past by UNSCEAR (1969 report, annex C, [U9]). and by others [B15, Ll3, LI2. 4]. In most cases, the yield of 193

32 easily identifiable dicentric chromosomes and acentric fragments. indced in G cells, was stdied against the dose of varios radiations. The asymmetric exchanges are not compatible with cell srvival over many generations: however. the yield and kinetics of indction are accepted as representative of symmetrical chromosomal exchanges. The latter may be relevant to present considerations. The sbject of dose-response relationships for indction of dicentrics has been thoroghly reviewed by Lloyd and Edwards [L36]. The review inclded re-calclation of the dose-response fnctions from original data (62 experiments) sing a standardized maximm likelihood method for fitting a linear-qadratic eqation. The main conclsions are: (a) (b) (c) (d) (e) (f) For gamma rays and high-energy electrons. the vale of a 1 in the region of dicentric per cell and gray, appears to be consistent with most observations: For x rays (18-25 kvp), a representative vale of a 1 of abot dicentric per cell and gray seems to apply: The a 2 coefficients for gamma rays. x rays and electrons show less variation than the individal reported vales of a 1 and they clster within the range of dicentric per cell and gray sqare: For fission netrons, no significant a 2 term was observed. Typical a 1 vales wold be 4-9 dicentric per cell and gray. The coefficient tends to increase with lowering of the particle energy. to a peak at abot 35 kev (abot 16 dicentric per cell and gray); In several stdies [K36, VI 1, Zl]. the yield of chromosome aberrations in hman lymphocytes was stdied after doses well below.5 Gy of x rays. When the data on dicentrics from these stdies were pooled [Z l], they fitted a linear fnction of dose passing throgh the origin, as wold be expected from the vales given above. The response for acentric fragments was similar. even thogh less reglar: At netron energies above 15 MeV, a significant vale of a 2 appears: (g) For protons with energies > 7.4 MeV, the a 1 coefficients are very similar to those for x rays. The a 2 vales are rather lower than for typical low-let radiation: (h) (i) For alpha particles, all dose-response relationships (exclding one stdy with possible methodological complications) are linear. bt a I vales scatter widely. owing probably to systematic dosimetric inaccracies; When the vale of a 1 is correlated with LET, the coefficient increases with the latter above 5 kev/,m to a maximm at abot 7 kev/µm, and then falls off sharply at still higher vales of the LET. The a 2 vale is roghly constant for low-let radiation. bt virtally absent for high LET particles The effects of dose rate and fractionation pon dicentric yield in lymphocytes may be smmarized as follows: (a) (b) For low-let radiation. the yield decreases with decreasing dose rate (or the extension of irradiation time) and with increasing fractionation intervals. This effect is attribted to the repair of interacting lesions. whose mean life appears to be abot 2 hors. althogh some of the lesions may have a mch longer life. Available experimental data and their theoretical interpretation sggest that, for very long irradiation times, the a 2 coefficient wold be zero and the a I wold remain naffected (linear relationship); For netrons. no effects of dose protraction or fractionation were seen The RBE of netrons verss x or gamma rays is inversely proportional co the sqare root of the netron dose. The limiting vale at low dose rates may be obtained from the ratios of the relevant a 1 coefficients This was also noted when whole blood was irradiated by adding tritiated water prior to initiation of blastic transformation. Doses between.2 and 4 Gy /3 radiation were delivered over 3 mintes or 24 hors. Whereas the dose-response crve for 3-mintes exposres was practically the same as that after acte x rays. the response was close to linearity after longer exposres (the linear component remained nchanged while the qadratic one decreased by a factor of abot 3 [P2]) Similar relationships were obtained for dicentrics in the lymphocytes of all mammalian species stdied so far (marmoset [B41]. Chinese hamster [B41], rabbit [BI. B2. S5]. pottoroo [S5]. swine [BI. B4L L 7]. sheep [L7]. goat [L7]. mose [823. B41], rhess monkey [B61]. and other primates [H9, M37]). With protraction or fractionation of x- or gamma-ray doses. the linear term prevails [S4. L9]. With irradiation times exceeding 24 hors the relationships for dicentrics after exposre to gamma rays tend to approach linearity For acentric fragments. a linear-qadratic relationship is sally observed. bt the reslts are less reprodcible [B2, LI3]. As these aberrations inclde both interstitial and terminal deletions. classified together. the relationship does not contradict linearity for single-break aberrations In conclsion. indction of chromosomal aberrations of the one- and two-break type. in somatic cells in G or G 1 is in broad agreement with the linear-qadratic model for low-let and with the linear models for high-let particles. This agreement refers to dose and to the effects of dose rate and fractionation. For single-break aberrations. a linear model shold fit the data. bt methodological problems make the demonstration difficlt. 4. Oncogenic transformation of cells in vitro (a) General 179. Stdies of transformation from the normal to the oncogenic state, on cells in vitro. are very relevant 194

33 to the prposes of the present annex. They allow investigation of the mechanisms of cancer initiation withot interference of other factors at the tisse and systemic levels. In addition. killing and transformation may be measred in the same target cells. allowing the relative importance of the two effects to be stdied at the same time. Morover, in vitro experiments are less expensive and time-consming than stdies on irradiated animals and are easier to analyse, owing to the Jack of competing risks and the higher precision of measrements. On the other hand, lack of close intercelllar contact and systemic reglation dring cltre growth is an artifact compared with the sitation in sit. and cold change-at least qantitatively-the reactions of cells stdied. Details of cell handling, sch as whether trypsinization has been applied to the cltres, or not. in the corse of the experiments, may also sbstantially modify the reslts. So far. observations on cell transformation have been limited almost exclsively to fibroblasts. whereas tmors of epithelial origin prevail among radiationindced cancers in man. -' -' I.LI C!l z >... > ~ =:, Vl ~ I.LI a. Vl 1- :z: ~ ~.. Vl :z: ~ I ! 25 kvp x rays o 43 kev netrons 95% confidence limits 1_6...,_~... ~.,...~~... ~~~.--~~ (b) Dose-response relationships 18. Transformed cells are assessed by scoring characteristic colonies recovered in a cltre. This directly measres the transformation freqency per srviving cell. Most experiments are reported in this way, bt. with some additional effort, information can be obtained on cell srvival as a fnction of dose. and on the plating efficiency. The transformation yield per initial cell at risk may be calclated by correcting the transformation freqencies per srviving cell for the respective srvival vales and plating efficiency. This latter measre of transformation can seldom be compted from the freqencies per srvivors reported in the papers, which often makes it difficlt to give a precise description of the dose-response relationships. Many featres of the dose-response relationships depend strongly pon the time interval between the establishment of a cltre (seeding of cells) and irradiation: new and established cltres shold therefore be dealt with separately. (i) Irradiation of freshly established cltres 181. Borek and Hall [B33] x-irradiated early-passage golden hamster embryo cells in vitro, 24 hors after trypsinization and seeding. with doses from. l to 6. Gy (a point at.3 Gy x rays was added later [B76]). A dose-response relationship per srviving cell is presented in Figre VII, showing a rise with dose p to l Gy. a platea between 1.5 and 3. Gy, and a decline thereafter. The dose-response relationship from the same and sbseqent experiments was also expressed per initial cell at risk [B29]. It is impossible to say whether a linear or an pwards concave line wold be a better fit of the experimental points along the ascending portion of the crve. A linear fit was assmed by the athors themselves bt a linearqadratic fit was also attempted by Chadwick and Leenhots [L3]. In a linear plot the line tends to pass throgh the origin, sggesting no threshold. Below DOSE (.Gy) Figre VII. Dose-response crves for transformation of hamster embryo cells by netrons and x rays. The crves were obtained by non-parametric least-sqares fitting. [B79].5 Gy of x rays and down to.1 Gy the yield of transformants appears to rise with a power of dose < I. The transformation freqency per Gy of x rays below I Gy is of the order of 1-2 per cell. which is several orders of magnitde greater than for any somatic mtation [U6] bt almost an order of magnitde lower than the total yield of chromosomal aberrations indced by comparable doses of x rays in mammalian lymphocytes [D4]. In a comparison of 3-kVp x and 6 Co gamma rays [B76] the RBE of the latter varied from.85 at abot I Gy to.5 at a few tens of mgy Borek. Hall and Rossi [B29] and Borek and Hall [B79] irradiated hamster embryo cells with monoenergetic netrons (.43 MeV) and compared the dose-response relationship with that for x rays. The crves representing the relationships. both per srviving cell or per initial cell at risk, were peaked in shape and roghly parallel. A fit to the netron data similar to that for x rays. bt shifted toward lower doses, was obtained (Figre VII) with no sggestion of a threshold. Below.1 Gy, the transformation freqency cold also be interpreted as rising with a power of dose lower than nity. The RBE of netrons. at transformation freqencies of 1-3 and 5 1-3, was abot 12 and abot 6, respectively. Measrable transformation was indced even by I mgy of netrons. Accelerated argon ions (429 MeV per atomic mass nit) were roghly as effective as netrons [B29] Data were also obtained for cell srvival [B29]. For x rays a sigmoid srvival crve (Do= 1.47 Gy; n = 6). and for 43-keV netrons a prely exponential crve (D =.5 Gy), were reported. The RBE of netrons varied approximately with the inverse sqare 195

34 root of the nerron dose, with vales from abot 3 at doses below.1 Gy to abot 5 at 1.5 Gy. At doses above 2-3 Gy of x rays, cell killing was appreciable. bt the decline of transformation yield per nit dose above I Gy of x rays and.25 Gy of netrons was explained by assming that transformed cells were preferentially killed by radiation. An alternative explanation of this phenomenon was also proposed [C7, L3] The dose-response relationship for x-ray-indced transformation of C3H lotl/2 cells irradiated 24 hors after seeding was also stdied [T7] and expressed per srviving cell (at cell densities below 3.8 cell cm- 2 : 1-kV x rays, to 15 Gy). The transformation freqency rose rapidly with dose, p to 4 Gy, and reached a platea at abot 15 Gy. At abot I Gy. the transformation freqency was abot.5 1-4, roghly 2 orders of magnitde lower than for hamster embryo cells [B33]. The doses reqired to doble the transformation yield along the ascending branch of the crves for these two types of cells were 1. and.1 Gy, respectively. The dose-response crve for C3HIOT1/2 cells showed no apparent threshold and was concave pwards, the effect rising with a power of dose > 1 and close to In a later stdy. Miller et al. [M23, M41) examined, nder a similar experimental protocol in the same C3HlTl/2 cells, the effect of single and split doses of 3-kVp x rays. The single-dose response crve had a complex, crvilinear shape. The transformation freqency was less than proportional to dose between.1 and.3 Gy, almost flat between.3 and abot I Gy, and then rising with the sqare. or higher. power of dose. A plot of the data at.1-1 Gy on a log-log scale, per srviving cell, is given in Figre VIII. Interpretation of the shape of this dose-...j...j w <.!) z - > > c:r:: ::, VI c:r::.j Q.. VI... z i I.L. single doses VI z split doses... ~ DOSE. (Gy) Figre VIII. Dose-response relationships for the nmber of transformants per srviving C3H1T1/2 cells, following single or spilt doses of 3-kVp x rays. In the case of split doses, the radiation was delivered In two eqal exposres, separated by 5 hors. [M23] response relationship is very difficlt, particlarly in the dose range below I Gy Yang, also, stdied the transformation of C3 HI OT 1 /2 cells by x rays and energetic silicon ions (initial energy 32 MeV per atomic mass nit. residal range in water 4.5 cm, dose average LET= 88 kev/µm) [Y5]. This radiation was more effective than 225 kvp x rays (RBE not given). Cells kept in a highly conflent state for some time after irradiation were able to repair part of the transformation damage indced by x rays, bt not that indced by accelerated silicon ions When C3H IOTl/2 cells were irradiated with netrons (35-Me V d+ - Be). the reslts were as in Figre IX [B79]. The dose-response crve is shifted towards lower doses with a similar dose-independent platea as for x rays. The RBE increased with decreasing dose and was similar for both transformation and cell killing Lloyd et al. [Ll 5] stdied transformation in C3H lotl/2 cells by irradiation with 5.6-MeV alpha particles, showing a higher effectiveness as compared with x rays (the relationship of irradiation- and seeding-time is not available). For alpha particles, the dose-response relationship is of a grossly crvilinear (concave pward) character, the effect rising with approximately the cbe of the dose. When expressed per srviving cell. the maximm transformation rate was seen at abot 2 Gy C3H JOTI/2 cells were transformed by 1-kVp x rays, fast fission netrons (mean energy.5 MeV. 8-2% gamma contribtion) and cyclotron netrons (mean energy 38 MeV, 8% gamma contribtion). The dose ranges were: Gy,.5-5 Gy, and Gy, respectively [B89]. The cells were irradiated 36 hors after seeding. For fission netrons, the response per srvivor was most effective and the initial portion of the crve was only slightly concave pwards. X rays were least effective and 38-MeV netrons prodced crves very similar to those of x rays. For both radiations. the dose-response crves were grossly concave pwards at their ascending portions. A platea of effect at high doses was reached for all radiations tested. It is worth mentioning that for each radiation the srvival crves had prononced sholders. which were roghly proportional to the degree of crvilinearity of the dose-transformation crves. The RBE relative to x rays at the transformation freqency of was 1.2 and 3.8 for 38-MeV and.5-mev netrons, respectively. 19. Transformation of BALB/3T3 cells by 238 P alpha particles and x rays was stdied by exposre of freshly seeded single cells or conflent cell monolayers [R39]. The dose ranges were Gy and.5-5 Gy for alpha particles and x rays, respectively. When calclated per srviving cell the rise of transformation freqency with dose was exponential (steeper than proportional). However. when the reslts were expressed as transformants per exposed cell, for alpha particles the highest yield was indced already by a dose of.5 Gy, with a slight decline thereafter. This I96

35 ...I...I w ~... z... > > c::: =:, Vl ~ w.. Vl 1- z ~.. Vl z ~ I- 3 kvp x rays + o 35 MeV d... Be netrons 95% confidence limits DOSE (Gy) Figre IX. Pooled data obtained for the fraction of transformants per srviving C3H1T1/2 cell following Irradiation with x rays or netrons. [B79] may be attribted to highly efficient cell sterilization by alpha rays. A high effectiveness of the alpha particles. relative to x rays, was noted. bt the precise shapes of the crves at the lower end of the dose scale cannot be established. Little stdied the transformation by x rays of freshly seeded cells (2 hors prior to exposre) of BALB/3T3 mose [LIO]. Althogh there was some indication of an pward concave crve, the statistical ncertainty of the data does not allow precise conclsions Golden Syrian hamster embryo fibroblasts, seeded abot I 2 hors before, were exposed for 17 hors to varying concentrations of methyl 3 H-thymidine [L33]. The nmber of transformants per srviving cell (and mtants at HGPRI locs) showed a proportionality to the concentration of the sbstance. Tritiated ridine. which is not incorporated into DNA. was ineffective in indcing both oncogenic transformation and mtations BALB/3T3 embryo fibroblasts. synchronized in the S-phase, were incbated for 16 hors with the DNA precrsors 3H-thymidine and 1 2s1-iododeoxyridine ( dUrd) [L34]. The incorporation of the radionclides into celllar DNA was proportional to concentration in the cltre medim. The transformation freqency per srvivor. verss concentration of activity per cell, was linear in both cases-withot sggestion of a threshold-p to a freqency of abot _ The ratio of the initial slopes of 12SJ-dUrd verss 3 H-thymidine dose-response crve was abot 25. In the same experimental series, the transformation by x rays was somewhat irreglar, with an approximately exponential increase of the transformation freqency with dose over the whole range of doses (.5-6 Gy). The high effectiveness of 12 5I-dUrd was explained by the release of a clster of low energy Ager electrons (abot 25 per decay). leading to a very high energy deposition in the vicinity of the DNA doble strand. (ii) Lare irradiation of cltres 193. Dose-response relationships for in vitro transformation of cell line C3H I OT 1/2 by actely delivered x rays (5-kVp,.18 mm Al filtration) and netrons (fission, mean energy.85 Me V) were reported by Han and Elkind [HJ]. In contrast with the experiments reviewed above, cells were irradiated 48 hors after seeding. When the transformation rate was expressed per initial cell at risk, the decline of the yield beyond the maximm had the same slope (or D ) as for the killing of non-transformed cells. sggesting that sterilization by high doses of normal or transformed cells was similar. On linear co-ordinates (Figre X). the ascending portions of the dose-response crves (per initial cell at risk) for x rays was concave pwards (at least above Gy) bt details cannot be established from the graphs. For netrons, the ascending portion of the crve appears to be close to linearity. The RBE of netrons increased with declining doses, and the maximm attainable transformation freqency by netrons was higher by abot 5% than that for gamma rays (per initial cell at risk). In contrast with the data of Miller et al. [M23] (Figre VIII) there was a significant flattening of the yield above 4 Gy of x rays. when the data were expressed per srviving cell [Hl4] (see Figre XI) Watanabe et al. [W 18] irradiated golden hamster embryo cells 72 hors after establishing the cltre. The exposres varied from 5 to 6 R at exposre rates of 5, 75 and 6 R/min. Additional exposres (1-5 R) were applied at the dose rate of 5 R/min. The 197

36 :,.:: V'I "' < _, _, (x 1-4 ) L,..J _, :::... :z: "' I.LI V'I... z 6 < ~... V'I... z < 4 "' DOSE (Gy) 5 kvp x rays netrons Figre X. Changes in transformation freqency of C3H1T1/2 cells after single exposres to x rays or fission-spectrm netrons. [H1] a: :::: a: > ::, V, a: w >- z:... ::, O'... i.. "' z: ~ < ~ Gy min V'I -1 :z: Gy min ~ D,OSE (Gy) Figre Xl. Freqency of neoplastic transformation of C3H1T1/2 cells expressed per srviving cell as a fnction of dose of gamma rays delivered at high or at low dose rate. Error bars represent the standard errors of the data pooled from varios experiments. [H14] logarithm of transformation freqency per srvivor was approximately linear p to 1 R, and had a very shallow increase from 2 R onwards (semi-platea). From the graphical presentation of data. it appears that p to 75 R, at 5 R/min. the dose-response relationship is linear, with no sggestion of a threshold Terasima et al. [T24] stdied C3HIOTI/2 cells, x-irradiated (.23 to 9.3 Gy) in the platea phase of growth (contact inhibited). The dose response (per srviving cell) was very similar to that fond by others for the same cell line. After a steep exponential increase, p to abot 1.8 Gy, there was a gradally slower increase of the transformation freqency with a platea above 6-8 Gy. Recent data on in vitro transformation of C3H JOTl/2 cells by 31-MeV protons were pblished by Bettega et al. [B3J, showing that the transformation freqency per srviving cell had a marked change in slope at abot 2 Gy. (i) (c) Effects of dose fractionation and protracrion; repair of sb-transformation and potential transformation damage Irradiation of freshly established c/rres 196. Irradiation of cltres seeded 24 hors before exposre with two doses of Gy of x rays at varios intervals (zero to five hors) showed that srvival of C3H JOT 1/2 fibroblasts nearly dobled when the interval exceeded 2 hors, with no frther increase for longer intervals [T4]. For doses of 1.5, 3. and 8. Gy in two eqal fractions at five-hor intervals. the transformation freqency per srviving cell remained naltered at the lowest dose, bt declined significantly (p =.5) at the two higher doses Hamster embryo cells in cltre (seeded 24 hors earlier) showed a dobling of the effect after split, as opposed to single, doses of.5 or.75 Gy [B9]. At the same time, the srvival of irradiated cells increased only marginally. When the experiment was extended [B64] by stdying the effect on the transformation freqencies of splitting 1.5, 3 and 6 Gy of x rays, no enhancing effect of 2 X.75 Gy. as opposed to 1 X 1.5 Gy. was noted. Fractionation of 3 and 6 Gy led to a redction of the total effect Miller and Hall [M23] stdied dose fractionation on C3H IOTI/2 fibroblasts. plated 24 hors before x-ray treatments (.3 to 8. Gy). Their reslts confirmed the sparing effects of splitting the dose above 1.5 Gy (with a possibly incidental deviation of the points at 4 Gy) and an enhancement of transformation below that dose. The experiment was repeated [M41], and the dose range expanded to inclde.1 and 1 Gy. The reslts are presented in Figre VIII. In addition. a dose of I Gy was split into 2. 3 or 4 eqal fractions [H2l]. An almost proportional (2-, 3- and 4-fold) enhancement of transformation was observed. Also. when a dose of I Gy of 6 Co gamma rays was delivered at a low dose rate (over 6 hors) the effect was significantly enhanced, by a factor of abot 3, by comparison with acte exposre (over 1 min) [H21]. In conclsion, sing this experimental protocol. in the range.3-1 Gy there is an enhancement of the effect by splitting the dose into two eqal fractions within 5 hors or by protracting the dose over a similar interval. At 2 Gy, there is no effect. There is a decline of the yield after fractionation at higher doses. Approximately the same trend is seen when the data are expressed per initial cell at risk. 198

37 199. LiHle [LIO] reported dose-fractionation experiments on BALB/3T3 cells seeded 24 hors before the first exposre with essentially identical reslts. The fractionated doses ranged from.1 to 3.5 Gy, and similar observations were also made in still another stdy by Szki et al. [S44]. It may be conclded, therefore. that enhanced transformation (per srvivor) below 1.5 Gy of x rays after fractionation within 5 hors is a phenomenon observable in all cell lines stdied so far, irradiated in freshly established cltres. Sch an effect might be dedced from the crves in Figre VIII [M23]. In fact. if an independent action of the two dose fractions is assmed, the decrease above 2 Gy. where the response rises with 2 or with a higher exponent, is easily nderstood. Similarly. below I Gy. over the platea. where the effect is roghly independent of dose, fractionation shold increase the yield, as observed. 2. Figres VIII and IX show clearly that a linear extrapolation down from intermediate ( I Gy) and high (2-3 Gy) doses of x rays wold lead to an nderestimate of the real transformation freqency, particlarly with respect to the effect of split doses in the low-dose region. The greatest nder-estimate, by a factor of 4-6. wold be encontered when the effect of single doses in the range 1-3 Gy is linearly extrapolated to predict effects of split total doses of.1-.3 Gy. The relevance of this finding to tmor indction in vivo has been commented pon by several athors [B76, H21. M41]. (ii) Late irradiation of cltres 21. Data on fractionation were reported by Han and Elkind [HI] for x rays (.75 Gy) and fission netrons (3.8 Gy) on C3H IOTl/2cells exposed 48 hors after seeding. The transformation rate after doses of this magnitde was on the declining part of the crve when expressed per initial cell at risk. Fractionation of the netron dose within O to 16 hors had little. if any. effect on single-dose srvival and cased a slight redction of the transformation yield. perhaps within the limits of experimental error. With x-ray fractionation times p to 5 and 1 hors, the freqency increased and then declined to the level of the nonfractionated exposre at abot 16 hors. The increase was within a factor of 2 and its statistical significance ncertain. When expressed per sr,;vor, the transformation freqency after x irradiation declined steadily to a platea at hors fractionation interval, reflecting essentially an expected and considerable effect pon srvival of the non-transformed cells. This observation. therefore. confirmed other findings [M23. M4I) that fractionation of high x-ray doses in this cell line redces the transformation freqency when expressed per srviving cell. 22. Han and Elkind also irradiated C3HITl/2 cells. 48 hors after plating. with 6Co gamma rays at high (1. Gy min-1) and low (.5 Gy min-1) dose rate [H 14]. The range of doses varied between.2 and 13 Gy. The transformation freqency per srvivor (Figre XI) was consistently redced at the lower dose rate. as was cell sterilization. The redction in transformation yield was by a factor of abot 5 above 6 Gy and by a factor of abot 3 at 2 Gy, and was still present at doses of.5 and.2 Gy [E9. H26]. 23. Hill et al. [H3] stdied the effects of gammaray fractionation in the same system. Throghot the dose range tested (.25-3 Gy). fractionation of the dose into 5 daily fractions reslted in a significant redction of transformation freqency. Up to 1.5 Gy delivered nfractionated. and p to 3 Gy fractionated regimes. the dose response cold be fitted by a straight line passing thogh the origin (Figre XII). The yield (x 1-4 ) 8 I Gy/min O. 5 Gy/min SLOPE ( Gy - I ) "" a 7 CORR, COEFFICIENT ~ > SLOPE RATIO (1').31 a: =:, 6 a, "'... Q.... s z... =:,... c,::... 4 ts,,... J,,.,.,,,,,.,! a... "' z 2 < c,:... I Gy/mln.5 Gy/min 2 3 DOSE (Gy) Figre XII. Transformation freqency of C3H1T1/2 cells per srviving cell against dose of gamma rays. Open circles: acte exposre at 1 Gy mln- 1 ; Dashed line: continos exposre at 1 mgy mln- 1 ; Closed circles: five fractions at.5 Gy mln- 1. given at 24-hor Intervals. [H3] of transformants per nit dose was 3 times lower after fractionated than after single doses. and very similar to the dose-response crve for the lower dose-rate exposre referred to above [E9. H26]. Increasing the fractionation intervals beyond 24 hors, and redcing the dose per fraction, led to a frther slight redction of the effect. This sggested that there may be a limit beyond which frther dose protraction cannot redce the transformation freqency. Trypsinization per se, or delivery of the dose within 7 days of growth in cltre, did not affect the yield of transformants per nit dose, provided they took place at least 4 hors after seeding. Conflence of the cells was exclded as a possible reason for the observed fractionation effect. These reslts of dose fractionation and protraction below 1.5 Gy are in striking contrast with those observed when the cells were irradiated 24 hors after seeding [H21, M23]. The difference was attribted to atypical conditions of celllar growth early after plating and to the action of radiation on parasynchronos cells [H3]. 24. Watanabe et al. [W 18) irradiated golden hamster embryo cells. sb-cltred 72 hors earlier. with x rays 199

38 at varios exposre rates: 5, 75 and 6 R/min. Throghot the whole range of exposres tested (5-6 R), the lower dose rates prodced a redced yield of transformed cells. even within the 5-1 R range where the dose-response relationship was practically linear. At eqal srvival levels (which were inversely related to dose rate). the transformation freqency was higher in cells irradiated at higher than at lower dose rates. Terasima et al. [T24] stdied the effect of fractionating x-ray doses (.93, 1.86 and 3.72 Gy) in C3HTIOl/2 cells irradiated in the platea phase. Eqal fractions were spaced at intervals of 3. 1 and 15 hors. There was significant redction of the yield at all doses tested, as compared with single doses, the maximm redction being achieved already at the 3-hor interval. 25. Hill et al. [H26. H34] worked with fissionspectrm netrons (mean energy.85 MeV) at, arios dose rates in C3H!OTl/2 cells irradiated 48 hors after seeding. There was no discernible effect pon cell srvival of netrons delivered at a low rate ( mgy min- 1 ) verss a high rate ( Gy min- 1 ). At the same time, the low-rate netron irradiation enhanced significantly the transformation freqency at doses below I Gy (by a factor of abot 9 at.25-.l Gy) (Figre XIII), an effect that cannot ls 15 ~ ~ Bl C:.... >- ~ 1 &... C:... a,- i 5 a... V') ~... SLOPE (Gy-l) CORR. COEFFICIENT SLOPE RATIO (\).2.38 Gy/min Gy/min 53.o,o Gy/min.86 Gy/mfn DOSE (Gy) Figre XIII. Oncogenlc transformation of C3H1T1/2 cells exposed to fission-spectrm netrons al high and low dose rate. Dashed lines Indicate the linear regressions fitted to the initial portions of the crves. [H34] 1. response crves at low doses was abot 8. The increased effectiveness of the low dose rate and of the fractionated netron irradiation was explained either by the indction of more efficient error-prone repair or by facilitation of the expression of ''sb-effective transformation damage", whatever this means [H37]. A recent stdy on transformed mammary epithelial cells transplanted in vivo to virgin female BALB/c mice sggests that the effect of fractionated netron exposres may be analogos to chemical promotion and is de to an enhanced expression of the transformation damage [U28]. The effects of fractionation schemes applied to the cells in an asynchronos growth indicate that x and gamma irradiation-by analogy with sblethal damage-reslt in error-free. reparable. sb-transformation damage. The transformation damage indced by netrons either does not ndergo repair, or is sbject to error-prone repair [E 1, H37]; however. why the latter shold be more effective at protracted or fractionated low doses is not clear. 27. Terzaghi and Little [T5] fond that irradiating C3H!OTl/2 cells with x rays in a density-inhibited state, and delaying plating p to 48 hors sbstantially increased srvival rates. Delay for 2 to 4 hors cased marked increase of the transformation rate. Frther delay of plating redced the yield of transformants to very low levels after 48 hors. Terasima et al. [T24]. however, did not observe any initial increase of the transformation freqency. as the potential transformation damage fell over the first 3 hors of delayed plating. The rate of potential transformation repair in C3H 1 OT I /2 cells depended strongly on a particlar batch of feta! calf serm sed in the cell cltre medim, both before and after release from conflence [TI I]. Lack of serm was accompanied by a high degree of repair (8% over 6 hors) after keeping the irradiated cells (x rays Gy) in the non-proliferating state. Some batches of serm totally prevented the occrrence of repair. No repair of potential transformation cold be fond after irradiation with alpha particles [R39] and accelerated heavy ions [Y5]. 28. When potentially lethal damage was stdied by holding these cells in conflence after irradiation for varying lengths of time. a prononced repair was observed for x rays and none for alpha particles [R39]. Eqally. no repair of potential transformation damage was seen for p to 22 hors after alpha particle exposre. whereas the x-ray-indced potential damage was efficiently repaired. be explained on biophysical gronds. Above doses of Gy. the effect of the dose rate disappeared and dose-response relationships converged to a platea of transformants per srviving cell. 26. The inflence of dose fractionation (5 fractions over 4 days) of fission-spectrm netrons. delivered at a rate of O. l Gy min- 1 was stdied by the same athors. Over the whole range from.13 to 1.12 Gy. an enhancement of effects by fractionation was observed. The ratio of linear slopes of the dose- 29. The reslts of other experiments [L23] sggest that two separate processes may be involved in the recovery of damage: first. DNA repair leading to enhanced srvival: and, second, a slower error-prone repair process responsible for changes in DNA and leading to mtation and transformation. It shold be stressed that the freqencies of transformation reported for most in vitro experiments apply to actively proliferating cell poplations and may not be representative of sitations where the overwhelming majority of cells is in the resting state, as in most organs in vivo. 2

39 (d) Other factors affecting oncogenic transformation in vitro 21. Nmeros factors affect the yield of in vitro oncogenic transformation. Among these, a basic role is played by both the time and the density at which cells are seeded Stdies on three cell lines (hamster embryo. C3HIOTJ/2 mose embryo and mose BALB/3T3 cells), have shown [B34, Kl2, LIO, T6] that several post-irradiation cell divisions-4 to 6, depending on the cell line stdied-are necessary for expression or fixation of the transformed state. Of these, at least one mst take place in the first 24 hors post-irradiation Terzaghi and Little [T7] investigated the inflence of cell density on the transformation freqency per srviving cell. Above 4 viable cells per dish (abot 5 cell cm- 2 ). there was a steep decline from platea vales of transformation observed at lower cell densities. The athors sggested that this effect cold be the reslt of an incomplete expression of transformation de to early conflence, which lowers the nmber of cell divisions dring exponential growth below that necessary for fll expression of transformation [87 L B72, L26]. Han and Elkind observed, in the same cell line, a similar relationship between the density of seeded cells and the transformation rate after x-ray and netron irradiation [HI]. Direct celllar contact cannot operate at sch cell distances, bt a diffsible mediator sbstance might prevent occrrence or growth of the transformed clones [Ll5] Lloyd et al. [LI 6] experimented on co-cltivation of transformed C3H JOT 1/2 cells with nontransformed ones. When transformed cells were seeded at low densities. together with normal cells, at some ratios of the two cell types (1:5 to 1:5. respectively). expression of the malignant state cold be completely prevented. Adding non-transformed cells in greater nmbers again elicited this response. The experiment points clearly to a very complex natre of the interaction between transformed and non-transformed cells. even in vitro The data by Kennedy et al. [K26] are of importance becase they throw some light on the complexity of the phenomena leading to oncogenic cell transformation in vitro and cold modify sbstantially the crrent interpretation of this radiation effect. These athors stdied the inflence of resspending C3H IOTI/2 cells after irradiation. once conflence was reached. When diltions at the time of re-seeding were progressively increased, the nmber of originally irradiated and inoclated cells per plate decreased to very low vales. In spite of that, after a constant dose of 4 Gy of x rays. the nmber of transformed foci per plate remained practically constant. This indicates that the nmber of transformed foci per plate is apparently independent of the nmber of cells initially irradiated. It is interesting to note that similar reslts were obtained when the effect of inoclm size was stdied on the appearance of tmors, after in vitro irradiation of monodispersed thyroid or mammary cells injected into the fat pads of rats [G3]. These observations are in direct conflict with the expectation that-if initiation of a cell is a rare phenomenon-the probability of observing a tmor shold rise in proportion to the size of the inoclm It has been sggested [K26, G3] that exposre to radiation reslts in some change of the celllar state in many or, after high doses, in all cells. Gold postlates that initiation is not an all-or-none phenomenon, bt that its intensity rises with dose p to a satration level, which is reached in C3HlTl/2 cells at 5-6 Gy after 6Co gamma irradiation or 3 Gy of netrons. The change is probably transmitted to the whole progeny of srviving cells. The natre of initiation-genetic or epigenetic-is nknown, bt the latter wold presmably be sggested by these observations. This interpretation is not easily reconciled with the demonstrated direct involvement of DNA in the carcinogenic process [814. L33]. Only a second rare phenomenon (perhaps mtation), wold lead to a phenotypically recognizable transformation at conflence of C3HITl/2 cells. The probability of this second phenomenon shold be in some proportion with the degree of initiation A report by Hall et al. [H28] challenged the hypothesis that commitment of irradiated cells to the transformed state takes place at conflence. By re-plating cltres at different times after irradiation and testing for distribtion of transformed foci among the plates. the athors cold show that commitment takes place within I week of exposre. This observation is in agreement with a stdy [891] on in vitro oncogenic transformation of C3HJOTl/2 cells by 7 P,8a-dihydroxy,9a, J Oa-epoxy, 7.8,9. I -tetrahydrobenzo(a)-pyrene. The reslts of this stdy showed that acqisition of the ability to form transformed foci occrred within 2 days of exposre to a carcinogen. Hall's experiments [H28] also showed that increasing the density of cells in cltre probably sppressed expression of the transformed state. These reslts do not contradict the fact that at 4 Gy of x rays the probability of initiation per cell is close to I. bt they, have been interpreted to challenge the postlated nonstochastic natre of initiation [K26] The presence of thyroxin in the cltre medim at a physiological concentration was shown to be a prereqisite for transformation to occr in Syrian hamster embryo and C3HITl/2 cells [B73. G26. G27]. Thyroxin is reqired for protein synthesis only within a few hors of irradiation and its action can be reversed by an agent blocking protein synthesis. e.g.. cycloheximide Other factors were shown to enhance transformation in cltre dring or after irradiation. Extensive investigations were made on a well-known promoter of in viva carcinogenesis, the active ingredient of croton oil, 12--tetradecanoyl-phorbol-13 acetate (TPA). By adding TPA to the cltres within 96 hors of irradiation, Kennedy et al. [KIO] showed enhancement of the x-ray transformation freqency (or its expression) in C3H IOTI/2 mose embryo cells. 21

40 The effect was particlarly marked after doses of.5-1. Gy. These data indicate that promoting agenrs can increase the levels of x-ray-indced transformation in vitro, jst as they enhance carcinogenesis in vivo. It was also shown [K27] that the most effective period of exposre to TPA is that of exponential growth. However. a delay of TPA application, e.g.. addition to cltres re-seeded after conflence, did still lead to a significant enhancement of oncogenic transformation When TPA was added for six to seven weeks (at.1 µg/ml of medim) to cltres of C3H1Tl/2 cells irradiated with graded doses of x rays ( Gy), there was not only an absolte increase of transformation freqency, bt also a change in the shape of the dose-response relationship. from crvilinear (pward concave) after x rays alone to linear (Figre XIV). The _J _J.J.J _J 6 co ~... > :: 5.J c.. >- :z 4.J ::, O'.J :: LI.. 3 :z... I- ~ ::e: :: LI.. V, :z ~ :: I DOSE (Gy) Figre XIV. Inflence of post-irradiation Incbation with the tmor promoter TPA on x-ray transformation of mose C3H1T1/2 cells. O x-lrradlatlon onl,v; t::,. TPA (.1 µg/ml) added for the entire period of expression. [L26] action of TPA was especially prononced at doses in the low and intermediate range [L26]. A very similar effect of TP A on the shape of the dose-response crves for x-ray-indced and netron-indced transformation of C3HITl/2 cells was also seen in another stdy [H22]. 22. Nmeros mechanisms of action for TPA have been advocated, inclding inhibition of the cell-to-cell commnication [H22, TIS]: enhancement of chromosomal rearrangements [K28]; prodction of shortlived radicals [B76]: and action via mechanism of celllar differentiation [B74, S43]. Increase of celllar proliferation is a phenomenon that often accompanies promotion [F 14], bt it is dobtfl whether it is at all necessary. or sfficient [D13, K27]. The TPA promoting effect is apparently not related to any of the known DNA repair processes [K27] Enhanced transformation of varios cell lines was also noted after addition to the medim of the epidermal growth factor, a natral hormone-like polypeptide [FIS]: bromodeoxyridine [R3]; high concentrations of inslin [U29]; cortisone [K3]: and interferone [B46]. The particlar batch of feta! calf serm sed may also have a profond inflence on the transformation freqency eventally observed [TI l]. A new promoter, dihydroteleocidin 8, an antibiotic derivative recently discovered, has a promoting capacity per nit mass abot 1 times higher than that oftpa [H3I]. A synergism between x rays and a food pyrolysate prodct 3-amino-1-methyl-SH-pyrido-( 4-3b )indol (Trp-P-2), itself a mtagen and carcinogen isolated from boiled meat and fish, has also been shown [876] in Syrian hamster embryo cells. Hyperthermia (43 C for 6 mintes or 45 C for 15 mintes before irradiation) cased some increase of the x-ray-indced transformation freqency in C3HITl/2 cells [C33]. The natre of most enhancing agents mentioned above sggests that the mechanisms of action are epigenetic rather than likely to case strctral changes in the genome itself There are also sppressing (inhibiting) factors. Kennedy and Little [K35] and Little [L26] described sppression by the protease inhibitors antipain and lepeptin. In other stdies [B65, G28] it was shown that antipain added to cltres before irradiation enhanced the yield of transformants. while addition shortly after exposre redced the transformation freqency. and addition 1 or 2 days later was ineffective. Inhibition of radiation-indced oncogenic transformation was also observed when inhibitors of poly(adp-ribose) synthesis (benzamide. 3-aminobenzamide) were added to cltres of C3HIOTl/2 and golden hamster embryo cells [B96] An inhibition (by a factor of 3) of transformation in C3H IOTl/2 cells was observed by adding. 24 hors before exposre. a non-toxic derivative of vitamin A, trimethyl-methoxy-phenyl, analoge of N ethyl retinamide (Ro-H-143) and other retinoids [B73]. The retinoids irreversibly sppressed oncogenic transformation when present in cltre for only a few days after irradiation. Later addition of TP A to the cltre remained ineffective. The retinoids may act by sppressing the carcinogen-indced progression of the neoplastic process [B76] Selenim componds [B73], actinomycin D [K31] and lymphotoxin [D19. E6, R44] also inhibit radiation-indced transformation. When speroxide dismtasc was added to irradiated C3H lotl/2 cltres from irradiation to expression of transformation, the freqency of radiation-indced transformants was 22

