BIOLOGICAL EFFECTIVENESS OF LOW-ENERGY PHOTONS AND ELECTRONS FOR EVALUATING HUMAN CANCER RISK

1 NCRP DRAFT SC 1-20 REPORT Council Draft (12-5-16) BIOLOGICAL EFFECTIVENESS OF LOW-ENERGY PHOTONS AND ELECTRONS FOR EVALUATING HUMAN CANCER RISK December 2016 Note: Copyright permission is being sought for the figures and tables requiring such permission prior to their use in the final NCRP publication. National Council on Radiation Protection and Measurements 7910 Woodmont Avenue, Suite 400, Bethesda, Maryland 20814 2 1

3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 Preface The dependence of biological effectiveness on energy is an unresolved question in evaluating the risk of human cancer from exposure to low linear-energy transfer (LET) radiation (i.e., photons and electrons). This dependence is relevant for estimating the level of cancer risk from exposure to low-let radiation at lower energies in mammography, other medical imaging procedures, and various other occupational and public radiation exposures. The National Academies/National Research Council (NA/NRC, 2006) indicated that the biological effectiveness of lower-energy low-let radiation based on chromosomal aberration data and biophysical considerations may be two or more times greater than for higher-energy low-let radiation. However, the biological systems used in the experiments and the biophysical analysis provide only indirect evidence and may not be strictly applicable to human cancer. Therefore, the assessment in this Report was undertaken. This Report draws on evaluation by specialists in microdosimetry, deoxyribonucleic acid (DNA) damage, cellular radiobiology, animal studies, and human epidemiology of the available evidence relevant to the variation in biological effectiveness regarding the risk of human cancer for low-let radiation at lower energies. Probability density functions (PDFs) were derived for the biological effectiveness observed for the endpoints studied in each specialty area (line of evidence) for defined lower-energy groups. Using these PDFs and evaluation of the relevance of the data from each line of evidence to the risk of cancer in humans, guidance is provided on the biological effectiveness regarding the risk of human cancer of low-let radiation for the defined lower-energy groups. This Report was prepared by Scientific Committee 1-20 on the Biological Effectiveness of Low-LET Radiation as a Function of Energy. Serving on Scientific Committee 1-20 were: Steven L. Simon, Chairman National Cancer Institute Bethesda, Maryland 2

37 NCRP SC 1-20 Leslie A. Braby Texas A&M University College Station. Texas Dudley T. Goodhead Medical Research Council Oxford, United Kingdom David C. Kocher Oak Ridge Center for Risk Analysis Oak Ridge, Tennessee Jerome S. Puskin U.S. Environmental Protection Agency Washington, D.C. Members Polly Y. Chang SRI International Menlo Park, California Stephen C. Hora Center for Risk and Economic Analysis of Terrorism Events Los Angeles, California Kiyohiko Mabuchi National Cancer Institute Bethesda, Maryland David B. Richardson University of North Carolina Chapel Hill, North Carolina 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 James D. Tucker Wayne State University Detroit, Michigan Eliseo Vano Complutense University Madrid, Spain NCRP Secretariat Marvin Rosenstein, Staff Consultant Cindy L. O Brien, Managing Editor Laura J. Atwell, Office Manager James R. Cassata, Executive Director (2012 2014) David A. Smith, Executive Director (2014 2016) Kathryn D. Held, Executive Director and Chief Science Officer (2016 ) The Council expresses appreciation to the Committee members for the time and effort devoted to the preparation of this Report, and to the Centers for Disease Control and Prevention for the financial support provided during its preparation. John D. Boice, Jr. President 3

57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 CONTENTS Preface... 2 1. Executive Summary... 8 1.1 Overview... 8 1.2 Microdosimetry... 11 1.3 Deoxyribonucleic Acid Damage... 12 1.4 Cellular Radiobiology and Animal Studies... 14 1.5 Human Epidemiology... 15 1.6 Analysis of Research Threads... 16 1.7 Findings... 18 1.8 Conclusions and Recommendations... 22 2. Introduction... 24 2.1 Background for Present Study... 25 2.1.1 Recommendations for Purposes of Radiation Protection... 25 2.1.2 Assumptions for Purposes of Cancer Risk Assessment... 27 2.1.3 Previous Reviews by NCRP... 29 2.2 Importance of Present Evaluation... 30 2.3 Term to Describe Modifying Factor to Represent Biological Effectiveness... 32 2.4 Specification of Reference Radiation... 34 2.5 Use of Effectiveness Ratio in Cancer Risk Assessments... 36 2.6 Approach to Evaluation of Biological Effectiveness... 38 3. Spectral Characteristics of Representative Low-LET Radiations... 43 3.1 Production of Energetic Secondary Electrons by Incident Photons... 44 3.2 Representative Spectra of Incident Photons... 48 3.3 Spectra of First-Collision Electrons and Tritium Beta Particles... 53 3.4 Spectra of Lower-Energy Electrons Produced by First-Collision Electrons... 59 4. Line of Evidence: Microdosimetry... 63 4.1 Lineal Energy Distributions Produced by Photons... 67 4.2 Prediction of R i Based on f(y)... 76 4.3 Evaluation of the PDF of R i... 84 4

92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 5. Line of Evidence: Deoxyribonucleic Acid Damage... 91 5.1 Deoxyribonucleic Acid Damage from Ionizing Radiation... 91 5.2 Photon and Electron Damage to Deoxyribonucleic Acid... 94 5.3 Experimental Data on RBE for DNA Double-Strand Breaks... 99 5.4 RBE of DNA Double-Strand Breaks from Theoretical Simulations... 107 5.5 Enhancement of RBE for Slow-Rejoining Double-Strand Breaks... 113 5.6 Enhancement of RBE for Complex Double-Strand Breaks from Simulations... 118 5.7 Role of DNA Base Damage... 123 5.8 Estimation of Probability Density Functions for R i of Double-Strand Breaks and also with Enhancement for Biological Severity... 125 5.8.1 PDFs of R i for 15 to 30 kev Photons... 125 5.8.2 PDFs of R i for 1.5 kev Photons... 127 5.8.3 PDF of R i for 40 to 60 kev Photons... 130 5.8.4 PDF of R i for >60 to150 kev Photons... 132 5.8.5 PDF of R i for Tritium Beta Particles... 132 5.8.6 Summary of Recommended PDFs... 135 5.8.7 Relevance of R i for Initial DNA Double-Strand Breaks... 135 6. Line of Evidence: Cellular Radiobiology and Animal Studies... 137 6.1. Introduction... 137 6.2 Structural Chromosome Aberrations... 137 6.2.1 Studies Using Conventional Giemsa Staining... 137 6.2.2 Difficulties with Studies Using Giemsa Staining... 146 6.2.3 Studies Using FISH and mfish... 147 6.3 Micronuclei... 151 6.4 Cell Survival and Cell Killing... 154 6.5 Cell Mutation... 159 6.6 Cellular Transformation... 162 6.7 Effects in Cells from Low-Energy Radionuclide Emissions... 162 6.8 Other Considerations of Cellular Effects... 170 5