41 greatly redced, whether or not the drg was present dring x-ray exposre [M49]. Qantitatively. the effects of oxygen and misonidazol pon oncogenic transformation and cell killing of C3H lot 1/2 cells are very similar [897]. (e) Conclsions 225. The oncogenic transformation in vitro of hamster embryo cells and several established mose fibroblast lines by ionizing radiation shows nmeros similarities in respect of dose-response relationships, effects of dose fractionation, and inflence of cell density pon transformation freqency The mechanisms of transformation are still imperfectly nderstood and several methodological qestions are not yet flly resolved. The sbject has been thoroghly reviewed [B7 l. B76, G34, L26. Y3]. Transformation is probably a complex mlti-stage phenomenon that incldes initiation, progression and final expression of malignant foci. A nmber of promoting and sppressing factors may affect the freqency of transformation and therefore the shape of the dose-response relationships For these reasons, analysis of dose-response relationships for low- and high-let radiation mst, for the time being, remain descriptive. After single doses of x rays or netrons, the dose-response crves have some featres in common with those seen in viva for many experimental tmors. For example, when expressed per initial cell at risk. transformation shows an initial rise of the freqency with dose, then the crve reaches a peak and the yield declines at still higher doses. The lower efficiency of gamma rays at the low. as opposed to high, dose rates in cltres established at least 4 hors before irradiation is also in accordance with most observations in vivo. The same applies to the RBE vales of netrons and alpha particles, which are definitely higher than nity and tend to decline with increasing dose. in agreement with the general nderstanding of celllar radiobiology [Bll, K7]. Netron dose fractionation pon oncogenic transformation in vitro prodces effects that are similar to those seen recently in some-bt not all-experiments in vivo. There are. however, differences, particlarly when comparisons are made with cells plated shortly before irradiation. The most important is the rise of the transformation freqency with a power of dose less than 1 at intermediate doses (< 1.5 Gy). This shape of the crve may be linked with the enhancing effect of fractionation of x ray doses in the same region. a phenomenon observed in all cell lines tested so far If enhancement of transformation by fractionation applied in vivo, the complex natre of the low-let dose-response relationship for single doses might imply an nder-estimation of cancer indction when risk coefficients for man are extrapolated from the intermediate down to the lowest doses. Similarly, in the case of fractionated exposre. there might be an even greater nder-estimation. owing to the enhancement prodced by dose fractionation and protraction in some transformation experiments with low-let radiation. However. in the light of present knowledge-which is still insfficient-sch conclsions are prematre, for the following reasons: (a) (b) (c) No dose-response relationship has been observed for tmors in animals with a platea below l Gy after single acte doses of x or gamma rays. It is also qite clear that no enhancement of tmor indction by fractionation or protraction of external low-let irradiation has ever been noted at doses below I Gy (with one possible exception. see IV.B); The dependence of the dose-response relationships on the length of the cltre period before irradiation shows that the irreglar shape of the dose-response crves and the enhancement of fractionation in freshly plated cltres reslt from irradiation of para-synchronos cells and may therefore be considered an artifact [H37]: The effects of low-let dose fractionation and protraction on established cell cltres are in general agreement with observations related to other effects at the celllar level (indction of mtations. chromosome aberrations, cell sterilization). B. TUMOURS IN EXPERIMENTAL ANIMALS 229. In its 1977 report [U6]. UNSCEAR reviewed experimental data on radiation-indced tmors. inclding an examination of the dose-response relationships then available. The essential conclsions were that the pecliarities of each tmor model were sch as to prevent large generalizations. There were difficlties in interpreting tmor indction crves on the basis of simple mechanisms of action. in view of the complex interplay of primary and secondary contribting factors. With few exceptions, the data came from observations at doses above.5 Gy. Data were insfficient to define, nambigosly, dose-effect relationships in the low and sometimes also in the intermediate dose region. 23. With the exception of the mammary tmor of the Sprage-Dawley rat. dose-incidence crves obtained at high dose rates with low-let radiation showed a slope that increased with increasing dose, not incompatible with a linear-qadratic trend. With high-let radiation, on the other hand, dose-effect relationships tended to be more linear and their initial slopes showed relatively little change with dose rate and fractionation. However, with low-let radiation. a decrease in dose rate Jed mostly to a decrease of the oncogenic effect, following some inverse fnction of the exposre time. It appeared difficlt to qantify this decrease becase the shape of the dose-indction relationships was often altered by the change in dose rate. It was clear, however, that the sparing effect of low dose rate or fractionation was higher for lowthan for high-let radiation. Occasional departres from this general scheme were attribted to the pecliarity of the model systems tested, rather than to real exceptions of established radiobiological mechanisms. 23

42 231. The National Concil on Radiation Protection and Measrements [NI] evalated the inflence of dose and its distribtion in time on the indction crves for tmors. with different radiations and a wide range of dose rates. The ratio of the linear nonthreshold crves fitted to the high- or to the low-doserate experimental points was taken to be a measre of the "dose-rate effectiveness factor" (DREF). DREF vales referring to 1 different experimental tmor systems varied from I.I to 6.7, the majority of the vales clstering between 3 and 5. This analysis Jed to the conclsion that in most cases the risk of low-let radiation wold be significantly overestimated if nonthreshold linear extrapolation of the data for highdose-rate were applied to low-dose-rate irradiation. Moskalev et al. [M61, M68] came to the same conclsion when reviewing the available scientific literatre. New data that have become available since 1977 are reviewed below. 1. Myeloid lekaemia 232. Robinson and Upton [R7] re-analysed part of the.original data on radiation effects in RF mice [UI4-UI6]. correcting for competing risks. Abot 2 male mice, irradiated with 25-kV x rays at doses between zero and 4.5 Gy. were selected. Early cases of death (myeloid lekaemia. M, and thymic lymphoma, T) or late cases (reticlm cell sarcoma, L, or others, R) were analysed separately by a non-parametric Kaplan-Meier srvival fnction and its logarithmic transform (the cmlative force of mortality, cm. F.M.). Models were set p for treatment of these two categories, on the assmption of independence between the varios cases of death. For cases M and T, there was a significant decrease of the latent period with dose p to 3 Gy. When the effect of dose on the integral tmor rate of M (which corresponds to the cm. F.M.) was stdied. the form of the relationship for this case of death peaked, with a maximm between 2 and 3 Gy and a frther decline p to 4.5 Gy. depending on the age at irradiation of the animals (which was 5-6 weeks for grop A and 9-1 weeks for grop B) (Figre XV). The model for fitting the experimental data was sch that the estimate of the ltimate vale of the cm. F.M. corresponded to the nmber of lekaemogenic cells per animal. These were assmed to have a linear ~ :: 1.2 :::, s.8 ~ "'....6 s.4 ~ ::,.2 Hyeloid lekaemia 2 DOSE (Gy) 3 O yong mice (A 8 l older mice ( Figre XV. Final cmlative tmor rate ± standard error verss dose tor myelold lekaemia In x-lrradlated RF mice. [R7] 4 qadratic dose dependence for indction, and linearqadratic kinetics for killing. Following Barendsen [B7]. the athors took an aifa 2 qotient of.5 Gy and a (J if (J 2 qotient of 3 Gy. The fit of the data actally prodced a (J ii /3 2 qotient of 2.4. which is in reasonable agreement, considering the assmptions involved. It was conclded that the data were consistent with the postlated linear-qadratic dose-response model Petoyan and Filyskin [P22] tested another model on x-ray- and netron-indced (high dose rate) lekaemia in RF mice [U 16]. The basic assmptions of the model were: (a) (b) (c) (d) (e) Initiation in the bone marrow stem cells is a two-track phenomenon. perhaps a symmetrical chromosomal translocation; The sensitivity of initiated and non-initiated stem cells to killing by radiation is the same: Most of mitotic cell death is de to asymmetric chromosomal exchanges, which have a fnctional dose dependence basically similar to that of the symmetrical translocations. Therefore. the shape and parameters of the dose-response fnctions for indction of symmetrical exchanges and cell death shold be similar for a given cell line; The promoting inflence of radiation is broght abot throgh an enhancement of mitotic activity and. therefore, the degree of promotion shold be inversely proportional to the post-irradiation srvival of stem cells; The probability of scoring an overt malignancy dring the remaining life span is a complex fnction of the "spontaneos" incidence of a given malignancy The parameters of the model were defined by selecting from independent sorces the srvival fnctions of bone marrow stem cells irradiated with x rays and netrons. The model was fitted to experimental data on the incidence of lekaemia in the RF mose [Ul6] by the maximm likelihood method. sing a single set of two adjstable parameters; the same set was applied to x rays and netrons. A good fit was obtained, by visal inspection, and the dose-response crve showed a slight pward concave crvilinearity and linearity of the initial ascending part of the crve for x rays and netrons, respectively. Linear extrapolation from 1.5 Gy to the low-dose region wold lead to a small over-estimate of the risk for x rays. bt wold be adeqate for netrons. The assmption that the dose-response fnctions for the initiation and cell killing are similar cold be criticized on the basis of available information (B9, BI I]. bt the postlate may be accepted as a first approximation to reality Mole et al. [M2. M52. M64-M66] stdied the indction of acte myeloid lekaemia by x rays, gamma rays and fission netrons in male CBA/H mice that have almost a zero incidence of this disease. With little intercrrent mortality from competing cases of death, this tmor model is also less sensitive to systematic bias from sch cases than that on RF mice [R7]. 24

43 236. Mice were exposed to one of ten doses of x rays (from.25 to 6 Gy, delivered at the rate of.5 Gy min- 1 ) and then followed ntil they died [M52, M64]. Median srvival in all grops was very similar. There was essentially no association between latency and dose. The reslts were fitted to a for-term polynomial of the general form I= (a 1 D + a 2 D 2 ) exp- (jj 1 D + P 2 D 2 ) as well as to for simplifications with three or two parameters only. All fits were acceptable in a statistical sense (P for goodness of fit= ), bt only two eqations had all parameters positive and significantly, larger than zero (p <.5). These were: I= a 1 D e-p"dand I= a 2 D 2 e-p1d. The former was rejected on the grond that cell srvival depending solely on the sqare of dose is normally not fond. The observed incidence cold best be fitted by the latter relationship, as shown in Figre XVI. In another recent stdy of the indction of myeloid lekaemia in CBA male mice by whole-body x irradiation (25 kvp, H\!L 1.5 mm C) [D3], the age-corrected incidence after doses of l, 3, 5 and 7 Gy was essentially sperimposable on the data by Mole [M52] (see Figre XVI). A prely exponential srvival for haemopoetic cells is probably a simplification of reality, bt srvival crves for bone marrow stem cells sally show only a small sholder at low doses. For in vivo irradiated bone marrow stem cells McClloch and Till [MI6] fond D =.95'Gy and n = l.5. Mean vales for 17 measrements of D for spleen-colony-forming nits in different mose strains irradiated in vivo with x rays were in the range Gy [H32]. The vale of P 1 ( Gy- 1 ) fond by Mole [M52] is compatible with this range A contribtion of a dose-linear term (a 1 D) to the dose-response relationship in Figre XVI cannot be exclded on prely statistical gronds. The a/a 2 qotient estimated for the complete for parameter model was negative, bt the pper 95% confidence limit was 1.29 Gy: and therefore, vales of.3-.5 Gy cannot be exclded as very nlikely. Moreover, when the data were tested on three models with a nmber of adjstable parameters greater than two. one or two of the parameters were always insignificant. From the simplified two-parameter models. one had to be chosen on the basis of additional information, and this implies a constraint a posteriori even if none had been chosen a priori. In smmary, it appears that complete parameters for cell initiation and srvival cannot be estimated, independently and simltaneosly, solely from the data. However. when a likely shape of the srvival fnction is assmed. the kinetics of initiation for low-let radiation appears, in this case [M2. M52, M64], to be concave pwards with a prononced D 2 component Indction of myeloid lekaemia in CBA/H mice was also stdied after brief (.25 Gy min- 1 ) and protracted exposre to 6Co gamma rays ( 1.5, 3. and 4.5 Gy) [M65]. Protracted exposres were delivered either as daily fractions (.25 Gy min- 1 5 days per week for 4 weeks) or at a constant rate of mgy min- 1 The latter two modes of exposre did not differ in their effectiveness and gave a rather constant response at 5-6% incidence for doses above zero. The shape of the dose-response crve is very similar to that observed for lekaemias indced by fractionated x-ray treatment for ankylosing spondylitis [S49]. The response after acte irradiation was higher by a factor of 2.2, 2.8 and 5 for the three doses. respectively. The most effective gamma-ray dose for lekaemia indction was higher than that of x rays Irradiation of CBA/H male mice with fission netrons was performed at high rate (exposre times. 2-2 mintes) with 7 air-midline kerma vales in the range from.2 throgh 2 Gy [M66]. The observed incidence of acte mveloid lekaemia was fitted by the eqation I(D) = (4:5 ± 1.25) 1-1 e-n. 1 ± o.2s1 (P =.25). Neither a prely linear model withot correction for cell killing. nor the dose-sqared model that fitted the x-ray data. cold be satisfactorily interpolated to the netron data (P for goodness of fit in either case < 1-4 ). 3 I ~ et - 2 ~ LU ~ ::, w...j J w >- ~ zo2 e-e1 2 3 DOSE (Gy) Figre XVI. Dose-response relationship for Incidence of myelold lekaemia after brief exposres of male CBA mice to 25-kVp x rays. Closed circles, data frqm Mole [M52]; Open circles, data from DI Malo et al. [3]. 25

44 24. In experiments by Ullrich and Storer [U2], specific-pathogen-free (SPF) RFM/Un mice of both sexes were irradiated at 1 weeks of age with 137 es gamma rays at.45 Gy min- 1, with doses ranging from.1 to 3 Gy. Myeloid lekaemia was less freqent in the females. in which the age-corrected incidence reached significance over the control level at 1.5 Gy. Althogh both a linear and a linear-qadratic model provided a satisfactory fit to the data (P >.5 and>.8. respectively), the dose-sqared component was not significant and linearity predominated between zero and 3 Gy In male mice, the incidence was significantly higher than control, even at.5 Gy. and the form of the relationship was similar to that in females: it cold be described either by a linear (P >.5) or a linearqadratic model (P >.35), with a very small and non-significant dose-sqared term. The ratio of the linear slopes for the two sexes indicated that males were more ssceptible by a factor of abot 5. Lowdose-rate (.83 Gy per day) gamma exposre [U3] was very mch less effective in indcing myeloid lekaemia in the female mice; no significant increase above controls cold be reached even at 2. Gy. After acte and chronic netron irradiation, the incidence of the disease was very low and did not permit any detailed stdy of dose-response relationships. In either case, the peak incidence was observed at.47 Gy. bt the reslts w;re nor significantly different from controls (p >.5). 2. Thymic lymphoma 242. Ullrich and Storer [U2, U3, U5. U2] stdied the dose-response relationship and dose-rate effects for 137 es gamma rays and netrons in SPF 1-weekold RFM/Un mice. The response was assessed as agecorrected incidence, standardized to the distribtion of the age at death of control animals. In females, thymic lymphoma was efficiently indced by gamma rays at.45 Gy min- 1 : the incidence was significantly higher than in controls in all grops receiving.25 Gy or more. No simple model cold describe the response over the entire dose range (. 1-3 Gy). There was a steep rise in the region below.5 Gy. followed by a shallow rise at higher doses. Over the range of -.25 Gy (three points). a dose-sqared model provided an adeqate fit and linearity cold be rejected (P <.6). From.5 to 3 Gy the increase in incidence with dose was compatible with linearity In males, the significance of the difference above controls was reached only at l Gy. A linear model provided a satisfactory fit over the entire range of gamma-ray doses (.1-3 Gy; P>.4). Over -1.5 Gy, both a linear and linear-qadratic eqation (with D 2 coefficient not significant) provided a very good fit to the data (P >.95 and>.8, respectively). Ths. in males the dose-response relationship for acte gamma rays,vas predominantly linear over the entire dose range Lowering the dose rate to.83 Gy per day considerably decreased the incidence of thymic lymphoma in the females and changed the form of the crve from qadratic followed by linear to linearqadratic with a negative linear component. Simple linearity cold be rejected (P <.1). It is qite clear that in the RFM mose, after low-let exposre. there is no sggestion of the threshold-type response fond by Kaplan and Brown [K33] in e57bl mice. In these older data there was a statistically verified threshold in the dose-response relationship. As may be seen in Figre XVII (kindly provided by G. Walinder). fractionation of the dose into l or 8 fractions with one-day spacing. is withot effect pon the incidence of the lymphoma. When plotted against exposre (either simple or fractionated) the incidence may be fitted by a linear fnction of the form I= ad - b where a=.112. and b = (significantly different from zero); the coefficient of correlation is very high (.93) In experiments by Maisin et al. [M62]. 12-weekold male BALB/c mice were exposed to single or to fractionated doses of mes gamma rays ( 1 eqal doses separated by l day) in the dose range from.25 to 6 Gy. The dose-response crve for thymic lymphoma was of a threshold type. the actarial incidence rising above control only at 4 and 6 Gy. Single doses were more effective at 4 than at 6 Gy Other data on thymic lymphoma indction by x rays in neonatal (e57bl/6 JNrs X WHT /HC) F I mice were reported [S48]. Even thogh the incidence in females and males at three exposre points (2. 4 and 6 R) cold be fitted to a linear-qadratic eqation with a negative linear term (P >.8), the data adeqately fit a threshold-type response sch as that presented in Figre XVII for the e57bl mice Fast netrons (actely delivered doses at.5 and.25 Gy rnin-i and chronic doses at.1 Gy per day, with total doses between.1 and 3 Gy) indced thymic lymphoma in female RFM mice [U5]. In the range Gy, the RBE with respect to high-doserate 137 es gamma rays was between 3 and 4. For acte netron exposres. the dose-response crve was concave downward. In the range -.94 Gy, linearity cold be rejected (P >.1) and a good fit was obtained with the sqare root of the dose (P >.8). Up to.47 Gy, a linear fit was satisfactory (P >.75). For chronic netron exposre, a linear fit adeqately described the crve over the -.94 Gy range (P >.8); and the loss of effectiveness over the acte exposres amonted to abot 3%. However. there was a decrease of ssceptibility to thymic lymphoma pon acte exposre with increasing age, so that the lower efficiency of the low-dose-rate exposre may in part be de to this latter factor alone. The RBE of acte netron exposre relative to gamma rays varied proportionally to the inverse sqare root of netron dose Indction of thymic lymphoma was also stdied in male BALB/c mice irradiated at the age of 12 weeks with d(5)be netrons (modal energy abot 23 MeV) [M63] at doses from.2 to 3 Gy. The actarial incidence of thymic lymphoma fitted the same sigmoidal threshold-type crve fond for 137 es gamma rays [M62]. 26

45 >- w Nmber of 8 treatments 7 z 6 C: w Q. w z 5 w 4 z C57Bl[ ~ Interval between treatments (days) 1 w ::, C: _ Spontaneos freqency EXPOSURE (R) Figre XVII. Crde incidence of thymic lymphoma in C57BI mice given single and fractionated x-ray doses at 1 day interval [K33). Rhombard symbols represent the crde incidence of thymic lymphoma in neo-natal (C5781/6 JNrs x WHT/Ht) F 1 mice [S48]. 3. Other reticlar tmors 249. The so-called mose reticlm cell sarcoma is a grop of neoplasms showing a decline of the incidence with dose after whole-body x irradiation (CJ9, K25, U3, U5. U2]. Ullrich and Storer [U3, U5, U2] stdied the age-corrected incidence of reticlm cell sarcoma among SPF RFM/Un male and female mice irradiated with gamma rays (.45 Gy min- 1 or.83 Gy per day) and in RFM/Un female and BALB/c mice irradiated with fast netrons ( Gy min- 1 and.1 Gy per day). All treatments prodced a decrease of reticlm cell sarcoma with dose, netrons being more effective than gamma rays, and chronic exposres less effective than acte ones. The difference between actely and chronically delivered gamma rays was more prononced than that for netrons. There was a clear, inverse relationship between the increase of thymic lymphoma and the decrease of reticlm cell sarcoma. sggesting some link between varios radiation-indced reticlar neoplasms which cold greatly affect the observed dose-response crves. 25. Male mice (C578L/Cne x C3H/Cne)F 1 hybrids received lethal whole-body irradiation (9 Gy) and were resced from early death de to haematological failre by a homologos bone-marrow graft [C 18]. In these mice. the spontaneos incidence of reticlm cell sarcoma of abot 5% is redced in irradiated srvivors to a few per cent. Shielding of a part of the marrow dring exposre affords a similar protection against early post-irradiation death and effectively depresses the incidence of the tmor. It was sggested that whole-body irradiation sterilizes cells from which these neoplasms originate [Cl8. CJ9] and that the for these cells is similar to that for the bone-marrow stem cells [M2] Frther stdies [M2] showed that in vivo irradiation of the bone-marrow graft with doses of 2 and 4 Gy of x rays before transplantation into hosts given whole-body irradiation (9 Gy) provides the same protection against early death, bt leads to a significantly elevated incidence of reticlm cell sarcoma in animals given irradiated marrow. The incidence expressed per nmber of injected marrow cells rises with dose, amonting to , J and s tmors per cell irradiated with, 2 and 4 Gy. When related to the nmber of injected srviving stem cells, the incidence rose steeply from throgh and at, 2 and 4 Gy. respectively. It is an nproved assmption that the tmors originated from these cells. If this hypothesis cold be accepted, the dose-response wold display a prononced crvilinearity (concave pwards). the effect rising with a power of dose appreciably greater than l. and even the presence of a threshold dose cold not be rejected In the same animal model, stdies [M 19] were made on whole-body irradiated mice with shielding of one or two hind legs, from which shielding was removed for a variable fraction of irradiation time. ths exposing some of the protected marrow in sit 1 graded doses of x rays p to 8 Gy. The observed 27

46 incidence of reticlm cell sarcomas increased in these mice to a maximm at abot 5 Gy and then declined with frther increase of the dose. The final incidence of these neoplasms in all experiments with exogenos and endogenos bone-marrow protection is plotted in Figre XVIII, showing in all instances a crvilinear (pward conca, e) rise with x-ray dose. The form of these relationships is similar to that observed for many other radiation-indced tmors in animals irradiated in vivo.... -Iz ex w.. ~ 4 ex ~ 3...I...J w 3 2 a -I- '-' er: 6.~ 5 1 wb A 1 sh V 2 sh a exo DOSE (Gy) Figre XVIII. Final Incidence of reticlm cell sarcoma as a fnction of dose nder three different experimental conditions: (a) whole-body Irradiation (wb); (b) shleldlng of one (1 sh) or two (2 sh) hind legs; (c) syngenlc bone marrow transplantation (exo). [M19] 4. Tmors of the thyroid 253. An extensive stdy of thyroid tmor indction by 25-kVp x rays and 131 I in abot 3 6-week-old female Long-Evans rats was reported by Lee et al. [L29]. The stdy aimed at comparing the effectiveness of the two radiations delivered at different dose rates (2.8 Gy min- 1 for x rays; the maximm dose rates for in individal grops were estimated at.17,.69 and 1.6 mgy min- 1). Local doses of x rays to the thyroid were.94, 4.1 and 1.6 Gy. Total doses delivered by 13IJ were estimated at and 8.5 Gy. The role of pititary irradiation in the indction of thyroid cancer was also explored by delivering doses of 4.1 Gy of x rays to the pititary alone. or to the thyroid and the pititary together: findings were negative in this respect Tmors inclded were those that appeared later than 6 months in rats dying before the age of 26 months and in animals sacrificed at 24 months. The reslts were given as incidence corrected for the slightly enhanced mortality of irradiated verss control rats. Data on tmor mltiplicity and on tmors per animal were not presented A least sqare fit yielded the following fnctions describing the dose-response relationships (Figre XIX): 1() = 7.47 I-4 ( + 1) for x rays I(O) = 1.78 w-3 (O + IO)D-5348 for where O is the dose given in rad (.1 Gy). The exponents were significantly lower than nity (p <.1). indicating that the incidence of thyroid carcinomas in these rats rose with approximately the sqare root of the low-let dose. Ths. there was no difference in the effectiveness of the two radiations over the observed range of doses, bt a lower effectiveness of 1311 per nit dose (p to a factor= 3) cold not be exclded on statistical gronds [N 1). 24 g x rays... z 2 O -- iodine-131 a: "" :c 16 I-3: Ill...I :::: "" -z "" z I.I IX I.I Q DOSE (Gy) Figre XIX. Incidence of thyroid carcinoma as a fnction of mean thyroid dose from Injected lodlne-131 and localized x Irradiation, Error bars are estimates of the 95/o confidence llmlts. [L29) 256. For adenomas. the two dose-response relationships were slightly different (p <.1) and when fitted by exponential eqations they assmed the following forms: 1() =.245 exp (.27 ) for x rays 1() =.24 exp (.123 D) for 131J, where Dis the dose in rad (.1 Gy). At no dose point was there a significant difference between the effect of the two radiations. However, de to the large variability of the observations. some difference in the effectiveness cold not be exclded. Control incidence of medllary tmors was not affected by either radiation The downward concavity of the dose-incidence crve for carcinomas was attribted to killing of initiated cells. If an in vivo srvival crve of thyroid 28

47 folliclar cells with Dq = 4.9 Gy and = 4.5 Gy is assmed [D 15. M5 l), then the dose-response crve for initiation of carcinomas or adenomas in this experiment may be expected to rise with a power of dose either close to or greater than nity. The slope of the initial portion of the dose-incidence crve for carcinomas was estimated at Gy- 1, which is close to the estimate for hmans [S6) The presence of a strong linear component in the relationships is sggested by the absence of a clear dose-rate effect, at least in yong animals. This observation contradicts previos findings (reviewed in [D 16. W 12]) that. per nit dose, x rays are more efficient for thyroid tmor indction than The stdy reviewed here [L29] differs from the older ones in two important aspects: first, the range of doses is below lo Gy. and ths cell killing and other nonstochastic effects wold not be expected to prevail: second, the nmber of animals sed is large and gives the stdy good statistical credibility. Nevertheless. the findings refer only to one sex of one particlar strain of rats. These experiments do not spport the sggestion [R29] that irradiation of the pititary gland may give rise to increased incidence of thyroid cancers in children irradiated for treatment of tinea capitis. 5. Mammary tmors 259. The age-corrected incidence of mammary adenocarcinomas was stdied in SPF female BALB/c mice after high- and low-dose-rate exposre to gamma rays [U3. U25) as well as to high-dose-rate. Jow-doserate and fractionated exposres to fission netrons [U4. U25. U26]. After acte gamma-ray exposre (doses.1-2 Gy) the incidence increased rapidly over the range -.25 Gy. with a rather irreglar platea appearing at higher doses. The irreglar response was consistent with the models l(d) = D + 15 D 2 (P >.4) for the high dose rate and with I(D) = D (P >.4) for the low dose rate, in accordance with the linear-qadratic model. Ramzaev et al. [Rl] described a "spra-linear" response for indction of mammary adenocarcinoma in mice irradiated over ten generations throgh constant feeding of 137 Cs and 9Sr. 26. Acte irradiation of BALB/c mice with fission netrons ( Gy min- 1 ) [U25] yielded a doseresponse that was downward concave over the range -.5 Gy (the lowest dose being.25 Gy) and wold be poorly described by a linear model. A good fit was obtained by a model 1() = l/ (P >.85). In the range -. l Gy. the response cold be well fitted by a linear model I(D) = D When fission netron exposres were split into two eqal fractions delivered 24 hors and 3 days apart [U26], there was no difference with respect to the single dose-response for the 24-hor interval. bt for the 3-day interval there was an increase of abot 3-4% only for the highest dose tested (.5 Gy). In the latter case. the crve was still bending over in the range -.2 Gy. Protracted irradiation (.1-.1 Gy per day) Jed to a significant enhancement of the effect relative to the acte exposre. and over the whole range of doses. This was interpreted either as a promoting effect or as stimlation of progression of the growth of malignant clones. The maximm RBE of netrons was 33; this was calclated as a ratio of the linear slope for the gamma and netron crves (for single irradiation at low dose rates) Two-month-old female. Sprage-Dawley rats were followed by Shellabarger et al. [S37] for their entire life span after single exposres to 25-kVp x rays (.28,.56 and.85 Gy) and 43-keV netrons ( and.64 Gy). The animals were observed for appearance of adenocarcinomas and fibroadenomas. In all irradiated grops the tmor rate increased steeply with age and the effect of irradiation cold be described as a forward shift in time of the spontaneos age-specific tmor rate The response was expressed separately for carcinomas and adenomas as mortality-corrected prevalence and cmlative tmor rate [R(t.D)]. The fibroadenoma and total mammary tmor responses were approximately proportional to x-ray dose over all times, while netrons prodced a downward concave relationship. rising with a power of netron dose less than nitv. The increase in R(t,D) was significant even at I ~Gy of netrons and the forward shift of the prevalence. corrected for age-specific mortality, cased by this dose was eqivalent to 35 days. The R(t.D) for fibroadenomas at 8 days was less crvilinear than at I year after exposre. The distribtion of tmors was non-random among the animals (over-dispersion was noted relative to Poisson distribtion). The response for adenocarcinomas was statistically ncertain. owing to the small nmber of tmors involved, bt linearity is the best approximation to the dose-response relationship In this stdy [S37]. the RBE varied roghly with the inverse sqare root of the netron dose. At the highest doses stdied it exceeded 1: at the lower end of the dose scale. it approached a vale of 1. The implications of these observations were already discssed in chapter II In other experiments [V9]. female Sprage Dawley rats were actely exposed whole-body to fission netrons. or to protracted doses of.2..6 and.5 Gy (over 242 hors in one month). The prevalence of mammary tmors at JO months after exposre increased. relative to single irradiation. at all doses tested. Protraction of gamma-ray exposres (15, R) in a similar pattern redced significantly the prevalence seen after corresponding single exposres. The RBE vales for netrons, measred at 2% prevalence, were 16 and 68 for acte and chronic irradiation. respectively Breast tmors (histology not specified) in yong female Sprage-Dawley rats were stdied [G35] after acte (over one hor) and chronic (over 1 days) x irradiation, as well as after internal irradiation by tritim, administered as tritiated water (75 and 9% of total dose delivered over lo and 2 days, respectively). The doses of 2-kVp x rays varied from.29 to 2 Gy, 29

48 and those of tritim from.5 to 4 Gy. The competitiverisk-corrected tmor rate was calclated p to 45 days after the start of exposre and the integral vales were linearly related to dose for all three irradiation modes. The effectiveness of tritim per nit dose did not differ from that for the chronic x-ray exposre. The latter prodced effects similar to those of acte exposre at.6 Gy, bt was 25-3% less effective at abot 1.8 Gy X-ray-indced and netron-indced (.43 MeV) mammary tmors were stdied in weanling, virgin female ACI rats [S52], irradiated at days. The doses for x rays ranged from.17 to 3 Gy and for netrons from.1 to.36 Gy. The animals were irradiated either 2 days after implantation of a pellet containing 5 mg of diethylstilbestrol (DES) or withot prior hormonal treatment. Fibroadenomas and adenocarcinomas were diagnosed histologically and recorded separately. DES greatly increased the incidence of spontaneos and radiation-indced adenocarcinomas in this strain, which has a low spontaneos incidence of mammary neoplasms. Significant excess prevalence was detected in DES-treated rats even at the lowest netron dose; for x rays. the significance occrred at.17 Gy. The RBE varied inversely with the sqare root of the dose and approached a vale of 1 at the lowest end of the range Appearance of tmors in rats not treated with DES was followed, on the average, to abot 75 days after irradiation. For x rays, the cmlative tmor rate at 6 days (for the sm or for each type of tmor separately) rose almost linearly with the dose. The same applied to adenocarcinomas after netron irradiation, bt for fibroadenomas a downward concave relationship cold not be exclded. The RBE of netrons for the sm of tmors tended to decrease with dose over the range of doses tested, bt did not vary mch from an average vale of 1. The length of tmor-free life lost de to the sm of benign and malignant tmors was a linear fnction of dose, for both x rays and netrons, and for the whole range of doses tested When non-inbred female rats were irradiated externally, with single. increasing doses of fj rays from a 9 Sr + 9 Y sorce, the rise of the crde incidence of mammary carcinoma was linear with dose, p to 4 Gy [M68, M7 l]. These animals were also irradiated, starting at the age of 8 weeks. singly. with doses of 16 Gy (from the pblished text it appears likely that the doses refer to the srface of the body): daily (5 days a week) at.8 Gy per day for 4 weeks; or once every two weeks with.8 Gy over a period of 39 weeks. The observed life shortening differed significantly from the control vale. bt not between grops with varios irradiation schemes. There was an obvios acceleration of the appearance of mammary tmors after irradiation. and the fractionated schemes were slightly less effective, althogh this difference cold be acconted for by the fact that, on average, the fractionated exposres were delivered somewhat later in life. The crde incidence over lifetime was elevated in all irradiated grops to 1% from 4% in controls. This experiment [M7 l] spports the notion that fractionation of low-let radiation has little effect pon the incidence of mammary tmors in rats. 27. Other athors reported that mammary tmors were also indced in female Sprage-Dawley and in W AG/Rij rats by x rays. and by monoenergetic.5- Me V and 15-MeV netrons [B86, B92]. There was no evidence that a virs played any role in mammary carcinogenesis in these animals. The rats were 8 weeks old when irradiation was started with fractionated doses or given as single doses (absorbed doses in the glands). They were kept ntil their death. and the freqency of fibroadenomas and carcinomas was based on histological examination The probability of animals srviving withot evidence of a tmor was calclated according to the Kaplan and Meier life-table analysis. Weibll fnctions were fitted to these data and showed that irradiation cased an acceleration in the appearance of fibroadenomas and carcinomas in both strains of rats. X-ray fractionation (1 x.2 Gy at one-month intervals) in WAG/Rij rats was marginally less effective. in respect of the appearance time of mammary carcinomas. than a single dose of 2 Gy. Fractionation of.25 Gy of netron into 1 eqal fractions at onemonth intervals was also withot a significant effect relative to I X.25 Gy When dose-response relationships were calclated from actarial incidence data at 95 weeks of age, the reslts for carcinomas in W AG/Rij rats for x rays and.5 MeV netrons showed sch a wide scatter as to preclde neqivocal inferences abot the character of dose-response relationship. For fibroadenomas (Figre XX) the response for x rays (.25-4 Gy) was proportional to dose. The response for.5 and 15 MeV netrons ( or 1.6 Gy. respectively) showed a satration effect. In Sprage Dawley rats, the actarial incidence of fibroadenomas showed 1 x rays 8.5 HeV netrons ~ A 15 HeV netrons... z... C 6 z _, ::: : ~ ~ Ill Ill >< ABSORBED DOSE (Gy) Figre XX. Actarial Incidence at 92 weeks of age of mammary adenomas In WAG/RIJ rats as a fnction of dose for varios types of radiation. [B92] 21

49 satration effects at high doses of all radiations. The RBE of netrons was mch lower than that observed in other stdies on the same strain [S37, S52]. Carcinomas were so rare that no dose-response relationship cold be established When the prevalence of mammary adenocarcinomas was stdied at 1 months after x-ray exposre {2 R) in female Sprage-Dawley rats irradiated as. yong virgins or as pregnant, lactating or postlactating animals [H23], no effect of the physiological state pon the incidence cold be fond. This reslt is at variance with the age-effect seen in women (see chapter V) Nmeros stdies show that several factors may inflence the development of radiation-indced mammary tmors in varios strains of rats. Some of these stdies were considered in annex L of the 1982 UNSCEAR report [U24]. A strong inflence of genetic factors on netron-indced mammary carcinogenesis was reported by Vogel et al. [VIO]. After a single dose of.5 Gy from fission netrons. the highest (56%) prevalence at I year after exposre was seen in Sprage-Dawley and Long-Evans females; the lowest (5%) was observed in Wistar, and an intermediate one (25-29%) in Bffalo and Fischer-344 rats. Wide variation in ssceptibility to mammary tmor indction was also fond by Broerse et al. [B86, B92] for Sprage-Dawley. WAG/Rij and BN/BiRij rats. 6. Lng cancer 275. Lng adenocarcinomas were indced in BALB/c female mice by gamma irradiation at.45 Gy min- 1 and at.83 Gy per day in the dose range.1-t Gy [U2, U3, U25]. After acte exposre. the age-corrected incidence cold be related to dose by a linear [l(d) = D: (P >. 7)] or a linear-qadratic model [l(d) = D D 2 (P >. 7)]. At low dose rate, the dose-response relationship was linear with the same slope [l(d) = D (P > 8)] as at high dose rate in the linear-qadratic model. This model, therefore, seems to describe adeqately the response in qestion After single acte exposre to fission netrons (doses.25-2 Gy; Gy min- 1 ). the relationship was downward concave with a maximm at.5 Gy [U25]. Over the range -.5 Gy the data were consistent with a sqare-root model [(D) = VD (P >. 7)]. A linear model cold not be rejected bt was obviosly nlikely. ln the ranges -.2 Gy, the data were fitted by a linear model [I(D) = D (P >. 7)]. When the doses were split into two eqal fractions. no difference was seen for 24-hor fractionation [U26]. A slight increase of the incidences was noted for 3-day fractionation. bt only at the highest dose of.5 Gy. For this irradiation mode the dose-response became practically linear p to.5 Gy Protracted netron irradiation ( Gy per day and Gy per day) was more effective than the high-dose-rate exposre and, as in the case of 3-day fractionation more nearly linear over the entire dose range [U26]. Higher effectiveness for.5 Gy at chronic and at 3-day fractionated exposre is nlikely to reslt from repair of sb-transformation damage, as no repair was seen for fractionation over 24 hors. The effect was attribted to repoplation of alveolar target cells. The maximm RBE of netrons from the ratio of the linear slopes at low doses was When derived from the sqare root of the dose model, it eqalled 7.7/J/D 278. Chmelevsky et al. [C34] analysed the indction of lng carcinoma in male Sprage-Dawley rats which were 9-clays old at the start of an exposre to radon in eqilibrim with its short-lived daghter prodcts. The radon concentrations and the dration of daily or weekly exposre were varied. The total dration of the inhalation period was kept within one to six months. The dose was defined as a prodct of potential alpha energy concentration and exposre time, and expressed in working level months (WLMc): the range of dose covered zero to 14, WLM delivered at varios rates Animals were stdied at death for the presence and histology of lng neoplasms. In this species, lng cancer is almost never a case of death: therefore independence between disease and death cold be assmed and the data cold be treated as flly censored. The prevalence of neoplasms as a fnction of time was estimated for each exposre grop by non-parametric methods in the form ofa monotonically increasing fnction (by a most-likelihood fit). Three models were selected for fitting prevalence to dose: a time shift, an acceleration, and a proportional hazard model. The fit of the data to all three models was eqally good: none cold be discarded on statistical gronds. 28. The crde and mortality-corrected incidence of lng carcinomas was calclated from the prevalence fnction p to 9 days. The reslts of the actarial incidence are shown in Figre XXI. The initial slope (below 1 WLM) is practically linear in both cases. When the incidence was fitted to a fnction of cmlative exposre E of the type I(t:E) = K EP, the parameters obtained at t = 85 days were: K = (3.4 ± 1.4) 1-4 and p =.92 ±.7. When linearity was assmed to apply over this range of exposres. the proportionality coefficient was (2 ±.5) 1-4 per WLM. Perhaps by accident, these vales are very similar to those fond for man (chapter V). At doses above WLM, the response flattens off. At doses above WLM, the exposres at high dose rate(> 16 WLM per month) are less effective in indcing tmors than those delivered at lower rates. Toe data do not sggest a threshold, which. if present. wold have to be mch lower than 65 WLM. The character of the doseresponse relationship is basically similar to that seen 'The concentration of potential alpha energy of short-lived radon-222 daghters is sometimes expressed in nits of Working Level (WL). This level corresponds to any mixtre of these daghters which, pon complete decay, will emit alpha particles \\ith total energy of L3 1s MeV per litre of air (2.8 1-s J m-3). The Working Level Month (WLM) corresponds 1 exposre to a mean concentration of I WL for the reference period of 17 hors. I WLM = 17 WL hors= MeV h m-3 = J J h m

50 ... z... C z... C.1,- ::, ~ "' ~ C.t::.... o+ ST AT PH.1 -,...~~~~~~~~-,-~~~~~~~~.,...~~~~~~~~... ~~ OJMULATIVE EXPOSURE (WLM) Figre XXI. Adjsted Incidence of lng carcinoma In rats as a fnction of the cmlative exposre (WLM), calclated on the basis of three models: ST, time shift model; AT, accelerated time model; PH, proportional hazard model. Open symbols: exposre rate< 16 WLM per month; Solid symbols: exposre rate> 16 WLM per month. [C34] in ranim miners (chapter V). The loss of tmorfree life (the so-called "effect period" [U6)) was also stdied in these rats as a fnction of cmlative exposre. Agreement between the models is less close than for the mortality-corrected incidence: however, linearity seems to prevail p to WLM The actarial incidence of lng carcinomas was stdied [C38] in Sprage-Dawley rats. after wholebody irradiation with fission netrons. and evalated by methods of analysis similar to those described for radon-indced cancers [C34]. The netron doses ranged from.12 to 8 Gy. The doses p to 2.3 Gy were delivered within I day and the higher doses were protracted over periods from 14 to 42 days to secre early srvival of the animals. The time-related prevalence of plmonary cancers was fitted to shiftedtime and accelerated-time models and an actarial incidence was calclated which did not depend pon the model chosen (Figre XXII). The dose-response... "'... z!= ~... I- :::,.., "' cc Agre XXII. i e e i e shifted time model o accelerated time model NEUTRON ABSORBED DOSE (Gy) Adjsted Incidence of lng carcinomas In netron irradiated male Sprage-Dawley rats. [C38) relationship cold be linear p to abot.5 Gy. At higher doses, the crve does flatten, bt does not decrease (actally, the absence of a decrease is de to correction for life shortening). In the low-dose range, the. efficiency of netrons verss radon in indcing plmonary carcinomas exceeded 3 WLM/mGy. At higher doses, it was lower. Recalclation of this ratio in terms of lng doses, rather than exposres, indicates that the efficiency of netrons, relative to alpha particles, is abot 14 when expressed in terms of the bronchial dose and abot 2 in terms of dose to the plmonary parenchima [T3] An exhastive review of lng cancer indction by inhalation of varios radionclides in diff erem animal species was compiled by an ICRP Task Grop and pblished as ICRP Pblication 31 [13]. When the data from varios experimental series were combined for an analysis of dose-response relationships a very wide scatter was apparent. Whereas for varios alphaemitting transranic nclides a linear fit to the data did not appear nreasonable, a linear dose-response relationship did not adeqately describe the experimental points obtained for beta and gamma emitters. For these an pward concave dose-response relationship was typical. 7. Lng adenomas 283. Indction of benign Jng adenomas in RF mice was stdied [YI] by locally-delivered, acte doses of x rays of between 2.5 and 15 Gy. The dose-response crve for tmors recorded 12 months thereafter, when plotted on a log-log scale. had a slope of I below 7.5 Gy and a steeper increase of the tmor yield at higher doses. In terms of a linear-qadratic model, this wold imply indction of tmors by single-track kinetics in the lower portion of the crve 212

51 and two-track kinetics in the steeper region of the relationship. A split-dose experiment with fractionation inter.-als of 24 hors confirmed that the nmber of adenomas per animal was lowered by fractionation, bt only above 7.5 Gy, where repair of interacting lesions wold. theoretically, be expected Indction of lng adenomas by localized irradiation of the thorax with 3-kVp x rays and fast netrons was also stdied in I -12-week-old female SPF RFM/Un mice [U21). After.5 Gy of netrons, the nmber of tmors per mose at 6 months did not increase frther and it was assmed that by 9 months it shold have been complete. The doses of x rays ranged from I to 9 Gy and of netrons from.5 to 1.5 Gy. The level of mortality did not inflence the reslts sbstantially. For acte x irradiation, linearity cold be exclded and the dose-response crve cold be described by: (a) a linear-qadratic model with an initial negative linear slope (P >.9); and (b) a threshold model of the form P(t:D) = c + a (D - D*) 2 (P >.99). where D* is the threshold dose. estimated to be 2.47 Gy. For netrons, the average nmber of adenomas per mose rose abot linearly to.25 Gy. peaked at I Gy. and then declined by a factor of abot 2 toward 1.5 Gy. Over the range -.25 Gy. the rise was appr~imately linear (P >.5). A yroportionality to \/ D cold be rejected over this range (P<.5). Also a linear threshold model with threshold dose at abot.7 Gy described the response within the range -.25 Gy very well (P >.99). For the linear model, the RBE of netrons increased from a vale of 25 at.15 Gy netrons to one of 4 at.1 Gy Splitting of the dose of either radiation into two eqal fractions separated by 24 hors or 3 days was also tested [U22]. Under these conditions. there was a redction in tmor yield for x-ray doses above 4. Gy. and the response tended to become more nearlv linear. There was no difference between responses with doses split by 24 hors or 3 days. Recovery was complete by 24 hors, and amonted to 5% of the initial effect. After split netron irradiation. no recovery cold be seen. Linearity for all data combined over the range -.25 Gy cold not be rejected (P >.8), bt better fits were obtained with a dose-sqared model (P >.99) or with a model sing a thre~hold dose at abot.54 Gy. If it is accepted that decline of the tmor yield at high doses reflects killing of potentially transformed cells. then linearity over the range -.25 Gy becomes rather improbable. as alre~dy discssed in chapter ll. However. an al~~rna_ve explanation of the slight pwards concave crv1lmeartty in the -.25 Gy range cold be a possible inverse relationship between latency and dose. 8. Ornrian tmors 286. Indction of ovarian tmors by x rays and netrons was stdied by Ullrich et al. [U2] sing SPF RFM mice. An acte gamma-ray exposre (.4 Gy min- 1 ) indced varios types of ovarian tmors, and their age-corrected incidence increased rapidly over the dose range -.5 Gy. At higher doses, there was a platea of the response with a marginal rise of tmor incidence with increasing doses. Over the range -.5 Gy linear and dose-sqared non-threshold models cold be rejected (P <.1) while a linear-qadratic model with a negative linear component adeqately described the relationship (P >.25). Since. according to the athors, no biological meaning cold be attribted to sch a model, threshold models were tested and a linear one was again rejected (P<.1). However, a threshold dose-sqared model (threshold estimated to be.12 Gy) provided a satisfactory fit ( P >. 75) over the dose range p to.5 Gy. At.8 Gy per day, a significant incidence was seen at l Gy. where it corresponded to that seen after.25 Gy acte exposre. Again, non-threshold linear and dose-sqared models cold be rejected (P <. l} as well as the threshold dose-sqared one. An adeqate fit was obtained for a threshold linear model with a threshold dose of. 7 Gy and a linear-qadratic model with an initial negative linear slope (significantly negative at the 5% level) After netron irradiation (U5]. the ovarian tmor incidence in RFM mice was lower at. I Gy per day than at.5 and.25 Gy min- 1 for all dose levels. The decline was attribted partly to the redction of the dose rate and partly to an age-related decrease of ssceptibility. This sggests that a tre dose-rate effect was perhaps observed in this tmor system after high-let irradiation Ullrich [U23, U25, U26] stdied the age-corrected incidence of all ovarian tmors (granlosa cell tmors. lteomas, tblar adenomas) in BALB/c female mice, indced by gamma radiation and fission netrons. High-dose-rate gamma irradiation (in the range.1-2 Gy} increased the incidence steeply from control levels (abot 2.5%) at.1 Gy to 77% at.5 Gy. with a platea at higher doses. As previosly observed for RFM mice, the dose-response relationship was consistent with a threshold model, and the linear. dose-sqared and linear-qadratic models co~ld be rejected (P<.1). Irradiation of BALB/c mice with fission netrons (.25-2 Gy) led to a steep increase in the incidence of ovarian tmors from control level at.25 p to 76% at.5 Gy. with a slight decline towards 2 Gy. Over the dose range -.5 Gy, a variety of dose-response models cold describe the data, inclding the non-threshold models (linear-qadratic, dose-sqared) and a threshold-type model. The linear non-threshold model cold. however, be rejected (P <.1). A threshold model wold be consistent with both the postlated role of nonstochastic mechanisms involved and the data for gamma radiation. Becase of the app~rent thr~shold. at low doses any estimate of RBE 1s meanmgless. From.2 Gy pwards the RBE is close to nity [U25]. Splitting the netron dose at 24-hor and 3-day intervals was ineffective. the data being practically indistingishable from the single-dose series. At protracted exposres, the netrons were Jess effective than when delivered actely bt, for instance,.4 Gy yielded the same reslt whether delivered?ver 4 or 4 days. This indicates that the difference with regard to acte exposre is a tre dose-rate effect and not a change of ssceptibility to tmor indction with age. The observation is similar to that for the RFM mose, bt no explanation is available. 213

52 289. In other experiments [Y2], mice received a series of x-rav doses over a variable time T and the animals were f~llowed for incidence of ovarian tmors. The reslts were later re-analysed [KS] on the assmption that the cmlative net incidence (prevalence) is described by the eqation F{t) = f [a 1 D + q(t)d2] (4.1) where the form of the fnction is left ndecided and q(t) is a redction factor for recovery of the damage with protraction of irradiation. The optimal fit to the data was obtained for an assmed exponential decay of the sblesions T = 15 h. The optimm vale for a1 for the linear term was zero, and a 1 /a 2 qotients greater than.4 Gy in terms of the linear-qadratic model cold be exclded when data were corrected for time-dependent repair sing non-linear optimization. In essence, no linear component cold be fond, and incidence was proportional to the sqare of the dose over a wide dose range. 9. Pititary gland tmors 29. In female RFM mice, pititary tmors increased irreglarly with gamma-ray dose over the range.5 to 3 Gy (.45 Gy min- 1 ) and the agecorrected incidence was not significantly different from control. A linear-qadratic response adeqately described the data (P >.6), bt a linear response cold not be exclded (P >.2) [U2]. In male mice. the effect was so slight that the dose-response cold not be investigated. Lowering the dose rate (.83 Gy per day) cased a less efficient indction of pititary tmors in the same mice [U3]. If a linear-qadratic response was assmed at high dose rate, the linear component was very similar nder both conditions and the lower dose rate eliminated the contribtion of the dose-sqared component. Netrons indced a large nmber of pititary tmors with an RBE greater than 5. A dose rate of. I Gy per day appeared as effective as.5 and.25 Gy min- 1 p to.2 Gy, bt rather less effective at.47 and 2 Gy. 1. Harderian gland tmors 291. These tmors were indced by gamma rays in both male and female RFM mice, and the response in the two sexes was very similar. The data on ageadjsted incidence appeared to fit best a linearqadratic model (P >.25 and>.99 for females and males, respectively), bt linearity cold be exclded only for females (P >.5). The low-dose-rate gamma irradiation sbstantially redced the incidence of tmors, and a good fit (P >.9) to the eqation was obtained with the same linear coefficient as for high dose rate bt withot a dose-sqared term [U2, U3]. The same rmors showed a marked sensitivity to indction by netrons, with an approximately linear increase in age-adjsted incidence as a fnction of dose p to abot.5 Gy. with a platea at I Gy and a decline at 2 Gy. No significant effect of the dose rate was observed [U5]. When the age-adjsted incidence of harderian gland tmors was stdied in whole-body x-irradiated CBA mice [3) it was fond that the dose-response crve rose as a fnction of dose p to 3 Gy (there were, however. only 3 points) and then became flat. Correction of these data for cell killing (a srvival crve was determined independently) reslted in a dose-response crve per srviving cell which rose with the sqare of the dose between I and 7Gy. 11. Skin tmors 292. Skin cancer can be indced in experimental animals. mostly CBA mice and CD rats, by both densely- and sparsely-ionizing radiation, and relevant experiments were reviewed in the 1972 and 1977 UNSCEAR reports [U6, U7]. For both species. highly crvilinear dose-response relationships were reported for p particles and high-energy electrons [B59, Hl2, H 13) as well as for accelerated (4 MeV) helim nclei. which also prodced a crvilinear response in CD rats [B59]. From detailed considerations of depthdose distribtion verss efficiency of cancer indction. and from the association of the latter with damage to the hair follicles, it was conclded that radiationindced skin cancer in this latter strain is likely to originate from cells located in the follicles The indction of epidermal and dermal skin tmors was again stdied in 3-month-old female CBA/CaH mice (3125 exposed. 612controls), irradiated locally over one-third of their skin srface (mid-torso) by P particles emitted from a 2 ~TI sorce [H33]. The radiation neither penetrated the abdominal wall nor reached the bone marrow. Tmorigenic doses were taken to be 75% of the nominal dose at the body srface for epidermal tmors. and 65% for dermal ones. The nominal doses ranged from 5.4 to 26 Gy, and the dose rate varied from 2 to.17 Gy min- 1 Most tmors arose well within the irradiated area and only a few at the edge. Over 7% of all tmors were dermal (fibromas and fibrosarcomas). and abot 6% of the dermal and epidermal ones were malignant. It was confirmed histologically that fibromas cold become malignant. and mltiple tmors were noted in 12-15% of tmor-bearing mice irradiated at the highest rates. The incidence of tmors was expressed per nit area of skin. Redction of dose rate from 1-2 to Gy min- 1 did not alter the incidence appreciably; at.5-.8 Gy min- 1 the incidence of epidermal tmors was redced at all except the highest doses ( 12 Gy) The incidence of dermal tmors began to increase significantly with doses to the dermis of abot 16 Gy and rose steeply into a platea above 6 Gy. The incidence of epidermal tmors increased significantly and then rose steeply when the dose in the epidermis exceeded 21 Gy. The peak incidence was reached at 6 or 1-12 Gy and fell at still higher doses. At exposre rates of Gy min- 1 no epidermal tmors occrred and the incidence of dermal tmors was very mch redced. At lower rates, therefore. a range of high doses was ineffective in prodcing skin tmors. The dose-response crve resembled those typical of non-stochastic effects. with an apparent threshold at abot JO Gy. 214

53 295. The series irradiated at the highest dose rate ( Gy min- 1 ) had sfficient statistical power to distingish between varios dose-response models. The observed incidence was fitted to a series of models incorporating both initiation and cell-killing terms at varios powers of dose, as well as threshold doses [P21]. From a rather elaborate analysis it was conclded that only one non-threshold model cold not be discarded, and this was for indction of epidermal tmors with a very high exponent of dose (abot 4) in the indction term. The incidence above apparent thresholds sally rose very steeply with dose. The athors sggested that the apparent thresholds may be de to some relatively radioresistant factor in the skin that restrains potential tmor cells, and that this factor can be overcome by very high doses. and also perhaps by aging In CD rats. the incidence of tmors over the ascending part of the crve rose roghly with the forth power of the electron dose [B59. H12, Hl3] and with the sqare of the dose of alpha particles [B59]. However, these conclsions are only approximate, becase the data were insfficient for more detailed analysis. The vale of the RBE of alpha particles at abot 1 tmor per animal, relative to high energy electrons. was roghly 3. The peak incidence of skin tmors always occrred at several tens of Gy of sparsely-ionizing radiation Fractionation of the electron dose [B57, B58] showed that sboncogenic damage was repaired with a half-time of abot 4 hors. This observation sggested that the repair process was operating at the celllar level, as for sblethal damage. Experiments on the inflence of dose fractionation were also made with 1-MeV protons [B56): when fractions were separated by 24 hors. a redction of the effect was interpreted as showing complete repair of recoverable damage. It was conclded. therefore. that for skin tmor in CD rats 1-MeV protons have the sal characteristics of both high- and low-let radiation: an RBE of abot 3. typical for the former, bt recovery at fractionation and yield of the effect proportional to a power of dose greater than one, typical for sparsely-ionizing radiation It is difficlt to say to what extent skin cancer indction in rats is an adeqate model-qalitatively or qantitatively-for man. A relatively low sensitivity of hman skin to cancer indction by radiation was noted in annex G of the 1977 UNSCEAR report [U6] and in the BEIR III Report [C29). On the other hand, histogenesis of most skin cancers in rodents is different from those of the hman skin [LS). Albert. Brns and Shore [A4) sggested that radiationindced cancers in CD rats, and in children irradiated for tinea capitis. show similarities in the distribtion of the latent periods when time is normalized for both species to their mean life spans. However, the fractions of total skin srface irradiated were different in animals and in children, so that any similarity cold be fortitos On the basis of data in CD rats, Vanderlaan et al. [VI) proposed a mathematical model of skin cancer indction by ionizing radiation. The model is based on the hypothesis that radiation can initiate a cancer by converting normal cells (S 1 ) to potentially malignant cells by one of two rotes: a reversible rote involving two steps (S 1 - S 2 - S 3, the reversible stage being S 2 ) and an irreversible single-step process (S 1 - S3). It was assmed that the initiation process takes place in single cells bt its natre was not specified. Cells in the intermediate stage S 2 can revert to the normal state S I at a rate ;. assmed to be constant and independent of dose. The probability of developing carcinoma of the skin was assmed to be proportional to the nmber of irreversibly transformed cells S3. Expression of the latter as a fnction of time after irradiation was obtained from soltion of a pair of differential eqations. The soltion showed that the relationship of cmlative incidence verss dose was dominated by the dose-sqared component. In this model, the relative vales of the linear and sqared components are not predicted by LET or any other microdosimetric consideration. From experimental observations. Vanderlaan et al. [VI] conclded that, after a dose-independent tmor-free inten:al, throghot the remaining life span the tmors appear at an approximately constant rate. directly proportional to the dose. However. the experimental data show a considerable scatter for all radiations sed. 3. From fractionation experiments. a decrease of nrepaired damage with time was dedced, with a half-life of abot 4 hors. For estimation of the doserate effect on the yield of tmors. a dose-rate factor DRF (the ratio of acte to protracted dose for the same yield of S3 cells) was calclated, and it was predicted that the redction of the carcinogenic response at low dose rate wold be very prononced. This was confirmed in fractionation experiments with.8-me V electrons [A 18), showing that weekly exposre to fractions of 2 Gy per week (13 Gy in 65 weeks) yielded the same incidence as a single exposre to 15 Gy. 3 I. In conclsion, all the data discssed above sggest that linear extrapolation of skin cancer risk from the rates observed at high acte doses to the lowdose range wold greatly over-estimate the risk. To what extent these conclsions might apply to man is still. however, an open qestion. 12. Bone tmors (internal irradiation) 32. Althogh the data discssed in the following paragraphs are often referred to in terms of the average skeletal dose, it shold be borne in mind that the relevant dose for the effects in qestion is that delivered to the bone-lining cells. The crde incidence of bone sarcoma indced by 239P Cf, 252 Cf. 24 'Am, and 226 Ra was compared in C57BI/Do mice and albino mice [T22); the data seemed compatible with linearity of the dose-response relationships for all radionclides. However, the scarcity of data points for each series makes this conclsion at best tentative. It will be recalled that linearity of the dose-response was also seen earlier for bone sarcoma indction by 226 Ra in CFI mice [M69]. 215

54 33. The qestion of a practical threshold in bone sarcoma indction following internal irradiation by bone-seeking radionclides was raised originally by Evans [E4]. This isse was addressed again by a stdy of Raabe et al. [R32, R4] who tested the hypothesis on the basis of reslts on beagle dogs. Two grops of animals were given: (a) 226 Ra in 8 fortnightly injections of graded activities, the median age dring the injection period being abot 14.5 months, to simlate the hman experience with radim: and (b) 9 Sr by continos feeding at varios levels of 9 Sr/Ca ratio from mid-gestation throgh 54 days of extra-terine life, to simlate the hman contamination from the environment. 34. On inspection, the mortality data showed that at higher dose levels srvival declined as the dose rate to the skeleton increased. The mean latent period of osteosarcomas varied inversely with the time-averaged dose rate calclated for individal animals from periodic measrements of whole-body activity (Figre XXIII). When the srvival period t was plotted against the mean dose rate to the skeleton, the data cold be fitted to the relationship t = K o-s. where K and S were parameters determined by a least-sqares procedre. When a log-normal distribtion of times to death from osteosarcoma for a given dose rate,vas assmed (on the basis of empirically tested distribtions), K and its geometric deviation {± ag) for radim and strontim-9 were 25 (± 1.24) and 84 (± 1.72). respectively. The corresponding vales for S ± S.E. were.29 ±. I and. 77 ±.5. For 116 and 27 dogs dying with malignant bone tmors in grops given radim-226 and strontim-9, respectively. the fit to the fnction was good, with correlation coefficients of.92 and From inspection of the regression lines. it appeared that, if the relationship held down to the lowest dose rates, there mst be latent periods in the low-dose range exceeding the normal life span. A practical threshold was therefore defined as the dose rate corresponding to the intersection of the regression line log(t) verss Iog(D) (Figre XXIII) mins three geometric standard deviations of the mean srvival time t for control dogs. For dogs injected with 22 "Ra. the threshold dose rate so defined was.11 mgy per day, corresponding to a lifetime dose of.5 Gy and a probability of abot.7% of incrring a sarcoma towards the end of life. Corresponding vales were not given for 9Sr-contaminated animals bt they wold be expected to exceed the vales for radim by one or two orders of magnitde. 36. The S vales derived from dogs injected with radim were sed to fit the data on latency and skeletal dose rate in women dial painters and in RFM mice. There was a high degree of correlation between the species' average life span L (years) and K vales calclated for each species according to the relationship Km= [655 ± 57 (S.E.)] + [119 + I (S.E.)] L (4.2) It cold be conclded that a practical threshold for 2 26Ra in man is.39 mgy per day and a total dose of.8 Gy. The relative effectiveness of 9 Sr and radim, in terms of doses indcing eqal effects in dogs. was also estimated: it varied in some proportion to the average dose from 9 Sr to the skeleton. If this relation holds down to the lowest doses. a practical threshold of dose rate and cmlative dose for 9 Sr wold be at least one or two orders of magnitde greater than those for 226 Ra. 37. The same methodology was applied [R4] to a stdy of relative effectiveness for the indction of bone sarcoma in yong beagles injected with nine a emitting bone-seeking nclides. All nine stdies showed, with relatively good precision, a three-dimensional, log-normal, response relationship represented in two dimensions by the eqation t = Ko-s. S was fond to be, on average..29 (±. l S.E.) for all radionclides. 1.,, 3 >, - "' 2 :c... < 1... C :!: t =25(,- 29 Ra Ra Bone tmor from radim-226 o Bone tmor from strontim-9 1 Time to death (days) after beginning exposre Average dose rate (rad/day) ' ' ',... ' AVERAGE OOSE R~TE TO SKELETON, D (rad/day) 1 Figre XXIII. Comparison of the dose-response relationships for radlm-226 and strontlm-9 In beagle dogs that died of primary bone cancer. The dala show the median time, tl, of spontaneos deaths among nexposed control and the fitted fnctions In irradiated animals. [R32] 216

55 A practical threshold dose or dose rate, below which bone sarcoma deaths are very nlikely. is sggested at low dose rates for all these radionclides. jst as it was reported for 226 Ra [R32]. Mean skeletal doses at which the defined practical threshold was reached were lower than for 226 Ra (.5 Gy) by the following factors: 228 Ra, 3.; 241 Am, 6.4; 2 49 Cf. mcf. 2 53Es, 6.6: 239P. 9.; 228 Th. I. 7; 2J8P, If these factors were expressed in terms of the doses 1 the bone-lining cells, their spread wold be considerably redced. 38. The reasoning of Raabe et al. [R32] is based on an empirical linear regression of srvival times verss mean skeletal dose rate in animals dying with, or de to, malignant bone tmor. However. in the low-doserate grops and among the controls a majority of animals were still alive at the time when the experiments were evalated. There is some ncertainty in the data as to the mean srvival vales and their distribtion among control animals. The data cold also change somewhat as a reslt of frther follow-p. The derived practical thresholds mst therefore be treated as tentative at best and any final assessment mst be deferred. Moreover. any hypothesis regarding species-related interpolation of the risk neglects differences in ssceptibility to the indction of bone sarcoma among varios strains of mice and breeds of dogs. Alternative interpretations of the data were presented by Mays [M69]. However. the concept of a practical threshold for bone sarcoma is important for risk assessment. becase. if extrapolation of the basic relationships is correct. there mst be vales of the dose and activity at which no radiation-indced life shortening occrs, and therefore the somatic component of the index of harm [I 6] can be taken to be zero. 39. Osteosarcoma indced by P in 294 beagle dogs of both sexes was stdied following single injections at ages between 4 and 8 days [W 14]. The activity injected ranged p to abot.11 MBq/kg body weight. in 7 grops of animals [M 11]. Times at death de to osteosarcomas were fitted to the Weibll distribtion fnction and described in terms of cmlative hazard or hazard rate. in both cases as a fnction of time after injection. The power fnctions of time for cmlative hazard or hazard rate corresponding to R(t,D) and r{t.d) for tmors detected at death in individal injection grops were fitted to the data by a most likelihood method. The plots on a log-log scale were virtally parallel for all grops, and therefore the proportionality constants of the lines cold be analysed as a fnction of the injected activity of 239P. Below 37 kbq/kg. there was a strong inverse relationship between the injected activity of pltonim and the time at which a given vale of R(t.D) was reached. At the highest injection level the hazard was lower than in the preceding grop, probably de to cell killing or some other non-stochastic effect. When the highest activity level (.95 MBq/kg) was discarded. the data for R(t.D) (below 37 kbq/kg) gave an acceptable fit to a pre qadratic model (P =.19) and linearity cold be rejected (P <. I). A model containing both a linear and a qadratic component did not fit the data better than the qadratic one, where R(t,D) depended also on time after injection to the power of The same data were stdied in parallel [Wl4] by the generalized Marshall-Groer model for bone sarcoma indction [M4) (see V.E.2 for details). The soltion of the model showed lack of a linear component in the relationship between effect and cmlative dose. The best fit was obtained for a qadratic relationship between dose and R(t,D) for osteosarcoma. Since many animals injected with the lower activities are still alive. a linear contribtion to the dose-incidence relationship might still appear after frther follow-p. When the same data [PI 7) were treated in terms of a non-parametric proportional hazard relative risk model. both a linear and a qadratic fnction cold be fitted to the data and neither cold be exclded. However. if-as was legitimately done by Whittemore [WI4]-the data for the highest injection level are omitted. the reslts again spport a roghly qadratic rise of the relative risk with injection level An experiment has been completed [Wl9] on 16 yong beagle dogs. injected at the age of 17 ± 2 months with saline (controls. 44 dogs) or with increasing amonts of 226Ra (8 levels from abot 26 Bq to 37 kbq/kg). The average a-particle dose to the skeleton varied from.35 to 19 Gy. Most of the primary bone tmors were osteosarcomas. Bone sarcoma incidence was a monotonically increasing fnction of the dose p to 7 Gy. The dose-response fits well a linear non-threshold model over the six lowest dose ranges with a slope of 3.3% per Gy. No tmors were fond in the controls or in the.35 Gy grop, bt I tmor over 25 dogs was fond in the next grop with a mean skeletal dose of.85 Gy. No significant change in bone tmor incidence was seen among the three highest dose levels at which the incidence exceeded 9%. and the life span was considerably shortened. There was very little. if any, life shortening at the five lowest levels p to 9.5 Gy Liver tmors were indced in Chinese hamsters by 239 P citrate injection. and the crde incidence for the few dose points available sggested an initial approximately linear rise p to 2.7 Gy at the time of 5% srvival. At the highest dose of 14 Gy. the effectiveness per nit dose was mch lower [B87] The dose-response relationship of injected thorotrast was stdied in female Wistar rats for the sm of liver and spleen tmors [Wl5]. In some of the experimental grops. the preparation was enriched with 2 3 Th in order to stdy the inflence of the mass of injected colloidal thorim dioxide. For a constant dose rate (activity). an increase in mass by a factor of 1 proved more effective in respect of life shortening. When the crde percent incidence (I) of all benign and malignant liver tmors was plotted against the relative dose rate (D), a linear dose-response relationship was observed: I{D) = 3.3 ±.79 D with a correlation coefficient of.97. The I vale at D = does not differ essentially from the observed control incidence of 2. 7%. 217

56 1-t Combined tmor data and indirect inferences 314. Life shortening by radiation was reviewed in depth in the 1982 UN SC EAR report [ U24]. It was conclded that, in the incermediate- and low-dose region, life shortening is de essentially to a higher tmor mortality. This does not mean, of corse, that dose-response relationships for both endpoints are directly comparable, becase, even for the same radiation, the mean latent period of some tmors varies with dose, dose rate, age at exposre, etc. However, if life shortening is the major consideration, then comparison of reslts from \'arios dosage schedles for the same radiation and in the same strain of animals cold provide some indication as to what the comparative incidences and/or latent periods might be. This seems particlarly interesting for netrons. for which no hman data are available Life shortening and incidence of varios diseases. inclding Jekaemias and malignant tmors. were stdied in male BALB/c mice [M62] after single and fractionated doses of gamma rays split into 1 eqal fractions delivered at intervals of I day. The doses ranged from.25 to 6 Gy. The cases of death were ascertained by atopsy and histological examination, and actarial incidences were calclated, with correction for competing risks. In general. single exposres were more effective than fractionated ones with regard to life shortening. Over the whole range of doses, the incidence of all lekaemias and cancers combined appeared significantly higher after fractionation. However, the dose-responses were very flat and inclded more than 7% of spontaneos cancers. For all carcinomas and sarcomas, except lng carcinomas (for which the spontaneos incidence in BALB/c males is very high). a higher effectiveness of fractionated doses was seen only after doses above I Gy These observations mst be taken with some cation, de to the very high spontaneos tmor incidence; to the Jack of statistically significant differences between incidences of individal tmors after two modes of irradiation; to the generally flat doseresponse relationships; and to the fact that. as shown by Storer [S53]. this strain has a nmber of significant associations between varios tmors. Finally. the effect of single and fractionated gamma-ray doses pon life span was opposite to that seen for combined cancer incidence, showing. perhaps. that in this case a specific combination of indction and acceleration of some tmors cold have played a significant role Whereas incidence of all carcinomas and sarcomas combined (except lng carcinomas) showed a negative trend after single doses of gamma rays on the same strain [M63]. an almost linear increase was seen after single doses of d(5)-be netrons in the range from.2 throgh 3 Gy Life shortening of B6CFI mice was stdied by single, fractionated (short- and long-term), and lifelong exposres to gamma rays and fission (.. 85 MeV) netrons [T2. T21. T25]. When srvival data were compared for doses given in 24 weekly fractions or as single exposres. the effectiveness of gamma radiation was consistently decreased by fractionation. However, for netrons the same fractionation scheme increased the effect throghot the whole range of doses. down to.5 Gy. It was arged. therefore. that netron fractionation in the intermediate- and high-dose range led to an increased incidence of tmors, or their earlier appearance, or both Thompson et al. [T26] recently pblished new additional data on life shortening indced by single exposres of 86CF I female mice to fission netrons in doses down to.1 Gy. They conclded that the assmption of a response proportional to og.s grossly overestimated the life shortening indced by.1-.3 Gy of netrons and that the data p to.. 1 Gy of netrons were best fitted by a linear dose-response relationship. The limiting vale for the RBE fission netrons compared with gamma rays was in the region of 12 to 16. depending pon the method sed to analyse the data. No consideration was given in any of these pblications to RBE vales at low doses for fast netrons as compared with the standard reference radiation. i.e.. x rays of abot 2 kvp. 32. In another experiment [S57], life shortening was stdied in female BALB/c mice exposed to single doses of fission netrons (mean energy 1.3 MeV) or given into two eqal fractions at either J- or 3-day intervals (whole-body doses from.2 to 2 Gy). Protracted exposres from a 252Cf sorce were also sed at rates from J to 1 mgy per day (total.25 to.4 Gy). Life span was considerably shortened between zero and.5 Gy for both single and fractionated exposres. and then a platea ensed. The data were fitted by a linear and a sqare-root model. Both models adeqately described the reslts below.4-.5 Gy for single. fractionated and protracted exposres. with the exception of a 3-day protraction scheme where the sqare-root model cold be rejected and the linear one still provided a good fit. In contrast with the observations in B6CF 1 mice, no increase in effectiveness on life shortening was observed for fractionation and protracted exposres to netrons below.5 Gy. Only at.5 and 2. Gy was a significantly increased effect of fractionated doses observed Storer and Mitchell [S62] analysed pblished data from the Oak Ridge and Argonne laboratories on the shape of the dose-response crves for life shortening indced by single exposres to fission netrons at doses from.25 to.2 Gy and by fractionated exposres reslting in total doses p to.8 Gy. The response was fond to be proportional to ~ 9 to DJ, and not to the fnction ~ 5 which does provide a reasonable description of the dose-response crve at higher netron doses. The limiting RBE of fission netrons compared to gamma rays was fond to be if the response were proportional to ~ 9 and to be in the eqally likely event that the response were proportional to ~ The range of RBE in both cases was determined by the strain and sex of the mice. Ths. the most recent pblications [S62. T26] sggest that a linear dose-response model rather than a sqare-root model is appropriate at low netron doses. 218

57 C. SUMMARY AND CONCLUSIONS 322. On the assmption that cancer is initiated by lesions in the genome of a single and relatively atonomos cell. the kinetics of indction of point mtations and chromosomal aberrations. mostly symmetrical translocations, cold perhaps throw some light on the form of the dose-response relationship for cancer initiation. Stdies of cell transformation in vitro cold also provide some information in this respect. These end-points cannot, however. provide insight into other factors affecting tmor development at the tisse and systemic level: ths, the relevant conclsions mst be regarded as over-simplifications of the real sitation The dose-response relationship for point mtations indced by low-let radiation in mose male and female mose germ cells appears to be crvilinear (pward concave). It also appears to be greatly dependent on dose fractionation and protraction as a reslt, at least in part. of repair of primary damage at the celllar level. At very low dose rates. linearity and independence of dose rate have been observed. For high-let radiation the sal form of the crve is linear for all dose rates; fractionation and protraction effects are marginal or absent. Under these conditions, linear extrapolation from high and intermediate doses delivered at high dose rates. down to low doses and dose rates, \vold over-estimate the mtation rate by sparsely-ionizing radiation. For high-let radiation, a linear extrapolation provides an adeqate estimate of the risk The indction of somatic mtations in the stamen hair of the plant Tradescantia was stdied over a wide range of doses, dose rates and radiation qalities. This system provides direct and extensive information at very low doses of x and gamma rays and of netrons. The relevant dose-response relationships fit very well a linear-qadratic model for low LET radiation and a linear model for netrons Data on varios other types of somatic mtations are also available in several established and early-passage mammalian cell lines. For x- or gammaray doses delivered actely, the observed dose-response relationships are mostly crvilinear. with prononced fractionation effects in the few instances tested. Exceptions showing linear responses (e.g.. hman fibroblasts) are also fond, showing that the shape of the crves depends on the biological characteristics of the cells. High-LET particles indce mtations in a linear proportion to dose. Sb-mtational damage is readily reparable after low-let radiation bt not after high-let particles. Stdies in several cell lines show a simple inverse relationship between the indction of somatic mtations by low-let radiation and the logarithm of cell srvival. This mst imply that linearity of the observed response is accompanied by simple exponential cell srvival. Since for low-let radiation this is an exceptional finding in mammalian cells, the inverse relationship between srvival and mtation mst, as a rle, imply an pward concave dose-response for indction of somatic mtations. similar to that described for Tradescantia For low-let radiation. data on chromosomal exchanges in germinal and somatic cells are, almost withot exception, in agreement with a linear-qadratic model, both in respect of dose and of time-dose relationships. For terminal deletions. linearity wold, in theory. be expected for all types and dose rates of radiation. bt crvilinearity is often seen in some cells (e.g., lymphocytes). This is attribted to the difficlty in distingishing between terminal and interstitial deletions that have linear and qadratic relationships, respectively. with reslting crvilinearity of the response for both types combined. For high-let radiation the response is linear and practically naffected by dose protraction or fractionation Mch work has been done over the last few years on radiation-indced oncogenic transformation in vitro of several established and some early-passage mammalian cells. Nmeros factors affect the observed transformation freqency. These inclde: cell density at irradiation and post-irradiation: cell-cycle stage; time between seeding and exposre: phase of growth at exposre; mitotic delay; and promoting and inhibiting agents. The experimental model appears to be mch more complex than originally envisaged for cells in cltre, and cell-to-cell inflences are fond to be present Cells irradiated with a wide range of doses and types of radiation show dose-response crves that have some featres in common with those reported for tmor indction in animals. These are: an initial rise of the transformation freqency with dose p to a maximm and an ensing decline or platea: a higher transformation efficiency by high-let radiation: and a decline of the transformation yield by low. as opposed to high, dose rate of gamma rays in cells irradiated in cltre 4 hors or later after seeding. An enhancement of the oncogenic transformation in vitro by a redction of dose rate of fission netrons at low and intermediate doses is similar to that which has been described in some, bt not all, experiments on tmor indction (or life shortening) in animals However, several featres of the single-cell stdies are at variance with animal stdies. These are mostly for cells irradiated shortly after seeding: a shallow rise of the transformation rate over the range of intermediate doses. with a power of dose less than nity: an enhancing effect of low dose rate and dose fractionation for low-let radiation for a fractionation interval of abot 5 hors at doses below 1.5 to 2. Gy, with redced transformation at higher total doses. These reslts most probably indicate a nonhomogeneos sensitivity to transformation of the irradiated cells. de to the non-random distribtion of cell-cycle stages in freshly seeded cells. If so. they wold not be representative of most sitations in vivo. 33. The above reslts were sometimes interpreted to imply that linear extrapolation to zero dose from the transformation freqency at high doses cold lead to an nder-estimation of the effect at doses below.5 Gy, particlarly for fractionated and protracted exposres. For the time being, however. while taking note of the complexities referred to above and of the 219

58 enhancing effect of fractionation, UN SC EAR does not encorage a direct projection of the in vitro data to the sitation in vivo. particlarly when irradiation is performed early after seeding. Frther analysis of varios cltre conditions, and a better nderstanding of the mechanisms of initiation, oncogenic transformation and progression in vitro, is reqired for meaningfl extrapolations to tmor indction in vivo. It appears likely that reslts obtained from asynchronos cell cltres (i.e., those irradiated later than abot 4 hors after seeding) and those irradiated in density-inhibited state may more adeqately represent the sitation in vivo New dose-response relationships for a variety of tmors generally spport the conclsions reached in the 1977 UNSCEAR report. the most important being the specificity of the response for each tmor-host system. The new experiments enlarge the data base, particlarly at the lower end of the dose scale, down to 15 mgy of gamma rays and to a few tens of mgy for netrons. They jstify more specific statements than hitherto possible abot the form of some doseindction relationships For sparsely-ionizing radiation, most crves are pward concave and may be fitted by linear-qadratic or qadratic (lekaemia in CBA/H mice) models, bt in some cases approximate linearity may apply. This is so with mammary fibroadenoma and carcinoma in Sprage-Dawley, adenomas in WAG/Rij and carcinomas in AC rats. thymic lymphoma in RFM mice, and lng adenocarcinoma in BALB/c mice. Thymic lymphomas in most mose strains and tmors with a prononced hormonal dependence (e.g., ovarian tmors), show crves with prononced thresholds. They may reflect the role of non-stochastic mechanisms in the development of these tmors, sch as the need to inactivate a large proportion of hormonally-active cells in the ovary. or to kill a large proportion of cells in the thyms The incidence of the so-called reticlm cell sarcoma after whole-body irradiation sally declines with increasing dose in several mose strains. Animals protected from early radiation death. by either bonemarrow transplantation or partial marrow shielding, show a very low incidence of this neoplasia. It has been sggested that by relating the nmber of indced tmors in these animals to the nmber of irradiated marrow stem cells at risk (assmed to be the targets for tmor indction) it may be possible to estimate in vivo the incidence of tmors per cell at risk. The x-ray dose-response crves obtained nder these assmptions are crvilinear (pward concave) Indction of skin cancer in rats and mice by local low-let irradiation reqires high doses. The dose-response relationships are therefore grossly crvilinear (pwards concave) and apparent or real thresholds cannot be exclded. Fractionation and protraction of doses greatly redces the carcinogenic action of radiation in the skin of rodents Dose-rate stdies with sparsely-ionizing radiation almost invariably show a decreased incidence with decreasing dose rate. In some instances (lng adenoma in RFM and BALB/c mice and harderian or pititary tmors in RFM female mice), lowering the dose rate reslts in a more nearly linear dose-response relationship in which the dose-sqared component is sbstantially redced or abolished For netrons. the ascending slopes of the doseresponse crves are closer to linearity than for gamma rays. However, in some cases (lng adenoma in RFM mice). crvilinearity is sggested at low doses, even thogh linearity cannot be exclded. Factors sch as a dose-related latency may perhaps contribte to this crvilinearity. In other cases, netron-indced tmors rising with a power of dose Jess than nity have been confirmed at low doses. The downward concave shape may be acconted for by killing of potentially neoplastic cells For most tmors, either no effect, or an enhancement of the tmor yield by a lower netron dose-rate or fractionation, has been docmented. At low and intermediate netron doses, sch effects are Jess prononced than for x rays. At high doses. an enhancement of tmor incidence bv fractionation or protraction is often observed. bt th~ overall pictre is not as simple as celllar data might imply. Factors operating at the tisse level, or perhaps systemic inflences, cold modify the final tmor appearance to some extent Plmonary carcinomas indced by low doses of inhaled radon and of netron irradiation show an essentially linear non-threshold relationship with cmlative exposre or dose. respectively. High exposre rates of radon daghters at high total exposres are, however, Jess efficient than low rates. For netrons also, the effectiveness per nit dose decreases at high doses. These observations are in line with reslts of epidemiological stdies in miners exposed to radon daghter prodcts Analyses of the latent period of bone sarcoma after incorporation of 226Ra and 9 Sr in dogs and mice show an inverse relationship between mean latent period and mean dose rate to the skeleton. According to these observations, there may be a practical threshold for these tmors. Under these circmstances a linear extrapolation of the incidence race and of the length of life lost per indced bone sarcoma down to zero, from vales recorded at high doses. wold srely overestimate the risk at low doses. This also applies to tmors indced by P-emitting bone seekers. 34. Stdies of bone sarcoma indced by alphaemitting bone-seeking radionclides. particlarly 226 Ra. spport a linear activity-response relationship for crde incidence. However. for 239p, more refined statistical analyses of data from dogs injected with this nclide, and observed for longer times, have shown that a prely linear dose-response relationship between integral tmor rate (cmlative hazard) and administered activity is highly nlikely. The same seems to apply to the relationship between dose rate to cells at risk and tmor rate (hazard rate). The appearance of a linear component in the crves cannot be exclded 22