123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 6.9 Studies of Effects in Whole Animals... 170 6.9.1 Studies with External Radiation... 171 6.9.2 Studies with Radionuclides Incorporated in Animal Tissues... 171 6.10 Development of Probability Density Functions of R i... 174 6.10.1 Photons of Energy 1.5 kev (Al K )... 174 6.10.2 Photons of Energy 15 to 30 kev... 176 6.10.3 Photons of Energy 40 to 60 kev or >60 to 150 kev... 176 6.10.4 Tritium Beta Particles... 179 6.11 Summary... 179 7. Line of Evidence: Human Epidemiology... 182 7.1 Introduction... 182 7.2 Discussion of Epidemiologic Methods... 182 7.2.1 Representations of Excess Cancer Risks in Radiation Epidemiologic Studies... 183 7.2.2 Transfer of Risks between Populations of Different Nationalities... 184 7.2.3 Comparisons of Risks from Different Types of Exposures... 186 7.2.4 Other Complications in Comparing Estimated Risks in Different Populations... 186 7.3 Limitations of Available Epidemiologic Studies... 188 7.4 Comparisons of Results of Epidemiologic Studies... 190 7.4.1 Studies of All Cancers... 190 7.4.2 Studies of Leukemia Excluding Chronic Lymphocytic Leukemia... 192 7.4.3 Studies of Lung Cancer... 195 7.4.4 Studies of Breast Cancer... 197 7.4.5 Studies of Thyroid Cancer... 203 7.5 Summary and Conclusions... 209 7.6 Proposed Subjective PDF of R i for Lower-Energy Photons... 212 8. Elicitation of Judgments and Processing of PDFs of the Effectiveness Ratio... 214 8.1 Probability Elicitation Strategies... 216 8.2 Holistic Assessment... 217 8.3 Decomposition Modeling... 217 8.4 An Alternative Formulation for Decomposition Modeling... 223 8.5 Decomposed Probability Assessment... 223 8.6 Computation... 225 8.7 Sensitivity Analysis of Reasearch Thread Contributions... 226 6

158 159 160 161 162 163 164 165 166 167 168 169 170 171 8.8 Discussion... 226 9. Conclusions and Recommendations... 233 9.1 Conclusions... 233 9.2 Research Recommendations... 238 Appendix A: Input Data from the Expert Elicitations... 239 Glossary... 247 Abbreviations, Acronyms and Symbols... 257 References... 258 7

172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 1. Executive Summary 1.1 Overview This Report presents a review and evaluation of the present state-of-knowledge on the biological effectiveness of lower-energy photons and electrons compared with higher-energy photons and electrons for their potential to cause radiation damage that may lead to induction of cancer in humans. Considerable data have been collected for many years in various kinds of nonhuman biological systems suggesting that lower-energy photons and electrons may have higher biological effectiveness, so it has been recognized for several decades that there is a need to better understand the relative effectiveness for lower-energy photons and electrons in inducing cancer in humans. In this Report, a set of literature-based but subjectively-derived state-of-knowledge density functions are provided for a quantity named here effectiveness ratio ( ) and given the symbol L (see Glossary) when derived for five defined lower-energy groups (L) of photons or electrons. Here and elsewhere in this Report, the derived density functions are composites of alternative values representing the degree-of-belief of experts but are simplistically referred to as probability density functions (PDFs). The effectiveness ratio represents the effectiveness of lower-energy photons or electrons, relative to a reference radiation, for their potential to induce cancer in humans at low absorbed doses or low absorbed-dose rates 1 (referred to in this Report as low doses or low dose rates). The derived PDFs of ρ L s provided in this Report were developed through an evaluation of the available scientific evidence about the unknown true value of ρ L. The PDFs of ρ L s are intended only for use in human cancer risk assessments to account for the presence of uncertainty. The ρ L s are not to be confused with, equated to, or applied as either the: 1 For the purpose of this Report, for low linear-energy transfer (LET) radiation, a low absorbed dose is <100 mgy delivered acutely, and a low absorbed-dose rate is <5 mgy h 1 for any accumulated absorbed dose [as adopted at the present time by NCRP (2015)]. 8

200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 (1) formal scientific concept of relative biological effectiveness (RBE), as that term is rigorously defined for controlled experimental observations (see Glossary), or (2) radiation weighting factors (w R values) (see Glossary) that are used in the practical radiation protection system to determine the quantity equivalent dose (see Glossary). The subjectively-derived PDFs of ρ L s should be useful in evaluating cancer risks, either in populations or for individuals, resulting from relatively common-place radiation exposures where lower-energy photons or electrons are involved with statements of uncertainty. The extensive use of computed tomography (CT) and mammography in modern medicine and the presence of 3 H (tritium) in the environment are examples of exposure to lower-energy low linear energy transfer (low- LET) radiation. In evaluating information on the biological effectiveness of photons and electrons of various energies, it is useful to recognize that absorbed doses from irradiation by photons are due almost entirely to the energy deposited by secondary electrons that are generated by interactions of photons in tissue. Thus, information on the biological effectiveness of photons of a given energy is indicative of the biological effectiveness of the energy spectrum of electrons produced by those photons. This Report evaluated evidence for the biological effectiveness of lower-energy photons and electrons for five combinations of low-let radiation type and energies (referred to in this Report as lower-energy groups), namely: photons of energy about 1.5 kev; photons of energy in the range of 15 to 30 kev; photons of energy in the range of 40 to 60 kev; photons of energy in the range of >60 to 150 kev; and electrons produced in beta-particle decay of tritium. The reference radiation, with a defined biological effectiveness of unity, is assumed to be photons of energy in the range of 0.5 to 2 MeV or high-energy electrons of equivalent biological 9

231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 effectiveness. The lower-energy groups of low-let radiation considered in this Report were selected on the basis of the availability of information on biological effectiveness from radiobiologic and epidemiologic studies or microdosimetric calculations and the relevance of the lower-energy groups to common radiological exposures. The lowest photon energy group of about 1.5 kev is representative of ultra-soft x rays of energy less than about 5 kev which have been widely studied both experimentally and theoretically. Those studies suggest that the biological effectiveness of photons may be highest near 1.5 kev and that the dependence on energy may be most pronounced at the lowest energies. The photon-energy ranges of 15 to 30 kev and 40 to 60 kev are representative of spectra of x rays that are used in mammography and CT, respectively, and the highest energy range of >60 to 150 kev is representative of spectra of orthovoltage x rays that were used in radiology and sometimes as reference radiations in radiobiologic studies. The spectrum of low-energy electrons from beta-particle decay of tritium has a maximum and average energy of 18.6 kev and 5.7 kev, respectively. Various types of observational, experimental, and simulation-derived data can be used to evaluate the potential effectiveness of different radiations and energies to produce biological damage. However, as discussed at length in this Report, no single area of scientific inquiry has been able to provide sufficient and appropriate evidence to quantify the effectiveness of photons or electrons of various energies in inducing cancer in humans. For that reason, this Report adopted a four-stage strategy to fulfill its charge with the foundation being the collection, analysis and interpretation of published data on RBEs or other effect measures relevant to the estimation of L s for the five lower-energy groups. The first steps of the strategy used in this Report involved the collection and review of past and current evidence on biological effectiveness in each of four radiation-health related specialty fields (microdosimetry, DNA damage, radiobiologic studies in cell systems and laboratory animals, and human epidemiology) by individual experts (or pairs of experts) on the committee. The specialty fields produced data (lines of evidence) that in the terminology of the field of probability evaluation are referred to as research threads. Hence, this Report specifically considered evidence from four research threads (one being a combination of cell and animal systems). The purpose of the first step was to evaluate each research thread s contribution to understanding radiation damage. The particular metric of radiation damage that each literature 10

263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 publication reported on varied. In some cases, the metric studied was the experimentally defined RBE that was specific to a single set of conditions defined by radiation type and energy, reference radiation, biological system, and endpoint, with measured RBEs then extrapolated essentially to zero dose to give estimates of RBE at low doses and low dose rates (RBE M ) when feasible. The biological effectiveness derived from the literature for each research thread (which often required additional subjective judgment) was denoted, for purposes of this Report, by R i (see Glossary). The first step is reflected in the presentations of Sections 4 through 7, which are briefly summarized here. 1.2 Microdosimetry The basic assumption underlying the study of the effects of ionizing radiation on biological systems and, therefore, the risks of cancer resulting from radiation exposure, is that ionizing radiation initiates chemical changes that, in turn, initiate a sequence of biochemical and biological changes eventually leading to cancer. Both ionization and excitation of atoms or molecules can initiate chemical changes, but ionization can initiate a wider range of changes and is usually assumed to be the dominant source of DNA damage. As the biological consequences of irradiation are initiated by the chemical changes produced by electron interactions in tissue, any differences in the biological effects of the same dose of different radiations are assumed to be related to the differences in the track structures of those electrons. In order to characterize radiation damage, it is necessary to consider the probability density of either the energy of the electrons at the point of interest, the energy deposited in a small volume representative of the volume where the chemistry that initiates the radiation response occurs, or some other measure of the clustering of energy deposition along electron tracks. While no description of the track structure of a charged particle that correlates directly with the definition of RBE or the derived R i for a specific radiation has been found, these modifying factors appear to depend on the interaction between the characteristics of the energy deposition and the structure and biochemical processes functioning in the irradiated system. The radiation damage determined by any of the microdosimetric models reflects only the initial stage of the sequence of processes (i.e., the deposition of energy), which in turn initiates 11

295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 the biochemical changes leading to the health effect. The derived R i from microdosimetry is relevant to ρ L because: R i is an indicator of the energy-deposition pattern, which presumably is important to induction of cancer; the initial types of radiation damage, such as double-strand breaks (DSB) and free-radical production, are generally independent of the biological system or endpoint (effect) under study; and DNA damage and repair mechanisms may differ quantitatively but probably are similar qualitatively in in vitro and in vivo experimental systems and in humans. 1.3 Deoxyribonucleic Acid Damage It is generally recognized that damage to cellular deoxyribonucleic acid (DNA) can lead to a variety of long-term effects in cells that may lead to cancer. Ionizing radiation is an effective agent at creating DNA damage, especially clustered damage, both via direct interaction of the ionizing particle with the DNA and via production of reactive radicals in the immediate neighborhood of the DNA. DNA DSBs are of particular concern as a pathway to permanent alterations to the DNA sequence, with potential long-term health consequences; non-dsb clustered damage may also be relevant. A large amount of experimental evidence has shown that DSBs are produced linearly with absorbed dose of radiation over an enormous dose range from tens of milligray to thousands of gray. Moreover, analytical evaluations of the fraction of the total absorbed dose that is deposited by electrons of a given energy at a point in the irradiated medium indicate that a substantial fraction is delivered by low-energy electrons. For example, for 60 Co gamma rays, about 33 % of the absorbed dose is deposited by electrons of energy <5 kev and that this fraction increases to ~50 % for 220 kv 2 x rays. The corresponding fractions of absorbed dose that is deposited by electrons of energy <1 kev are ~27 % (for 60 Co gamma rays) and ~32 % (for 220 kv x rays). 2 Throughout this Report, the unit kilovolt (kv) refers to peak tube potential applied to the x-ray tube (Glossary). 12

325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 Some trends in the experimentally determined RBE values in the study of DNA DSB are: most RBEs for photons are greater than unity over the energy range from 0.28 kev (monoenergetic) to 80 kv (continuous spectrum of x rays) and none of the RBEs are significantly less than unity; there is a strong increase in RBE with decreasing photon energy over the lower-energy range (below ~5 kev); and there are indications of modestly raised RBEs at the higher energies (a few tens of kiloelectron volts), although only a few measurements are available (all for broad spectra of photons). From the data studied, photon-energy-specific subjective probability density functions (PDFs) are proposed for R i based on experimental and simulated RBEs for total DSBs and also with some additional weighting for DSBs of assumed greater severity (Section 5). It is assumed that the severity-weighted PDFs of R i are the most relevant from the DNA damage thread of evidence for estimating ρ L for cancer in humans. It is understood that there is an extended chain of events between initial DNA damage and long-term consequences such as cancer with the probability of such an outcome being strongly dependent on the particular mammalian species, tissue under consideration and individual genetics. The initial DNA damage is relevant to ρ L for cancer in humans because: the conformation of DNA, its local environment, and the fast physical and chemical processes of initial damage formation are similar across mammalian species, cell types, tissues, and genetics; and ratios of initial yields of DNA DSBs from different radiations, as represented by R i, are likely to be reflected in ratios of effects through the chain of subsequent events, especially when differences between spectra of DSB complexity are relatively small, as is the case for photons (and electrons) of different energies. 13

356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 1.4 Cellular Radiobiology and Animal Studies Section 6 presents a review of studies that investigated RBEs of lower-energy photons and electrons, including low-energy radionuclide emissions, in inducing various biological effects in cells and whole animals. Effects in cells include structural chromosome aberrations, micronuclei, cell survival and cell killing, cell mutation, and cell transformation. Studies in whole animals have provided information on RBEs for induction of cancers and other effects in whole organs. Studies reviewed in Section 6 show that RBE depends not only on the energy of a low-let radiation, but also on the biological endpoint, irradiation and experimental conditions, absorbed dose, and cell type. Estimates of R i for lower-energy photons and electrons based on studies in cells may be relevant to ρ L for cancer in humans because: DNA damage and fixation in the form of gene mutations in cells are generally considered to be involved in the multistep process of malignancy; large-scale chromosomal damage and recombination that results in alterations at the chromosomal level are known to be involved in the development of neoplasms; and numerical chromosome changes have been associated with tumorigenesis and have the potential for undesirable heritable aneuploidy, which is a signature of a number of cancer types in humans, in subsequent cell populations. Although there are potentially important differences between laboratory animals and humans (e.g., laboratory animals often are carefully inbred strains that may have atypical responses to radiation), most types of radiation-induced cancers observed in animals also occur in humans. In addition, there are similarities in physiology, biochemistry, metabolism, and the stepwise nature of tumor progression in animals and humans. Consequently, estimates of R i for lower-energy photons and electrons based on studies in animals generally are considered to be relevant to estimating ρ L for cancer in humans. 14

386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 1.5 Human Epidemiology In principle, findings from human epidemiologic studies should be the most relevant to an evaluation of the effectiveness of different energies of low-let radiations in inducing cancer in humans. In Section 7 of this Report, estimates of cancer risks in populations that were exposed to medical x rays are compared with risks in Japanese atomic-bomb survivors who were exposed mainly to high-energy photons to assess whether available data from epidemiologic studies are informative on the question of whether medical x rays are more effective in inducing cancer in humans. Comparisons based on data for all solid cancers excluding leukemias and data for leukemias or specific solid cancers (lung, female breast, and thyroid) were considered in evaluating the biological effectiveness of medical x rays. There are no epidemiologic data that can be used to assess the effectiveness of tritium beta particles in inducing cancer in humans. Epidemiologic studies face a number of challenges in attempting to assess the biological effectiveness of photons of different energies. In contrast to randomized and controlled experiments on cells and laboratory animals, there is the possibility that biases, in particular confounding by cancer risk factors other than radiation, may distort results of epidemiologic studies. An important limitation in all epidemiologic studies reviewed in this Report is the statistical imprecision of estimates of cancer risks, which is a consequence of the small numbers of estimated excess cancers in study populations associated with radiation exposure. On the basis of the comparisons of epidemiologic studies in Section 7, it is concluded that there is no definitive epidemiologic evidence to support a substantially increased, or decreased, cancer risk associated with exposure to lower-energy photons relative to higher-energy photons. There have been no informative epidemiologic studies to compare risks of all cancers or all cancers excluding leukemias from whole-body exposures to photons of different energies. Published studies of leukemias or lung, female breast, or thyroid cancers have provided limited information on the biological effectiveness of 60 to 250 kv x rays. Central estimates of R i obtained from most of those studies are in the range of about 1.0 to 1.7. However, all estimates have substantial uncertainty, and values as low as about 0.2 to 0.3 and as high as 3 to 4 or more, including values exceeding 10, are consistent with the available data, even when results that appear to be outliers (e.g., an estimate based on risks of thyroid cancer in the tinea capitis cohort) 15