59 in the follow-p of animals that are still alive bt at present the dose-sqared component is dominating The most recent data on radiation-indced life shortening in mice sggest that a linear dose-response model applies at low netron doses, while at intermediate and high doses the response follows a sqareroot fnction of the dose. Fractionation and protraction of the netron dose increases the effect of life shortening in some experiments. bt not in others; fractionation of low-let radiation redces the effect. To the extent that life span shortening in irradiated mice can be attribted to the indction of new tmors, these conclsions may also be applicable to radiation carcinogenesis. V. DOSE-RESPO:'\SE RELATIONSHIPS FOR TUMOURS IN MAN 342. Data on the incidence of radiation-indced hman cancer, at different levels of dose, may be tested for consistency with varios models in two ways: directly or indirectly Indirect inferences may be drawn from limited comparisons in the few instances where risk coefficents (FT,\ or FTA) are available at two ends of the scale: high and low doses (total or per fraction). Risk coefficient which are not sbstantially different nder sch circmstances will be taken to be consistent with a linear-qadratic or a linear model, becase in most cases the variability of data precldes a clear distinction between the two models. Theoretically, for a dose-sqared model, an appreciable difference wold be expected between Fn, or FTA at low and intermediate doses, bt examples of this kind ha\'e not been fond for the types of cancers re\'iewcd In direct stdies, attempts may be made to fit varios models to the epidemiological data O\'er a wide range of doses. There are often shortcomings to these data (dosimetric ncertainties, a narrow range of doses, pre-selection of grops. ill-defined controls, possible inflence of confonding variables. errors of sampling) and consistency with a model rarely, if ever, excldes a fit to alternative models. Ths. consistency of a set of data with a gi\'en model shold not be taken as evidence that only that model provides an adeqate description Mch information on dose-response relationships for radiation-indced malignancies (e.g., lekaemia, cancer of the breast) has been derived in the past from the follow-p of atomic bomb srvivors at Hiroshima and Nagasaki. These cohorts represent the statistically most powerfl set of data available for dose-response stdies, becase of the great nmber of person-years at risk. the wide range of doses, and the relative lack of selection of the samples The dose-response data in these grops were based on the so-called T65 dosimetry. Since 198 this dosimetry has been nder revision [K34, L27, L3, R4, R6] and a new dosimetry system (DS86) is being installed at the Radiation Effects Research Fondation in Hiroshima as of March I 986. Preliminary estimates based on somewhat more detailed analvs~s and the crrent state of knowledge (for a prelim inary review see [877)) differ sbstantiallv from those of T65 dosimetry. It appears now tha"t the gamma-ray component had been nder-estimated in Hiroshima bv a factor of 1.5 to more than 2. and the net;on component had been over-estimated bv a factor of abot 1 in Hiroshima and of abot 2.5 in Nagasaki. As a reslt, mch of the information based on T65 dosimetry will have to be re-assessed pon completion of the ongoing dosimetric revision and will be reviewed by UNSCEAR in the ftre. However. there is no qestion as to the qalitative validity of the previosly accmlated information abot the effects of age and sex pon the incidence of varios neoplasms in atomic bomb srvivors. The same applies to the clinical and pathological forms of diseases and to the distribtion of their latent periods, which are discssed here. (a) A. LEUKAEMIA I. Direct stdies Lekaemia in aromic bomb sn>ivors 347. From the information collected in the Life Span Stdy Sample it appears that the age at time of irradiation had a prononced effect on the magnitde of the risk when expressed in absolte and relati\'e terms. The effects of sex differences were also discssed in the 1977 UNSCEAR report [U6]. The ageadjsted lekaemia mortality rates reveal that the risk of lekaemia (all types) in males is significantly higher than in females. For both cities combined. the male : female ratio is and 1.56 for the TD65 kerma categories of and> I Gy. respectively. The distribtion of cases in time varied with the type of lekaemia and with age. as shown schematically in Figre XXIV [1). (b) Lekaemia in x-ray treated spondylitics 348. Smith and Doll [S49) stdied mortality throgh 197 in 14, l l I ankylosing spondilitis patients given a single corse of irradiation dring the period Of these patients. 98.5% were traced to their death, to the date of emigration from the United Kingdom. to the closing date of follow-p (I Janary 197) or to the end of a vear after a second treatment. whichever was the eariiest. The average period of follow-p of those with one corse of treatment was 16.2 years. Cases of death were obtained from death certificates The nmber of expected deaths in the cohorts were calclated from age- and sex-specific mortality rates for England and Wales in the respective calendar five-year intervals. An analysis of mortality revealed that the nmber of deaths was 66% more than 221

60 ALL FORMS OF LEUKAEMIA ACUTE LEUKAEMIA CHRONIC GRANULOCYTIC LEUKAEMIA AGE ATB A - < 15 B C D - > 45 E - > YEARS AFTER EXPOSURE Figre XXIV. Schematic model of the Inflence of age et the time of the bombing (ATB) end calender time on the lekeemogenic effect of radiation (heavily exposed srvivors). [1] expected in the patients ( 1759 verss 162 expected). The total excess of lekaemia, as well as that for neoplasms. was highly significant (p <.1), and did not differ among males and females (the latter made p only one-sixth of the poplation). 35. The only malignancies for which the doseresponse relationship has been stdied were lekaemias. The excess death rate was high. the overall observed/ expected (/E) ratio being (3.6 and 5. 1 in women and men, respectively). The time trends were characteristic. with the death rate rising even at 2 years after the first irradiation, peaking at 3 to 5 years, and declining to the control level more than 2 years after treatment. Chronic lymphatic lekaemia was apparently not increased by the treatment. The relative risk of lekaemia (/E ratios) was eqally increased in all age grops (age at first treatment). The excess risk coefficient per 1 5 person-years rose from 251, in those yonger than 25 years, to 1366 in patients treated at more than 55 years of age. In this instance the radiation-indced lekaemia incidence increased temporarily in proportion to the age-specific spontaneos incidence {at age of exposre). as represented schematically in Figre II. c and d As the distribtion of local absorbed and mean doses to the bone marrow among all patients is nknown, it had to be reconstrcted throgh an extensive phantom stdy in all patients who died with lekaemia and in a sample of all followed-p patients. The doses ranged from.5 to 7 Gy, with a rather niform distribtion of freqency. involving 9% of cases below 5 Gy. The proportion of cases above 6.5 Gy did not exceed 2%. The mean marrow dose was 2.49 Gy The excess lekaemia mortality rate, plotted verss the mean dose to the marrow is very erratic. and a zero response (independence of mortality rate on dose) may not be exclded on prely statistical gronds. However, de to the time characteristics of the mortality, and to the fact that no increase of lekaemia mortality was seen in non-irradiated spondylitics [R33], the non-zero effect at zero dose was discarded as very nlikelv. Three models, a linear. a linear with cell killing, and a qadratic with cell killing were fitted to the data by the method of maximm likelihood. Sch a method is not formally correct for the last model becase simple averaging of organ doses when a large fraction of marrow was nexposed, 222

61 mst lead to erroneos reslts when a dose-sqared dependence of initiation of the cells is assmed. The first model cold be discarded (P <.3). and the fit obtained for the second was satisfactory (P =.28) The coefficient of the linear slope in the second model gave a risk coefficient in accordance with most other estimates of lekaemia risk. However, owing to the erratic character of the dose-response, this vale mst be taken with cation. The cell-killing term cannot be taken in its sal meaning of mean lethal dose. becase absorbed doses to the irradiated fraction of the marrow were 2.5 times higher than the mean. Moreover, the fractionated treatment allowed sblethal repair and celllar repoplation to occr. 2. Indirect evidence 354. The risk of lekaemia was stdied in radiologists exposed to x and gamma rays either at low dose rates or at higher rates. bt in small fractions, accmlated over decades [M7. S8]. The risk was significantly elevated. The records of mortality of the cohort followed for 5 years showed excesses of 39 and 17 deaths per 1 6 person-years at risk de to lekaemia and to aplastic anaemia, respectively. Calclation of risk per nit dose to the marrow is affected by great ncertainty abot the doses accmlated. The diagnoses are also very ncertain. The BEIR Committee [B2] estimated the likely average vale to be abot 6 Gy, with a reslting annal risk coefficient of 7 lekaemia deaths per 1 6 Gy- 1 a-1. This vale is not significantly different from estimates for low dose and dose rate derived by UNSCEAR in annex G of its 1977 report [U6] from stdies on other actely-irradiated grops (spondylitics, females irradiated for indction of menopase, or other benign conditions) given doses and/or fractions above I Gy. If the above estimates of dose were in error in either direction by a factor of 2-3, the derived risk wold still appear similar to that dedced from stdies on acte irradiation, and therefore consistent with the linear or linear-qadratic model of indction of lekaemia In conclsion. therefore, it may be assmed, on the basis of the present fragmentary information, that the dose-response relationship for lekaemia after acte low-let irradiation is likely to follow a linearqadratic model. B. CANCER OF THE BREAST 356. The female breast is one of the most sensitive organs for cancer indction by ionizing radiation [B2, I I, M28, U6]. The data reviewed in annex G of the 1977 UNSCEAR report [U6] have been given either in terms of mortality or morbidity, bt the latter category will preferentially be discssed here. Epidemiological data relate to three categories of exposre: atomic bomb explosions, therapetic irradiation for benign conditions (e.g.. mastitis). and tberclosis patients ndergoing mltiple floroscopies for pnemothorax or pnemoperitonem. Since the 1977 UNSCEAR report [U6] new information has become available [TI 2, T23] and several attempts have been made to fit cancer indction models to the relevant epidemiological observations [B24, B25. Ll, L22] Dosimetric estimates were retrospective in all stdies. The ncertainties were particlarly great in the Nova Scotia Floroscopy Stdy [M39. M4], where sbjective jdgement made assessment of the likely confidence limits nfeasible [B2. B25]. This stdy, therefore, shold not be sed for discssion of consistency with dose-response models. The Swedish radiotherapy stdy [B4] is also limited by the absence of an appropriate control grop and by a possible association between fibroadenomatosis or chronic mastitis and increased cancer risk. The following discssion. therefore. is based principally on stdies of three series: atomic bomb srvivors (tentative). the Massachsetts floroscopy stdy, and the Rochester mastitis patients Several factors cold affect the reslts of sch stdies. if not appropriately acconted for. These wold mostly be age at irradiation, age at manifestation of the effect, length of the latent period, dration of the period at risk, and relationship between latency and dose Age at exposre is a major determinant of the breast cancer risk. The three stdies cited provide strong evidence of cancer indction in females exposed below age 3, with girls of 1-19 years having the greatest absolte risk. The coefficient of the relative risk increment was already highest in the lowest age bracket (-9 years) at the time of the bombing, bt the absolte attribtable risk coefficient for this age was lower by a factor of 2 when compared with that for the next age bracket. i.e., 1-19 years. The absolte and relative risk coefficients for ages 2-39 years remain approximately constant [T23]. The floroscopy data are scarce for ages 4-49 years. and absent for older ages at irradiation. The atomic bomb srvivors showed no significantly elevated cancer incidence when exposed at ages higher than 4. On the other hand. the Swedish radiotherapy data [B4] imply a radiation risk at all age grops. inclding 4-49 and greater than 5 years, bt statistically the estimates at older ages are ncertain. The possibility that concomitant ovary irradiation in atomic bomb srvivors at, or close to, menopase might have been responsible for this protective effect (de to sterilization) was sggested by Boice et al. [B25] and Kato and Schll [K39], bt this explanation mst remain hypothetical. It has more recently been sggested [T23] that differences in the distribtion of the risk at ages above 4 parallel the age-specific incidence of mammary cancer in the varios contries (Japan, United States of America, Sweden). Taken together. the data show that the risk is highest when the hormonal action affecting the gland is most intense [T23]. It shold be mentioned, however. that the follow-p of the yonger age cohorts is still incomplete. 36. The increased incidence becomes manifest in all grops at ages above 3 and particlarly at the ages of highest natral incidence rate of the disease. This 223

62 observation calls for the se of age-corrected incidences in the exposed and control grops. The so-called minimm latent period can be estimated approximately from all observations [B25], bt pitfalls of sch estimates have been discssed in chapter I. In the Massachsetts, Nova Scotia and Canadian floroscopy series, no excess of breast cancer cold be seen for the first 1 years after exposre. In atomic bomb srvivors, and in the Swedish radiotherapy patients, the increased risk-particlarly in those irradiated at ages above 3 years-cold be seen even at 5-9 years after irradiation. Ths, it appears that in yonger women (nder 3 years) the minimm latent period is perhaps close to 1-15 years, and still longer in irradiated children, whereas in the older women it may be shorter (abot 5 years). This observation is important for the selection of a period at risk in calclations of the excess incidence. All stdies [B26. Sl6. TI2. T23] show a contining excess cancer incidence p to follow-p times of 3-45 years in women older than 3 years at exposre. There is, as yet. no indication of a decreasing risk even at the longest follow-p of 45 years. There also appears to be no relationship of latency to dose [L32]. 1. Radiotherapy for mastitis 361. The Rochester stdy on women treated for mastitis [S 16] was reviewed in detail in annex G of the 1977 UJ\SCEAR report [U6]. Only those aspects dealing with the shape of the dose-response relationship are discssed here. By dividing the irradiated grop into five dose sb-grops, a dose-response relationship was obtained which was controlled for age and for interval since entry into the stdy, by means of the Mantel combination 1 2 techniqe, and by test of linear and qadratic trend. The interval at risk for the irradiated grop was from 1 to 34 years after exposre. In a first analysis, the average dose to both breasts was sed as the independent variable: the data are presented in Figre XXV. The difference between controls and irradiated women was significant (p =. I) and the linear trend 7. 2 test for association with dose was highly significant (P =.6). The incidence increased linearly p to abot 4 Gy, and then dropped al higher doses. No significant qadratic trend with a coefficient greater than zero occrred. The second analysis was based on doses to single breasts. The relative risk for the period I -34 years after irradiation was highly significant (P <.1), and amonted to 3.3 when controlled for age and interval since entry into the stdy. The reslts in terms of relative risk are presented in Table 1. together with reslts of the first analysis. Significance was obtained for the linear gradient (P <. I) and for a qadratic effect with a negative coefficient (P =.3). The latter reflects a decreased carcinogenic effect above 4 Gy. owing perhaps to cell killing or other inflences The possible effects of fractionating the dose were analysed.vithin a range of total doses below 3.5 Gy. The analysis compared breasts receiving one or two treatments (mean dose = 1.9 Gy) with three or more treatments (mostly 3-4. mean dose = 2.86 Gy). a:..., < > :z: V')..., c::: n. ~ 6 n. a: LU ~ 5 I V') < ~ 4 a,... s 3 LU z ~ 2 :z: l (8.) J (12.1) I (4) (8.5) I..&...--r ,,---"""T--...,...--, DOSE (Gy} Figre XXV. Breast cancer Incidence rate as a fnction of the average dose to both breasts in mastitis patients for the Interval 1-34 years after Irradiation. Error bars represent 8/oconfldence Intervals. Nmbers in parentheses are breast cancer cases in each dose grop. [S16] Single. non-irradiated breasts served as controls. For a risk interval of 1-34 years post-irradiation, the adjsted coefficient of the absolte (attribtable) risk of breast cancer was Gy- 1 a- 1 per breast in the 1-2 fraction grop, as compared with I Gy- 1 a-1 per breast in the 3-4 fraction grop. Ths, fractionation did not appear to redce the carcinogenic effect. althogh the validity of the comparison is limited by the difference between mean doses in the two sbgrops The annal absolte risk coefficient dring the period, sing the average dose to both breasts, and age- and interval-adjsted incidence rates (irradiated adjsted to controls), was calclated as Gy- 1 a- 1 For irradiation of a single breast, the age- and interval-adjsted estimate of the coefficient amonted to Gy- 1 a- 1. If this figre is dobled for the two breasts. it is consistent with the average for two breasts given above. 2. Pnemotherapy patients 364. The Nova Scotia series on breast cancer indction by repeated floroscopies [Ml, M39. M4] showed that repeated low doses of diagnostic x rays, when delivered to the gland in sfficient amonts, led to an increased risk of tmors. The doses were 224

63 estimated retrospectively, bt the margin of ncertainty was so large [B2. B25] that detailed discssion of the likely shape of the dose-response relationship is difficlt. However, the fact that the risk estimate was similar to that derived from other stdies sggests additivity of the effect and ths approximate linearity of the dose-response relationships A follow-p stdy of Boice and Manson [B26] on female tberclosis patients has been discssed already in annex G of the l 977 UNSCEAR report [U6]. The methodology of estimating the radiation dose to the breasts was described in detail in a separate report [B24] and warrants some discssion becase it is critical for the assessment of the doseresponse relationship. The data on the nmber of floroscopies were abstracted from the medical records and checked by interview. It was estimated that 63% of the patients faced the physician dring the floroscopic examination, 16% faced the x-ray tbe, 21 % had variable positions. and 12% reported being rotated dring examination Ot of the 15 former tberclosis physicians interviewed. 29% reported that they examined their patients with breasts directed to the x ray tbe. Sixtynine per cent condcted the floroscopy with the beam-shtters wide open and 81 % always scanned the opposite lng. The average time of examination was given as 15 s (3 to 6); 7-8 kvp and 5 ma were the sal parameters of x-ray machines employed: and I mm Al filtration was added in Exposre measrements were made on three representative types offloroscopy nits. and exposre rates were determined at the floroscope panel. Halfvale layers for a range of tbe voltages and added filtrations were also checked. The exposre measrements were translated into average conditions before and after Based on these data, estimates were made of absorbed dose in the breast, by means of a Monte Carlo radiation transport method in relation to an anthropomorphic phantom. The shape of the breast was geometrically simlated, separately for an adlt and for an adolescent breast. A typical elementary composi tion of the gland was assmed. The absorbed dose in the breast per I R skin exposre (free in air) was calclated for selected exposre conditions: 2 beam qalities, 2 patient orientations. 2 breast sizes and 4 field sizes. These calclations were approximately in agreement with TL dosimetry in a phantom Cmlative breast doses of all floroscoped patients were calclated from all the above information (medical records. physician interviews. patient contacts, exposre measrements and absorbed dose calclations). Females nder 17 years of age were assigned doses for the adolescent breasts and the older ones doses for adlt breasts. Individal orientations cold not be determined and it was assmed that 25% of all examinations were performed in the anteroposterior position. and both breasts were assmed to be in the nshttered beam 81 % of the time while floroscoping a nilateral lng collapse, and 19% of exposre time was assigned to the ipsi-lateral breast. Average doses were compted by adding the dose received by the directly exposed breast and the scatter dose received by the opposite breast, and dividing by 2. In a few cases of bilateral lng collapse, both breasts were assmed to be in the beam for 1% of the time Uncertainties affecting dose estimates based on so many variables, sometimes estimated from memory of 2 to 45 years previosly. mst be affected by considerable errors and shold therefore be taken with great cation. It is very difficlt to say whether average breast doses calclated by Boice et al. [B24, B26] are representative of the the real sitation and what might be the systematic errors involved. However, the shape of the dose-response relationship shold perhaps not be affected by a systematic overor nder-estimate of the real dose. 37. For the exposed women, the onset of the period of risk for breast cancer development was assmed to be the date of the first floroscopic examination, and, for the non-irradiated controls. the date of admission to the sanatorim. Expected breast cancer cases were determined with the se of the age- and calendarspecific incidence rate for the neighboring state of Connectict, where a cancer registry has been in existence since The nmber of expected breast cancers in the exposed grop was calclated by mltiplying the age- and calendar-specific women years (WY) at risk by the corresponding incidence rates in Connectict. Standard morbidity ratios and absolte risk coefficients were calclated. The end of the period at risk was regarded as: the date of cancer diagnosis: or the date of death from another case: or I Jly 1975 for those fond alive; or the last date at which the woman was known to be alive, for those Jost from the follow-p The nmber of breast cancers fond among the irradiated women was 41, as against 23.3 expected (excess significant: p =.6). Among the controls. 15 cancers were fond, against 14.1 expected. The observed/expected ratios were. correspondingly and 1.6. The absolte excess risk increased with increasing nmbers of floroscopies p to 1-149: a decrease after a greater nmber was sggested. Nevertheless, a linear increase cold not be exclded [x 2 (trend) = 4.1. P =.4; x2 (departre from linearity)= 6.9. P =.14] When the incidence rate was calclated as a fnction of the average dose to the breast, the reslts were as shown in Figre XXVI. The average dose to the breast per examination was estimated to be.15 Gy, and the mean cmlative dose to the breast amonted to 1.5 Gy [B24]. No decrease in the excess incidence per nit dose was apparent at doses above 4 Gy (the highest average was 5.74 Gy) and the doseresponse was consistent with linearity down to the lowest dose grop (.32 Gy average) [x 2 (trend) = 9.4. P =.22: x 2 (departre from linearity)= 1.6, P =.8]. Some differences between the relationships for dose and for nmber of floroscopies reslted from floroscopic examinations in different positions of the patients. in different periods. at varios ages and for different sizes of the breast. 225

64 > :JC a: w a.. w... ~ (x 1 5) 4 w 3 z w C = 2 C LJ!:::! C a: c:: C z c::.,., BREAST DOSE (Gy) Figre XXVI. Standardized incidence rate ol breast cancer in floroscoped pnemothorax patients per 1, women-years (WY) at risk, as a!nction of estimated cmlative breast doses. Error bars represent 8% confidence limits. [B24] 373. The breast cancer risk appeared after 15 years. and the excess risk was present for at least 4 years after initial exposre. The average age at cancer diagnosis was 48.4 years (the greatest risk appeared among women aged 4-49 years at diagnosis) and the mean latent period from exposre to diagnosis was 24.4 years. The radiation risk coefficient derived from sbjects of this stdy living 1 or more years after the first exposre was 62 excess cases per 1 6 womenyear and Gy (28- I 7 are the 9% Poisson confidence limits) In 198 Land et al. [L22] reported a comparative stdy fitting the same set of models to mammary cancer incidence data from 3 main stdies: atomic bomb srvivors, Rochester mastitis and Massachsetts floroscopy series. They assmed that the level of breast cancer response to irradiation may depend on age at exposre. To avoid confonding of dose and age. the dose-specific data were age-standardized sing the overall age distribtion of each series as an internal standard. It was assmed that the shape of the dose-response crve was not dependent on age. Linear, linear-qadratic and qadratic models, with or withot a cell-killing term. were fitted to the data. The linear component was omitted from the killing term as it did not improve the statistical fit. bt increased the statistical ncertainty in other parameters. This procedre is, however, biologically implasible. Another constraint was that no parameters shold be negative The model assmed that the control incidence rate, a, was eqally affected by presmptive cell sterilization. implying that all transformed or potentially transformed cells from which cancers wold later develop were already present at the time of exposre. This nproved assmption reqires that latent periods for spontaneos and radiation-indced breast cancers shold be the same, or longer for the former. and is in fact spported by the apparent lack of dependence of the latent period on dose, as shown by Land et al. [L32]. I 376. The reslts of the above fits are presented in Tables 2 and 3. according to the notation in eqations (3.6) and (3. 7). For the Massachsetts series, parametric constraints redced the linear+ killing model to linear, and the linear-qadratic + killing model to linear-qadratic. For the Rochester series, the linearqadratic model was redced to linear. For the mastitis series, acconting for cell killing improved the fit significantly (Table 3). Better reslts were obtained in this respect when single-breast data were sed from the mastitis series. where the linear+ killing model gave a sbstantially improved fit (P =.27) over the linear model (P =.12). This sggests that cell killing at 4-14 Gy may be a real effect in nfractionated or relatively nfractionated exposres. The qadratic and qadratic+ killing models did not fit the agestandardized data as well as did the models with the linear term a 1 D. The reslts show that low-dose risk estimates for breast cancer indction shold be based on models with a prononced linear term. The vales for the a I term are very close in the series, particlarly when differences in age distribtion are considered When age-standardized percentage incidences were plotted in exposed and nexposed women of the three cohorts against time after irradiation, in no grop was there evidence of an acceleration of breast cancer expression with dose. This reslt emphasizes two points: that expression of radiation-indced breast cancer may be inflenced by the same physiological factors that affect the normal age-specific incidence rate of this tmor; and that an inverse relationship between dose and latency is highly nlikely for this neoplasm (see also [K39 and L32]). C. TUMOURS OF THE THYROID 1. Direct stdies 378. In their most recent pblished report. Shore et al. [S6) pdated the follow-p to 1977 and reinvestigated several aspects of thyroid cancer indction by x rays in a series of sbjects irradiated.in childhood for alleged thyms enlargement [HI 8. HI 9, S38]. The sbjects stdied inclded 2856 irradiated and 553 controls; 2652 and 4823, respectively, were followed for five or more years after the treatment (on the average. for 3 years) Irradiation was performed with x-ray machines operating at varios parameters of voltage and filtrations that were probably not very critical for assessment of the dose to the thyroid (in all cases, the skin dose was very close to the absorbed dose in the gland). The important factor was whether the gland was inside or otside the primary beam. Individal dose estimates were made on this basis, with other corrections. For 241 sbjects, sch estimate proved impossible, and they were exclded from stdies involving analyses with dose. 38. Siblings of the irradiated sbjects were the primary control, bt comparisons were made also with cancer rates in pstate New York. The sbjects were located by varios means. and a qestionnaire 226

65 was sed to identify health problems, inclding tmors. Responses were obtained from 88% of the sbjects. Only medically verified neoplasms were sed in frther considerations, and pathology reports were obtained for all cancers and abot three-forths of the nodles (which were benign). From the total personyears, five years were sbtracted in each sbject as the likely shortest latent period There were only small differences in some parameters characterizing the two grops. A large excess of thyroid cancer (P<.1) was diagnosed in the irradiated grops (3 cases of varios histological forms). and only 1 case in the controls ( 1.43 expected). The relative risk for the carcinomas. adjsted for sex, ethnicity and post-irradiation interval, was 49.1 (9% confidence limits ) (p <.1) when related to controls, or 44.6 ( ) when related to the rate in the general poplation. There was also an excess of adenomas: 59 among irradiated against 8 among the controls (adjsted RR= 15 (9% confidence limits ) (p <.1). A separate comparison of cases with matched controls yielded similar odd ratios and p vales for cancer as given above. bt a somewhat lower estimate of relative risk for adenomas Doses in the thyroid ranged from.5 to 11 Gy. with a weighted average of 1.38 Gy and 62% of individal doses below.5 Gy. An absolte risk analysis of the dose-response relation (Figre XXVII) 25 >- a.. 2 :::.J a...j z ~ 15 - z -:::.J ~ 1 ex: - :::?i: I- (x 1-5 ) so (368) I THYROID DOSE (Gy) Figre XXVII. Thyroid cancer Incidence per person year (PY) as a fnction of the radiation dose In the thyroid. Rates adjsted for sex, ethnicity and Interval after Irradiation. Error bars represent 9% confidence limits. [S6] was ndertaken with 6 and 12 dose categories, controlling for sex, ethnicity, and interval since irradiation. A linear regression (Mantel regression coefficient) yielded a coefficient of 3.46 (±.82 S.E.) 1-4 Gy- I a- 1 (P <. I) (5.25 and 2.5 for females and males. separately). Fitting the data to a linearqadratic model, by iteratively weighted least sqare regression analysis, reslted in a highly significant linear term (P <.1) and a non-significant dosesqared term (P >.4). When the so-called ridge regression analysis was applied, the linear term retained its predictive power, bt the qadratic term was even farther removed from significance. The Cox regression analysis sed for calclation of RR-model coefficients yielded for cancer a strong linear component (P <. I) and an insignificant qadratic one (P =.8). The vale of the relative increment coefficient was.58% of the control freqency per Gy For adenomas. the risk expressed in absolte terms yielded a linear dose-response with a coefficient of 5.7 (± 1.2 S.E.) 1-4 Gy-1 a-1 (7.7 for males and 3.6 for females) significant at P <. I. In a linearqadratic fit. the linear coefficient was 3.8 (± 1.4) 1-4 Gy-l a- 1 ; that for the qadratic term was negative, -.4 (±.4) 1-4 Gy- 2 a- 1. and the relationship was downward concave. A strong linear component (P<.1) and a significant qadratic component (P <. l) with a negative coefficient. were obtained for the risk expressed in relative terms. A simple linear regression yielded a relative increment coefficient of.44% of the control freqency per Gy Of 2358 sbjects with doses below 6 Gy. 5 I% were irradiated once, 39% twice and in 1% of cases the dose was split into 3 to 11 fractions. No inflence of the dose per fraction, nmber of fractions and interval between fractions was fond for carcinomas or adenomas. Tmor appearance as a fnction of time seemed to decrease slightly beyond 25 years postirradiation. bt still after 4 years there were excess tmors. The data were not a good fit to the relative risk projection model. bt an adeqate fit was obtained to the absolte risk model. The latent periods of thyroid carcinomas and adenomas did not show any relationship with the dose In conclsion, these data. derived from a cohort irradiated in infancy. show a dose-response relationship with a strong contribtion of the linear terms, lack of demonstrable dose-sqared terms. and lack of dose-fractionation effects. A paradox was also noted. in that irradiation seemed to mltiply the sex-specific "spontaneos risk. while, at the same time, the reslts fitted an absolte, bt not a relative, risk projection model. 2. Indirect evidence 386. Among the stdies on thyroid cancer reviewed in annex G of the I 977 UNSCEAR report [U6] none seems to qalify for the derivation of the doseresponse relationship or for testing of varios models. However, when the risk coefficients for irradiated 227

66 children are considered (see Table 1 in annex G of the same report) they seem to fall within a rather close range, althogh the average estimates of dose for each grop vary from.65 to abot JO Gy. This wold sggest a dose-response relationship close to linearity, a sggestion that mst be taken with cation owing to the great ncertainties in retrospective dose estimates in all stdies, and lo the genetic, demographic and geographic differences between the series Modan [M25], and later Ron and Modan [R29, R45], followed p children irradiated for tinea capitis in their childhood. The last paper in the series [R45] smmarizes the reslts for the follow-p period All individals were irradiated between 1948 and 196 at an average age of 7. The majority (96%) were treated between I and 15 years. and over 5% between 3 and 8 years. Only 6% were irradiated more than once The final cohort stdy inclded 1,842 irradiated persons (IS) and non-irradiated tinea-free sbjects matched for age, sex, contry of origin and year of immigration to Israel (PC). A second control grop was composed of 54 siblings (SC), nonirradiated and tinea-free. All grops were eqally divided by sex. By the end of the three grops had accmlated , and 121,854 person-years at risk, respectively. The average follow-p interval was 22.8 years. The dose to the thyroid was estimated to range from.43 to.16 Gy. with a mean of.96 Gy The follov.ing nmbers of cancers were fond in the three grops: IS, 29: PC, 6: SC. 2. They were ascertained from the Israel Cancer Registry with an efficiency of recovery of abot 8%. The relative risk of thyroid cancer, estimated by sing the Mantel Haenszel method for cmlative data. amonted for the whole grop to 5.4 (95% confidence interval ). The absolte risk may be estimated betweeen 12 and Gy- 1 a- 1, depending pon the efficiency of detection assmed ( 1 or 8%. respectively). There was a clear inflence of sex pon the incidence. the relative risk and absolte risk ratios females:males being 1.2 and 3.8, respectively. Sbjects in the age bracket -5 years were twice as sensitive as those in the 9-15 years grop. A significant ethnic difference was also seen becase the children of Moroccan and Tnisian origin (contribting abot one-half of the total) had an absolte risk twice as high, and a relative risk 6-9 times higher than the rest of the cohort. 39. The FTA from this stdy is abot 3-4 times higher than that derived by Shore et al. [S6] for a cohort with an average dose of 1.38 Gy. This however, does not indicate that the risk is trly higher at the lower dose of abot.1 Gy, becase the tinea capitis sample is composed entirely of people of Jewish origin. From the reslts of Shore et al. [S38]. it follows that this factor alone cold accont for a twofold difference. and the contribtion of individals of Moroccan and Tnisian origin may have added another factor of 2. Any systematic bias in dose estimates de to inadvertent movements of children dring irradiation wold have reslted in higher, and not lower, doses to the thyroid In smmary, therefore, a similar vale of FTA for thyroid cancers in a cohort of children who received a thyroid dose of abot.1 Gy, and in another cohort with doses ranging from.17 to 7.5 Gy. strongly spports the conclsion that the dose-response relationship for thyroid cancer may be essentially linear. Concrrent irradiation of the hypophysis with higher doses (of the order of.5 Gy) has been sggested as a possible explanation for this niqe observation. However, only an increased secretion of thyroid stimlating hormone (TSH) wold be expected to enhance the risk of thyroid carcinoma, while irradiation of the hypophysis wold be expected to reslt in a decreased secretion. Data from an experiment aiming at answering this qestion are also in conflict with the explanation referred to above [L29]. A smaller grop of2213 children [S36] irradiated to indce epilation did not prodce an excess of thyroid cancer. However. becase of the low statistical power of the stdy, this reslt is consistent with reported vales of FTA 392. Holm [H 17] pblished a retrospective stdy on patients given for diagnosis of thyroid diseases in the Stockholm area in The poplation consisted of persons, of average age 44 years, followed for an average period of 18 years after administration of the nclide. Of these, 95% received when 2 years or older (286 males and 847 females). The diagnostic tests were administered for sspected thyroid tmor (32%): hyperthyroidism (44%); hypothyroidism (16%); or other reasons (8%). Ths. from the viewpoint of thyroid physiology and/or pathology. the poplation stdied was a selected one. Persons with cancers diagnosed within 5 years of administration of l3ij were exclded from the stdy as were those given any therapetic irradiation of the thorax, head and neck The mean activity administered was 2.2 MBq, and doses to the thyroid were calclated on the basis of a measred retention of abot 4% at 24 hors after administration. The assmed mean effective half-life of 131 I was 7 days and the mass of the thyroid was estimated in a representative sample (abot 1%) of the poplation nder stdy from scintigraphic measrements and palpation. In persons above 2 years, the mean weight of the thyroid was 5 ± 33 (S.D.) g, and in those below 2. JO± 5 g. The reslting mean doses were.58 and 1.59 Gy. The poplation average (adjsted for age) was.62 Gy The expected nmber of thyroid carcinomas was compted from contry-average age. sex and calendar-year specific rates derived from the Swedish Cancer Registry statistics. The assmed period at risk was the period of follow-p since administration of 131 I mins 5 or mins I O years. In the first case, the nmber of expected cancers was 8.3. in the second 5.9. The Swedish Cancer Registry was searched for names and identity code nmbers of all the patients enrolled in the stdy. Nine cancer diagnoses among them were identified and 8 confirmed. Ths, there was no excess cancer among the sbjects given diagnostic amonts of 131 I. 228

67 395. Holm et al. [HI 7) compared the nmber of cancers diagnosed with estimated nmbers calclated from UNSCEA R risk coefficients of 5-15, cases 1-6 Gy-l [U6) (derived predominantly from srveys on children) or from Hiroshima and Nagasaki srveys [U6] of Gy- 1 In the first case, the nmber of expected cancers wold be which is obviosly very different from the observed 9 cases. For the second alternative, the athors qote a figre of 19 cases. which they claim to be not significantly different from the 9 cancers diagnosed However, when the lack of the excess incidence in that stdy was compared [N JO] with the expectation in adlts, taking the risk for adlts to be one-half that in children [S6). to accont for the age-related lower sensitivity. the difference appeared statistically significant on two tests applied. This wold sggest that the effectiveness of 131 I for indction of thyroid carcinomas is lower, at least by a factor of 2. The qestion whether this is a tre dose-rate effect, and to what extent other factors play a role, cannot be answered at present Several factors, separately or in combination, cold be responsible for the difference between the predicted and the observed nmbers of cancers. First, the linearity of the dose-response relationship in Figre XXVII may be more apparent than real, becase the positive dose-sqared term for initiation and the negative term for cell killing may compensate each other. Secondly, the sbjects are a selected nhealthy poplation. with a high percentage of thyroid involvement, to whom specific rates of thyroid cancer indction. valid in the general poplation, may not apply. Thirdly, the treatment given after the diagnosis (srgery. thyroid hormones) cold have sppressed the risk. This appears rather nlikely, as more than half the patients were not given any treatment and the nine with diagnosed cancers were treated \\ith hormones (2 cases) and srgery (4 cases). which did not prevent the manifestation of tmors. Forthlv. a non-niformitv of dose distribtion from in p~thologically enlarged glands cold be another confonding factor. Fifth. it cannot be exclded that risk coefficients are being compared for poplations which may differ in their average iodine ptake and hence in activity of the thyroid and hypophysis (TSH stimlation). Finally. other dietary or environmentrelated factors cold frther confond the comparison In smmary, two stdies on hmans [R45, S6] arge persasively that the contribtion of the linear term is very strong in the dose-response relationship for indction of carcinomas of the thyroid. This is also the conclsion to be drawn from the experiments on rats by Lee et al. [L29] discssed in chapter IV. If this conclsion were tre. a dose-rate effect, shold not be expected. However, the reslts of Holm's stdy [H 17] are not consistent with this pictre. Frther investigations in this field are certainly called for. D. CANCER OF THE RESPIRATORY TRACT 399. Indction of lng cancer has been discssed repeatedly in many UNSCEAR reports [U6. U7, U9- U 12, U24) and risk estimates were given based on varios grops' findings ([U6]. annex G). Most of the data on lng cancer indced by external irradiation appeared nsitable for analysis of the dose-response relation. The most nmeros grop. the atomic bomb srvivors, is still affected by ncertainties in organ dosimetry. As a reslt, there are large statistical ncertainties in excess rates in varios exposre categories ([U6], annex G ). 4. The most interesting case for a discssion of the shape of dose-response relationships is the prolonged exposre of miners to radon and its daghter prodcts. The difficlties in this case are: (a) (b) (c) (d) The long period of exposre, often at a variable rate. Exposres toward the end of the follow-p period are increasingly less effective, in view of the long latent period of these tmors: Possible changes in ssceptibility with age, which is itself a variable correlated with the length of the exposre: Possible interaction of radon and its daghters with other factors, particlarly smoking habits of miners [A8, A 16. W 17]; The dosimetry. which is based either on measred concentration of potential alpha-energy in the air (WL)d or on estimates of radon concentrations, on the assmption of some degree of radondaghter prodct eqilibrim. Translation of these qantities into an absorbed dose in the cells at risk (basal layers of bronchial epithelim) is ncertain, bt a recent ICRP review [I4] gives. for mining conditions, a range of abot 3 to 8 mgy per WLM. The se of cmlative WLM as an indirect measre proportional to dose is probably adeqate for discssions of the derived dose-response relationships. 1. The risk expressed in absolte terms 41. In Newfondland, a grop of 212 florspar miners were exposed to an accmlated mean exposre to radon of abot 55 WLM. They were followed for abot 25 years and those who accmlated more than I WLM had a relative risk of cancer of the respiratory tract of 5.58 (95% C.L ). The absolte annal excess risk coefficient amonted to 6 ( ) cases per 1 6 PY per WLM. The excess lng cancer incidence rose roghly in proportion to the exposre [M5]. 42. Several stdies were pblished on the epidemiology of cancer of the respiratory tract among ranim miners in the United States. The 1977 UNSCEAR report [U6] reviewed those by Archer et al. [A8] and by Lndin [Ll8]. To sm p the sitation, the stdy by Archer [A8] referred to 3366 white and 78 non-white miners who had worked one or more months in ndergrond ranim mining before I Janary Mortality analysis was terminated on 3 September and only 1% of people dscc footnote c. 229