418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 are excluded. The available data also are not sufficiently precise to allow a distinction to be made between the biological effectiveness of lower-energy (40 to 60 kev) (e.g., as used for CT procedures) and higher-energy (>60 to 150 kev) (e.g., as used for orthovoltage x-ray procedures). Because the available epidemiologic data are only weakly informative regarding an R i for medical x rays with energies of photons of about 40 to 150 kev, the judgment of the experts for this research thread was that the subjective PDF of R i for photons of those energies ranges from 0.5 to 4.5, with a median value of 1.5. 1.6 Analysis of Research Threads As noted, the first part of the strategy used in this Report to evaluate L was the collection and review of past and current evidence in each of four radiation-health related specialty fields (microdosimetry, DNA damage, radiobiologic studies in cell and animal systems, and human epidemiology) by individual experts (or pairs of experts) on the committee. In the second step, the variation of the collected measurements and data for each research thread was analyzed and used to develop PDFs by the expert (or pair of experts) to describe the relative likelihood for that R i to take on given values. However, this step recognized that the data analyzed did not necessarily describe the likelihood of inducing cancer in humans but, rather, the data were metrics of damage specific to a particular biological system that (except for the research thread of human epidemiology) were for a nonhuman system. The third step was an intensive education component in which each subject matter expert (or pair of experts) informed the rest of the committee on the science and the quality and relevance of the collected data from a specific research thread in regard to how those data might help in estimating the effectiveness of one or more of the lower-energy groups in inducing cancer in humans. The third step was conducted via presentations and discussions during the multiple meetings of the full committee. 16

448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 In the final and fourth step described in Section 8, a process based on probability elicitation and Bayesian analysis 3 was implemented to aggregate the committee s understanding and views, with the result that expert-based, subjectively-derived PDFs were constructed to describe the state-of-knowledge of L for each lower-energy group. It should be noted that some of the derived PDFs of R i in the second step are expressions of variation (e.g., for DNA damage), while others are subjective measures of the uncertainty (state-of-knowledge) of the R i studied (e.g., for epidemiology), which is specific to a particular biological system and endpoint studied in the original investigations. Regardless of the nature of the analysis used to determine the PDFs of R i from the research threads, each PDF of R i from an analysis of a particular research thread was subsequently used as input data to the fourth step of the analysis. The fourth step of the strategy used in this Report, which was based on the field of probability elicitation and Bayesian analysis and decision theory, obtained input from multiple experts regarding the interpretation, the importance, and the relevance of the data to the primary goal of describing the uncertainty of L for photons or electrons of a specified lower-energy group. In this Report, the primary elicitation strategy is termed decomposition modeling. The decompostion probability assessment is a two-step elicitation process and while complex in terms of mathematics, is considered rigorous and traceable in terms of the logic required by the participants. The first of the two steps of this method was for each expert (or pair of experts) that was tasked to collect and evaluate the evidence for a particular research thread, to develop a PDF describing degree-of-belief of values of R i for that research thread for each lower-energy group based on the collected evidence. The second step was for each committee member to provide PDFs (in the form of first, second and third quartiles) for L for each lower-energy group, conditional on each research thread R i being known with certainty at either the first, second, or third quartiles of the PDFs of R i obtained in the first step. This second step allowed each committee member to assess the likelihoods for a range of density functions that might represent L for each lower-energy group. An aggregated density function (similar to a PDF but 3 Bayesian analysis or Bayesian decision theory is sometimes called Bayesian probability theory where probability is loosely interpreted as a degree-of-belief (Bernardo and Smith, 1994; Lee, 1989). 17

477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 representing degree-of-belief) of L for each of the five energy groups discussed in this Report was computed by methods discussed in Section 8. In addition to the decomposition modeling method, a holistic strategy was conducted for comparison purposes. Generally, the holistic strategy is more intuitive and simpler to understand. In the holistic modeling method, the experts were asked to evaluate the research thread PDFs of all R i s as an integrated or holistic set of data across the entire spectrum of research threads and to mentally integrate (combine) them into a single PDF of L for each of the five lower-energy groups. No formal method or record of each committee member s thought process was involved in the holistic elicitation strategy. For that reason, this method is more difficult to document. With details on the differences of the methods provided in Section 8, here it is sufficient to say that both strategies allowed committee members, as experts, to evaluate the data from the research threads for reliability and relevance to quantification of the ρ L s 1.7 Findings The findings from the review and analysis of the various research threads and the subsequent elicitation process resulted in the development of 10 PDFs of L s, which are graphically presented in Figure 1.1 [for the decomposition method (the preferred method); one PDF of L for each of the lower-energy groups] and Figure 1.2 (for the holistic method; one PDF of L for each of the lower-energy groups). These are Figure 8.3 and Figure 8.1, respectively, from Section 8. Selected quantiles for the 10 PDFs are given in Table 1.1. Inspection of these PDFs of L indicates the results from the two elicitation strategies were reasonably consistent, suggesting that a choice between the two methods is not particularly critical to the outcome of the analysis. The two lowest-energy groups [i.e., about 1.5 kev, and the tritium beta-particle decay spectrum (average energy, 5.7 kev; maximum energy, 18.6 kev)] produced PDFs centered at higher L values, consistent with observations from decades of specialized experiments in various biological systems. 18

508 about 1.5 kev 15 to 30 kev 40 to 60 kev 70 to 120 kev Tritium 2.5 2 1.5 1 0.5 0 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 509 510 511 512 513 514 515 516 517 518 0.5 Fig. 1.1. Aggregated probability density functions (PDFs) for L for the five lower-energy groups derived from the decomposed elicitation strategy. NOTE: The label for the x-axis is L ; the label for the y-axis is f( L ). NOTE: Will change the label 70 to 120 kev to >60 to 150 kev. NOTE: The color codes for the various lower-energy groups will be made consistent for Figures 1.1 and 1.2. 19

519 520 1.5 kev 15 to 30 kev 40 t0 60 kev 70 to 120 kev Tritium 1.4 1.2 1 0.8 0.6 0.4 0.2 521 522 523 524 525 526 527 528 529 0 0 1 2 3 4 5 6 Fig. 1.2. Aggregated probability density functions (PDFs) for L for the five lower-energy groups derived from the holistic elicitation strategy. NOTE: The label for the x-axis is L ; the label for the y-axis is f( L ). NOTE: Will change the label 70 to 120 kev to >60 to 150 kev. NOTE: The color codes for the various lower-energy groups will be made consistent for Figures 1.1 and 1.2. 20

530 531 532 533 Table 1.1 --- Five quantiles of the subjective PDFs of L for five lower-energy groups derived by the decomposition method (the preferred method) and the holistic method. The 0.5 quantile row presents the median values of the PDFs for each of the five lower-energy groups. Photons about 1.5 kev Photons 15 to 30 kev Photons 40 to 60 kev Photons >60 to 150 kev Tritium Beta- Particle Decay Electrons METHOD Quantile Decomposition Holistic Decomposition Holistic Holistic Decomposition Decomposition Holistic Decomposition Holistic 0.05 1.4 0.9 1.0 0.7 1.0 0.7 1.0 0.7 1.1 0.9 0.25 1.9 1.7 1.2 1.3 1.1 1.1 1.2 1.1 1.4 1.5 0.5 2.4 2.4 1.5 1.8 1.2 1.4 1.3 1.3 1.7 2.0 0.75 2.8 3.1 1.7 2.4 1.3 1.9 1.4 1.7 2.0 2.6 0.95 3.5 4.1 2.2 3.4 1.6 2.8 1.7 2.5 2.5 3.8 534 21

535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 On average, the ratio of the 95 % to 5 % quantiles for the PDFs were about 2.1 and 4.2 for the decomposed and holistic methods, respectively, reflecting greater uncertainty for the holistic method. This observation likely reflects that the process leading to the PDFs by the decomposition method does not require as much mental integration of the data from each research thread and is also consistent with the more rigorous methodology of the decomposition method. The width of each PDF (e.g., from the 5 % to 95 % quantile) emphasizes the uncertainty associated with the findings from this elicitation exercise and is a representation of the present state-of-knowledge of ρ L for each of the defined five lower-energy groups. 1.8 Conclusions and Recommendations The analysis in this Report demonstrates that decades of study and experimentation have not led to the development of data sets from which definitive deductions can be made about the effectiveness of lower-energy photons and electrons in inducing cancer in humans relative to higher-energy photons, especially the high-energy photons that exposed Japanese atomic-bomb survivors. Nevertheless, the derived PDFs of L when viewed in their entirety suggest a decrease in L with increasing energy for photons and electrons. The outcomes of this analysis are subjectively-derived PDFs of each L which describe the present state-of-knowledge as based on the literature reviewed in this Report and the committee s interpretation of the relevance and quality of the literature data. The decomposition strategy represents the preferred method since it is viewed as more rigorous and traceable in terms of participant reasoning and decisions. The decomposition strategy required judgments about the research threads individually, rendering it to be a sequence of tasks and potentially less subject to individual biases. However, as discussed in Section 8.5, some factors (e.g., lack of conditional independence between the research threads) could result in PDFs that are unrealistically narrow in the decomposition method. The second elicitation strategy (i.e., the holistic assessment) is relatively straight forward but required committee members to mentally integrate their knowledge produced by the several research threads, a difficult and complex cognitive task. The relatively minor differences in the PDFs from the two elicitation strategies 22

567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 for the same lower-energy group suggest that the results were not strictly a product of the analysis method but are reflective of the present state-of-knowledge and the limitations of the data analyzed. This Report does not recommend or endorse point values for ρ L because of uncertainty in the state-of-knowledge. Rather, PDFs of ρ L are provided to represent the range in which likely true values are believed to lie given our present understanding. The PDFs of L as presented, regardless of the choice of the elicitation method, should be a useful contribution to the capability for conducting probabilistic risk assessments for human cancer from exposure to photons and electrons at low doses or low dose rates. It is concluded here that the derived PDFs are the best information presently available and should be used to propagate uncertainty in probabilistic human cancer risks assessments for exposure to such sources of ionizing radiation. This Report emphasizes that present-day data as well as present-day theory can not precisely determine ρ L values, and that the PDFs of ρ L presented are subjective evaluations for each of the lower-energy groups defined for this Report. Given the need for knowledge of ρ L for certain assessments of cancer risks due to exposure to lower-energy photons or electrons found in some important occupational, medical, or environmental exposures, there is a clear need for additional research. However, caution is recommended that further research to elucidate ρ L should include careful planning to justify the particular biological system chosen, the endpoint to be studied, and the test and reference radiations. The ideal data to evaluate ρ L would be evidence from human epidemiologic studies. Such studies are, however, difficult to conduct and typically suffer from many uncertainties and confounding factors. Consideration of further epidemiologic studies is recommended if it is possible to rigorously focus on lower-energy photons and electrons in comparison to higherenergy photons and electrons, with great attention given to control of confounding factors. 23

598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 2. Introduction Models to estimate risks of health effects from exposure to ionizing radiation usually require assumptions about the biological effectiveness of different radiation types (e.g., photons, electrons, alpha particles, or neutrons). The biological effectiveness of a particular radiation type can be represented by a factor that modifies an estimate of absorbed dose in an organ or tissue of interest to give a biologically significant dose on which the risk of a health effect (e.g., cancer) in that organ or tissue is assumed to depend. 4 The modifying factor used in health risk assessments is analogous to w R used in radiation protection. However, it is distinct from w R, which is used to estimate equivalent doses but is not intended for use in estimating health risks from known exposures (ICRP, 1991; 2007). 5 This Report presents an evaluation of information on the biological effectiveness of photons and electrons of various energies. This evaluation focuses on cancer as the health effect of primary concern in humans; it does not consider the effectiveness of photons and electrons in inducing noncancer effects. An essential aspect of the evaluation in this Report is an assessment of the state-ofknowledge in the effectiveness of photons and electrons of various energies in inducing cancer in humans. The objective of this assessment is to obtain quantitative representations of the current state-of-knowledge of the biological effectiveness of those radiations. Such an assessment is necessary when the intent of a risk assessment is to fully represent uncertainties in estimates of cancer risk (NCRP, 2012). For example, uncertainties in the effectiveness of radiation types of concern must be taken into account when an upper limit of confidence of an uncertain estimate 4 An exception to the need to account for the biological effectiveness of a radiation type of concern involves models to estimate risks of lung cancer from exposure to radon on the basis of estimates of the risk per unit exposure to short-lived, alpha-emitting decay products in air in J h m 3 that are obtained from epidemiologic studies of underground miners (e.g., ICRP, 1993; NA/NRC, 1999). Such models do not require estimates of absorbed dose from alpha particles or the biological effectiveness of those radiations. 5 The biologically significant dose obtained by modifying an absorbed dose from a particular radiation type by a factor that is appropriate for use in estimating health risks from known exposures is not given a name, mainly because estimation of such a dose is not of direct interest to risk assessment. In effect, the factor of concern to this Report modifies an estimate of the risk associated with a given absorbed dose of a reference radiation with a defined biological effectiveness of unity to give an estimate of the risk associated with the same absorbed dose of a radiation type of interest (Section 2.5). 24

622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 of risk and the associated probability of causation/assigned share (PC/AS) of a diagnosed cancer is used in programs to compensate individuals whose cancers might have been caused by exposure to ionizing radiation (Neton et al., 2008). In evaluating information on the biological effectiveness of photons and electrons, it is useful to recognize that absorbed doses from irradiation by photons are due almost entirely to the energy deposited by secondary electrons that are generated by interactions of photons in tissue. Thus, information on the biological effectiveness of photons of a given energy is indicative of the biological effectiveness of the energy spectrum of electrons produced by those photons. 2.1 Background for Present Study Authoritative organizations and other investigators have evaluated information on the effectiveness of photons or electrons of various energies in inducing stochastic biological effects. Although many studies have suggested that the biological effectiveness of those radiations is greater at lower energies than at higher energies, the possibility that the effectiveness of photons and electrons in inducing cancer in humans might be greater at lower energies has not often been taken into account in radiation protection or in cancer risk assessments. 2.1.1 Recommendations for Purposes of Radiation Protection Prior to the late 1950s, it was generally assumed for purposes of radiation protection (control of exposures) that the biological effectiveness of photons and electrons, which are low- LET radiations, is the same at all energies (e.g., NCRP, 1953). Since that time, only for lowenergy electrons was a greater biological effectiveness considered for use in radiation protection by ICRP and NCRP or adopted by regulatory authorities in the United States. In the late 1950s, ICRP (1959) and NCRP (1959) recommended that, for purposes of radiation protection, it should be assumed that the biological effectiveness of electrons (including positrons) of energy 0.03 MeV is a factor of 1.7 greater than the effectiveness of x and gamma radiation and higher-energy electrons. This recommendation applied, for example, to low-energy electrons emitted in beta-particle decay of 3 H (tritium). However, ICRP later judged that an 25