68 enrolled were lost from the srvey. The death certificates of those deceased before the closing date were obtained. There was a sbstantial excess of deaths (437 observed verss expected) among which there were 7 cancers of the respiratory system against expected (p<.1 ). The diagnoses were ascertained sing all sorces of evidence (clinical findings, x rays. atopsy. histology, biopsy, cytology, sptm analysis). Nmeros cancers occrred after 1968, or were evalated as probable ones. These were not sed in the assessment of risk or for characterization of the dose-response relationship [B2]. 43. The exposre was assessed by radon daghter measrements (nearly 43. from 1951 throgh 1968 in approximately 25 mines). Measrements taken in a mine over yearly periods were averaged, bt, in many instances. exposre over a nmber of years cold only be estimated approximately. particlarly in the early years of highest exposre. In addition, a fraction of miners had a history of hard-rock mining, before the work in ranim mines, which also reqired approximate assessment. Cmlative radon daghter exposre vales were calclated with the above approximations for each miner and for each month of his employment. 44. Person-months of risk of dying by single calendar year. for 5-year age grops, and for each time interval after the start of ndergrond ranim mining, were calclated by a modified life table techniqe. These vales were related to cmlative radon daghter exposre in WLM. Individals who were lost to follow-p were considered to be alive throghot these analyses. A miner often contribted person-months to more than one exposre category if dring the corse of employment he passed from one category to the next. The expected nmber of deaths was calclated by applying the specific rates for age, race, calendar year and case of death for white United States males to the appropriate nmber of person-years at risk. 45. The nmber of expected and observed cancers of the respiratory tract as a fnction of cmlative exposre in WLM dring the whole period of observation was converted to incidence rate by BEIR [82], weighing the incidence in each grop of white miners by its inverse variance. The fit was as shown in Figre XXVIII. Consistency with linearity is evident and the slope is abot 3 I-6 person-year WLM- 1 for the whole range of exposre. 46. There were confonding variables, like cigarette smoking, metropolitan verss non-metropolitan residence, possible mis-classification of miners into exposre categories. and dependence of WLM calclations on qality of the work histories, which cold have distorted the data to some extent. It is difficlt to see what their combined effect cold have been. However. the athors sggest that the inflence of these variables shold not basically alter the trend of the dose-response relationship. Also, the excess of cancer cases even in the WLM category is beyond dobt. >-.. : w.. 15 w I- ~ 4 BOO w z w - "' w - z 1 X w White miners 5 Non-white miners Figre XXVIII. Dose-response relationships for lng cancer In ranim miners in the United States. Error bars Inclde 9/o range based on Poisson statistics. Modified from [82]. 47. The mortality stdy of the ranim miners from the United States was extended to September 1974 [A 16] with essentially the same methodology as in a previos stdy [A8]. Lng cancer rates were calclated for eight exposre categories and for the time after starting ranim mining. The total nmber of person-years at risk was : the average period of work in ranim mines was 9 years: the mean accmlated exposre was 82 WLM; and the average follow-p period was 19 years. The incidence rate increased approximately linearly p to 48 WLM, then the slope of the response decreased to give a lower risk per WLM. a featre possibly related to cell killing, which was also seen in the series from Czechoslovakia [S3]. Miners exposed to higher rates were sally followed for longer times. and most miners were middle-aged when they entered the occpation. Cigarette smoking. and constittional factors sch as statre and eye pigmentation were fond to be correlated with the effect. Statre was thoght to be correlated to lng volme in the sense that. for the same work at comparable oxygen consmption, a smaller lng volme means enhanced ventilation rate and higher lng doses. More recently, ranim miners from the United States have been followed throgh 1977 [W 13). bt the data. as reported, do not provide new information, as all ranim miners examined by the United States Pblic Health Service medical srvey team were already considered in the stdy by Archer et al. [A 16]. 23

69 48. The data from the United States miners were tested by Myers and Stewart (M43] for their fit to varios dose-response fnctions. The fit was made to ~m eqation of the form ita(wlm) =a+ bei where ITA is the annal net incidence rate (1-6 a- ) and E the cmlative exposre in WLM. The j exponent was fixed at 1.,.5 and.25, and maximm likelihood estimates were obtained in each case for a and b. Allowing all parameters to prodce a maximm likelihood estimate led to a preferred vale of j not significantly different from nity and to a positive vale of a, indicating that an excess cancer of the Jng may obtain even in the absence of radon exposre. A linear fit to the data was fond to be the best. with b = WLM-1 a The latent period from the start of the exposre to diagnosis was examined as a fnction of the age at the start [A 16]. It declined from abot 45 years to abot 1 years in those who started mining at 2 or 6 years. respectively. Smokers sally had a shorter indction time (see annex L of the 1982 UNSCEAR report [U24] for a discssion of the inflence of tobacco smoke). There was also a shorter latent period associated with higher exposre rates, which appeared to contribte to shortening more than cmlative exposre. 41. Sevc, Knz and Placek [S3] reported on lng cancer incidence among ranim miners in Czechoslovakia. The nmber of miners was not specified. bt the grop was similar in the nmber of people to that stdied by Lndin et al. [LI 8, Ll9] in the United States. which consisted of 3366 workers. According to [15), the stdy was based on abot 6. person-years at risk. and a follow-p in the range of years, JO years of average working period in the mines and a mean cmlated exposre of 31 WLM. Fifty-six per cent of these miners started ranim mining between I 948 and 1952 inclsive (grop A). the rest between 1953 and 1957 inclsive (grop B). Miners were also sbdivided into 3 age grops at the start of exposre:< 29 years (45%), 3-39 years (28%), and> 4 years (27%). The age strctre in grops A and 8 was very similar. Each age grop was sbdivided into eight exposre sbgrops. the exposres being expressed as WLM. The nmbers in each sb-grop were approximately eqal. Exposre estimates were based on abot 12, radon concentration measrements made since Estimates of potential alpha energy concentration were made from radon measrements by choosing average vales of eqilibrim between radon and its decay prodcts, typical for given ventilation conditions and practices (which were available from records) and radon daghter measrements which were introdced in 196. Attempts were made to estimate the variability of dose measrements. The rates of exposre accmlation in the three age grops were comparable In each sb-grop of miners (belonging to each exposre category and age grop), the expected rate of Jng cancer was calclated by year from the beginning of exposre p to 31 December 1973, on the basis of age- and calendar-specific Jng cancer mortality rates for Czechoslovakia. A statistically significant excess (at 5% level) started to appear in the WLM category.r The presmption of linearity for the doseresponse relationship in grop A (start of the exposre follow-p years) cold not be rejected at the 5% level of significance The same data [S3] were fitted by Myers and Stewart [M43] to qasi-threshold fnctions of probit and logistic response on the logarithm of cmlative WLM. When reslts were tested by means of z 2 distribtion for significance of deviations from these assmed fnctions, the z 2 vales were tenable at P = However, the qasi-threshold was less than 3 WLM. Ths, the mathematical analysis, sing maximm likelihood as the criterion. does not provide any reason to reject linearity, for which the x 2 test gives a P vale of Cmlative exposre E (WLM) and lng cancer incidence (!TA) in grop A were also fitted by Jacobi [J4] to a fnction of the form ITA = a E e-"e. A better fit was obtained than that reslting from assmption of linearity throghot the whole range of exposres. The most likely estimates of a and k were 32 excess cases 1-6 WLM-1 and WLM-1. respectively The excess cmlative cancer incidence in the varios age grops shows that for the older age grop the incidence was significantly elevated (5% level) at 1 WLM or more. For the two yonger grops, this significance was reached at exposres above 15 WLM. The risk per nit exposre (WLM) increased with age, with some irreglarity of response in miners who were older than 4 at the beginning of the exposre. In the yonger age grops, linearity. with different slopes, cold not be rejected at the 5% level of significance. The differences in incidence between the grops disappeared above 3 WLM Cigarette smoking did not appear to be a confonding variable becase it was not dependent on exposre categories, althogh it was correlated with age. From random sampling, the freqency of smoking was similar to that in the contry for which specific mortality rates were sed for comptation of the expected nmbers of cancers. Also, hard-rock mining was not a complicating factor in assessment of the exposre In another paper, Knz et al. [K22] stdied the excess lng cancer incidence in miners in Czechoslovakia over the period and related their findings to different time distribtions of exposre. The previosly mentioned grop A composed of miners who started work between 1948 and 1952 was followed for 26 years on average. The cohort was divided into five exposre categories and also, according to dration of exposre, as short (4.-7.9: mean 5.6 years), intermediate ( ; mean 9.5) and Jong (> 12: mean 14). The same grop was also divided alternatively, according to exposre rate. into grop II with nearly constant accmlation rate; grop I for which the earlier rates of exposre were higher than tmore recently, it has been reported (SSJ] that the excess is already significant at an average exposre of 72 ± 2 WLM. 231

70 later ones; and grop Ill where the rate of accmlation increased with time and reached highest vales in recent years The additional lng cancer incidence for the grop as a whole, when related to dration of exposre and its cmlated vales, is shown in Figre XXIX. For the grop as a whole, a linear exposre- "'... z :c 15 ~... "' c.. "'... 1 z <: "' z ::, _, _, <: z 5 8!:: C <: EXPOSURE (WU4) Figre XXIX. Relationship between additional lng cancer freqency and cmlated exposre In all ranim miners from Czechoslovakia. Error bars represent 95% confidence limits. [K22] response relationship cannot be rejected, bt there is some sggestion of satration at the highest vales of WLM. When divided into three grops in relation to the dration of exposre. the two grops with shorter exposre showed satration very clearly. The early exposre rates in both grops with shorter dration of exposre at which a significant deviation from linearity occrs corresponds to abot 5 WLM a- 1 The lower incidence reslted from a diminished excess of smallcell ndifferentiated cancer. The grops with niform and decreasing rate of accmlation showed linearity of the response, whereas that with the increasing rate showed a breakdown of the response in the highest exposre category (> 4 WLM). This decrease of the incidence was also de to diminished occrrence of small-cell ndifferentiated cancer. A similar inflence of exposre rate to radon daghters pon lng cancer incidence was observed in experiments on rats by Chmelevslcy et al. ( C34] (already discssed in chapter IV) Since all Czechoslovak miners started exposre in those accmlating their exposre over a shorter period did not have a shorter follow-p, and therefore the lower incidence in the two grops of 9.5 and 5.6 years of mean exposre dration cannot reslt from the shorter time available for manifestation of the effect (trncation de to long latency and short followp). A more likely explanation wold be a killing effect on the transformed cells, which becomes manifest at an exposre rate of abot 5 WLM per year. However, a shorter srvival of miners at high exposres was also sggested [J7]. Ths. lng cancer indction by alphaparticles of radon daghter prodcts can be described by eqation (3.1). with the proviso that the proportionality constant of the cell-killing term is dose-rate dependent. The slopes of the exposre-response relationships for different times of accmlation in the low-dose region were very similar. The effect of exposre rate cannot be easily separated from the effects of time corse in grops that differ in respect of the rate of accmlation of WLM. Both the exposre rate and latency effects may play a role in the decreased incidence for the 4 WLM category of grop Ill Horacek and Sevc [H25] stdied the relationships between incidence (per 1 OJ. ITA.) of varios histological forms of respiratory cancer in ranim miners from Czechoslovakia and the exposre (E, in WLM). For all miners, a fnction of the type ITA =. 16 E gave a good fit to the data. and absence of a threshold cold not be exclded. For epidermoid-type lng cancer, the relationship was different. and a regression line of the form ITA = ae + c =. 15E-11.2 cold be fitted p to 7 WLM (c was significantly different from zero). Ths, the analysis sggests the presence of a threshold of abot 75 WLM. There was also a small excess-nrelated to exposre-of other histological types of lng cancer the statistical significance of which cannot be assessed from the report [H25]. 42. It is interesting to note that basically the same pictre was obtained for the miners whose exposre began at ages less than 4 years. For those above 4 years. the relationship for epidermoid-type lng cancer was again linear with threshold (IrA =.13 E -21.7), sggesting a threshold of a similar magnitde. bt a steeper linear rise of incidence with exposre p to 7 WLM. For small ndifferentiated cancer, the excess at abot 6 WLM was already close to cases and rose mch less steeply to cases at abot 3 WLM. with a decline to abot cases at abot 5 WLM. Other histological types also showed an increased incidence to abot cases at 3 WLM. with a decline to very low levels at 5 WLM: however. the form of the ascending branch of the crve cannot be derived from the report The relationships of these findings to smoking cold not be stdied, de to Jack of relevant information. The reasons why the two main histological types of plmonary cancer shold have different forms of the dose-response relationship (presence or lack of threshold) and different age dependence, are nknown. However, there is a casal link between development of epidermoid lng cancer and metaplastic changes in the bronchial epithelim (S58, S61]. The intensity with which noxios factors (e.g.. tobacco smoke. radon exposre) act pon exposed individals is probably related to the speed with which metaplasia develops. Therefore. at lower exposre rates, development of metaplasia and epidermoid cancer may simply be delayed. The percentage of the small ndifferentiated and epidermoid cancers varied strongly with nmeros factors, sch as calendar time. age at diagnosis, and iength of exposre. The epidermoid cancers sally appeared later, and at higher exposres, than the oat-cell ndifferentiated carcinomas. All 232

71 these circmstances call for great cation in accepting a threshold-type relation for epidermoid lng cancer in ranim miners, particl~rly becase the analysis of a similar type of cancers indced in rats exposed to radon decay prodcts [C34) (see IV.8.6) did not sggest any threshold above a few WLM. A detailed stdy of the latent periods for the two types of cancer cold give some insight into the problem By far the largest stdy on the mortality of miners, inclding that from lng cancer, was recently reported by Millier et al. [M7] from Ontario, Canada. Ot of several grops of miners engaged in mining varios minerals. excess oflng cancer mortality was fond in three: ranim miners, gold miners, and those mining a mixed assortment of minerals. The dose-response stdy described below refers only to ranim miners. for whom radon daghters in the air cold be envisaged as the sole casative factor. The grop inclded all fll- and part-time ndergrond ranim miners with mining experience of one month or more. They were followed from l Janary 1955 throgh 31 December The cohort nmbered 15,984 men, with an average follow-p of 15 years, average dration of exposre of 2 years. mean accmlated exposre of 6 WLM. and a total of person-years at risk. The exposres to 22 2 Rn daghters were calclated in WLM from area monitoring and occpancy data from 1955 throgh 1967 and, starting in 1968, from personal exposre records. For each miner. pper and lower best estimates were derived. called special'' and "standard WLM", respectively. It was postlated that the most likely tre exposre wold fall within these bondaries Deaths among cohort members de to respiratory cancers were identified via the Canadian Mortality Data Base, tilizing death certificates. The expected nmber of deaths was calclated sing age and calendar mortality rates for the male poplation of Ontario. Smoking habits were not taken into accont, bt. according to the athors, there were good reasons to sppose that the miners did not differ in this respect from the reference poplation. Among the ranim miners there was a highly significant excess of cancers of the respiratory tract ( 121 observed against 7.5 expected; P < 1-9) When the absolte attribtable risk was calclated in 6 exposre grops according to a linear model, the intercept did not differ significantly from zero, irrespective of whether a zero. 5- or JO-year lag in exposre against effect was assmed. The slopes of!ta as a fnction of WLM varied between WLM- 1 for "special" WLMs and WLM-l for the "standard'' WLMs (at 3 different lag intervals). The slopes were significant at P <.1. An example of a dose-response relationship is presented in Figre XXX. In terms of the absolte attribtable risk, the slope was significantly greater. by a factor of abot 6. for those who lived to age 55 or more than for those who died at yonger ages. Ths. all dose-response relationships were well described by a non-threshold linear fnction. A single relative risk model described the relationship for all attained ages (see paragraph 428).,.: VI a:.....j co (x 1-4 ) 16 < 12 I- :::, co a: a I- I- <... I- :::, 4..J VI co < EXPOSURE (WLM) Figre XXX. Dose-exposre relationships for lng cancer incidence In ranim miners from Ontario, Canada, who had no prior exposre in gold mines. Exposre is expressed in WLM "standard" nits. [M7] 425. Preliminary data were reported from Sweden on lng cancer indction in iron-ore miners exposed to relatively low levels of radon daghters. A typical rate of exposre is 2 to 6 WLM per year of ndergrond work, presmably rather constant over the last 6 years [R35, R36]. The grop consists of 1435 miners born between 188 and 1919, known to be alive on I Janary 192. The mortality analysis dealt with the cohort alive on I Janary 1951 (I 294 men) and covered the period throgh 31 December The analysis [R36] is almost a life-long followp. Details and accracy of exposre evalation are not known. The estimates of exposre ranged from 2 to 3 WLM, the average being abot 9 WLM. Detailed histories of smoking were available for each miner. and the expected nmbers of lng cancers were calclated from the calendar- and age-specific cancer rates for smokers and non-smokers in Sweden. The interesting featre of this stdy is the very long latent period, with a minimm of 2-29 years after the start of mining and a mean of abot 4 years From this series. only limited data have been pblished [R41] concerning the dose-response relationship. The earlier report [R36] states that "... the preliminary data show a reasonable dose-response relationship over the range of cmlative exposre from 2 to 3 Working Level Months (WLM). The significant point is that in the lowest dose range we do get a significant excess of lng cancer at abot 25 WLM. So we have some indication that at these low doses and dose rates there is a significant lng cancer risk as the linear dose-response wold have predicted. ' 2. The risk expressed in relath e terms 428. The lng cancer risk of Canadian ranim miners was also expressed in terms of a relative increment verss exposre. The intercept of the 233

72 fnction was not significantly different from zero for all assmed exposre- risk intervals (.5 and I O years). The slopes for relative risk verss exposre varied from.51 to.54. and the coefficient for increment of spontaneos risk per WLM assmed a single vale of I.3%. The slope was significant at P <.1. Ths, a non-threshold linear model. with a constant slope for all attained ages, was adeqately described in the data Lng cancer mortality of the ranim miners in the United States was re-appraised [W 17) sing Cox's proportional hazard model for mortality rates. The analysis was based on 194 lng cancer deaths recorded throgh 31 December 1977 in the cohort of 3362 white and 78 non-white miners in the Colorado platea, who worked at least 2 months ndergrond. The data on exposre were the same as sed by Waxweiler [W 13). Smoking histories were sed to estimate the total nmber of cigarettes smoked ntil the exposre ct-off (l 968). 43. When considered in terms of the relative risk (RR). varios specific models were tested which treated the exposre to 222 Rn decay prodct and smoking as additive or synergistic factors. The best fit to the data by the maximm likelihood method was obtained by a synergistic model of the form: RR= (I+ a WLM) {I+ b packs) (5.1) where WLM and packs are accmlated exposre to 222Rn daghters and estimated total nmber of cigarette packages smoked p to 1 years before the miners' crrent age, respectively. The fit yielded a= WLM- 1 and b = pack- 1 and was significantly improved over that obtained when exposre to 2 22 Rn decay prodcts was taken alone. Several additive models cold be rejected. For exposre to 222Rn decay prodcts alone the dose-response model is a non-threshold linear one An ICRP Task Grop [15) re-calclated the lng cancer risk among United States, Czechoslovak and Canadian miners in terms of the relative risk. The ratios of the coefficient for the linear slope of the exposre-response relationships. i.e.. the relative increment of spontaneos lng cancer risk per nit of exposre are as follows: United States..5-1.; Czechoslovakia, and Canada per cent per WLM. These vales are mch closer to each other than the respective vales of the coefficient FTA (in terms of the absolte attribtable risk) of and WLM per year. The relative risk per WLM also becomes slightly less at high accmlated exposres Data on indction of lng cancer are practically limited to prolonged exposre to alpha particles emitted by short-lived radionclides attached to dst particles or as free ions in air passage ways (trachea. bronchi). Moreover, the available data refer to miners among whom a majority are tobacco smokers. The latter factor may act synergistically with radiation in indcing bronchial neoplasms ([U24]. annex L). There were also other factors in the working environment of miners whose co-carcinogenic or promoting effects pon lng cancer indction cold not be completely discarded. Attempts at dissociating these from radon decay prodcts and tobacco effects have been so far nsccessfl. Whatever the limitations of the five stdies discssed, they imply basic agreement with expectations of the linear non-threshold model. irrespective of whether the risk is expressed in absolte or relative terms. E. BONE SARCOMA 433. The available information on indction of bone sarcoma by bone-seeking radionclides has already been reviewed by UNSCEAR ([U6]. annex G). which also discssed many experiments performed on animals ([U6), annex I). Relevant new information on animals is discssed in chapter IV of the present annex. 1. Direct stdies 434. A grop of dial painters who acqired varios amonts of 22 6 Ra and 228Ra has now been followed for more than five decades [E3. E4. Rl8-R2. R34) and has provided knowledge of average doses to bone and to endosteal tisse, in which the cells at risk are thoght to be located. Average doses from radim isotopes in man can be calclated with reasonable accracy for grops of sbjects who acqired their body brdens 5 or 6 years ago [R 18-R2]; however, the precision of individal estimates is difficlt to appraise The 22 6 Ra and Ra body brdens were measred in most of the individals (in some of them on nmeros occasions) and the initial skeletal ptake was estimated sing Norris' [N7] empirical retention fnction for radim in man. The average skeletal doses of alpha radiation from both isotopes of radim and their daghter prodcts were compted by integrating the retention fnction from the mid-point of exposre to the end of follow-p. and dividing the released energy by the average skeletal mass A large grop of these sbjects (dial painters) has been followed for more than 45 years and the estimated cmlative incidence shold be very close to the complete life-long effect of the radionclides. It is still ncertain whether the grop might be selected for dose, i.e., whether or not the sbjects with symptoms are groped preferentially in higher dose categories where skeletal effects (inclding bone-sarcoma) occr more freqently [M29) Rowland et al. [R34, R46] analysed the doseresponse relationships for bone sarcoma in this poplation. Female dial painters were selected from the total grop and were followed ntil 31 December Among 1468 individals with a measred body content of radim, 42 bone sarcomas were fond. whereas.4 were expected on age- and time-specific rates for white females in the United States. However, among all recognized sarcoma cases, the body brdens were not measred in abot one-third of cases [R46]. Excess of sins and mastoid carcinomas was also reported and they were thoght to arise from irradiation of the lining epithelim by 222Rn and decay prodcts accmlating in the cavities. Their nmber 234

73 was too small and dosimetry too ncertain for any reasonable discssion of dose-response relationships The nmbers of cases per person-year at risk were calclated as a fnction of the systemic intake of radim (activity of 226 Ra activity of 228Ra; it had been shown [R 18) that for indction of osteosarcomas in man 228 Ra is 2.5 times more effective per nit activity than 226 Ra). All cases with systemic intakes greater than abot I O kbq were sed for the fit. The breakdown of cases into 14 intake grops was made in sch a way that 3 strata covered each order of magnitde (Figre XXXI) from abot 1 kbq p to abot I GBq. Mean intakes were calclated either as simple weighted means or as roots of mean weighted sqares. The former,vas sed when linear. and the latter when qadratic, fnctions were fitted to the independent variable. Vales of and I were taken as expected (control) incidence rates for the two modes of treatment mentioned above Varios forms of general initial activity (A) incidence-rate expression ita = (a + a 1 A + a 2 A 2 )e-/l,a were fitted to the data and tested by 1. 2 statistics for goodness of fit. Fitting of 13 data points was carried ot by a general weighted least-sqares procedre. A qadratic initiation+ killing expression gives the best fit (P =.73). Arbitrary selection of a positive a 1 still gave an acceptable fit (P =.5) of a 1 = s. The parameters of the eqation ita = (a + a 2 A 2 ) e-/l,a are: a 2 = (7.±.6) 1-s and /J 1 = (I.I ±.1) 1-3, where A is the initial systemic intake of activity. 44. The pre qadratic and linear-qadratic fnctions give a similar response at higher intakes, bt differ markedly in the region of abot 4 MBq intake. To change the response from qadratic to linear while retaining a good fit, three additional sarcomas wold have to appear in the years to come in the low-dose region where no sarcoma has yet been seen. This, the athors jdge as qite improbable (althogh the activity distribtion among nmeasred sarcoma cases is not known). Arbitrary changes of intake ranges did not alter the fact that dose-sqared pls killing fnctions always gave the best fit to the data; only in isolated categorizations of the intake cold a linear fit not be discarded When the dependence between radim intake and time of sarcoma appearance was examined, an inverse relationship was fond between intake and maximm or mean (bt not minimm) indction period. The nmbers of observed and predicted bone sarcomas as a fnction of time after intake were in good agreement, sggesting that the risk is rather constant over 65 years after intake. No bone sarcoma has been observed among the 168 measred cases (all categories) with intakes of less than abot 2 MBq. The conclsion is, therefore, that the life-span probability of bone sarcoma indction below this level of intake mst be very small. The athors calclated that after an intake of abot 4 kbq 22 6 Ra over one year. the corresponding lifetime risk of bone sarcoma was and s per year, for the qadratic and linear-qadratic dose-response fnctions, respectively..5 >- a.. er..j a...j I- c::c c:::.j z.j Cl... z...j c::c :::> z c::c INITIAL SYSTEMIC ACTIVITY (µci 226 Ra µci 2 28 Ra) Figre XXXI. A semi-logarithmic plot of bone sarcoma Incidence rate as a fnction of systemic Intake for female dial painters employed before 195, showing a dose-sqared exponential fll The shaded band Indicates the range covered by the fitted fnction when the coefficients are allowed to vary by ± 1 S.D. Error bars represent the binomial standard errors of the observed Incidences. [R34] 235

74 442. Schlenker [S5], assmed a binomial distribtion to calclate the probability of observing a nmber of bone sarcomas greater than predicted by the relationship ita = (a + a 2 A 2 ) -exp (J 1 A in the zero range, i.e.. for the activity A below abot 4 MBq where no tmors were actally seen. For the dose-response crve itself, the probability was 5%. For pairs of crves delineating the 68% and 95% confidence limits (Figre XXXII) there is a corresponding chance that the tre nmber of tmors lies between the two predictions. The crves enveloping the 95% confidence interval span a factor of 9 and their widest distance gives a feeling for the ncertainty in the estimated dose-response relationship at low activities of radim systemic ptake Becase of the inverse relationship between activity and latent period, it might be expected that the pward concavity of the dose-response crve wold reslt from the choice of the variable (rate per year) and wold vanish when expressed as cmlative incidence. In annex H of the 1972 UNSCEAR report ([U7]. Figre IX). the reslts of the follow-p throgh 1969, expressed in this way, showed a downwards concave dose-response relationship. In 1972, Mays and Lloyd [MI I] investigated the dose-response relationship for all United States sbjects containing radim in their bodies, ths contining the combined MIT and the Argonne series pdated to May In all, 762 individals below 2 Gy average skeletal dose were inclded, 48 of whom developed bone sarcomas. The dose-incidence relationship had a linear trend with a slope of.46% per Gy. However, the probability of seeing the deficit of cases below JO Gy as observed, on the basis of the regression analysis. ::.- Cl. er: w Cl. w I- ~ w z w C z _J c:( :::> z <( s Envelopes., 9% 68% 95%.,,. / / / /,,,,,,,,, / / was less than.1. Mole [ M29] plotted. as a fnction of dose. the cmlative incidence of bone sarcoma per 1 3 individals for the grop of dial painters who started work before 193 [M27] and who were followed throgh 1972 [AIO]. He assmed eqal effectiveness of doses from both radim isotopes. as qoted in [A JO]. Fnctions of incidence!ta against dose {D) in Gy. of the form!ta= a 2 o2e-/l,d and!ta= a 1 De-//,D gave fits of comparable goodness (P =.26 and.45, respectively). The fit to a linear eqation is ITA(D) = (111 ± 22)1-4 D e-<b.5± 2-2llo- 3 (5.2) characteristic of the linear model for high-let radiation. which cold be sed for interpretation of the data on incidence. The slope gives the risk of bone sarcoma from the absorbed dose from radim alpharadiation. Very low vales of (1 1 cannot be interpreted in terms of a very high lethal dose to the transformed cells for a variety of reasons. inclding the long dration of exposre with possible repoplation of the irradiated cell pool Schlenker (S5] approached the same problem theoretically by calclating the life-time probability lta of bone sarcoma. corrected for the presence of competing risks. expressed as a fnction of timeaverage tmor rate from 22 6Ra Ra derived from Rowland's eqation referred to above (Rl8]. Ths, IOU!TA= f da ita e-{ (ITA + SJ da (5.3) A where A is the age at exposre: a is the age in years; i,.a is the time average tmor rate, eqal to zero /,,, / /,,,,,, /,,,,,,,,,,,,,.,,,,,, / /,,,,,, ; /,,,,,,,,,,,,,,, I I I I I I I I I 1 I I I INITIAL SYSTEMIC ACTIVITY (µci 226 Ra µci 228 Ra) Figre XXXII. Plot of bone sarcoma Incidence rate In female dial painters, as a fnction of the Initial Intake of radim. Three pairs of crves denote probabilities that the tre nmber of bone sarcomas, In the region of Intake where no tmors were observed, falls within the calclated envelopes [S5]. The shaded area covers two standard deviations on either side of the doseresponse crve ita = (a + a~2) e-p,a, aa calclated from the standard errors of the parameters [R34]. 236

75 ntil a five-year latent period has passed and constant thereafter: and Sa is the death rate from all other cases as a fnction of age For single 22<,Ra intakes at ages 2, 45 and 7 years, the reslts are presented in Figre XXXIII, <I'., z: z: AGE INITIAL SYSTEMIC ACTIVITY {µci 226 Ra µci 228 Ra) Figre XXXIII. Predicted Ille-time Incidence ol bone sarcoma in the presence of competing risks, as a!nction ol a single Intake of 226,228Ra. [SSO] together,vith 68% confidence limit envelopes for a 7-year-old. For other ages. the envelopes are very similar. lt is qite obvios that an approximately linear dose-response for cmlative incidence!ta against intake of activity into the blood might be the prevailing trend over the range from abot 2 kbq to abot 8 MBq. This conclsion. of corse. excldes neither the possibility of a.. practical threshold" nor the absence of a life-shortening effect sch as that referred to above. nor that a dose-sqared relationship may be fitted into the envelopes The linear trend of bone sarcoma incidence (ITA) against alpha-ray dose in the hman skeleton is in accordance with the blk of the evidence from animal experiments. as reviewed in chapter IV. However, more refined analyses of bone sarcoma integral tmor rates, indced in beagles by 239 P injections, spport a dose-sqared dependence of the rate [Pl7, W!4]. For radionclides emitting low-let particles like 9 Sr. an pwards concave dose-response relationship was invariably observed [Ml2-M 14] Several thosand patients in the Federal Repblic of Germany were given repeated intravenos injections of Ra in the corse of therapy for tberclosis and ankylosing spondylitis. In contrast with 2 26 Ra. 22-IRa, with its short half-life of 3.64 days, decays on the srface of bone where it becomes deposited early after injection de to rapid ionic exchange. The fraction of the mean dose to the skeleton delivered by 22 4 Ra to the target cells is mch higher than that for 226 Ra. The effectiveness of the mean skeletal dose is therefore abot 7 times higher for : 2-1Ra. The patients were of both sexes and inclded adlts as well as jveniles. The most recent report on these patients throgh I 98 [M57] incldes 2 I 8 jveniles and 68 adlts. Bone sarcomas had appeared in 35 patients injected as jveniles and in 18 adlts. The follow-p times ranged from O to 36 years; deaths acconted for most of the short follow-p times, whereas 582 patients were followed for more than 19 years and 78% were still alive in 198. The mean skeletal doses from a-particles were estimated to be I.98 and 2.5 Gy in jvenile and adlt, respectively: the corresponding injection spans averaged I 1 and 6 months. Virtally all bone sarcomas are ascribed to radiation, as only.2 cases wold be expected, based on general poplation rates The shortest appearance times were 3.5 and 5 years in jvenile and adlt patients. respectively; the corresponding average vales were I.4 ± 5.1 and 11.6 ± 5.2 years, and were not significantly different. The time distribtion was similar to that for lekaemias in atomic bomb srvivors, as described for those exposed within 15 m of grond zero by Bizzozero et al. [882] for which the average appearance time was 9 years. The peak of bone sarcoma appearance was 6-8 years. No additional cases were discovered since the previos report in 1974 [S21]. It may be conclded, therefore, that by 25-3 years after injection the vast majority of bone sarcomas have already become manifest There was some tendency towards a lower risk of bone sarcoma per nit dose with advancing age at injection. bt this cold be an artifact de to ncertainties in dosimetric assmptions, a possibly non-linear response, or a more prononced fractionation of injected activity in jveniles. Fractionated administration appeared more effective: a protracted weekly administration over a few years carries a 5-times greater risk per nit dose than single injection [S 18]. This observation is in agreement with stdies on mice injected with 22 4 Ra in varios fractionation schemes [L2, M67]. Original data from reference [S 18] were examined [P8] for partial correlations between the increment of incidence per nit dose. average injection span and the average skeletal dose. Significant positive correlations were fond between the span and dose. and also between the incidence increment and the span. The correlation between the increment of incidence and the fractionation span itself was non-significant (in fact it was negative) when the above partial correlations were acconted for. Ths. the sggestion of an increased effectiveness of fractionated verss single administration of 224 Ra, based pon hman experience. appears nfonded. The dose-response relationship for bone sarcoma incidence in the treated patients was compatible with linearity. The slopes for jveniles and adlts differ by a factor not greater than 2 (Figre XXXIV, [S21]). However. a dose-sqared trend of the effect cold not be rejected on statistical gronds. No carcinomas of the head sinses were observed in patients injected with 224 Ra de to lack of sfficiently long-lived radon nclides in the thorim decay series. 237

76 c w z w c:,... z :;; c::: < 2 V'l w :z cc 1 Adlts o Jveniles 452. Marshall and Groer assmed that there are abot I O 11 cells at risk in the endostem of a "reference man". These cells are sbject to sarcoma initiation in two steps, bt also to cell killing, the killing being by a factor of 1s more probable than initiation ( for the cells mst be of the order of I Gy). Ths, doses which prodce bone sarcoma (at a rate of a few tens of mgy per day) shold also completely sterilize the poplation of cells at risk before any transformation cold have occrred. To eliminate this difficltv, the athors assmed that cells within a -1 µm dis{ance from the bone srface, if killed by alpha particles, can be replaced at a rate of.1 per day from a stem-cell compartment which itself is beyond the range of 5. 7 MeV alpha particles. A conceptal otline of the model is shown in Figre XXXV. At all stages of malignant development, STEM CELLS SKELETAL DOSE (Gy) Figre XXXIV. Bone sarcoma Incidence verss skeletal dose from 224Ra. For lhe linear non-threshold lines shown lhe cmlative risk of bone sarcoma dring years postexposre is. 7% (adlts) and 1.4% Qvenlles) per nit average skeletal dose. [S21] 45. Schlenker [S5] calclated the dose-response relationship for 224Ra-indced osteosarcoma cases. For incidence ITA (ncorrected for competing risks) in sbjects older than 16 years at injection. the fnction had the form ITA = l - e " At the widest separation, the 95% confidence limits span a range of 75. reflecting mch less information at low doses for 2? 4 Ra than for Ra. This can be explained by the smaller nmber of cases and shorter follow-p times. 2. The model of Marshall and Groer 451. Theories of bone sarcoma indction were developed in the past [B2 l, M26] bt the most comprehensive one based on hman epidemiological data has been presented by Marshall and Groer [M4]. They based their model on several observations. The first is that a linear dose-response relationship gave a poor fit to the incidence rate of bone sarcoma verss dose (bone averaged) from n 6Ra and 228Ra. as discssed in detail by Rowland [Rl8-R2]. There were no cases of osteosarcoma below 9 Gy, and pre linearity of incidence rate verss dose was therefore highly nlikely on statistical gronds. They postlated, therefore, that two events prodced by alpha radiation in a cell at risk are necessary to initiate the malignant transformation. ~INITIATION INITIATION PRc»IOTION steosarcoma cell Figre XXXV. Schematic representation of the Marshall and Groer [M4] model of bone sarcoma Indction by alpha particles. An original endosteal cell Is Initiated twice and promoted to become an osteosarcoma cell; cells killed at any stage before promotion are replaced by stem cells. The rate of killing, F.., Is 15 limes larger than the rate of Initiation Fa (F = endosteal dose rate). The rates of replacement p and ol promotion l are Independent of radiation dose. the cells arc at risk of killing, which is proportional to their nmbers (M, M 1, M 2 ) and to the dose rate F. The dose rate in trn is proportional to the content of radim per nit mass of the diffse radim component in bone. The flly transformed cells (M 3 ) promoted to celllar division are not assmed to be at significant risk for cell killing de to a rapid mltiplication of the clone and escape of the tmoros cells from the volme irradiated by alpha particles To smmarize. the model incorporates the following parameters: (a) (b) (c) (d) (e) (f) initiation probability per cell per nit dose (a): killing probability per nit dose (k): nmber of cells at risk (s): promotion probability per nit time().); replacement probability of killed cells (p): growth time of a tmor to become recognizable (g); 238

77 (g) dose rate (F) proportional to activity per body weight (c). These parameters were incorporated in a set of differential eqations to give a probability of appearance, P(t), of an osteosarcoma per nit time per individal and nit activity of radim reaching the blood stream. The vales of p, a and s were held fixed, and initiation probability, promotion rate, tmor growth time and dose rate were adjsted throgh a soltion of the model by best fit of the data obtained from observation of individals with radim body brdens followed throgh December 1973 at the Argonne National Laboratory [A I OJ From the soltion of the model for man, it follows that the time of sarcoma appearance after radim ptake shold increase with decreasing vales of the intake q reaching a constant platea at abot 3-35 years at the level of 4 kbq per kg body weight. The theory predicts [G23, M4] an enhanced effect on bone sarcoma indction by fractionated administration of a short-lived bone-seeking alpha emitter. The existence of sch a phenomenon was, in fact. sggested when bone sarcoma incidence was stdied after injections of 224 Ra. When the isotope was administered over one month, the incidence was lower than that seen after prolonged (average 2 years) administration of the nclide (P =.16) [SIS). The lower sarcoma incidence in this case shold reslt, according to Marshall and Groer (G23. M4], from sppression of the poplation of viable endosteal cells M by very high dose rates delivered over short intervals (the empirical basis to spport this contention is open to critiqe, see paragraph 449) The theory fits satisfactorily the preliminary data on the indction of bone sarcoma by 226 Ra in Utah beagles. in which the cmlative incidence reaches a platea of abot 92%, a factor of higher than in man. The parameters of the theory, which give a good fit to the experimental sitation. were rather similar to those for man. with the exception of higher promotion and initiation rates (both roghly by a factor of 1). Reasons for this interspecies difference in initiation probability are not clear Marshall and Groer [M4] correctly stated that fit of the model to the measred data does not prove correctness of the postlates. Some reslts of the theory were not envisaged a priori. e.g., the platea of incidence rate at high doses (Figre XXXVI). This point of agreement enhances the credibility of the theory. However, in the light of the evidence that the cmlative incidence of 226Ra-indced bone sarcoma. both in man and in beagles, can be reasonably approximated by a linear model, the basic assmption of two-hit initiation is open to qestion. A model similar to that of Marshall and Groer. bt postlating a one-hit initiation has been proposed [P23]. The model explains the non-linear part of the activityincidence rate crve by the promoting action of alpha radiation exerted via repoplation of the target cell pool (osteogenic cells). 1 ~ 1! Data (S.E.) Model t a r----+-"""t- 1 1 INTAKE (µci/kg body) Figre XXXVI. The cmlative Incidence of osteosarcomas In man verss the total Intake (µcl per kg body weight) of radlm- 226 pls radim-228. Up to abot 1 µcl per kilogram body weight systemic Intake the slope of the line, according to the model of Marshall and Groer [M4], Is close to 1. F. INTERSPECIES COMPARISONS 457. In nmeros instances in the corse of the review of experimental (chapter IV) and hman data (chapter V), qalitative similarities have been pointed ot between animal and hman data that may seflly be smmarized here. From sch comparisons. the pattern seems to emerge that the shape of doseresponse relationships may be basically similar in varios species Ths, the linear non-threshold type of doseresponse relationship for female mammary carcinoma indced by x rays finds its conterpart in the reslts of stdies on at least for strains of rats: (a) the Sprage Dawley females irradiated externally by x rays and/or gamma rays. and internally by fj particles from tritim: the linearity applies to adenomas bt the reslts for adenocarcinomas are also consistent with the notion: (b) females of the WAG/Rij rat strain and the AC strain irradiated externally with x rays showing linearity for adenomas and carcinomas, respectively: and (c) non-inbred female rats irradiated from a /J-particle sorce also showing a linear rise of incidence of mammarv carcinomas. The dose-response crve for mammary ~arcinomas indced in BALB/C mice by gamma rays shows irreglarities that defy any possible comparison A strong linear component is manifest in the non-threshold dose-response relationship for x-ray indced thyroid carcinoma in man; effects of dosefractionation are virtally absent. Data on Long Evans rats. even if the nmber of animals is limited and the observations were terminated 2 years after irradiation, are compatible with linearity. They also show a Jack of a prononced dose-rate effect, as shown by the comparison of x-ray and 131 I doseresponse crves. The absence of an excess of thyroid cancers in hmans exposed to diagnostic qantities of 131 I is not consistent with the above observation, bt 239