654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 assumption of a biological effectiveness of unity for photons and electrons of any energy has a degree of precision appropriate for purposes of radiation protection (Dunster, 1969). This assumption has been used by ICRP and NCRP ever since (ICRP, 2007; NCRP, 1993), even though more recent evaluations of available information have suggested that the biological effectiveness of tritium beta particles and photons at energies of orthovoltage x rays 6 or lower, relative to higher-energy photons, could be greater than unity (CNSC, 2010; HPA, 2007; Kocher et al., 2005; ICRP, 2003; ICRU, 1986; NCRP, 1990; Straume and Carsten, 1993; UNSCEAR, 2015). The assumption of a biological effectiveness of 1.7 for tritium beta particles (ICRP, 1959; NCRP, 1959) is still incorporated in some environmental regulations in the United States. That assumption was used by the U.S. Environmental Protection Agency (EPA) in deriving the current maximum contaminant level (MCL) for tritium in public drinking water supplies (EPA, 1977; 2000b), and it is incorporated in U.S. Nuclear Regulatory Commission (NRC) guidance on evaluating compliance with design objectives for limiting releases of radionuclides from operating light-water-cooled nuclear power reactors (NRC, 1977). 7 EPA s MCL for tritium in public drinking water supplies also is incorporated in current EPA standards for cleanup of contaminated groundwater or surface waters (EPA, 1990). No other EPA or NRC standards or guidance on radiation protection of workers or the public incorporate an increased biological effectiveness of tritium beta particles, nor do EPA s current methods of estimating lifetime cancer risks to the U.S. population from exposure to ionizing radiation (Eckerman et al., 1999; EPA, 2011). 6 Orthovoltage x rays are produced by generators operating at peak tube potentials of about 180 to 300 kv. 7 The MCL for tritium is a concentration of 740 Bq L 1, which was estimated to correspond to an annual equivalent dose to the whole body of 40 μsv from consumption of 2 L d 1 of drinking water by an adult (EPA, 1977). NRC s design objectives for limiting releases from operating nuclear power plants that apply to tritium are annual equivalent doses to the whole body of 30 μsv from all radionuclides combined in liquid effluents and 50 μsv from all radionuclides combined in gaseous effluents (NRC, 2015). 26

677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 2.1.2 Assumptions for Purposes of Cancer Risk Assessment Assumptions about increases in the biological effectiveness of lower-energy photons and electrons and their uncertainties have been developed in recent years by various investigators for purposes of estimating cancer risks from known exposures of individuals or populations. On the basis of several lines of evidence, Kocher et al. (2002; 2005) developed probability distributions [referred to in this Report as probability density functions (PDFs)] to represent uncertainties in the effectiveness of photons of energy 30 to 250 kev or <30 kev, relative to photons of energy >250 kev, in inducing cancer in humans. Information used in that analysis included: radiobiologic data reviewed by NCRP (1990), primarily data on induction of dicentric chromosome aberrations in human blood lymphocytes; comparisons of estimated cancer risks in epidemiologic studies of populations exposed to medical x rays or higher-energy gamma rays; radiobiologic studies in which the biological effectiveness of fission neutrons relative to orthovoltage x rays or higher-energy gamma rays could be compared; and a calculation of the energy dependence of the effective quality factor for photons (ICRU, 1986). The PDFs developed by Kocher et al. (2002; 2005), which have 95 % subjective confidence intervals (CIs) of (1.0, 4.7) at photon energies of 30 to 250 kev and (1.1, 6.1) at energies <30 kev, are incorporated in risk models in the Interactive RadioEpidemiological Program (IREP), which is used to evaluate PC/AS of diagnosed cancers in exposed individuals in U.S. compensation programs (Kocher et al., 2008; Land et al., 2003). 8 In a later paper, however, Trabalka and Kocher (2007) discussed more recent data on induction of dicentric chromosome 8 Cancer risk models in IREP, including representations of the effectiveness of various radiation types in inducing cancer in humans, were intended to provide unbiased estimates of cancer risk, PC/AS, and their uncertainties that represented the state-of-knowledge at that time. 27

704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 aberrations in human blood lympocytes which indicate that the primary data set used by Kocher et al. (2002; 2005) probably provides invalid estimates of the biological effectiveness of lowerenergy photons. 9 Information on the biological effectiveness of tritium beta particles and its uncertainty has been evaluated in several recent studies. A review of radiobiologic data by the U.K Health Protection Agency (HPA, 2007) led to a recommendation that a biological effectiveness of two, relative to high-energy gamma rays, should be assumed in epidemiologic studies and retrospective risk assessments for individuals. In a subsequent analysis of the data reviewed by HPA (2007), Little and Lambert (2008) developed PDFs of the biological effectiveness of tritium beta particles with 95 % CIs of (0.96, 1.39) relative to orthovoltage x rays and (2.04, 2.33) relative to high-energy gamma rays. If those PDFs are assumed to be normal, their ratio gives a 95 % CI of the biological effectiveness of orthovoltage x rays relative to high-energy gamma rays of (1.6, 2.3). Other evaluations of uncertainties in the biological effectiveness of tritium beta particles relative to high-energy gamma rays gave 95 % CIs of (1.0, 3.5) (Hamby, 1999), (1, 2.5) (Harrison et al., 2002), and (1.2, 5.0) (Kocher et al., 2002; 2005). The PDF developed by Kocher et al. (2002; 2005), which was assumed to apply to electrons of energy 15 kev except certain Auger electrons emitted by radionuclides, is incorporated in cancer risk models in IREP (Kocher et al., 2008; Land et al., 2003). That PDF also was used in an assessment of uncertainties in lifetime cancer risks associated with EPA s MCL for tritium in public drinking water supplies (Kocher and Hoffman, 2011). However, a concern about all the reported 95 % CIs is that their upper bounds were based to a significant extent on data for induction of dicentric chromosome aberrations in human blood lymphocytes that probably are invalid, as noted above. Computer simulations show that lower-energy photons and beta particles deposit a greater proportion of their energy in the form of ionization clusters than do higher-energy gamma rays, 9 On the basis of a review of data evaluated by Kocher et al. (2002; 2005) and more recent microdosimetry calculations and radiobiologic data, UNSCEAR (2015) noted that [g]iven the current state-of-knowledge, it might be said that the [biological effectiveness of] orthovoltage X-rays and other photons of similar energy [relative to higher-energy photons] could be in the range of about 1 4, and that the [biological effectiveness of] lower-energy photons could be somewhat higher. However, UNSCEAR (2015) did not perform a formal analysis of the biological effectiveness of photons of various energies and its uncertainties. 28

730 731 732 733 734 735 736 737 738 739 740 741 such as those dominating the energy spectrum to which atomic-bomb survivors were exposed. The clusters are mostly produced at the ends of tracks, where the electron kinetic energy has been reduced to a few kiloelectron volts. Such clusters are expected to produce complex damage in DNA, which is difficult for the cell to repair properly (Section 5.1). Based on the assumption that RBE is roughly proportional to the energy deposited by electrons of a few kiloelectron volts, Bellamy and Eckerman (2013) and Bellamy et al. (2015) estimated RBEs of lower-energy photons and electrons relative to 1 MeV electrons, which should have about the same biological effectiveness as the spectra of photons from the atomic bombings in Japan. With this approach, the authors derived RBEs of approximately 1.5 for x rays commonly used for diagnostic medical procedures. They also derived RBEs for low-let radiations emitted by radionuclides included in ICRP (2008). Most of those RBEs were only slightly greater than one, but a few estimates exceeded 1.5; most notably, the estimated RBE for tritium was approximately two. 742 743 2.1.3 Previous Reviews by NCRP 744 745 746 747 748 749 750 751 752 NCRP Report No. 104 (NCRP, 1990) presented an evaluation of radiobiologic data on the effectiveness of various radiation types, mainly fission neutrons and alpha particles, in inducing stochastic effects. That evaluation included data on the effectiveness of orthovoltage x rays relative to higher-energy gamma rays from decay of 60 Co for induction of dicentric chromosome aberrations in human blood lymphocytes. The data on dicentrics reviewed by NCRP (1990) were used in the assessment of biological effectiveness by Kocher et al. (2002; 2005) discussed in Section 2.1.2. NCRP (1990) did not include recommendations on use of the available data to define quality factors (Q values) or w R values for purposes of radiation protection. 753 754 755 756 757 758 759 760 NCRP Report No. 171 (NCRP, 2012) includes discussions of information on the biological effectiveness of lower-energy photons and electrons and their uncertainties. Presentations for photons emphasize radiobiologic data more recent than the data used by Kocher et al. (2002; 2005). The problem described in Section 2.1.2 with the data for dicentrics that were used by those investigators also is discussed. Presentations for electrons include discussions of PDFs of the biological effectiveness of tritium beta particles that were developed by various investigators and summarized in the Section 2.1.2. NCRP (2012) does not include recommendations on PDFs 29

761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 that should be used to represent uncertainties in the biological effectiveness of lower-energy photons or electrons for purposes of estimating cancer risks from known exposures. 2.2 Importance of Present Evaluation This Report is concerned with evaluating the biological effectiveness of lower-energy photons and electrons relative to higher-energy photons and electrons for purposes of estimating cancer risks from known exposures. The effectiveness of lower-energy photons relative to higher-energy photons in inducing cancer in humans is an important concern in assessing risks from use of computed tomography (CT) and mammography in diagnostic medicine. Concerns about risks from CT arise from its increasing use and the higher doses per CT scan compared with conventional diagnostic x-ray procedures (Brenner and Hall, 2007). In 2006, for example, the collective effective dose from CT in the United States was nearly 50 % of the collective effective dose from natural background radiation (NCRP, 2009). When fluoroscopy and conventional radiography are included, the collective effective dose from all uses of x rays in medicine was 72 % of the collective effective dose from natural background (NCRP, 2009). The risk of breast cancer in females from use of mammography is a concern due to the large number of procedures and low energies of the x rays (e.g., Brenner and Amols, 1989). 10 In 2006, for example, the number of mammography procedures in the United States was nearly 35 million, and the typical absorbed dose to the breast per procedure was 3.6 mgy (NCRP, 2009). A substantial increase in the biological effectiveness of mammography x rays relative to highenergy photons could have important implications for breast cancer screening programs that consider the risks and benefits of mammography scans in women of various ages (e.g., Heyes et al., 2006; 2009). The biological effectiveness of lower-energy photons also is important in programs to compensate individuals whose cancers might have been caused by radiation exposure. For example, in the compensation program for nuclear energy workers in the United States, estimated organ doses from external exposure to photons in workplaces are apportioned among the three energy bins (<30, 30 to 250, and >250 kev) assumed in the assessment of biological 10 Mammography x rays are produced by generators operating at peak tube potentials of about 29 kv. 30

792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 effectiveness by Kocher et al. (2002; 2005) on the basis of information on a claimant s conditions of exposure (Merwin et al., 2008; NIOSH, 2007). Doses from occupationally related medical x-ray examinations also are taken into account (Shockley et al., 2008). Assumed increases in the biological effectiveness of lower-energy photons can affect decisions on awarding compensation. Tritium is an important source of radiation exposure in some occupational settings (CNSC, 2010; HPA, 2007), and a consideration of the biological effectiveness of tritium beta particles is important in epidemiologic studies of exposed worker populations. Another important consideration for purposes of cancer risk assessment is that radionuclides that emit low-energy Auger electrons with energies similar to or less than the average energy of tritium beta particles (e.g., 67 Ga, 99m Tc, 123 I, 125 I, and 201 Tl) are frequently used in diagnostic nuclear medicine. Tritium is naturally occurring in the environment, albeit at very low levels (NCRP, 1979). However, planned and unplanned releases from nuclear power plants into surface waters (NRC, 2008; Richards et al., 2006) have led to concerns about their impacts on public health (Makhijani and Makhijani, 2009). Of particular interest is the question of whether EPA s MCL for tritium in public drinking water supplies (Section 2.1.1, footnote 4) would meet the objective of limiting lifetime cancer risks from intakes in drinking water to no more than about 10 4 (EPA, 2000a; 2000b). 11 EPA s estimate of the lifetime cancer risk associated with the MCL for tritium, which was based on an assumed consumption rate of drinking water of 2 L d 1 and no increase in biological effectiveness of tritium beta particles, was about 0.6 10 4 (EPA, 2000b). A recent analysis that took into account uncertainties in estimates of age-specific risks of cancer, excluding skin cancer, in the U.S. population from exposure to high-energy photons (NA/NRC, 2006), age-specific consumption rates of drinking water, and age-specific ingestion dose coefficients gave a mean lifetime cancer risk associated with the MCL for tritium and 95 % CI of about 0.5 (0.1, 1.4) 10 4, again assuming no increase in biological effectiveness of tritium beta particles (Kocher and Hoffman, 2011). These estimates indicate that the mean lifetime cancer risk and upper confidence limit would meet EPA s objective of about 10 4 if the biological effectiveness of tritium beta particles were no more than about two. 11 EPA considers that lifetime cancer risks up to about 3 10 4 meet this objective (EPA, 2000b). 31

823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 2.3 Term to Describe Modifying Factor to Represent Biological Effectiveness Most investigators have referred to the factor that modifies an estimate of absorbed dose from a particular radiation type for the purpose of estimating cancer risks from known exposures as the relative biological effectiveness (RBE). Use of that term is understandable when estimates of RBE obtained in radiobiologic studies have been the primary source of data on the effectiveness of different radiation types in inducing biological effects and to date a consensus on use of an alternative term is lacking. However, use of the term RBE to describe the modifying factor to be used in cancer risk assessments has conceptual difficulties, owing to the definition of RBE. An RBE is used to represent the influence of radiation quality, which is often described by the LET of a radiation in tissue, on effects in biological systems. As given, for example, in NCRP Report No. 104 (NCRP, 1990), the RBE for a particular radiation (A) is defined as: Dose of reference radiation required to produce a specific level of response RBE (A), (2.1) Dose of radiation A required to produce an equal response with all physical and biological variables, except radiation quality, held as constant as possible. In this definition, dose is the physical quantity absorbed dose. Important physical variables include the magnitude of doses and dose rates, and important biological variables include the system under study, its physiologic condition, and its environment. The definition of RBE does not depend on the dose-response relationships for the two radiations being the same, or that the dose-response function be a proportional (linear) relationship. In general, it is observed that RBE depends on the biological system and specific response (effect) under study. RBE also depends on dose or dose rate whenever the dose-response relationship for either radiation is nonlinear. The definition of RBE indicates that it is a radiobiologic quantity (that is, an RBE is a result of a specific radiobiologic study under controlled conditions). For example, the definition implies that if an RBE for a particular radiation type and biological effect obtained from a study in cells or animals is used to infer the biological effectiveness of that radiation in humans, the quantity that represents the latter is not an RBE even if the effects of concern in the radiobiologic study and in humans are the same. Similarly, if an RBE for an effect in human cells or tissues 32