78 the relevant sample is highly selected and the observations made on this cohort do not necessarily in\'alidate the basic similarities between the doseresponse relationships from the hman species and the rat. 46. Cancer of the respiratory tract indced in several grops of miners, exposed to daghter prodcts of mrn, consistently yields a non-threshold linear dose-response relationship p to a cmlative exposre of abot 3 WLM, with some decline of the risk coefficient (FTA) at higher exposres or exposre rates > 5 WLM per year. A very similar relationship is observed in male Sprage-Dawley rats made to inhale 222 Rn decay prodcts or exposed to fission netrons. The analogy holds when the data from the two species are analysed in terms of both the absolte and the relative risk model Dose-response relationships for bone sarcoma indced by long-lived isotopes of radim in man, beagles and mice are compatible with linearity between dose and cmlative incidence over the initial part of the crve. In all three species a decline of the tmor yield is seen at high and very high doses in the skeleton. In the three species there is also a prononced inverse relationship between dose (or initial activity in the body) and the latent period In smmary, it appears that the shape of the dose-response relationship for radiation-indced cancer of some organs in man generally follows the pattern seen in a nmber of experimental animal species. This qalitative finding will have to be verified in the light of ftre data. The similarity may not be srprising for those types of hman malignancies that have a reasonable conterpart in other mammalian species. in view of the basic similarity of patho-physiological phenomena on which rests the rationale of sing experimental animals to predict the response of man. However, the reglarities pointed ot do not imply that data from experimental animals may be sed to derive risk coefficient applicable to man, becase, so far. sch similarities may not be shown to apply to the qantitative expression of the carcinogenic effect in varios species. G. SUMMARY AND CONCLUSIONS 463. Epidemiological data sefl for an analysis of the shape of the dose-response relationships for radiation-indced cancer are limited. This applies to external and internal, to instantaneos and prolonged exposres. Cation shold therefore be exercised in formlating broad generalizations. Several factors are known to affect the magnitde of the neoplastic response. Among those identified. age at irradiation. sex and genetic or ethnic characteristics are of importance. bt the extent to which these factors may affect the form of the dose-response crves is almost totally nknown. For some tmors. e.g., cancer of the lng, the characteristics of the environment and the living habits may consistently affect the expression of malignancies With the Hiroshima and Nagasaki atomic bomb dosimetry still ncertain, the epidemiological information on lekaemia incidence in the srvivors cannot yet be flly tilized. Information on indction of lekaemia derived from the follow-p of spondylitis patients is insfficient to allow firm conclsions as to the most likely form of the dose relationship. particlarly becase lekaemia incidence in this grop may have been inflenced to an nknown extent by dose-fractionation effects. Therefore. frther discssion of dose-response relationships mst be postponed ntil new information on the atomic bomb srvivors becomes available Regarding breast cancer, there is no sggestion of a threshold for low-let radiation; either a pre linear relationship with dose or one with a small dosesqared term might apply. This conclsion is spported by the small or absent effect of dose fractionation and by the lack of a detectable relationship of latency to dose. For this tmor, there is, however, a prononced variation of the risk with the age at exposre. as women in the age range of -3 years are at the highest risk Dose-response relationships for carcinoma of the thyroid have a strong linear component and the data do not sggest a threshold. There appears to be no inverse relationship between latency and dose and this is consistent with the observed linearity. Infants and children are more ssceptible to cancer of the thyroid than grown-p persons. Dose-rate effects cannot be exclded. and data on patients given 1 > 1 I for diagnostic prposes might be interpreted in this way; however. other interpretations are also possible, particlarly since dose-rate effects have not been seen in recent ~nimal experiments. The indction of benign thyroid adenomas is linearly related with x-ray doses at low doses. bt with a decrease of the slope at doses above I Gy The available data on lng cancer indced by exposre to radon and its decay prodcts in mines shmv. withot exception. linear exposre-response relationships, with a decreasing slope at the highest exposre rates (above 5 WLM per year): there is no indication of a threshold. This type of relationship has also been consistently observed for airborne shortlived a-emitting radionclides in animal experiments. For low-let radiation the form of the dose-response crve is essentially nknown Alpha-emitting bone-seeking radionclides. sch as ~ 2 6Ra and 2 28 Ra. may indce bone sarcoma in man with a significantly elevated freqency at mean skeletal doses above lo Gy. Plotting the incidence or incidence rate against cmlative dose reslts in relationships that may be interpreted as either linear or qadratic. In the latter case, below abot JO Gy there is a significant deficit of tmors when a linear interpolation to zero dose is attempted and IrA fits very well a dose-sqared model with a cell-killing term. In this case. linearity may be rejected. It appears impossible at present to decide whether linear or qadratic models, or a combination of both. provide better fits to incidence UrA) verss dose from radim 24

79 isotopes over the initial portion of the crve. Animal experiments sggest linearity. Bone sarcomas indced in man and animals by a-emitting bone seekers show a prononced inverse relationsnip of latency to dose. The deficit of tmors in man, at the lowest end of the scale mentioned above, cold therefore reflect an apparent "practical" threshold, and a decreasing tmor-related life shortening with lowering of the dose A dose-response relationship for bone sarcoma indced in man by short-lived a-emitters (2 2 ~Ra) is compatible with linearity, even thogh alternative models may not be exclded. The latent period for bone sarcoma appears to be rather short; the age at the beginning of exposre has no strong inflence. There are no hman stdies of bone sarcoma cased bv low-let irradiation. However, irradiation of e~perimental animals by P-emitting bone seekers ( 45 Ca, 9 osr, 89 Sr) invariably reslted in pwards concave relationships and a prononced inverse relationship between dose and latent period. It wold be expected that the same kind of relationship might also apply to man. 47. In smmary, the model applying to the doseresponse relationship for lekaemia is nknown and mst remain so ntil new dose-incidence data for atomic bomb srvivors become available. The reviewed evidence sggests that incidence of lng. mammary and thyroid cancers is probably related with dose by a generalized linear or linear-qadratic relationship. However. dose-incidence information is not available for Jng cancers after low-let and for breast and thyroid after high-let irradiation In a comparison of data from experimental animals. reviewed in chapter IV, and from hman epidemiological stdies in the present chapter. UNSCEAR has identified a nmber of tmors for which the dose-response relationships appear to follow the same basic pattern in varios species. Sch a comparison will have to be verified as more data are accmlated. In any case. the qalitative similarity of the indction kinetics does not jstify any simple qantitative extrapolation of incidence data from animals to man. VI. SUMMARY 472. In order to estimate the absolte risk of radiation-indced cancer (nmber of indced tmors per million people exposed per nit dose, i.e.. cases 1-6 Gy- 1 ), or the relative increase of tmor freqency per nit dose over the natral" incidence at low doses and dose rates. two types of information are reqired: first, empirical data on the incidence of varios forms of malignancy at relatively high doses where observations have actally been made; second, knowledge of the mathematical form of the relationships linking the incidence of cancers with the dose of radiation. Sch data wold allow predictions to be made of the cancer incidence at doses (and perhaps also at dose rates) very mch lower than those at which direct observations in man have been gathered When the incidence of a given tmor in exposed animal or hman poplations is followed as a fnction of increasing dose. several observations may sally be made. At relatively low doses (at abot.1 Gy of sparsely-ionizing radiation), only seldom (and then mostly in controlled animal experiments) can a statistically significant increase of cancer or lekaemia be shown. At higher doses (p to a few Gy, bt with considerable differences between different tmors). the incidence of sch malignancies may be shown statistically to exceed the level observed in nonexposed control poplations, following some increasing fnction of the dose. At still higher doses (many Gy), the incidence gradally starts to fall off becase of cell killing. Dose-response relationships of this type, going throgh a maximm at some intermediate dose, are often fond in experimental animals. They are termed ''biphasic" A common interpretation attached to sch a shape advocates the concrrent presence of two different phenomena: (a) a dose-related increase of the proportion of normal cells that are transformed into malignant ones: and (b) a dose-related decrease of the probability that sch cells may srvive the radiation treatment. Both these phenomena are normally operating in the region of doses where data are available, bt to a different degree for varios doses and different types of cancer. What may happen at the low doses. where direct information is lacking, may only be inferred from a combination of empirical data and theoretical assmptions, linked together into some models of radiation action The models referred to are simplified semiqantitative representations of complex biological phenomena. Present knowledge of the mechanisms of carcinogenesis, inclding radiation carcinogenesis, is inadeqate to design comprehensive models acconting for all physical and biological factors known to inflence the indction of cancer. To avoid some of the complications involved. UNSCEA R sggests that the range of doses over which extrapolations may meaningflly be performed shold be limited to low and intermediate doses. Under these conditions, it seems likely that no serios distortions wold reslt from other radiation effects, called non-stochastic, that are observed when doses exceed fairly high thresholds, typical for each tisse and each effect The formlation and analvsis of models of radiation carcinogenesis mst reiy on a few basic assmptions, as follows: (a) The observed dose-response relationships for clinically visible tmors in vivo approximately reflect the relationships between dose and freqency of the initial triggering events in the cells where cancer will eventally arise, despite host reactions and the effect of latency. which may modify sch relationships to some degree. This assmption is based on the overall similarity of the dose-response crves for cancer indction with those of varios other celllar effects of radiation. UNSCEAR postlates this concept simply as a working hypothesis: 241

80 (b) (c) (d) Cancer initiation is basically a nicelllar process occrring at random in single cells. This is also a working hypothesis that has not yet definitely been proved. However. evidence lo the contrary, i.e., that cancer initiation takes place in several cells, is not entirely convincing. The ni-celllar theory of cancer indction is compatible with the notion that some, still ill-defined, inflences reslting from irradiation of other neighboring cells or other organs may modify the probability of an initiated cell to develop into an overt malignancy. Firm biological evidence in favor of this latter notion is very fragmentary: Absence of a dose threshold is characteristic of many, if not all, tmors. For some animal tmors (e.g., tmors of the ovary or thymic lymphoma of the mose) threshold-type doseresponse relationships are observed. In other cases (e.g., tmors of the skin) cancer is only indced with great difficlty, i.e., after high doses of radiation. In still others (i.e., epidermoid lng cancer in man) the data are nclear, owing perhaps to a short follow-p of the patients. However, in spite of these exceptions, absence of a threshold dose for the development of cancer is assmed by UNSCEAR as a working hypothesis for the time being; Ssceptibility of an irradiated animal or hman poplation to tmor indction is assmed to follow a nimodal distribtion. Althogh genetic predispositions to the development of some forms of malignancy are well docmented (see annex A), efforts to show that sch phenomena apply also w radiation-indced hman cancer have not been sccessfl so far. Therefore, pending frther stdies, the same nimodal distribtion of ssceptibilities LO the indction of cancer in irradiated and non-irradiated poplations is also provisionally accepted as a working proposition On the above assmptions, it is possible to infer likely shapes for the dose-response relationships of radiation-indced cancer. Sch inferences rely on the analysis of varios other radiation effects observed at the celllar level. These effects involve the cells' genetic material, which is also thoght to be the primary target for cancer initiation. The prodction of mtations and chromosomal aberrations in somatic and germinal cells, and the oncogenic transformation in vitro of mammalian cell lines, are examples of sch effects. If cancer indction in vivo involves mechanisms similar to, or related with. those nderlying the effects listed above, one wold expect all these phenomena to respond similarly in respect of changes in dose, dose rate and fractionation. As sch similarities have actally been observed, it may be possible to extrapolate the shape of dose-response relationships between sch effects and the phenomenon of cancer indction Three basic non-threshold models of radiation action as a fnction of dose have been discssed with respect to sch celllar effects and co cancer indction: the linear, the linear-qadratic and the pre qadratic. Notwithstanding some exceptions, these may provide a general envelope for a variety of radiation-indced end-points at the celllar level, as well as for tmor indction in experimental animals and hman poplations The vast majority of dose-response crves for indction of point mtations and chromosomal aberrations by sparsely-ionizing x and gamma rays may be described by a linear-qadratic model. For the same end-points, when cell killing is corrected for, a linear model sally applies to densely-ionizing netrons or alpha particles. As a rle. for a nmber of chromosomal strctral abnormalities. crvilinearity (pward concavity) is observed for low-let radiation, while, for the same effects and a wide range of doses. linearity prevails for high-let particles. Linearity of the dose-response for somatic mtations and terminal chromosomal deletions has been fond in some cell lines, even for low-let radiation. bt these findings are relatively infreqent. 48. Approximate estimates of proportionality constants linking the chromosomal effects with the dose or its sqare may be obtained experimentally. They allow the freqency of sch effects to be predicted at low doses and dose rates from observations at higher doses. For cancer indction, however, only fragmentary information spports the notion that similar qantitative relationships with the dose might apply. UNSCEAR has estimated that if the risk of tmor indction at I or 2 Gy of sparsely-ionizing radiation (at high dose rate) were extrapolated linearly down to zero dose, this procedre wold over-estimate the risk by a factor of perhaps p to 5 in typical sitations Over the last few years mch information on radiation-indced oncogenic transformation of mammalian cells has become available. The canceros natre of the transformed cells is shown by the fact that after transformation in vitro they are able to form malignant tmors pon transplantation into an animal of the same species. Transformation in vitro is therefore regarded as a model. albeit a simplified one, of radiation carcinogenesis at the celllar level. Cells exposed in vitro to sparsely-ionizing radiation 24 hors after seeding are transformed according to complex kinetics that cannot be fitted to models sed for other celllar effects. Moreover, fractionation of the dose (below 1.5 Gy total) enhances transformation, which effect is contrary to what wold be predicted by a linear-qadratic model. If similar phenomena wold apply to the indction of cancer in vivo. extrapolation to low doses and dose rates might reslt in a significant nder-estimate of the risk. Frther research is needed to elcidate these phenomena, bt several experiments indicate that the observations on in vitro systems referred to above may reslt from non-typical conditions of celllar growth dring the early periods after establishment of the cltre. Actally, irradiation of non-dividing cells or cells nder exponential conditions of growth {which are thoght to be more representative of an asynchronosly dividing cell poplation) prodces reslts that are compatible with those obtained for other celllar effects; ths, for example, high-dose-rate gamma irradiation reslts in a greater freqency of transformation than Iow-doserate exposre. 242

81 482. There are indications that when cells are irradiated with netrons. low-dose-rate or dose fractionation may increase the rate of transformation, even at low doses: however. whereas some observations on tmor indction in experimental animals clearly spport these findings. others do not. In other experiments, enhanced transformation by netron fractionation or protraction were seen only at intermediate and high doses. In view of the scarcity of sch data. and of the ncertainties regarding the mechanisms involved, frther research is needed before enhancement of cancer indction by netron fractionation and protraction (relative to single or high-doserate exposre) may be accepted for the prpose of risk assessment. Sch a possibility shold. however. be kept in mind even thogh the theoretical basis to explain sch phenomena is ncertain at present Recent experimental findings on radiationindced tmors in experimental animals have not sbstantially changed the main conclsions reached in annex I of the 1977 UNSCEAR report. Most data spport the notion that dose-response relationships for x and gamma rays tend to be crvilinear and concave pward at low doses. Under these conditions, tmor indction is dose-rate dependent, in that a redction of the dose rate, or fractionation, redces the tmor yield. A linear extrapolation of the risk from doses delivered at high rates to zero dose wold ths, as a rle, over-estimate the real risk at low doses and dose rates. However, in one experimental mammary tmor system (matched by epidemiological data on hman breast cancer) x and gamma radiation prodced a linear dose-response with little fractionation and doserate dependence. This was recognized as an exception in the 1977 UNSCEAR report. It appears likely that a similar dose-response relationship cold also apply to the indction of thyroid cancers, bt the data are too limited to permit any definite conclsion to be drawn in this respect For densely-ionizing netron irradiation, tmor indction in animals follows, in general. a very nearly linear crve at the lower end of the dose scale and shows little dependence on dose rate. In some cases, however, enhancement pon fractionation (and possibly protraction) has been noted. Above abot.1 Gy or so, the crve tends to become concave downward. markedly in some cases. Under these conditions, a linear extrapolation of the risk down to zero dose from intermediate or high doses and dose rates wold involve a \'ariable degree of nder-estimation Having reviewed existing data on dose-response relationships for radiation-indced tmors in man, UNSCEAR considers that this whole matter mst be treated with cation becase at the present time observations are fragmentary. Those for netrons are totally absent, and definitive data for atomic bomb srvivors at Hiroshima and Nagasaki are still not available. For example, dose-response information for sparsely-ionizing radiation have not been reported for lng and bone tmors. and data for densely-ionizing radiation have not been reported for thyroid and mammary cancer. For sparsely-ionizing radiation, in some cases (lng. thyroid, breast), the data available are consistent with linear or linear-qadratic models. For breast cancer, however, predominant linearity may apply. as the incidence is little affected by dose fractionation. Linearity of the response for lng cancer after irradiation by alpha particles from radon decay prodcts does not contradict the above statement, becase with alpha particles the dose-sqared component is practically absent. Some dobts still remain, however, as to osteosarcoma indced by bone-seeking alpha- or beta-emitting radionclides. However. in spite of the fragmentary character of the hman data, a general pictre may be emerging from which several tentative conclsions can be derived For sparsely-ionizing radiation, linear extrapolation from abot 2 Gy down wold not overestimate the risk of breast and possibly thyroid cancer. 1 t wold slightly over-estimate the risk of lekaemia. and it wold definitely over-estimate the risk of bone sarcoma. For lng cancer. lack of direct evidence does not permit any assessment to be made of the magnitde of the over-estimate For densely-ionizing radiation, the risk of lng cancer from accmlated exposres to radon decay prodcts at low dose rates from 3 WLM down (roghly corresponding to 2 to 5 Sv) wold not be over- or nder-estimated by linear extrapolation. However. extrapolation from observations made at higher cmlative exposres might reslt in a significant nder-estimation de to observed flattening (satration) of the dose-response crve in this region. It shold be stressed that absolte risk coefficients derived for male miners, of whom a high proportion are smokers, shold not be applied to the general pblic withot de corrections for varios factors (intensity of smoking. Jng ventilation rate. presence of other contaminating polltants, etc.) that are thoght to increase the risk in the miners The incidence of bone sarcoma after internal irradiation by alpha particles from long-lived boneseeking radionclides is distorted by the existence of a prononced inverse relationship between accmlated dose and latent period. This reslts in an apparent threshold at low doses, and in a redced loss of life span per indced tmor, as a fnction of decreasing dose, down to zero loss at low doses. If this is a correct explanation for the pward concavity of the dose-response relationship. a linear extrapolation from a few tens of gray mean skeletal dose down to the gray or milligray region wold grossly over-estimate the risk No hman data for indction of breast cancer and lekaemia by densely-ionizing radiations are available at present and, therefore, no direct inferences can be made abot risk extrapolation to the low-dose domain. On the basis of general knowledge. if the risk at intermediate doses cold be derived from data on sparsely-ionizing radiation (sitably corrected for the higher effectiveness of the densely-ionizing particles), a linear extrapolation down to low doses might either nder-estimate or correctly estimate the real risk in these cases. 243

82 49. For radiation-indced cancers of other organs, only experimental data are available. For sparselyionizing radiations pward concave crvilinear doseresponse relationships with prononced dose-rate and fractionation effects are sally fond. If similar crves shold apply to cancers in man, a linear extrapolation of risk coefficients (obtained at the intermediate dose region after acte irradiation) to the low dose and low dose rates. wold very likely overestimate the real risk. possibly by a factor of p to 5. For densely-ionizing radiation, shold relevant vales become available, a linear extrapolation wold probably nder-estimate the risk Upon close inspection. some reglarities seem to emerge that may indirectly help in assessing the character of dose-response relationships in man. A similarity of the shape of the relationships was noted between man and experimental animals for tmors of several organs for which reasonably good information exists. These are: mammary and thyroid cancers (low-let radiations), and lng and bone cancers (high-let radiations). Shold this pattern be confirmed, knowledge of epidemiogical stdies in man at intermediate or high doses, and of the shape of the dose-response relationships from several animal species, might make it possible to assess the bias introdced by linear extrapolation of the risk coefficients to low doses. VII. RESEARCH NEEDS 492. At the conclsion of this analysis, it appears qite obvios that there exists a wide intellectal gap between the abstract and general natre of the models discssed and the very complex biological reality, particlarly in respect of the moleclar mechanisms of tmor indction that are gradally being ncovered. Assigning a biological meaning to the nmerical constants ofche models. searching for the compatibility between the mechanisms identified and the models' predictions, and formlating new models based on hard data rather than on a priori generalizations. are the formidable tasks facing this type of research today. Any advancement in the field of radiation carcinogenesis will presmably come in parallel with progress in tmor biology, of which the former is only a special aspect. Within these very general gidelines, specialized stdies in the following sbjects cold be valable and UNSCEAR wold like to recommend them to the attention of the interested research worker: (a) 244 Stdies on the natre of the celllar and sbcelllar changes involved in the transformation of normal into canceros cells. These shold aim at establishing and possibly identifying the stages in the indction process that have been postlated. Within these stdies. the mechanisms of oncogenic transformation on a wide range of cell types in vitro, and the inflence of the main radiobiological variables on the form of the relevant dose-effect relationship. shold have a special place, becase the biological end-point in qestion (b) (c) is probably very relevant to the initiation of cancer. Stdies on the relationship between transformation of cells to a potentially malignant state in vitro, and the prodction of malignancies in vivo. shold be particlarly prsed. Mechanisms by which sb-transformation and potential transformation damage is repaired in cells after exposre to low- and high-let radiation shold also be explored. In experimental animals, a better nderstanding of the pathogenesis of radiation-indced tmors shold allow the formlation of qantitative models more firmly fonded than those available so far. More attention than in the past shold be paid to the end-points tested and to the experimental design which will allow a correct statistical analysis of the experimental data. In particlar, stdies shold aim at elcidating the mechanisms by which radiation interacts with oncogenes and the genetic material. Recommendations in respect of experimental work on animals were formlated in annex 1 of the 1977 UNSCEA R report and. in view of the rather slow rate of accmlation of the data for sch long-term effects, many of these recommendations are still valid at present. Particlar aention shold be devoted to stdies in the range of low doses. Comparative stdies wold be sefl of dose-response relationships for those tmors that presmably represent adeqate models for hman malignancies. and especially those for which qantitative information on dose-response relationships is likely to become available in man in the not too distant ftre. Stdies cold seflly be devoted to the RBE of different types of low- and high-let radiation. particlarly in those instances where reliable epidemiological information does not exist, as is now the case for fast netrons. A sbject which remains to be explored in greater depth is that of effects of exogenos and endogenos factors on the shape of dose-response relationships for radiation-indced cancer, pani clarly the inflence of the immnological state and the effects of promoting and inhibitory factors. Another important sbject concerns the mechanisms by which downward-concave doseresponse crves and enhanced fractionation (or protraction) effects are seen after varios doses of netrons. In hman poplations, the follow-p of grops already nder stdy shold be contined as a matter of high priority. New stdies on grops that might potentially yield frther information shold also be considered and, if warranted. careflly ndertaken. The topics on which frther information wold be reqired are, in general, the following: (i) The shape of the dose-response relationships for a wide spectrm of tmors and an extended range of doses, with special attention to the correct assessment of doses, as well as its distribtion in space and time; (ii) The time distribtion of the latent periods for tmors with long latencies;

83 (iii) The existence and the possible form of a relationship between the length of the latent period and the magnitde of dose; (iv) The inflence of age, sex, genetic factors and some habits (e.g., smoking) on the qantitative characteristics of dose-response relationships and the reslting variability in risk estimates; (v) The adeqacy with which relative and absolte risk models for different cancers permit projection of the risk for long periods after irradiation. T a b 1 e Observed () and expected (E) breast cancers w\th relat\ve r\sk estimates as a fnct\on of radiation dose for the \nterval 1-34 years after radiation treatment for mastit\s [ Sl 6) A. Analys\s of persons w\th average dose to both breasts Breast cancer Dose range Mean dose ( Gy) ( Gy) PY a [ RR ~I Q_/ r;_/ Q/ l D l l l Control B. Analysis of single breasts with total dose per breast Breast cancer Dose range Mean dose ( Gy) ( Gy) PY a [ RR ~I Q_/ r;/ Q/ B l l Control tl !I For benign tmors no latent period was assmed, so the person-years at risk inclde the fll interval 1-34 years after irradiat\on for the single breast analyses. Q_/ Observed cancers. ~/ Expected vales were calclated by the combination x2 method, w\th adjstment for age at irrad\ation and yearly lntervals since 1rrad\at1on, and were standard\zed so that /E =RR= l 1n the control grop. &/ Relat\ve r1sk. tl Observed cancers and person-years at risk for control grop 1n this analysis inclde 392 non-\rrad\ated breasts (with three cancers) of women w\th nilateral \rradiat\on. 245

84 T a b 1 e 2 Crve-flttlng analyses of breast cancer data from Massachsetts floroscopy ser1es [L22] (All parameters are scaled by a factor of 16) Massachsetts floroscopy serles Model Parameter Goodness of flt Estlmate ~, :I: s.o. X. 2 df p.48 Unear a1 56± Llnear-qadratlc a1 45±3 a2 29± 76 l Unear + ldlllng a, 45±23 a2 Q_/ Llnear-qadratlc a, 45±86 + k 1111 ng a2 29±57 a2 Q_/ Qadratlc a2 13± Qadratlc + k11 llng a2 22±14 a2 (l.6±2.2) ~I The dlmenslons of the parameters are as follows: a 1 : 1-6 a-1 Gy-1 a 2 : 1-6 a-1 Gy-2 a 2 : 1-2 Gy-2. Q_/ The best-flttlng parameter wold be negatlve; the vale of zero reslts from the prlor constralnt that the parameter be non-negatlve. 246

85 T a b l e 3 Crve-f1tt1ng analyses of breast cancer data from Rochester mast\t1s ser\es [l22] (All parameters are scaled by a factor of 16) Mean doses bo both breasts Crves for dose to single breast Model Parameter Goodness of f1t Goodness of f1t Estimate Estimate ~I ± s.o. ± S.D. x2 df p x2 df p Unear a.1 56± ± L\near-qadrat1c a.1 56±43 39±54 a.2 Q/ !1./ L1near + k1111ng a.1 86±29 98± (1.8±1.4) (3.4±1.6) L1near-qadrat1c a.1 32±114 78±148 + k 1111 ng a.2 37±69 1± (4.8±4.) (4.2±4.9) Qadratic a2 11 O:t: t Qadrat1c + k1111ng a.2 55±17 47± (5.6:tl.5) (6.6±2.4) ~I See footnote a/ 1n Table 2.!1.I The best-f1tttng parameter wold be negative; the vale of zero reslts from the pr\or constraint that the parameter be non-negat1ve. 247

86 REFERENCES Al Aalen,. Nonparametric inference in connection with mltiple decrement models. Scand. J. Stat. 3: ( 1976). A2 Abrahamson, S. and S. Wolff. The analysis of radiation-indced specific locs mtations in the mose. Natre 264: (1976). A3 A4 A 7 AS Abrahamson. S. Mtation process at low or high radiation doses. p. 1-8 in: Biological and Environmental Effects of Low-Le\'el Radiation, Vol. I. IAEA. Vienna, Albert, R.E., F. Brns and R. Shore. Comparison of the incidence and time patterns of radiation-indced skin cancers in hmans and rats. p in: Late Biological Effects of Ionizing Radiation. Vol. II. IAEA, Vienna, I 978. Anderson, T. W. Radiation exposres of Hanford workers: a critiqe of the Mancso. Stewart and Kneale report. Health Phys. 35: ( 1978). Archer, V.E., J.K. Wagoner and F.E. Lndin. Lng cancer among ranim miners in the United States. Health Phys. 25: (1973). AIO Argonne National Laboratary. Exposre data for radim patients. Appendix A (p ), ANL-75-3, part II ( 1975). AI3 Arlett. C.F. and J. Potter. Mtation to 8-azaganine resistance indced by gamma-radiation in a Chinese hamster cell line. t\ft. Res. 13: (1971). Al4 Asqith. J.C. The effect of dose-fractionation on gamma-radiation indced mtations in mammalian cells. Mt. Res. 43: 91-IOO (1977). Al6 Archer, V.. E.P. Radford and. Axelson. Factors in exposre-response relationships of radon daghter injry. Conference/Workshop on Lng Cancer Epidemiology and Indstrial Applications of Sptm Cytology, November Colorado School of Mines Press, Golden, Colorado A 17 Adams, L.M. and R.L. Ullrich. Srvival of mammary epithelial cells from virgin female BALB/c mice following in vivo gamma irradiation. Radiat. Res. 91: 49 only ( 1982). Al8 Albert, R.E., F.J. Brns and E.V. Sargent. Skin tmorigenesis by mltiple doses of electron radiation in the rat. (Abstract) Radiat. Res. 87: 453 only (1981). BJ B2 B3 B4 B5 Bajerska. A. and J. Liniecki. A variability of proliferation kinetics of rabbit lymphocytes in vitro. Mtat. Res. 6: (1979). Bajerska, A. and J. Liniecki. The yield of chromosomal aberrations in rabbit lymphocytes after irradiation in vitro and in viva. Mtat. Res. 27: (1975). Bettega. D.. P. Calzolari, P. Pollara et al. In vitro cell transformation indced by 31 MeV protons. Radial. Res. 14: I ( 1985). Baral, E., L.E. Larsson and B. Mattsson. Breast cancer following irradiation of the breast. Cancer 4: (1977). Barendsen, G. W. Mechanism of action of different ionizing radiations on the proliferative capacity of mammalian cells. p. 167 in: Advances in Theoretical and Experimental Biophysics, Vol. I. (A. Cole, ed.). M. Dekker, New York, Barendsen, G. W. Radiosensitivity of tmors and related cells in cltre. Biomedicine 26: ( 1977). B7 Barendsen, G. W. Fndamental aspects of cancer indction in relation to the effectiveness of small doses of radiation. p in: Late Biological Effects of Ionizing Radiation. Vol. II. IAEA, Vienna, I Barendsen, G. \V. RBE-LET relations for indction and reprodctive death and chromosome aberrations in mammalian cells. p in: Proceedings of the Sixth Symposim on Microdosimetry, Brssels. May 1978 (J. Booz and H.G. Ebert, eds.). Harwood Academic Pb!., London B JO B 11 Barendsen, G. W. Characteristics of cell srvival crves for different radiations in relation to iso-effect crves for fractionated treatments of a rat rhabdomyosarcoma. p in: Cell Srvival after Low Doses of Radiation: Theoretical and Clinical Implications. The Institte of Physics and John Wiley and Sons, London, I 975. Barendsen, G. W. Inflence of radiation qality on the effectiveness of small doses for indction of reprodctive death and chromosome aberrations in mammalian cells. lnt. J. Radiat. Biol. 36: ( 1979). BI 2 Barendsen, G. W. Qantitative biophysical aspects of response of tmors and normal tisses to ionizing radiations. Crr. Top. Radiat. Res. Qart. 9: (1973). 814 Barrett, J.C., T. Tstsi and P.O.P. Ts'o. Neoplastic transformation indced by a direct pertrbation of DNA. Natre 274: (1978). BI 5 Bachinger, M. Strahlenindzierte Chromosomen Aberrationen. p im: Handbch der Medizinischen Radiologic (D. Olsson et al., eds.), Band 4/3. Springer Verlag, Berlin, B16 Bam,.T.W. Poplation heterogeneity hypothesis on radiation-indced cancer. Health Phys. 25: ( 1973). BI7 Beebe, G.W.. H. Kato and C.E. Land. Stdies of the mortality of A-bomb srvivors. 6. Mortality and radiation dose Radiat. Res. 75: ( 1978). BI8 Beebe, G.W., C.E. Land and H. Kato. The hypothesis of radiation-accelerated aging and the mortality of Japanese A-bomb victims. p in: Late Biological Effects of Ionizing Radiation. Vol. I. IAEA, Vienna, Bedford, J.S. and H.G. Griggs. The estimation of srvival at low doses and the limits of resoltion of the single-cell-plating techniqe. p in: Cell Srvival after Low Doses of Radiation: Theoretical and Clinical Implications. The Institte of Physics and John Wiley and Sons, London, B2 BEIR Report. The Effects on Poplation of Exposre to Low-Level of Ionizing Radiation. Report of the Advisory Committee on the Biological Effects of 248

87 Ionizing Radiation. National Academy of Sciences, National Research Concil, Washington, D.C., Brch, P.R.J. Radiation carcinogenesis: a new hypothesis. Natre 185: ( 196). 823 Boer, P. de, P.P. van Bl et al. Chromosomal radiosensitivity and karyotype in mice sing cltred peripheral blood lymphocytes and comparison with this system in man. Mtat. Res. 42: (1977). 824 Boice, J.D., M. Rosenstein and E.D. Trot. Estimation of breast doses and breast cancer risk associated with repeated floroscopic chest examinations of women with tberclosis. Radial. Res. 73: (1978). 825 Boice, J.D. jr., C.E. Land, R.E. Shore et al. Risk of breast cancer following low-dose radiation exposre. Radiology 131: ( 1979). B26 Boice. J.O. and R.R. Monson. Breast cancer in women after repeated floroscopic examinations of the chest. J. Natl. Cancer Inst. 59: ( 1977). B27 Boag, J. \V. The statistical treatment of cell sn ival data. p in: Cell Srvival after Low Doses of Radiation: Theoretical and Clinical Implications. The Institte of Physics and John Wiley and Sons, London, B28 Bond, V.P., C.J. Shellabarger, E.P. Cronkite et al. Stdies on radiation-indced mammary gland neoplasia in the rat. V. Indction by localized radiation. Radiat. Res. 13: ( 196). B29 Borek, C., E.J. Hall and H.H. Rossi. Malignant transformation in cltred hamster embryo cells pro dced by x-rays, 43 kev mono-energetic netrons and heavy ions. Cancer Res. 38: ( 1978). B33 Borek, C. and E. Hall. Transformation of mammalian cells in vitro by low doses of x rays. Natre 243: ( 1973). 834 Borek, C. In vitro cell transformation by low doses of x-irradiation and netrons. p in: Biology of Radiation Carcinogenesis (J.M. Yhas et al., eds.). Raven Press, New York, Botwell, R.K. The fnction and mechanisms of promotion of carcinogenesis. Crit. Rev. Toxicol. 4: (1974). 841 Brewen, J.G., R.J. Preston, K.P. Jones et al. Genetic hazards of ionizing radiations: cytogenetic extrapolations from mose to man. Mtat. Res. 17: (1973). B42 Bridges, B.A. and J. Hckle. Mtagenesis of cltred mammalian cells by x-irradiation and ltraviolet light. Mtat. Res. 1: (197). 843 Broerse, J.J., S. Knaan, D.W. van Bekkm et al. Mammary carcinogenesis in rats after x- and netronirradiation and hormone administration. p in: Late Biological Effects of Ionizing Radiation, Vol. II. IAEA, Vienna, I 978. B45 Broerse. J.J. and G. W. Barendsen. Relative biological effectiveness of fast netrons for effects on normal tisses. Crr. Top. Radiat. Res. Qart. 8: (1973). 846 Broty-Boye, D. and J.B. Little. Enhancement of X ray-indced transformation in C3H-TIO-l/2 cells by interferon. Cancer Res. 37: ( 1977). B47 Brown, J.M. The shape of the dose-response crve for radiation carcinogenesis. Extrapolation to low doses. Radiat. Res. 71: 34-5 ( 1977). 848 Bres, A.H. Biological hazards and toxicity of radioactive isotopes. J. Clin. Invest. 28: (1949). B49 Bres, A. H. Critiqe of the linear theory of carcinogenesis. Science 128: ( 1958). B56 Brns, F.J., P. Strickland, M. Vanderlaan et al. Rat skin tmor incidence following single and fractionated exposres to proton radiation. Radial. Res. 74: ( 1978). B57 Brns. F.J. and M. Vanderlaan. Split-dose recovery for radiation-indced tmors in rat skin. lnt. J. Radiat. Biol. 32: 135- I 44 ( 1977). B58 Brns, F.J., R.E. Albert, J.P. Sinclair et al. The effect of a 24 hor fractionation interval on the indction of rat skin tmors by electron radiation. Radial. Res. 62: (1975). 859 Brns, F.J., R.E. Albert and R.D. Heimbach. RBE for skin tmors and hair follicle damage in the rat following irradiation with alpha particles and electrons. Radiat. Res. 36: (1968). B6 Bsch, H. (ed.) The Moleclar Biology of Cancer. Academic Press, New York, B6I Bl, P.P. van. A comparative stdy of the freqency of radiation-indced chromosome aberrations in somatic and germ cells of the rhess monkey (maccaca mlatta ). M tat. Res. 36: ( 1976). 862 Bragin, Y. N. and J.V. Filyshkin. The basis of the theory of dal action of radiation (in Rssian). Third All-Union Conference on Microdosimetry, Jne I 979. Abstracts, Moscow, B63 Borek, C. X-rav indced in vitro neoplastic transformation of h~an diploid cells. Natre 283: (198). 864 Borek, C. Neoplastic transformation following split doses of x rays. Br. J. Radio!. 5: ( 1979). B65 Borek, C., R. Miller, C. Pain et al. Conditions for inhibiting and enhancing effects of the protease inhibitor antipain on x-ray indced neoplastic transformation in hamster and mose cells. Proc. Natl. Acad. Sci. U.S.A. 76: (1979). B7I Barrett, J.C., B.D. Crawford and P.O. Ts'o. The role of somatic mtation in a mltistage model of carcinogenesis. in: Mammalian Cell Transformation by Chemical "Carcinogens". (N. Mishra et al., eds.). Senate Press, New Jersey, 198. B72 Barrett, J.C. Cell transformation, mtation and cancer. in: The Use of Mammalian Cells for Detection of Environmental Carcinogens, Mechanisms and Application. Gann. Monograph on Cancer Res. (C. Heidelbeyer et al., eds.). (in press). 873 Borek, C. Hormone and ntritional factors as reglators of X ray-indced transformation. (Meeting abstract.) Long Term Normal Tisse Effects of Cancer Treatment, Bethesda, Borek, C.. R.C. Mller, C. Geard et al. In vitro modlation of oncogenesis and differentiation by retinoids and tmor promoter. (Meeting abstract). p. I 13 in: Cocarcinogenesis and Biological Effects of Tmor Promoters. German Cancer Research Center, Heidelberg, Borek, C. Radiation oncogenesis in cell cltre. p in: Advances in Cancer Research. (G. Klein and S. Weinhose. eds.), Vol. 37, Academic Press, New York Bartlett, D.T. A review of Japanese bomb dosimetry. Radiat. Prot. Dosim. 2: (I 982). 249

88 B79 Borek, C. and E.J. Hall. Indction and modlation of radiogenic transformation in mammalian cells. p in: Radiation Carcinogenesis: Epidemiology and Biological Significance. (J.D. Boice and J.F. Frameni, eds.). Raven Press, New York, B82 Bizzozero,.J., K.G. Johnson and A. Ciocca. Radiation related lekemia in Hiroshima and Nagasaki New Engl. J. Med. 274:! (1966). B83 van Bekkrn, D.W. and P. Benovelzen. The concept of gene transfer misrepair mechanism of radiation carcinogenesis may challenge the linear extrapolation model of risk estimation for low radiation doses. Health Phvs. 43: (1982).. B84 Barlow, R.E., D.J. Bartholomew, J.M. Bremner et al. Statistical inference nder order restrictions: the theory and application of isotonic regression. John Wiley and Sons, New York, B85 Brooks. A.L., S.A. Benjamin, R.K. Jones et al. Interaction of 1 44 Ce and partial hepatectomy in the prodction of liver neoplasms in the Chinese Hamster. Radiat. Res. 91: (1982). B86 Broerse, J.J., LA. Hennen, M.J. van Zwieten et al. Mammary carcinogenesis in different rat strains after single and fractionated irradiations. p in: Netron Carcinogenesis (J.J. Broerse and G.B. Gerber, eds.) EUR-884 ( 1982). B87 Brooks, A.L., S.A. Benjamin, F.F. Hahn et al. The indction of liver tmors by 2 J 9 P) particles in the Chinese hamsters. Radiat. Res. 96: (1983). B88 Bird, R.P. Biophysical stdies with spatially correlated ions. 3. Cell srvival stdies sing diatomic deterim. Radiat. Res. 78: ( 1979). 889 Balcer-Kbiczek, E.K. and G.H. Harrison. Oncogenic transformation of C3H/!Tl/2 cells by x rays, fastfission netrons and cyclotron-prodced netrons. lnt. J. Radiat. Biol. 44: ( 1983). B9 Borek. C. and E.J. Hall. Effect of split doses of X rays on neoplastic transformation of single cells. Natre 252: I (1974 ). B91 Backer, J.M., M. Boerzig and J.B. Weinstein. When do carcinogen-treated l OT l /2 cells acqire the commitment to form transformed foci? Natre 299: (1982). B92 Broerse, J.J., LA. Hennen, M.J. van Zwieten et al. Dose-effect relations for mammary carcinogenesis in different rat strains after irradiation with X rays and mono-energetic netrons. p in: Proceedings of an IAEA/WHO Symposim on the Biological Effects of Low-Level Radiation with Special Regard to Stochastic and Non-Stochastic Effects. IAEA, Vienna, B93 Boice, J.D., N.E. Day, D. Andersen et al. Cancer risk following radiotherapy of cervical cancer: a preliminary report. p in: Radiation Carcinogenesis: Epidemiology and Biological Significance (J.D. Boice and J.F. Frameni, eds.). Raven Press, New York, B95 Brch, P.R.J. Problems with the linear-qadratic doseresponse relationship. Health Phys. 44: (1983). B96 Borek, C., W.F. Morgan, A. Ong et al. Inhibition of malignant transformation in vitro by inhibitors of poly(adp-ribose) synthesis. Proc. Natl. Acad. Sci. U.S.A. 81: (1984). B97 Borsa, J., M.D. Sargent. M. Einspenner et al. Effects of oxygen and misonidazde on cell transformation and cell killing in C3H IOTl/2 cell by x rays in vitro. Radiat. Res. 1: ( 1984). B98 Brenner D.J. and M. Zaider. Modification of the theory of dal radiation action for anenated fields. ll. Application to the analysis of soft x-ray reslts. Radiat. Res. 99: I ( 1984). B99 Bond, V.P. The conceptal basis for evalating risk from low-level radiation exposre. p in: Critical Isses in Setting Radiation Dose Limits. NCRP Proceedings No. 3 (1982). Cl C2 C7 C9 Chadwick, K.H. and H.P. Leenhots. The effect of an asynchronos poplation of cells on the initial slope of dose-effect crves. p in: Cell Srvival after Low Doses of Radiation: Theoretical and Clinical Implications. The Institte of Physics and John Wiley and Sons, London, Chadwick, K.H. and H.P. Leenhots. A moleclar theory of cell srvival. Phys. Med. Biol. 18: ( 1973). Chadwick, K.H. and H.P. Leenhots. Dose-effect relationship for malignancy in cells with different genetic characteristics. p in: Late Biological Effects of Ionizing Radiation. Vienna, IAEA, Chen, P.C., M.F. Lavin, C. Kidson et al. Identification of ataxiateleangiectasia heterozygotes, a cancer prone poplation. Natre 274: (1978). CIO Ch, E.W.Y. Mammalian cell genetics. III. Characterization of x-ray-indced forward mtations in Chinese hamster cell cltres. M tat. Res. 11: (197 I). Cl5 Clifton, K.H. and J.J. Crowley. Effects of radiation type and dose and the role of glcocorticorids gonadectomy and thyroidectomy in mammary tmor indction in mammotropin-secreting pititary tmorgrafted rats. Cancer Res. 38: (1978). Cl6 Coggle, J.E. and D.M. Peel. The relative effects of niform and non-niform external radiation on the indction of lng tmors in mice. p in: Late Biological Effects oflonizing Radiation. IAEA, Vienna, Cl8 Covelli, V., P. Metalli and B. Bassani. Decreased incidence of reticlm cell carcoma in whole bodv irradiated and bone marrow shielded mice. Br. i Cancer 31: (1975). Cl9 Covelli, V., P. Metalli, G. Briganti et al. Late somatic effects in syngeneic radiation chimaeras: II. Mortality and rate of specific diseases. Jnt. J. Radiat. Biol. 26: 1-15 (1974). C2 Cox, R., J. Thacker, D.T. Goodhead et al. Mtation and inactivation of mammalian cells bv varios ionising radiation. Natre 267: ( 1977). C21 Cox. R. and W.K. Masson. Changes in radiosensitivity dring the in vitro growth of diploid hman fibroblasts. Int. J. Radiat. Biol. 26: (1976). C22 Cox, R. and W.K. Masson. X-ray srvival crves of cltred hman diploid fibroblasts. p in: Cell Srvival after Low Doses of Radiation: Theoretical and Clinical Implications. The Institte of Physics and John Wiley and Sons, London, C23 Cox, R., J. Thacker and D.T. Goodhead. Inactivation and mtation of cltred mammalian cells bv alminim characteristic l trasoft x rays. II. Dose re;ponse of Chinese hamster and hman diploid cells to alminim x rays and radiations of different LET. lnt..j. Radiat. Biol. 31: (1977). C24 Cox, R. and W. K. Masson. Do radiation-indced thioganine-resistant mtants of cltred mammalian cells arise by HGPRT gene mtations or X-chromosome rearrangement? Natre 276: (1978). 25

89 C29 Committee on the Biological Effects of Ionizing Radiation. The Effects on Poplations of Exposre to Low Levels of Ionizing Radiation. National Academy of Sciences. Washington, D.C C32 Cairns, J. The origin of hman cancers. Natre 289: (1981). C33 Clark, E.P., G.M. Hahn and J.B. Little. Hyperthermic modlation of X-rav-indced oncogenic transformation in C3HIOTI/2 cells. Radiat. -Res. 88: ( 1981). C34 Chmelevskv, D., A.M. Kellerer. J. Lafma et al. Plmonary-neoplasms in the Sprage-Dawley rat after radon inhalation-a maximm likelihood analysis. Radiat. Res. 91: (1982). C36 Chmelevsky, D., A.M. Kellerer, J. Lafma et al. Maximm likelihood estimation of the prevalence of nonlethal neoplasms. An application to radon daghter inhalation stdies. Radiat. Res. 91: (1982). C37 Cathers, LE. and M.N. Gold. Hman mammary cell srvival following ionizing radiation. Int. J. Radiat. Biol. 44: 1-16 (1983). C38 Chmelevsky, D., A.M. Kellerer, J. Lafma et al. Comparison of the indction of plmonary neoplasms in Sprage-Dawley rats by fission netron and radon daghters. Radial. Res. 98: (1984). D3 Di Majo, V., ~- Coppola, S. Rebessi et al. Doseresponse relationship of radiation-indced Harderian gland tmors and mveloid lekaemia of the CBA/ Cne mose. J. Natl. e;;ncer Inst. 76: (1986). 4 Dolphin, G.W. A review of in vitro dose-effect relationships. p. 1-8 in: Mtagen-Indced Chromosome Damage in Man (H.J. Evans and D.C. Lloyd, eds.). Edinbrgh University Press, Edinbrgh, D7 D8 D9 Drand, R.E. and R.M. Stherland. Effects of intercelllar contact on repair of radiation damage. Exp. Cell. Res. 71: 74-8 ( 1972). Drand, R.E. and R.M. Stherland. Intercelllar contact: its inflence on the Dq of mammalian srvival crves. p in: Cell Srvival after Low Doses of Radiation: Theoretical and Clinical Implications. The Institte of Physics and John Wiley and Sons, London, Oilman, V.M. A mtation-metabolic model of cancer development and of the progression of the malignant process (in Rssian). p in: Voprosy Onkologii, Vol. 22, No. 8. D13 Di Paolo, J.A., A.J. De Marinis, C.H. Evans et al. Expression of initiated and promoted stages of irradiated carcinogenesis in vitro. Cancer Lett. 14: (1981). D 14 Darby, S.C. and J.A. Reissland. Low levels of ionizing radiation and cancer-are we nderestimating the risk? J.R. Statist. Soc. A: Part 3,144: (1981). DJS DeMott, R.K., R.T. Mlcahy and K.H. Clifton. The srvival of thyroid cells following irradiation: a directly generated single-dose srvival crve. Radial. Res. 77: ( 1979). Dl6 Doniach, I. Effects inclding carcinogenesis of ljlj and x-rays on the thyroid of experimental animals: a review. Health Phys. 9: (1963). DJ 7 Doll, R. and R. Peto. Epidemiology of cancer. p in: Oxford Textbook of Medicine (D.J. Wheatherall et al., eds.). Oxford University Press, Oxford, Day, N.E. Radiation and mltistage carcinogenesis. p in: Radiation Carcinogenesis: Epidemiology and Biological Significance. (J.D. Boice and J.F. Frameni, eds.). Raven Press, New York, Di Paolo, J.A., C.H. Evans, A.J. De Marinis et al. Inhibition of radiation-initiated and -promoted transformation of Syrian hamster embryo cells by lymphotoxin. Cancer Res. 44: (1984). E2 E3 E4 E6 ES E9 Elkind, M.M. The initial part of the srvival crve. Implications for low dose, low dose-rate radiation responses. Radiat. Res. 71: 9-23 ( 1977). Evans, R.D. The effects of skeletally deposited a-ray emitters in man. Br. J. Radio!. 39: ( 1966). Evans, R.D., A.T. Keane and M.M. Shanahan. Radiogenie effects in man of long term skeletal a-irradiation. p in: Radiobiology of Pltonim (B.J. Stover and W.S.G. Jee, eds.). J.W. Press, Salt Lake City, Evans, C.H. Lymphotoxin, an immnological hormone with anticarcinogenic and anti-tmor activity. Cancer Immnol. Immnother. 12: (1982). Evans, J.H. and Vijayalaxmi. Indction of8-azaganine resistance and sister chromatid exchange in hman lymphocytes exposed to mitomycin C and X rays in vitro. Natre 292: (1981). Elkind, M.M., A. Han and C.K. Hill. Error-free and error-prone repair in radiation-indced neoplastic cell transformation. p in: Radiation Carcinogenesis: Epidemiology and Biological Significance (J.D. Boice and J.F. Frameni, eds.). Raven Press, New York EIO Elkind, M.M. Repair processes in radiation biology. Radiat. Res. 1: ( 1984 ). FI F2 F3 F4 F5 F6 F7 FS Faber, M. Radiation carcinogenesis and the significance of some physical factors. p in: Radiation Indced Cancer. IAEA, Vienna, Fertil, B., P. Deschavanne, B. Lachet et al. Srvival crves of neoplastic and non-transformed hman cell lines: statistical analysis sing different models. p in: Proceedings of the Sixth Symposim on Microdosimetry, Brssels, May 1978 (J. Booz and H.G. Eben, eds.). Harwood Academic Pblishers. London, Fialkow, P.J. Clonal origin of hman tmors. Cancer Reviews 458: ( 1976). Fialkow, P.J. Clonal origin and stem cell evoltion of hman tmors. p in: Genetics of Hman Cancer (J.J. Mlvihill et al., eds.). Raven Press, New York, Fialkow, P.J., R. W. Sagebiel, S.M. Gastler et al. Mltiple cell origin of heriditary nerofibromas. N. Engl. J. Med. 284: (1971). Fialkow, P.J. The origin and development of hman tmors stdied with cell markers. N. Engl. J. Med. 291: (1974). Filyshkin, J. V. Microdosimetric determination of radiation qality factors. Sov. At. Energy 4: (1976). Finkel, M.P., B.O. Biskis and P.B. Jinkins. Toxicity of radim-226 in mice. p in: Radiation-Indced Cancer. IAEA, Vienna, FIO Friedman, J.M. and P.J. Fialkow. Viral tmorgenesis in man: cell markers in condylomata acminata. Int. J. Cancer 17: (1976). 251

90 Fl2 Filyshkin, I.V., U,1. Petoyan. Mathematical model of carcinogenic action of radia1ion. Radiobiologiya 24: (1984). (in Rssian). Fl3 Farber, E. and R. Cameron. The seqential analysis of cancer development. Adv. Cancer Res. 3 I: (198). Fl4 Farber, E. Chemical carcinogenesis. New Eng. J. Med. 35: (1981). FIS Fisher, P.B., R.A. Mfson, LB. Weinstein et al. Epidermal growth factor, like tmor promoters, enhances viral and radiation-indced cell transformation. Carcinogenesis 2: (1981). Fl6 Fry, R.J.M., R.D. Ley, D. Grbe e1 al. Stdies on the mltistage natre of radiation carcinogenesis. p in: Carcinogenesis, Vol. 7 (Hecker et al., eds.). Raven Press, New York, FIS Frankenberg-Schwager M., D. Frankenberg, D. Blocher et al. The linear relationship between DNA doblestrand breaks and radiation dose (3 MeV electrons) is converted into a qadratic fnction by celllar repair. lnt. J. Radial. Biol. 37: (198). FI 9 Freeman, M.L.. E. Sierra and E.J. Hall. The repair of sblethal damage in diploid hman fibroblasts: a comparison between hman and rodent cell lines. Radiat. Res. 95: ( 1983). F2 Farber, E. The mltistep natre of cancer development. Cancer Res. 44: ( 1984). F2 l Filyshkin I. V. and l.m. Petoyan. An assessment of carcinogenic action of radiation at the celllar level. Radiobiologiya 22: 78 I-786 ( I 982). (in Rssian). F22 Fry. R.J.M. Relevance of animal stdies to the hman experience. p in: Radiation Carcinogenesis: Epidemiology and Biological Significance. (J.D. Boice and J.F. Frameni, eds.). Raven Press, New York, GS G9 Gilbert, E.S. and S. Marks. An analysis of the mortality of workers in a nclear facility. Rad. Res. 79: (1979). Goodhead. D.T. Inactivation and mtation of cltred mammalian cells by alminim characteristic ltrasoft x rays. III. Implications for theory of dal radiation action. lot. J. Radiat. Biol. 32: 43-7 (1977). GlO Goodhead, D.T. and J. Thacker. Inactivation and mtation of cltred mammalian cells by alminim characteristic ltrasoft x rays. I. Properties of alminim :,; rays and preliminary experiments with Chinesehamster cells. Int. J. Radiat. Biol. 31: (1977). Gil Goodhead, D.T., J. Thacker and R. Cox. The cont1ict between the biological effects of ltrasoft x rays and microdosimetric measrements and application. p in: Proceedings of the Sixth Symposim on Microdosimetry, Brssels, May 1978 (J. Booz and H.G. Eben, eds.). Harwood Academic Pblishers, London, G 12 Gorlin, R.J. Monogenic disorders associated with neoplasia. p in: Genetics of Hman Cancer (J.J. Mlvihill et al., eds.). Raven Press, New York, G 14 Gold, M.N. and K.H. Clifton. Evidence for a niqe in sit component of the repair of radiation damage. Rad. Res. 77: (1979). G 15 Gold, M.N.. R. Jirtle, J. Crowley et al. Reevalation of the nmber of cells involved in the netron indction of mammary neoplasms. Cancer Res. 38: (1978). G 16 Gray, L.1-1. Celllar radiation biology. p. 725 in: Proceedings of the 18th Annal Symposim on Fndamental Cancer Research. University of Te:,;as and 1\1.D. Anderson Hospital. Williams and Williams, Baltimore, G 17 Groer, P.G. Dose-response crves from incomplete data. p in: Late Biological Effects of Ionizing Radiation, Vol. II. IAEA, Vienna, G 18 Groer, P.G. Dose-response crves and competing risks. Proc. Natl. Acad. Sci. U.S.A. 75: ( 1978). Gl9 Groer, P.G. Correction to dose-response crves and competing risks. Proc. Natl. Acad. Sci. U.S.A. 76: 1524 ( 1979). G23 Groer, P.G. and J.H. Marshall. A model for the indction of bone cancer by radim-224. p in: Biological Effects of radim-224 (W.A. Mller and H.G. Ebert, eds.). Martins Nijhoff Medical Division, The Hage/Boston G24 Glass, H.B. and R.K. Ritterlff. Mtagenic effect of a 5-r dose of x rays in Drosphila melanogaster. Science 133: 1366 ( 1961 ). G25 Gofman, J.W. The qestion of radiation casation of cancer in Hanford workers. Health Phys. 32: ( 1979). G26 Gernsey, D.L., C. Borek and LS. Edelman. Crcial role to thyroid hormone in X-ray-indced neoplastic transformation in cell cltre. Proc. Natl. Acad. Sci. U.S.A. 78: I ( 1981). G27 Gernsey, D.L., A. Ong and C. Borek. Thyroid hormone modlation of X-ray-indced in vitro neoplastic transformation. Natre 288: ( 198). G28 Geard, C.R., M. Rtledge-Freeman. R.C. Miller et al. Antipain and radiation effects on oncogenic transformation and sister chromatic exchanges in Syrian hamster embryo and mose C3H IOTl/2 cells. Carcinogenesis 2: I (1981 ). G29 Goodhead, D.T., R.J. Mnson, J. Thacker et al. Mtation and inactivation of cltred mammalian cells exposed to beams of accelerated heavy ions. IV. Biophysical interpretation. lnt. J. Radiat. Biol. 37: (198). G3 Gold, M.N. Radiation initiation of carcinogenesis in vivo: a rare or common celllar event. p in: Radiation Carcinogenesis Epidemiology and Biological Implications. Proceedings of a Symposim. (J.D. Boice and J.F. Frameni. eds.). Raven Press, New York, G31 Geard, C.R., R.D. Colvett and N. Rohrig. On the mechanics of chromosome aberrations: a stdv with single and mltiple spatially associated protns. ~1tat. Res. 91: (198). G32 Goodhead, D.T. An assessment of the role of microdosimetr)' in radiobiology. Radial. Res. 91: ( 1982). G33 Gerrero, I., A. Villasante, V. Corces et al. Activation of a c-k-ras oncogene by somatic mtation in mose lymphomas indced by gamma radiation. Science 225: (1984). G34 Goodhead. D.T. Dedction from celllar stdies of inactivation. mtagenesis and transformation. p in: Radiation Carcinogenesis: Epidemiology and Biological Significance. (J.D. Boice and J.F. Frameni, eds.). Raven Press, New York,

91 G35 Gragtmans, N.J., D.K. Myers, J.R. Johnson, et al. Occrrence of mammary tmors in rats after exposre to tritim /J rays and 2 kvp x rays. Radiat. Res. 99: (1984). G36 Gan, J.J. Statistical analyses of the relative risk. Environ. Health Perspect. 32: (1979). HI H2 H6 H8 H9 Han, A. and M.M. Elkind. Transformation of mose C3H IOTl/2 cells by single and fractionated doses of x-rays and fission-spectrm netrons. Cancer Res. 39: I ( 1979). Hall, E.J. Biological problems in the measrement of srvival at low doses. p in: Cell Srvival after Low Doses of Radiation: Theoretical and Clinical lmplications. The Institte of Physics and John Wiley and Sons. London, I 975. Haynes, R.H. The inflence of repair processes on radiobiological srvival crves. p in: Cell Srvival after Low Doses of Radiation: Theoretical and Clinical Implications. The Institte of Physics and John Wiley and Sons, London, Hill, R.P., B.F. Warren and R.S. Bsh. The effect of intercelllar contact on the radiation sensitivity of KHT sarcoma cells. Radiat. Res. 77: ( I 972). Hirai, M. and S. Nakai. Dicentric yields indced by x-radiation and chromosome arm nmber in primates. Mta!. Res. 43: (1977). H 1 Homsey, S. The radiation response of hman malignant melanoma cells in vitro and in vivo. Cancer Res. 32: ( 1972). HI I Hlse, E.V., R.H. Mole and D.G. Papworth. Radiosensitivity of cells from which radiation-indced skin tmors are derived. lnt. J. Radiat. Biol. 14: ( 1968). Hl2 Hlse, E.V. and R.H. Mole. Skin tmor incidence in CBA mice given fractionated exposre to low-energy P particles. Br. J. Cancer 23: ( I 969). H 13 Hlse, E. V. Incidence and pathogenesis of skin tmors in mice irradiated with single external doses of low energy P particles. Br. J. Cancer 21: (1967). H 14 Han, A. and M.M. Elkind. The effect of radiation qality and repair processes on the incidence of neoplastic transformation in vitro. p in: Radiation Research (S. Okada et al., eds.). J.A.A.R., Tokyo, H 15 Hoel, D.G. and H. E. Wal brg. Statistical analysis of srvival experiments. J. Natl. Cancer Inst. 49: (1972). Hl7 Holm, LE. Incidence of malignant thyroid tmors in man after diagnostic and therapetic doses of iodine-131. Epidemiologic and Histopatiologic stdy. Radimhemmet, Karolinska Hospital, Stockholm 198. H 18 Hempelmann, L.H., J. Pifer, G. Brke et al. Neoplasms in persons treated with x rays in infancy for thymic enlargement. A report of the third followp srvey. J. Natl. Cancer Inst. 38: (1967). H 19 Hempelmann, L.H., W. Hall, M. Phillips et al. Neoplasms in persons treated with x rays in infancy: forth srvey in 2 years. J. Natl. Cancer Inst. 55: (1975). H2 Htchinson, G.B., B. MacMahon, S. Jablon et al. Review of report by Mancso. Stewart and Kneale of radiation exposre of Hanford workers. Health Phys. 37: (1979). H21 Hall, E.J. and R.C. Miller. The how and why of in vitro oncogenic transformation. Radiat. Res. 87: (1981). H22 Han, A. and M.M. Elkind. Enhanced transformation of mose C3HIOTl/2 cells by 12--Tetradecanoylphorbol-13-acetate following exposre to x-rays or to fission-spectrm netrons. Cancer Res. 42: ( 1982). H23 Holtzman, S., J.P. Stone and C.J. Shellabarger. Radiation-indced mammary carcinogenesis in virgin, pregnant, lactating and postlactating rats. Cancer Res. 42: 5-53 ( 1982). H25 Horacek, J. and J. Sevc. Histological types of plmonary carcinoma after long-term exposre to radiation. Stdia Pnemo!. Phtiseol. Cechoslov. 4: (198). H26 Hill, C.K., F.M. Bonagro, C.P. Myers et al. Fissionspectrm netrons at redced dose-rates enhance neoplastic transformation. Natre 298: (1982). H27 Haynes, R.H. The interpretation of microbial inactivation and recovery phenomena. Radiat. Res. Spp. 6: 1-29 ( 1966). H28 Hall, E.J., H.H. Rossi, M. Zaider et al. The role of netrons in cell transformation research. II. Experimental. p in: Netron Carcinogenesis. Proceedings of a Symposim. (J.J. Broerse and G.B. Gerber, eds.). EUR-884 (1982). H29 Hoel, D.G. and H.E. Walbrg. Jr. Statistical analysis of srvival experiments. J. Natl. Cancer Inst. 49: (1972). H3 Hill, C.K., A. Han, F. Bonagro et al. Mltifractionation of cobalt-6 gamma rays redces neoplastic transformation in vitro. Carcinogenesis 5: (1984). H31 Hirakawa, T., T. Kaknaga, H. Fjiki et al. A new tmor-promoting agent, dihydroteleocidin-b, markedly enhances chemically-indced malignant cell transformation. Science 216: (1982). H32 Hendry, J.H. and B.l. Lord. The analysis of the early and late response to cytotoxic inslts in the haemopoietic cell hierarchy. in: Cytotoxic Inslt to Tisse (Plonen and Hendry, eds.). Chrchill and Livingstone, Edinbrgh, H33 Hlse, E. V., S. Y. Lewkowicz, A.L. Batchelor et al. Incidence of radiation indced skin tmors in mice and variations with dose rate. Int. J. Radiat. Biol. 44: (1983). H34 Hill, C.K., A. Han and M.M. Elkind. Protracted exposres to fission-spectrm netrons increase neoplastic transformation. in: Proceedings of the Seventh International Congress of Radiation Research. (J.J. Broerse, G. W. Barendsen et al., eds.) Manins Kijhoff Pblishers, Amsterdam, Also: lnt. J. Rad. Biol. 46: (1984). H35 Hall, T.D. The interpretation of exponential doseeffect crves. Bll. Math. Biophys. 15: ( 1953). H36 Hill, C.K., B.A. Carnes, A. Han et al. Neoplastic transformation is enhanced by mltiple low doses of fission-spectrm netrons. Radial. Res. 12: ( 1985). H37 Han, A., C.K. Hill and M.M. Elkind. Repair processes and radiation qality in neoplastic transformation of mammalian cells. Radiat. Res. 99: (1984). 11 International Commission on Radiological Protection. Recommendations of the International Commission on Radiological Protection. ICRP Pblication 9. Pergamon Press, Oxford, International Commission on Radiological Protection. Recommendations of the International Commission 253

92 on Radiological Protection. ICRP Pblication 26. Pergamon Press, Oxford, International Commission on Radiological Protection. Biological effects of inhaled radionclides. ICRP Pblication 31. Annals of the ICRP. Pergamon Press, Oxford, International Commission on Radiological Protection. Limits for Inhalation of Radon Daghters by Workers. ICRP Pblication 32. Pergamon Press, Oxford International Commission on Radiological Protection. Exposre and Jng cancer risk from indoor exposre to radon daghters. A Task Grop Report. to appear in Annals of the ICRP, International Commission on Radiological Protection. Bases for developing a nified index of harm. ICRP Pblication 45. Annals of the ICRP JI Jablon, S. and H. Kato. Stdies of the mortality of A-bomb srvivors. 5. Mortality and radiation dose, Radiat. Res. 5: (1972). 14 Jacobi, W. Carcinogenic effects of radiation on the hman respiratory tract. In: Radiation Carcinogenesis. (A.C. Upton, ed.). Elsevier North-Holland, New York, J6 J7 JS Kl K2 K3 K4 K5 K6 K7 K8 Jirtlc, R.L., P.M. Delca and M.N. Gold. The sn ival of parenchymal hepatocytes exposed to 14 t\1e V netrons. Radiat. Res. 91: (1982). Jacobi, W. Expected lng cancer risk from radon daghter exposre in dwellings. p in: Proc. lnt. Conf. on Indoor Air Qality. Stockholm. Agst Swedish Concil for Bilding Research. Stockholm Jacobi, W., H.G. Paretzke and F. Schindel. Lng cancer risk assessment of radon-exposed miners on the basis of a proponional hazard model. p in: Occpational Radiation Safety in Mining. Proceedings of the International Conference. Vol. I. (H. Stocker, ed.). Canadian Nclear Association, Toronto, I 985. Kaplan, E.L. and P. Meier. Non-parametric estimation from incomplete observations. J. Am. Stat. Assoc. 53: (1958). Kellerer, A.M. Statistical and biophysical aspects of the gamma-ray srvival crve. in: Cell Srvival after Low Doses of Radiation: Theoretical and Clinical Implications (J. Alper, ed.). The Institte of Physics and John Wiley and Sons, London, Kellerer, A.M., E.J. Hall, H.H. Rossi et al. RBE as a fnction of netron energy. II. Statistical analysis. Radiat. Res. 65: 172- I 86 ( 1976). Kellerer. A.M. and H.H. Rossi. The theory of dal radiation action. Crr. Top. Radiat. Res. Qart. 8: (I 972). Kellerer, A.M. and H. H. Rossi. A generalized formlation of dal radiation action. Radial. Res. 75: (1978). Kellerer, A.M. and H.H. Rossi. Biological implications of microdosimetry II: spatial aspects. p in: Proceedings of the Forth Symposim on Microdosimetry, Verbania-Pallanza, September EUR-5122 (1974). Kellerer, A.M. and H.H. Rossi. RBE and the primary mechanism of radiation action. Radial. Res. 47: (1971). Kellerer, A.M. Radiation carcinogenesis at low doses. p in: Proceedings of the Sixth Symposim on Microdosimetry, Vol. I, Brssels, May 1978 (J. Booz and H.G. Ebert, eds.). Harwood Academic Pblishers Ltd., London, KIO Kennedy, A.R., S. Modal. C. Heidelberger et al. Enhancement of x-ray transformation by 12--tetradecanoylphorbol-13-acetate in a cloned line of C3H mose embryo cells. Cancer Res. 38: (1978). Kl2 Klein, J.C. The se of in vitro methods for the stdy of x-ray-indced transformation. p in: Biology of Radiation Carcinogenesis (M.J. Yhas et al.. eds.). Raven Press, New York, K16 Knaap, A.G.A. and J.W.I. Simons. A mtational assay system for L5 l 78 Y mose lymphoma cells sing hypoxanthineganine phosphoribosl transferase (HG PRT) deficiency as marker. The occrrence of a long expression time for mtations indced by x rays and EMS. Mtat. Res. 3: (1975). Kl7 Kneale, G.W., A.M. Stewart and T.F. Mancso. Reanalysis of data relating to the Hanford stdy of the cancer risks of radiation workers. p in: Late Biological Effects of Ionizing Radiation, Vol. I I. IAEA, Vienna, K2 Kndson, A.G., Jr. Germinal and somatic mtations in cancer. p in: Hman Genetics (S. Amendares and R. Lisker, eds.). Excerpta Medica, Amsterdam K21 Korner, I., M. Walicka, W. Malz et al. DNA repair in two L5178 Y cell lines with different x-ray sensitivities. Std. Biophys. (GDR) 61: (1977). K22 Knz, E., J. Sevc, V. Placek et al. Lng cancer in man in relation to different time distribtion of radiation exposre. Health Phys. 36: ( 1979). K24 Kai, H.B. Effects of protracted gamma irradiation on cells from a rat rhabdomvosarcoma treated in vitro and in vivo. p i~: Cell Srvival after Low Doses of Radiation: Theoretical and Clinical Implications. The Institte of Physics and John Wiley and Sons, London, K25 Kasga, T., T. Sado. Y. Noda et al. Radiation-indced tmors in C57BLf/6JNrs [SPF] and C3Hf/HeMsNrs [SPF] strain male mice. p in: Late Biological Effects of Ionizing Radiation, Vol. II. IAEA, Vienna, K26 Kennedy, A.R., M. Fox, G. Mrphy et al. Relationship between x-ray exposre and malignant transformation in C3H I OT 1/2 cells. Proc. Natl. Acad. Sci. U.S.A. 77: ( 198). K27 Kennedy, A.R., G. Mrphy and J.B. Little. Effect of time and dration of exposre to 12--Tetradecanoylphorbol-13-acetate on X ray transformation of C3HlTl/2 cells. Cancer Res. 4: (198). K28 Kinsella, A.R. Do chromosomal rearrangements represent a rate-limiting step in carcinogenesis? (Meeting abstract). J. Spramol. Strct. Sppl. 15: 214 (1981). K29 Kinsella, A.R. and M. Radman. Tmor promoter indces sister chromatid exchanges: Relevance to mechanisms of carcinogenesis. Proc. Natl. Acad. Sci. U.S.A. 75: (1978). K3 Kennedy, A.R. and R.R. Weichselbam. Effects of dexamethasone and cortisone with X-ray irradiation on transformation of C3H!OT 1/2 cells. Natre 294: ( 198 I). K3 I Kennedy, A.R. and J.B. Little. Actinomycin D sppresses radiation transformation in vitro. Int. J. Radiat, Biol. 38: ( 198). K32 Kneale, G. W., T.F. Mancso and A.M. Stewart. Hanford radiation stdy III: a cohort stdy of the 254

93 cancer risks from radiation to workers at Hanford ( deaths) by the method of regression models in lifetables. Brit. J. lnt. Med. 38: (1981). K33 Kaplan, H.S. and M.B. Brown. A qantitative doseresponse stdy of lymphoid tmor development in irradiated C57 Blade mice. J. Natl. Cancer Inst. 13: {1952). K34 Kerr. G.D. Findings of a recent ORNL review of dosimetry for the Japanese atomic-bomb srvivors. ORNL/TM-878 (1982). K35 Kennedy, A.R. and J.B. Little. Effects of protease inhibitors on radiation transformation in vitro. Cancer Res. 41: (1981). K36 Kcerova, M., A.J.B. Anderson, K.E. Bckton et al. X-ray-indced chromosome aberrations in hman peripheral blood lecocytes: the response to low levels of exposre in vitro. lnt. J. Radial. Biol. 21: (1972). K37 Kalbfleisch, J.D. and R.L. Prentice. The statistical analysis of failre time data. John Wiley and Sons, New York, 198. K38 Kellerer, A.M. and D. Chmelevsky. Analysis of tmor rates and incidences. A Srvey of Concepts and Methods. p in: Proceedings of the Eropean Seminar on Netron Carcinogenesis. (J.J. Broerse and G.B. Gerber, eds.) EUR-884 (1982). K39 Kato, H. and W.J. Schll. Stdies of the mortality of A-bomb srvivors. 7. Mortality, : Part I. Cancer Mortality. Radiat. Res. 9: (1982). K4 Kato, H., C.C. Brown, D.G. Hoel et al. Stdies of the mortality of A-bomb srvivors. 7. Mortality, : Part II. Mortality from cases other than cancer and mortality in early entrants. Radiat. Res. 91: ( 1982). K4l Kellerer, A.M., Y.M.P. Lam and H.H. Rossi. Biophysical stdies with spatially correlated ions. 4. Analysis of cell srvival data for diatomic deterim. Radiat. Res. 83: ( 198). K42 Kato, H., K. Kopecky and D. Preston. Radiation mortality among A-bomb srvivors Technical Report. (Personal commnication) LI L3 L4 L6 L7 L8 L9 Land, C.E. and D.H. McGregor.- Breast cancer incidence among atomic bomb srvivors: implications for radiobiologic risk at low doses. J. Natl. Cancer Inst. 62: ( 1979). Leenhots, H.P. and K.H. Chadwick. An analysis of radiation-indced malignancy based on somatic mtation. lnt. J. Radiat. Biol. 33: ( 1978). Lea, D.E. Action of radiations on living cells. University Press, Cambridge, Lehnert, S. Relation of intracelllar cyclic AMP to the shape of mammalian cell srvival crves. p in: Cell Srvival after Low Doses of Radiation: Theoretical and Clinical Implications. The Institte of Physics and John Wiley and Sons, London, Leonard, H. and G.B. Gerber. The radiosensitivities of lymphocytes from pigs, sheep, goat and cow. M tat. Res. 36: ( 1976). Lever, W.F. Histopathology of the skin. Pitman Medical Pb!. Co. London/ A. Lippincott Co. Philadelphia, Liniecki, J., A. Bajerska, K. Wyszynska et al. Gamma radiation-indced chromosomal aberrations in hman lymphocytes; dose-rate effects in stimlated and nonstimlated cells. Mtat. Res, 43: (1977). LIO Little, J.B. Qantitative stdies on radiation transformation with the A31-l l mose BALB/3T3 cell line. Cancer Res. 39: (1979). Ll2 Lloyd, D.C., R.R. Prrott, G.W. Dolphin et al. Chromosome aberrations indced in hman lymphocytes by netron irradiation. lnt. J. Radiat. Biol. 29: (1976). LB Lloyd, D.C., R.J. Prrott, G. W. Dolphin et al. The relationship between chromosome aberrations and low-let radiation dose to hman lymphocytes. lnt. J. Radiat. Biol. 28: 75-9 ( 1975). L15 Lloyd, E.L., H.A. Gemmell, C.B. Henning et al. Transformation of mammalian cells by alpha particles. Int. J. Radiat. Biol. 36: ( 1979). Ll6 Lloyd, E.L., H.A. Gemmell and C.B. Henning. Spression of transformed foci indced by alpha radiation of C3HT 11/2 cells by ntransformed cells. ANL , Part II ( 1979). LIS Lndin. F.E., J.K. Wagoner and V.E. Archer. Radon daghter exposre and respiratory cancer, qantitative and temporal aspects. U.S. Department of Health, Edcation and Welfare, Pblic Health Service, National Institte of Occpational Safety and Health, National Institte of Environmental Health Sciences joint monograph no. I. Springfield, Ll9 Lndin, F.E., V.E. Archer and J.K. Wagoner. An exposre-time-response model for lng cancer mortality in ranim miners: effects of radiation exposre. age and cigarette smoking. p in: Proceedings of the Conference on Energy and Health, Alta, Utah. Jne (N.E. Breslow et al. eds.). SIAM Pblishers, Philadelphia, L2 Lz, A., W.A. Mller, W. Gossncr et al. Estimation of tmor risk at low dose from experimental reslts after incorporation of short-lived bone-seeking alpha emitters radim-224 and thorim-227 in mice. p in: Biological and Environmental Effects of Low-Level Radiation. IAEA, Vienna, L22 Land, C.E., J, D. Boice, R.E. Shore et al. Breast cancer risk from low-dose exposres to ionizing radiation: reslts of parallel analysis of three exposed poplations of women. J. Natl. Cancer Inst. 65: (198). L23 Little J.B., H. Nagasanra and A.R. Kennedy. DNA repair and malignant transformation: effect of x irradiation, I 2--tetradecanoylphorbol-13-acetate, and protease inhibitors on transformation and sisterchromatid exchanges in mose IOTl/2 cells. Radiat. Res. 79: (1979). L24 Land, C.E. and J.E. Norman. Latent periods of radiogenic cancers occrring among Japanese A-bomb srvivors. p in: Late Biological Effects of Ionizing Radiation, Vol. II. IAEA, Vienna L26 Little, J.B. Inflence of noncarcinogenic secondary factors on radiation carcinogenesis. Radiat. Res. 87: (1981). L27 Loewe. W.E. and E. Mendelsohn. Revised dose estimates at Hiroshima and Nagasaki. Health Phys. 41: (1981). L29 Lee, W., R.P. Chiaccierini, B. Shleien et al. Thyroid tmors following iodine-131 or localized x-irradiation to the thyroid and the pititary glands in rats. Radiat. Res. 92: ( 1982). L3 Loewe, W.E. A-bomb srvivor dosimetry pdate. Paper presented at the Third International Symposim on Radiation Protection, Inverness, Scotland, I

94 L32 Land, C.E. and M. Toknaga. Indction period. p in: Radiation Carcinogenesis: Epidemiology and Biological Significance. (J.D. Boice and J.F. Frameni. eds.) Raven Press. New York L33 Lin, S.L., M. Takii and P.O.P. Ts'o. Somatic mtation and neoplastic transformation indced by (methyl3h) thymidine. Radiat. Res. 9: 142- I 54 (1982). L34 Lemorte, P.K.. S.J. Adelstein and J.B. Little. Malignant transformation indced by incorporated radionclides in BALB/3T3 mose embryo fibroblasts. Proc. Natl. Acad. Sci. U.S.A. 79: (1982). L35 Little, J.B. B.N. Grossman and W.F. O'Toole. Indction of bronchial cancer in hamsters by polonirn-21 alpha radiation. Radiat. Res. 43: (!97). L36 Lloyd, D.C. and A.A. Edwards. Chromosome aberrations in hman lymphocytes: effect of radiation qality, dose and dose rate. p in: Radiation Indced Chromosome Damage in Man. Alan R. Liss Inc., New York, MI MacKenzie, I. Breast cancer following mltiple floroscopies. Br. J. Cancer 19: 1-8 (1965). M2 Major, I.R. and R.H. Mole. Myeloid lekaemia in x-ray irradiated CBA mice., Natre 272: (1978). M3 M4 M5 M6 M7 MS l\ 19 Mancso, T.F., A. Stewart and G. Kneale. Radiation exposres of Hanford workers dying from cancer and other cases. Health Phys. 33: ( 1977). Marshall, J.M. and P.G. Groer. A theory of the indction of bone cancer by alpha radiation. Radiat. Res. 71: (!977). Morrison, H.!.. R.M. Semenciw, Y. Mao et al. Lng cancer mortality and radiation exposre among the Newfondland florspar miners. p in: International Conference on Occpational Radiation Safety in Mining (H. Stocker, ed.). Canadian Nclear Association, Toronto.!985. Marx, J.L. Tmor promotion: carcinogenesis gets more complicated. Science 2 I: (! 978). Matanoski, G.N., R. Seltser, P.E. Sartwell et al. The crrent mortality rates of radiobiologists and other physician specialists: specific cases of death. Am. J. Epidemiol. 11: (1975). Mavneord, W.V. The time factor in carcinogenesis. Th~ 1977 Sievert Lectre. Health Phys. 34: (1978). Mayneord, W.V. and R.H. Clarke. Carcinogenesis and radiation risk: a biomathematical reconnaissance. Br. J. Radio!. Spplement 12: l l-12 (1975). MJO Mayneord, W.V. and R.H. Clarke. Time and dose in carcinogenesis. RD/B/N-394 (1978). Ml I Mays, C.W., G.N. Taylor. W. Stevens et al. Bone sarcomas at low doses of a-radiation in beagles. p in: C (1979). M12 Mays, \\I.C. and R.D. Lloyd. Predicted toxocity of 9Sr in hmans. p in: Proceedings of the Second International Conference on Strontim Metabolism. CONF (1972). MI3 Mays, C.W. and R.D. Lloyd. Bone sarcoma risk from 9Sr. p in: Biomedical Implications of Radiostrontim Exposre (M. Goldmann and L.K. Bstad, eds.). Symposim Series 25, CONF-7121 (1972). MI4 Mays, C.W. and R.D. Lloyd. Bone sarcoma incidence verss alpha particle dose. p in: Radiobiology of Pltonim (B. Stover and \V.S.S. Jee, eds.). J. W. Press. Salt Lake City,! 972. Ml6 McClloch, E.A. and J.E. Till. The sensitivity of cells from normal mose bone marrow to gamma-radiation in vitro and in vivo. Radiat. Res. 16: (1962). M 17 McGregor, D.H., C.E. Land and K. Choi. Breast cancer incidence among atomic bomb srvivors. Hiroshima and Nagasaki J. Natl. Cancer Inst. 59: ( 1977). M18 McNlty, P.J., C.H. Naman, A.H. Sparrow et al. Inflence of x-ray dose fractionation on the freqency of somatic mtations indced in Tradescantia stamen hair. Mtat. Res. 44: (1977). Ml9 Metalli, P., V. Covelli and G. Silini. Dose-incidence relationships of reticlm cell sarcoma in mice. Observations and hypotheses at the celllar level. p in: Late lliological Effects of Ionizing Radiation. Vol. II. IAEA, Vienna, M2 Metalli, P., G. Silini, V. Co\'elli et al. Late somatic effects in syngeneic radiation chimaeras: III. Observations on animals repoplated with irradiated marrow. lnt. J. Radiat. Biol. 29: ( l 976). M21 Millar, B.C., E.M. Fielden and J.L. Millar. Interpretation of srvival-crve data for chinese hamster cells, line V-29, sing the mltitarget, mltitarget with initial slope and a, /J eqations. Int. J. Radiat. Biol ( 1978). M22 Miller, D.R. A note on independence of mlti\ ariate lifetimes in competing risks models. Am. Stat. 5: (1977). M23 Miller, R. and E.J. Hall. X-ray dose fractionation and oncogenic transformations in cltred mose embryo cells. Natre 272: 58-6 ( 1978)..tvl25 Madan, B., D. Baidatz, H. Mast et al. Radiationindced head and neck-tmors. Lancet!: (1974). M26 Mole, R.H. Bone tmor prodction in mice by strontim-9: frther experimental spport for twoevent hypothesis. Br. J. Cancer 17: (!963). M27 Mole, R.H. Personal commnication. M28 Mole, R.H. The sensitivitv of the hman breast to cancer indction by ionizi~g radiation. Br. J. Radio!. 51: (1978). M29 Mole, R.H. Carcinogenesis by thorotrast and other sorces of irradiation, especially other a-emitters. Environ. Res. 18: (1979), M3 Mole, R.H. Letter to the editor. Lancet 1: (1978). M31 Mole, R.H. Ionizing radiation as a carcinogen: practical qestions and academic prsits. Br. J. Radio!. 48: (1975). M32 Mole, R.H. Pathological findings in mice exposed to fission netrons in the reactor CLEEP. p in: Biological Effects of Netron and Proton Irradiations. IAEA, Vienna, M33 Momeni, M.H. Competitive radiation-indced carcinogenesis: an analysis of data from beagle dogs exposed to radim-226 and strontim-9. Heailh Phys. 36: (1979). M35 Mlvihill, J.J. Genetic repertory of hman neoplasia. p in: Genetics of Hman Cancer (J.J. Mlvihill, R.W. Miller and J.F. Frameni, eds.). Raven Press. New York, M36 Mnson, R.J. and D.T. Goodhead. The relation between indced mtation freqency and an examina- 256

95 tion of experimental data for ekaryotes. M tat. Res. 42: (1977). M37 Mramats, S. and. Matsoka. Comparative stdies of radiation-indced chromosome aberrations in several mammalian species. p in: Biological and Environmental Effects of Low-Level Radiation, Vol. I. laea, Vienna, 1976.!v139 Myrden, J.A. and J.E. Hiltz. Breast cancer following mltiple floroscopies dring artificial pnemothorax treatment for plmonary tberclosis. Can. Med. Assoc. J. 1: (1969). M4 Myrden, J.A. and J.J. Qinlan. Breast carcinoma following mltiple floroscopies with pnemothorax treatment of plmonary tberclosis. Ann. Royal Coll. Phys. Srg. Can. 7: 45 ( 1974). M41 Miller, R., E.J. Hall and H.H. Rossi. Oncogenic 1ransformation of mammalian cells in vitro with split doses of x rays. Proc. Natl. Acad. Sci. U.S.A. 76: (1979). M43 Myers, D.K. and C.G. Stewart. Some health aspects of Canadian ranim mining. AECL-597 (1979). t-.145 Mole, R.H. Bone tmor prodction in mice by strontim-9: frther experimental spport for a twoevent hypothesis. Br. J. Cancer 17: ( 1963). M46 Mller, H.J. Advances in radiation mtagenesis throgh stdies on Drosophila. p in: Proc. 2nd lnt. Conf. Peacefl Uses of Atomic Energy. Vol. 22, Geneva M47 Mole, R.H. Carcinogenesis as a reslt of two independent rare events. p in: Celllar Basis and Aetiology of Late Somatic Effects of Ionizing Radiation (H.J. Harris ed.). Academic Press, London M49 Mller, R.C., R. Osman, M. Zimmerman et al. Sensitizers, protectors and oncogenic transformation in vitro. lnt. J. Radiat. Oncol. Biol. Phys. 8: ( 1982). M5 Moolgavkar, S.H. and A.G. Kndson. Mtation and cancer: A model for hman carcinogenesis. J. Natl. Cancer Inst. 66: (1981). M51 Mlcahy, R.T., M.N. Gold and K.H. Clifton. The srvival of thyroid cells: in vivo irradiation and in sit repair. Radial. Res. 84: ( 198). M52 Mole, R.H. Dose-response relationships. p in: Radiation Carcinogenesis: Epidemiology and Biological Implications. (J.D. Boice and J.F. Frameni, eds.), Raven Press, New York, M57 Mays, C.W. and H. Spiess. Bone sarcomas in 2 ~Ra patients. p in: Radiation Carcinogenesis: Epidemiology and Biological Significance. (J.D. Boice and J.F. Frameni. eds.), Raven Press, New York, M58 Mahler, P.A., M.N. Gold, P.M. De Lca et al. Rat mammary cell srvival following irradiation with 14.3 MeV netrons. Radiat. Res. 91: (1982). M6 Mahler, P.A., M.N. Gold and K.H. Clifton. Kinetics of in sit repair in rat mammary gland cells. Radiat. Res. 91: 49 only ( 1982). M61 Moskalev, Y.i., L.A. Bldakov, V.N. Jljin et al. Stdy of the dose-effect relationship from the standpoint of radiation hygiene. Med. Radiologia 4: ( 1982) (in Rssian). M62 Maisin, J.R., A. Wambersie, G.B. Gerber et al. The effects of a fractionated gamma irradiation on life shonening and disease incidence in BALB/c mice. Radiat. Res. 94: ( 1983). M63 Maisin. J.R., A. Wambersie, G.B. Gerber et al. Life shortening and disease incidence in BALB/c mice following a single d(5)-be netron or gamma exposre. Radiat. Res. 94: (1983). M64 Mole, R.H., D.G. Papworth and M.J. Corp. The doseresponse for x-ray indction of myeloid lekaemia in male CBA/H mice. Br. J. Cancer 47: (1983). M65 Mole, R.H. and I.R. Major. Myeloid lekaemia freqency after protracted exposre to ionizing radiation: experimental confirmation of the flat doseresponse fond in ankylosing spondylitis after a single treatment corse with x rays. Lekaemia Res. 7: (1983). M66 Mole, R.H. and J.A.G. Davids. Indction of myeloid lekaemia and other tmors in mice by irradiation with fission netrons. p in: Netron carcinogenesis (J.J. Broerse and G.B. Gerber, eds.). EUR-884 (1982). M67 Mller, W.A., A. Lz, E.H. Schaffer et al. The role of time-factor and R BE for the indction of osteosarcomas by incorporated short-lived bone-seekers. Health Phys. 44 Sppl. I: ( 1983). M68 Moskalev, Y.I. and V.N. Strelcova. Radiation carcinogenesis from the standpoint of radiological protection. Energoizdat, Moscow, 1982 (in Rssian). M69 Mays, C. W. Discssion of paper by O.G. Raabe [R4] in: Health Phys. 44: Sppl. I: (1983). M7 Mller, J., W.C. Wheeler, J.F. Gentleman et al. Stdy of mortality of Ontario miners. p in: Proc. Int. Conf. on Occpational Radiation Safety in Mining. Vol. I. (H. Stocker, ed.). Canadian Nclear Association, Toronto M71 Moskalev, Y.I. and I.A. Milovidova. Indction of mammary tmors in rats by single and fractionated /J irradiation from strontim-9 and ynrim-9. Radiobiologija 23: (1983). NI N2 N3 N4 N5 N6 N7 N8 N9 National Concil on Radiation Protection and Measrements. Inflence of dose and its distribtion in time on dose-effect relationships for low-let radiations. NCRP-R-64 (198). Naman, C.H., A.G. Underbrink and A.H. Sparrow. Inflence of radiation dose rate on somatic mtation indction in Tradescantia stamen hairs. Radiat. Res. 62: (1975). Nelson, W. Theory and applications of hazard plotting for censored failre data. Technometrics 14: ( 1972). Neyman, J. Pblic health hazards from electricityprodcing plants. Science 195: ( 1977). Nowell, P.C. The clonal evoltion of tmor cell poplations. Science 194: (1976). Nowell, P.C. and L.J. Cole. Hepatomas in mice: incidence increased after gamma irradiations at low dose rates. Science 148: (1965). Norris. W.P., T.W. Speckman and P.F. Gstafson. Stdies on metabolism of radim in man. Am. J. Roentgenol. Radim Ther. Ncl. Med. 73: ( 1955). Norris, G. and S.L. Hood. Some problems in the cltring and radiation sensitivity of normal hman cells. Exp. Cell Res. 27: 48 (1962). Nolibe, D., R. Masse and J. Lafma. The effect of neonatal thymectomy on lng cancers indced in rats by Pltonim Dioxide. Radiat. Res. 87: 9-99 (1981). 257

96 N 1 National Concil on Radiation Protection and Measrements. Indction of thyroid cancer by ionizing radiation. NCRP Repon No. 8 (1985). 1 Okada, S., H.B. Hamilton, N. Egami et al. A review of thirty years stdy of Hiroshima and Nagasaki atomic bomb srvivors. J. Radial. Res. (Tokyo): 16 (Sppl.): (1975). P2 P7 Parfenov Y. D. tvlathematical dose-effect models in stdies of cancerogenesis. Vopros. Onkol. 32(1): 9-22 (1986) (in Rssian). Paterson, M.C., A.K. Anderson, B.P. Smith et al. Enhanced radiosensitivity of cltred fibroblasts from ataxia teleangiectasia heterozygotes manifested by defective colony forming ability and redced DNA repair replication after hypoxic gamma-irradiation. Cancer Res. 39: ( 1979). P8 Pochin, E.E. Personal commnication ( 1986). PI I Polednak, A.P. Bone cancer among female radim dial workers. Latency periods and incidence by time after exposre: brief commnication. J. Natl. Cancer Inst. 6: ( 1978). Pl5 Pck, T.T., D. Morokovin, P. Marcs et al. Action of x rays on mammalian cells. II. Srvival crves of cells from normal hman tisse!i. J. Exp. Med. 16: (1957). P 17 Peterson, A. V., R.L. Prentice and P. Marek. Relationship between dose of injected '.:J9p and bone sarcoma mortality in yong adlt beagles. Radiat. Res. 9: (1982). Pl8 Peto, R., M.C. Pike, N.E. Day et al. Gidelines for simple, sensitive significance tests for carcinogenic effects in long-term animal experiments. p Annex to!arc Monographs on the Evalation of the Carcinogenic Risk of Chemicals to Hmans, Sppl. 2. Long-term and short-term Screening Assays for Carcinogena: A Critical Appraisal.!ARC. Lyon, 198. P 19 Peto, R. Gidelines on the analysis of rnmor rates and death rates in experimental animals. Br. J. Cancer 29: (1974). P2 Prosser, J.S., D.C. Lloyd, A.A. Edwards et al. The indction of chromosome aberrations in hman lymphocytes by exposre 1 tritiated water in vitro. Rad. Prot. Dosim. 4: (1983). P21 Papworth, D.G. and E.V. Hlse. Dose-response models for the radiation-indction of skin tmors in mice. lnt. J. Radiat. Biol. 44: (1983). P22 Petoyan, l.m. and I. V. Filyshkin. Theoretical model of radiation-indced cancer. Radiobiologiya 24: (1984) (in Rssian). P23 Rl R2 R3 R4 Petoyan, J.M. and I. V. Filyshkin. A mathematical model of the carcinogenic effect of osteotropic alpha emitters. Radiobiologiya 25: ( 1985). (in Rssian). Ramzaev, P.V., M.N. Troitzaia, A.P. Ermolaieva et al. Spra-linear effects of small doses of ionizing radiation. p in: Proc. 8th All-Union Conference on Radiation Hygiene. Leningrad, (in Rssian). Reif, A.E. Radiation carcinogenesis at high doseresponse levels: a hypothesis. Natre 19: (1961). Reissland, J.A. An assessment of the Mancso stdy. NRPB-R79 (1978). Reassessment of Atomic Bomb Radiation Dosimetry in Hiroshima and Nagasaki. Proceeding of a Workshop held at Nagasaki on Febrary Radiation Effects Research Fondation. Hiroshima, R5 R6 R7 R8 R9 Richold, M. and P.O. Holt. The effect of differing netron energies on mtagenesis in cltred Chinese hamster cells. p in: Biological Effects of Netron Irradiation. IAEA, Vienna, Reassessment of Atomic Bomb Radiation Dosimetry in Hiroshim;i and Nagasaki. Proceeding of a Workshop held at Nagasaki on 8-9 November Radiation Effects Research Fondation. Hiroshima, Robinson, V.C. and A.C. Upton. Competing-risk analysis of lekaemia and non-lekaemia mortality in x-irradiated. male RF mice. J. Natl. Cancer Inst. 6: (1978). Rose, D.M. An investigation of dependent competing risks. Unpblished dissertation, University of Washington, Rosenblatt, LS., N.H. Hetherington, M. Goldman et al. Evalation of tmor incidence following exposre to internal emitters by application of the logistic doseresponse srface. Health Phys. 21: ( I 975). Rl2 Rossi, H.H. and A.M. Kellerer. Biological implications of microdosimetry. I: temporal aspects. p in: Proceedings of the Forth Symposim on Microdosimetry, Verbania-Pallanza, September EUR (1974). Rl5 Rossi, H.H. Biophysical implications of radiation qality. p in: Radiation Research, Biomedical, Chemical and Physical Perspective (. Nygaard et al., eds.). Academic Press, New York, R 16 Rossi. H.H. Interrelation between physical and biological effects of small radiation doses. p in: Biological and Environmental Effects of Low-Level Irradiation. IAEA, Vienna, RI 7 Rossi, H.H. and A.M. Kellerer. Radiation carcinogenesis at low doses. Science 175: 2-24 (1972). Rl8 Rowland, R.E., A.F. Stehney and H.F. Lcas, Jr. Dose-response relationships for female radim dial workers. Radiat. Res. 76: (1978). Rl9 Rowland, R.E., P.M. Failla, A.T. Keane et al. Some dose-response relationships for tmor incidence in radim patients. p in: ANL-776 ( 197). R2 Rowland, R.E.. P.M. Failla, A.T. Keane et al. Tmor incidence in radim patients. p. 1-8 in: ANL-786 (1971). R21 Rowland, R.E. The risk of malignancy from internally deposited radioisotopes. p in: Radiation Research, Biomedical, Chemical and Physical Perspectives (.F. Nygaard et al., eds.). Academic Press, New York, R24 Rssell. W.L. Mtation freqencies in female mice and the estimation of genetic hazards of radiation in women. Proc. Natl. Acad. Sci. U.S.A. 74: ( 1977). R25 Rssell, W.L. Discssion of the paper by S. Abhamanson. p. 17 in: Biological and Environmental Effects of Low-Level Radiation, Vol. I. laea, Vienna, R26 Rssell, W.L. Criticism of a crrent model for estimating genetic risks of radiation. p in: Biology Division Annal Progress Report, I Jly 1974 to 3 Jne ORNL-572 ( 1975). R28 Raabe, O.G., S.A. Book and N.J. Parks. Bone cancer from radim: canine dose response explains data for mice and hmans. Science 28: (198). R29 Ron. E. and B. Modan. Benign and malignant thyroid neoplasms after childhood irradiation for tinea capitis. J. Natl. Cancer Inst. 65: 7-11 (198). 258

97 R3 Raaphorst, G.P., E.L. Azzam and J. Borsa. Enhancement by BUdR of x-ray-indced cell killing and oncogenic transformation in cltred C3H IOTl/2 cells. (abstract) Radiat. Res. 87: 498 ( 198 J ). R31 Rbin, H. ls somatic mtation the major mechanism of malignant transformation? J. Natl. Cancer Inst. 64: (198). R32 Raabe. O.G., N.J. Parks and S.A. Book. Doseresponse relationships for bone tmors in beagles exposed to 226Ra and 9Sr. Health Phys. 4: (1981). R33 Radford, E.P., R. Doll and P.G. Smith. Mortality among patients with ankylosing spondylitis not given X-ray therapy. New Engl. J. Med. 297: (1977). R34 Rowland, R., A.F. Stehney and H.F. Lcas. Doseresponse relationships for radim-indced bone sarcomas. Health Phys. 44: 15-3 J ( J 983). R35 Radford, E.P. Radon daghters in the indction of lng cancer in ndergrond miners. p in: Banbry Report 9: Qantification of Occpational Cancer. Cold Spring Harbor Laboratory, New York, R36 Radford, E.P. Radiogenic cancer in ndergrond miners. p in: Radiation Carcinogenesis: Epidemiology and Biological Significance. (J.D. Boice and J.F. Frameni, eds.), Raven Press, New York, R37 Rossi, H.H. Consideration on the time factor in radiobiology. Radiat. Environ. Biophys. 2: 1-9 ( 1981 ). R38 Rossi, H.H. Biophysical stdies with spatially correlated ions. I. Backgrond and theoretical considerations. Radiat. Res. 78: ( 1979). R39 Robenson, J.B., A. Koehler, J. George et al. Oncogenic transformation of mose BA LB/3T3 cells by pltonim-238 alpha particles. Radiat. Res. 96: (1983). R4 Raabe, O.G., S.A. Book and N.J. Parks. Lifetime bone cancer dose-response relationships in beagles and people from skeletal brdens of ::'6Ra and 9 Sr. Health Phys. 44: Sppl. I: ( 1983). R41 Radford, E.P. and K.G. St.Clair Renard. Lng cancer in Swedish iron miners exposed to low doses of radon daghters. New Engl. J. Med. 31: (1984). R42 Robertson, M. Oncogenes and mltistep carcinogenesis. Brit. Med. J. (Clin. Res.) 287: (1983). R43 Rossi, H.H. and E.J. Hall. The mlticelllar natre of radiation carcinogenesis. p in: Radiation Carcinogenesis: Epidemiology and Biological Significance. (J.D. Boice and J.F. Frameni, eds.). Raven Press, New York, R44 Ransom, J.H., C.H. Evans, A.E. Jones et al. Control of the carconogenic potential of 99mTechnetim by the immnologic hormone lymphotoxin. Cancer Immnol. Jmmnother. 15: (1983). R45 Ron, E. and B. Modan. Thyroid and other neoplasms following childhood scalp irradiation. p in: Radiation Carcinogenesis: Epidemiology and Biological Significance. (J.D. Boice and J.F. Frameni, eds.). Raven Press, New York, R46 Rowland, R.E. and H.F. Lcas, Jr. Radim-dial workers. p in: Radiation Carcinogenesis: Epidemiology and Biological Significance. (J.D. Boice and J.F. Frameni, eds.). Raven Press, New York, S2 Sanders. B.S. Low-level radiation and cancer deaths. Health Phys. 34: (1978). S3 S4 S5 S7 SS SIO Sevc, J., E. Knz and V. Placek. Lng cancer in ranim miners and long-term exposre to radon daghter prodcts. Health Phys. 3: (1976). Scott, D., H. Sharpe, A.L. Batchelor et al. Radiationindced chromosome damage in hman peripheral blood lymphocytes in vitro. I. RBE and dose rate stdies with fast netrons_ Mtat. Res. 17: ( 1969). Scott, D. and T.R.L. Bigger. The relative radiosensitivities of hman, rabbit and rat-kangaroo chromosomes. Chromosoma 49: (1974). Segaloff, A. and H.M. Pettigrew. Effect of radiation dosage on the synergism between radiation and estrogen in the prodction of mammary cancer in the rat. Cancer Res. 38: (1978). Seltser, R. and P.E. Sartwell. The inflence of occpational exposre to radiation on the mortality of American radiologists and other specialists. Am. J. Epidemiol. 81: 2-22(1965). Shellabarger, C.J. and R.W. Schmidt. Mammary neoplasia in the rat as related to dose of partial-body irradiation. Radiat. Res. 3: (1967). SI I Shellabarger, C.J., J.P- Stone and S. Holtzman. Rat differences in mammarv tmor indction \vith estrogen and netron irradiatio~. J. Natl. Cancer Inst. 61: (1978). SI2 Shellabarger, C.J. Modifying factors in rat mammary gland carcinogenesis. p in: Biology of Radiation Carcinogenesis (J.M. Yhas et al., eds.). Raven Press. New York, S13 Shellabarger, C.J., V.P. Bond, E.P. Cronkite et al. Relationship of dose of total body cobalt-6 radiation to incidence of mammary neoplasia in female rats. p in: Radiation-Indced Cancer. IAEA, Vienna, S14 Shellabarger, C.J., E.P. Cronkite, V.P. Bond et al. The occrrence of mammary tmors in the rat after sblethal whole-body irradiation. Radial. Res. 6: (1957). SI5 Shellabarger, C.J., R.D. Brown, A.R. Rao et al. Rat mammary carcinogenesis following netron- or x irradiation. p I in: Biological Effects of Netron Irradiation. IAEA, Vienna, S16 Shore, R.E., L.H. Hempelmann, E. Kowalk et al. Breast neoplasms in women treated,~ith x rays for acte post-partm mastitis. J. Natl. Cancer Inst. 59: (1977). Sl7 Sinclair, \V.K. The shape of radiation srvival crves of mammalian cells cltred in vitro. p in: Biophysical Aspects of Radiation Qality. IAEA, Technical Report Series No. 58. Vienna, I 966. SIS Spiess, H. and C.W. Mays. Protraction effect on bone sarcoma indction of radim-224 in children and adlts. p in: Radionclide Carcinogenesis (C.L. Sanders et al., eds.). U.S. Atomic Energy Commission, I 973. S19 S2 Sparrow, A.H., A.G. Underbrink and H.H. Rossi. Mtations indced in tradescantia by small doses of x-rays and netrons analysis of dose-response crves. Science 176: (1972). Sparrow, H.H., L.A. Schairer and R. Villalobos Pietrini. Comparison of somatic mtation rates indced in tradescantia by chemical and physical mtagens. M tat. Res. 26: ( 1974). 259

98 S21 S25 Spiess, H. and C.\V. Mays. Bone cancers indced by "'~Ra (Th X) in children and adlts. Health Phys. 19: (197). Szki, F. H. Hoshi and M. 1-Iorikawa. Repair of radiation-indced lethal and mtational damage in Chinese hamster cells in vitro. Jap. J. Genetics 54: (1979). S26 Stherland. R.}-1. and R.E. Drand. Cell contact as a possible contribtion to radiation resistance of some tmors. Br. J. Radio!. 45: (1972). S27 S33 Stherland, R.M. and R.E. Drand. Radiation response of mlticell spheroids: an in vitro tmor model. Crr. Top. Radial. Res. Qart. 11: (1976). Szmiel, J. Reqirements for potentiation of radiation effect by a platinm complex. lnt. J. Radiat. Biol. 33: ( 1978 ). S36 Shore, R.E., R.E. Albert and B.S. Pasternack. Followp stdy of patients treated by x-ray epilation for tinea capitis. Arch. Environ. Health 31: (1976). S37 S38 S39 Shellabarger, C.J., D. Chmelevsky and A.M. Kellerer. Indction of mammary neoplasms in the Sprage Dawley rat by 43 kev netrons and x rays. J. Natl. Cancer Inst. 64: ( 198). Shore, R.E., E.D. Woodard, B.S. Pasternack et al. Radiation and host factors in hman thyroid tmors following thyms irradiation. Health Phys. 38: ( 198). Sax, K. The rime factor in x ray prodction of chromosome aberrations. Proc. Natl. Acad. Sci. U.S.A. 25: ( 1939), S4 Sax. K. An analysis of x-ray indced chromosomal aberrations in Tradescantia. Genetics 25: ( 194). S43 S44 S45 Stras, D.S. Somatic mtation, celllar differentiation and cancer casation. J. Natl. Cancer Inst. 67: (1981). Szki, N., M. Watanabe and M. Horikowa. Stdies of radiation-indced carcinogenesis. p. 568 (Meeting abstract). Proc. of the Jap. Cancer Assoc. Tokyo, 198. Siminovitch, L. On the natre of hereditable variation in cltred somatic cells. Cell 7: 1-11 (1976). S46 Strame, T. and R.L. Dobson. Implications of new Hiroshima and Nagasaki dose estimates: cancer risks and netron RBE. Health Phys. 41: (1981). S48 Sasaki, S. and T. Kasga. Life-shortening and carcinogenesis in mice irradiated neonatally with x rays. Radial. Res. 88: ( 1981 ). S49 Smith, P.G. and R. Doll. Mortality among patients with ankylosing spondylitis after a single treatment corse with X rays. Brit. Med. J. 284: ( 1982). S5 Schlenker, R.A. Risk estimates for bone. p in: Critical Isses in Sening Radiation Dose Limits. Proceedings of the Seventh Annal Meeting of the NCRP. NCRP Proceedings 3 ( 1982). SS! Sevc, J., V. Placek and P. Vernerova. Malignant rmor in lngs and inhalation radiation exposre. Prac. Lek. 34: (1982). (in Czech). S52 Shellabarger, C.J., D. Chmelcvsky, A.M. Kellerer et al. Indction of mammary neoplasms in the ACI rat by 43-keV netrons, X-rays and diethylstilbestiol. J. Natl. Cancer Inst. 69: (1982). S53 Storer, J.B. Associations between tmor types in irradiated BALB/c female mice. Radiat. Res. 92: (1982). S54 Sanders, C.L., G.E. Dagle, W.C. Cannon et al. Inhalation carcinogenesis of high-fired 238 P2 in rats. Radial. Res. 71: ( 1977). S55 Scarpelli, D.G. Recent developments toward a nifying concept of carcinogenesis. Ann. Clin. Lab. Sci. 13: (I 983). S56 Sandberg, A.A. A chromosomal hypothesis of oncogenesis. Cancer Cytogenet. 8: ( 1983). S57 Storer, J.B.. and R.L. Ullrich. Life shortening in BALB/c mice following brief, protracted or fractionated exposres to netrons. Radiat. Res. 96: ( 1983). S58 S59 S6 S6 I Saccomano, G., V.E. Archer,. Aerbach et al. Age factor in histological type of lng cancer among ranim miners. A preliminary report. p in: International Conference on Radiation Hazards in Mining: Control, Measrements and Medical Aspects. (M. Gomez, ed.). Soc. of Mining Engineers of American Institte of Mining, Metallrgical and Petrolem Engineers, New York, Schll, W.J. Atomic bomb srvivors: patterns of cancer risk. p in: Radiation Carcinogenesis: Epidemiology and Biological Significance. (J.D. Boice and J.F. Frameni, eds.). Raven Press, New York, Shore, R.E., E. Woodard, N. Hildreth et al. Thyroid tmors following thyms irradiation. J. Natl. Cancer Inst. 74: (1985). Saccomano, G. Cancer of the lng in ranim miners. p in: International Conference on Radiation Hazards in Mining (M. Gomez, ed.). Society of Mining Engineers of American Institte of Mining. Metallrgical and Petrolem Engineers, New York, S62 Storer, J.B. and T.J. Mitchell. Limiting vales for the RBE of fission netrons at low doses for life shortening in mice. Radial. Res. 97: ( 1984). T3 Thorn, M. Personal commnication (1986). T4 T5 T6 T7 TB T9 Terzaghi, M. and J.B. Little. Oncogenic transformation in vitro after split-dose x-irradiation. Int. J. Radial. Biol. 29: ( 1976). Terzaghi, M. and J.B. Little. Repair of potentially lethal radiation damage in mammalian cells is associated with enhancement of malignant transformation. Natre 253: ( 1975). Terzaghi, M. and J.B. Little. Oncogenic transformation in vitro by x rays: inflence of repair processes. p in: Biology of Radiation Carcinogenesis (J.M. Yhas et al., eds.). Raven Press, New York, Terzaghi, M. and J.B. Little. X-irradiation-indced transformation in a C3H mose embryo-derived cell line. Cancer Res. 36: (1976). Thacker, J. The involvement of repair processes in radiation-indced mtation of cltred mammalian cells. p in: Proceedings of the 6th International Congress of Radiation Research, Tokyo, May (S. Okada et al., eds.). Jap. Assoc. for Radiation Research, Tokyo, Thacker, J.. A. Stretch and M.A. Stephens. The indction of thoganine resistant mtations of chincse hamster cells by gamma rays. Mtat. Res. 42: ( 1977). TIO Thacker, J. and R. Cox. Mtation indction and inactivation in mammalian cells exposed to ionizing radiation. Natre 258: (1975). 26

99 TI I Terasima, T., M. Yazkawa and M. Kimra. Radiationindced transformation of lotj/2 mose cells in the platea phase: post-irradiation changes and serm dependence. Gann. 72: (1981). Tl2 Toknaga, M.. J.E. Norman, Jr., M. Asano et al. Malignant breast tmors among atomic bomb srvivors, Hiroshima and Nagasaki, J. Natl. Cancer Inst. 62: (1979). T15 Trosko, J.E. and C.C. Chang. Environmental carcinogenesis: an integrative model. Qart. Review of Biol. 53: (1978). TI 7 Tanooka, H. and K. Tanaka. Evidence for single-cell origin of 3-methyl-cholantrene-indced fibrosarcomas in mice with celllar mosaicism. Cancer Res. 42: (1982). TI8 Thacker, J., A. Stretch and D.T. Goodhead. The mtagenecity of a-particles from pltonim-238. Radiat. Res. 92: (1982). TJ9 Thacker, J. and A. Stretch. Recovery from lethal and mtagenic damage dring post-irradiation holding and low dose-rate irradiations of cltred hamster cells. Radiat. Res. 96: (1983). T2 Thomson, J.F., F.S. Williamson, D. Grahn et al. Life shortening in mice exposed to fission netrons and gamma rays. I. Single and short-term fractionated exposre. Radiat. Res. 86: (1981). T21 Thomson, J.F., F.S. Williamson, D. Grahn et al. Life shortening of mice exposed to fission netrons and gamma rays. II. Dration-of-life and long-term fractionated exposres. Radial. Res. 86: ( 1981 ). T22 Taylor, G.N., C.W. Mays, R.D. Lloyd et al. Comparative toxicity of Ra-226, P-239, Am-241, Cf-249 and Cf-252 in C57BI/Do black and albino mice. Radiat. Res. 95: (1983). T23 Toknaga, M.. C.E. Land, T. Yamamoto et al. Breast cancer among atomic bomb srvivors. p in: Radiation Carcinogenesis: Epidemiology and Biological Significance. (J.D. Boice and J.F. Frameni, eds.). Raven Press, New York, T24 Terasima, T., M. Yaskawa and M. Kimra. Neoplastic transformation of platea-phase mose IOTl/2 cells following single and fractionated doses of x rays. Radiat. Res. 12: (1985). T25 Thomson, J.F., F.S. Williamson and D. Grahn. Life shortening in mice exposed to fission netrons and gamma rays III. Netron exposres of 5 and 1 rads. Radiat. Res. 93: ( 1983). T26 Thomson, J.F., F.S. Williamson and D. Grahn. Life shortening in mice exposed to fission netrons and r rays. V. Frther stdies with single low doses. Radiat. Res. 14: (1985). U2 U3 U4 US U6 Ullrich, R.L. and J.B. Storer. Inflence of gamma-ray irradiation on the development of neoplastic disease. II. Solid tmors. Radiat. Res. 8: ( 1979). Ullrich, R.L. and J.B. Storer. Inflence of gamma-ray irradiation on the development of neoplastic disease. III. Dose rate effects. Radiat. Res. 8: ( 1979). Ullrich, R.L., M.C. Jernigan and J.B. Storer. Netron carcinogenesis. Dose and dose-rate effects in BALB/c mice. Radiat. Res. 72: ( 1977). Ullrich, R.L., M.C. Jernigan. G.E. Cosgrove et al. The inflence of dose and dose-rate on the incidence of neoplastic disease in RFM mice after netron irradiation. Radiat. Res. 68: (1976). United Nations. Sorces and Effects of Ionizing Radiation. United Nations Scientific Committee on the U7 U9 U JO Effects of Atomic Radiation 1977 report to the General Assemblv, with annexes. United Nations sales pblication No. E.77.IX.l. New York, United Nations. Report of the United Nations Scientific Committee on the Effects of Atomic Radiation to the General Assembly, with annexes. Volme I: Levels. Volme II: Effects. United Nations sales pblication No. E.72.IX.17 and 18. New York, United Nations. Report of the United Nations Scientific Com mi nee on the Effects of Atomic Radiation. Official Records of the General Assembly, Twentyforth Session, Spplement No. 13 (A/7613). New York, United Nations. Report of the United Nations Scientific Committee on the Effects of Atomic Radiation. Official Records of the General Assembly, Twenty-first Session. Spplement No. 14 (A/6314). New York, Ul I United Nations. Report of the United Nations Scientific Committee on the Effects of Atomic Radiation. Official Records of the General Assembly, Nineteenth Session, Spplement No. 14(A/5814). New York, Ul2 United Nations. Report of the United Nations Scientific Committee on the Effects of Atomic Radiation. Official Records of the General Assembly. Seventeenth Session, Spplement No. 16 (A/5216). New York, Ul3 Underbrink, A.G. and A.H. Sparrow. The inflence of experimental endpoints, dose, dose rate, netron energy, nitrogen ions, hypoxia chromosome volme and ploidy level on RBE. p in: Biological Effects of Netron Irradiation. IAEA. Vienna, Ul4 Upton, A.C., V.K. Jenkins, H.E. Walbrg Jr. et al. Observations on, iral, chemical and radiation-indced myeloid and lymphoid lekaemia in RF mice. Natl. Cancer Inst. Monogr. 22: (1966). U 15 Upton, A.C., F.F. Wolff, J. Frth et al. A comparison of the indction of myeloid lekaemias in x-irradiated RF mice. Cancer Res. 18: (1958). UI6 Upton, A.C., M.L. Randolph and J.W. Conklin. Late effects of fast netrons and gamma rays in mice as inflenced by the dose rate of irradiation: indction of neoplasia. Radial. Res. 41: (197). U 17 Upton, A.C. Radiobiological effects of low doses. Implications for radiological protection. Radiat. Res. 71: (1977). U 18 Upton. A.C. The interplay of virses and radiation in carcinogenesis. p in: Radiation Research. Biomedical, Chemical and Physical Perspectives (. Nygaard et al., eds.). Academic Press, New York, U 19 Underbrink, A.G., A.M. Kellerer, R.E. Mills et al. Comparison of x-ray and gamma-ray dose response crves for pink somatic mtations in tradescantia clone 2. Radial. and Environ. Biophysics 13: 295 ff ( 1976). U2 Ullrich, R.L. and J.B. Storer. Inflence of gammairradiation on the development of neoplastic disease in mice. I. Reticlar tisse tmors. Radiat. Res. 8: (1979). U21 Ullrich, R.L., M.C. Jernigan and L.M. Adams. Indction of lng tmors in RFM mice after localized exposres to x rays or netrons. Radiat. Res. 8: (1979). U22 Ullrich, R.L. Effects of split doses of x rays on netrons on lng tmor formation in RFM mice. Radiat. Res. 83: (198). 261

100 U23 Ullrich, R.L. Tmor indction in BALB/c mice after fractionated or protracted exposres to fission-spectrm netrons. Radiat. Res. 97: ( 1984). U24 United Nations. United Nations Scientific Commit1ee on the Effects of A1omic Radiation 1982 Report to the General Assembly, with annexes. Ionizing Radiation: Sorces and Biological Effects. United Naiions sales pblication No. E.82.IX.8. New York, U25 Ullrich, R.L. Tmor indction in BALB/c female mice after fission netron or gamma irradia1ion. Radiat. Res. 93: (1983). U26 Ullrich, R.L. Tmor indction in BALB/c mice after fractionated or protracied exposres to fission-spectrm netrons. Radial. Res. 97: (1984). U27 Underbrink, A.G., F.M. Edwards, W.R. Lower e1 al. Absence of detectable fractionation effects in tradescantia somatic mtations for an x ray dose of 5 rad. Radiat. Res. IOI: (1985). U28 Ullrich, R.L. The rate of progression of radiationtransformed mammary epithelial cells is enhanced after low dose rate netron irradiation. Radiat. Res. 15: (1986). U29 Umeda. ~I. K. Tanaka and T. Ono. Effect of inslin on the transformation of BALB/3T3 cells by x irradiation. Gann 74: ( 1983). VI V2 Vanderlaan, M.. F.J. Brns and R.E. Albert. A model describing the effects of dose and dose-rate on tmor indction by radiation in rat skin. p in: Biological and Environmental Effects of Low-Level Radiation. Vol. II. IAEA, Vienna, Vilenchik, M.M. T.M. Tretyak, V.M. Lobachev et al. The similarity of circlar dichorism spectrm of DNA of aged animals or irradiated by ; rays. Dokl. Akadcmii Natl. SSSR 259: (1981). (in Rssian) \13 Vilenchik. M.M. Spontaneos instability and plasticity of DNA in vivo: recombination between the nclear and mitochondrial DNA and its biological imponance. Usp. Sovr. Biologii 99(2): ( 1985). V4 V5 V6 V9 Vogel. H.H. and R. Zaldivar. Netron-indced mammary neoplasms in the rat. Cancer Res. 32: ( 1972). Vogel. H.H. Jr. High-LET irradiation of Sprage Dawley female rats and mammary neoplasm indction. p in: Late Biological Effects of Ionizing Radiation, Vol. II. IAEA, Vienna, Vilenchik. M.M. Modification of carcinogenic and antilmor radiation effects (biomedical aspects). Medicina. Moscow (in Rssian) Vogel. H.H. Jr. and 1-1.W. Dickson. Mammary neoplasia following acte and protracted irradiation with fission netrons and 6 Co gamma-rays. (Abstract). Radiat. Res. 87: (1981 ). VJO Vogel. 1-1.H. Jr. and J.E. Trner. Genetic component in rat mammary carcinogenesis. Radiat. Res. 89: ( 1982). VJ I \'lpis, N., G. Panetta and L. Tognacci. Radiationindced chromosome aberrations in radiological protection. Dose-response crves at low dose-levels. lnt. J. Radiat. Biol. 29: ( 1976). W6 Weichselbam. R.R. J. Epstein, J.B. Little et al. In vitro celllar radiosensitivi1y of hman malignant tmors. Er. J. Cancer 12: (1976). W7 W8 W9 Wcichselbam, R., J. Epstein and J.B. Little. In vitro radiosensiti\'ity of hman diploid fibroblasts derived from patients with nsal clinical responses to radiation. Radiology 121: (1976). Whittemore, A.S. Qantitative theories of oncogenesis. Adv. Cancer Res. 17: (1978). World Health Organization. Cancer incidence in five continents. IARC. Lyon, W 12 Walinder, G.. C, Y. Jonsson and A.M. Sjoden. Doserate dependence of the goitrogen stimlated mose thyroid. A comparative investigation of the effects of roentgen, DIJ and 13JJ irradiation. Acta Radio!., Ther. 11: (1972). Wl3 Waxweiler, R..I., R..1. Roscoe, V.E. Archer el al. Mortality follow-p throgh 1977 of the white ndergrond ranim miners cohort examined by 1hc United States Pblic Health Service. p in: Proc. lnt. ConL on Radiation Hazards in Mining. (M. Gomez, ed.), Society of Mining Engineers, New York, \VJ4 Whittemore, A.S. and A. McMillan. Osteosarcomas among beagles exposed to P-239. Radiat, Res. 9: 41-56(1982). Wi5 Wesch, A., G. \'an Kaick, W. Riedel et al. Recent reslts of the German Thorotrast Stdy-statistical e\'alation of animal experiments with regard 1 the non-radiation effects in hman 1horotrastosis. Health Phys. 44. Sppl. I: (1983). WJ6 Wach, B., A.G. Underbrink, J. Hczkowski et al. Effects of dose fractionation in Tradescantia stamen hairs after high and imcrmcdiate doses of x-irradiation. Radial. Res. 9: ( 1983). Wl7 Whittemore, A.S. and A. McMillan. Lng cancer mortality among U.S. ranim miners: a reappraisal. J. Natl. Cancer Inst. 71: ( 1983). \VIS Watanabe, M., M. Horikawa and. Nikaido. Indction of oncogenic transformation by low doses of x rays and dose-rate effect. Radial. Res. 98: (1984). WJ9 Wrenn, M.E. G.N. Tavlor, W. Ste\'ens et al. Smmary of dosimetry, pathology and dose-response for bone sarcomas in beagles injected with 226 Ra. p in: Research in Radiobiology. University of Utah, School of Medicine, Radiobiology Division. Annal Report C Salt Lake City, Utah, YI Y2 Y3 Y5 Z I Yhas. J.M. Dose-response crves and their modification by specific mechanisms. p in: Biology of Radiation Carcinogenesis (J.M. Yhas et al.. eds.), Raven Press. New York, Yhas..J.M. Reco\'ery from radiation-carcinogenic injry to the mose O\'ary. Radial. Res. 6: (1974). Yang, T.C.1-1. and C.A. Tobias. Radiation and cell transformation in \'itro. p in: Ad\'ances in Biological and Medical Physics, Vol. 17, 198. Yang, T.C., J. Howard, L. Craise et al. Effects of energetic silicon ions. UV radiation and X rays on neoplastic cell tranformation and mtation. Radiat. Res. 91: (1982). Ziemba-Zoltowska,.. E. Bocian,. Rosiek et al. Chromosome aberrations indced bv low doses of X rays in hman lymphocytes in vitr~. lnt. J. Radiat. Biol. 37: ( 198). 262 I