855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 other than induction of cancer is used to infer the effectiveness of a particular radiation type in inducing cancer in humans, the quantity that represents the latter is not an RBE. Doses and dose rates in most radiobiologic studies are substantially higher than those of concern in most occupational or public exposures to ionizing radiation. For purposes of radiation protection and most cancer risk assessments, the quantity of interest from radiobiologic studies is assumed to be an RBE for stochastic effects at low absorbed doses and low absorbed-dose rates 12 (referred throughout this Report as low doses and low dose rates). That quantity is denoted as RBE M, which is estimated as the ratio of the slopes of fitted dose-response functions for the radiation of interest and the reference radiation as the dose approaches zero (ICRP, 2007); the subscript M denotes a maximal value to reflect that RBE often increases with decreasing dose. However, the notation RBE M, although widely used, does not represent an RBE as that term is strictly defined, essentially because it is an extrapolation of measured values. The effectiveness of some radiation types in inducing cancer in humans can be investigated using results of epidemiologic studies. For example, the risk of a particular cancer type in a population exposed to medical x rays can be compared with the risk in Japanese atomic-bomb survivors who were exposed mainly to high-energy gamma rays to estimate the biological effectiveness of the x rays. However, the quantity inferred from such a comparison is not an RBE, mainly because doses and dose rates as well as attributes of the study populations that can influence risk, such as ethnicity, age, or state of health, usually are not the same. The examples described above illustrate that an extrapolation of an RBE to conditions different from those under which it was estimated is not an RBE as that term is strictly defined. Another reason why the modifying factor of interest to this Report is not an RBE as that term is strictly defined relates to how it is used in cancer risk assessments. Whereas an RBE is defined as a ratio of doses at the same level of response, the modifying factor of interest in estimating cancer risks essentially is a ratio of responses (risks) at the same dose (Section 2.5). The two 12 For the purpose of this Report, for low linear-energy transfer (LET) radiation, a low absorbed dose is <100 mgy delivered acutely, and a low absorbed-dose rate is <5 mgy h 1 for any accumulated absorbed dose [as adopted at the present time by NCRP (2015)]. 33

884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 ratios are the same only if the dose-response relationships for the radiation of interest and a defined reference radiation are linear, in which case RBE is independent of dose. This condition is not often met in radiobiologic studies of photons or electrons, but is more commonly met at low doses and low dose rates of photons in epidemiologic studies. However, epidemiologic studies do not provide estimates of RBE as that quantity is strictly defined. To address the concern that the modifying factor to represent the biological effectiveness of different radiation types for purposes of cancer risk assessments does not conform to the definition of RBE, Kocher et al. (2002; 2005) proposed that the term radiation effectiveness factor be used. However, use of that term has not been adopted by consensus of national and international authorities that assess cancer risks in exposed populations, such as the United Nations Scientific Committee on the Effects of Atomic Radiation (UNSCEAR) and the National Research Council s Committee on the Biological Effects of Ionizing Radiation (BEIR). 13 Given that the modifying factor of interest to this Report is not an RBE and that the term radiation effectiveness factor proposed by Kocher et al. (2002; 2005) has not been adopted by consensus, this modifying factor is referred to in this Report as an effectiveness ratio, denoted in general by ρ, without the intention of giving the modifying factor that name. This term is used in this Report for purposes of identification only; it indicates a modifying factor that represents the effectiveness of specified energies of lower-energy photons or electrons in inducing cancer in humans relative to a defined reference radiation. The effectiveness ratio ρ as used in this Report is intended to apply at low doses and low dose rates. Use of the effectiveness ratio ρ in cancer risk assessments is described in Section 2.5. 2.4 Specification of Reference Radiation Specification of a reference radiation with a defined biological effectiveness of unity is necessary in evaluating information on the biological effectiveness of photons or electrons of 13 In Publication 92, ICRP emphasized the importance of making a clear distinction between estimates of RBE and quantities derived from those estimates and recommended, as proposed by Kocher et al. (2002), that the term REF be used to identify modifying factors that represent values of RBE in computations of PC/AS (ICRP, 2003; paragraph 86). 34

912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 various energies. Several reference radiations have been used in radiobiologic studies. The most common are high-energy gamma rays (often from decay of 60 Co or, less frequently, 137 Cs) or orthovoltage x rays with average energies substantially less than the energies of 60 Co or 137 Cs gamma rays. X rays generated at lower peak tube potentials (e.g., 120 kv) and high-energy electrons from beta-particle decay of certain radionuclides also have been used as reference radiations in some radiobiologic studies of the effectiveness of lower-energy photons or electrons. It is now recognized that the most appropriate reference radiation for purposes of cancer risk assessment is high-energy photons, such as gamma rays from decay of 60 Co (ICRP, 2003). This choice is based on the consideration that estimates of cancer risks in humans are derived primarily from studies of Japanese atomic-bomb survivors who mainly received acute exposure to high-energy gamma rays. Since 60 Co gamma rays were the reference radiation in many radiobiologic studies, the choice of high-energy photons as the appropriate reference radiation provides a desirable consistency between estimates of cancer risks in humans and much of the data on the biological effectiveness of lower-energy photons and electrons. It also is recognized, however, that the mean energy of 60 Co gamma rays of 1.25 MeV (ICRP, 2008) is lower than the energies of about 2 to 5 MeV (average energy of about 3 MeV) that contributed most of the dose to atomic-bomb survivors (Egbert et al., 2007; Santoro et al., 2005). The biological effectiveness of 60 Co gamma rays and the spectra of gamma rays at Hiroshima and Nagasaki presumably is not the same, but the magnitude of the difference is not well established. 14 Nonetheless, of the commonly used reference radiations in radiobiologic studies, 60 Co gamma rays should most closely approximate the biological effectiveness of gamma rays at Hiroshima and Nagasaki. The choice of high-energy photons as the appropriate reference radiation requires an assumption about the highest energy at which the effectiveness of photons in inducing cancer in 14 On the basis of microdosimetric calculations, Bellamy and Eckerman (2013) and Bellamy et al. (2015) estimated that the biological effectiveness of 60 Co gamma rays is only about 5 % greater than the biological effectiveness of gamma rays at Hiroshima and Nagasaki. In a previous analysis of RBEs for induction of dicentric chromosome aberrations in human blood lymphocytes, Straume (1995) estimated that the biological effectiveness of 60 Co gamma rays could be about twice the effectiveness of gamma rays at Hiroshima and Nagasaki. However, the uncertainty in that estimate is large and, as noted in Section 2.1.2, it is doubtful that the data on dicentrics used by Straume (1995) provide valid estimates of biological effectiveness. 35

938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 humans could depart from unity. On the basis of a calculation of the energy dependence of the effective quality factor in a 1-μm diameter sphere (ICRU, 1986), Kocher et al. (2002; 2005) assumed that the reference radiation is photons of energy >250 kev. However, calculations for different target diameters reviewed by ICRP (2003) suggest that other choices could be justified. An assumption about the highest energy of photons at which the biological effectiveness could depart from unity affects a similar assumption about electrons, given that doses from irradiation by photons are due almost entirely to ionizations produced by energetic secondary electrons. Another consequence of this relationship is that high-energy electrons also are an appropriate reference radiation in radiobiologic studies or calculations to estimate the biological effectiveness of lower-energy photons or electrons (e.g., Bellamy and Eckerman, 2013; Bellamy et al., 2015). 2.5 Use of Effectiveness Ratio in Cancer Risk Assessments An effectiveness ratio (ρ) to represent the effectiveness of lower-energy photons or electrons in inducing cancer in humans, relative to high-energy photons as the reference radiation, is intended to be used in estimating cancer risks from known exposures in the following ways. In estimating risks of any cancer type except leukemias (and possibly bone cancer and some skin cancers) from exposure to low-let radiations, a linear dose-response relationship usually is assumed (e.g., Land et al., 2003; NA/NRC, 2006). The excess risk associated with a given absorbed dose of lower-energy photons or electrons, denoted by, then is estimated as: In Equation (2.2): R ρ DDREF γ,h DT. (2.2) 36

966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 D T is the mean absorbed dose (Gy) of the lower-energy photons or electrons in an organ or tissue in which a cancer type of interest is induced; R γ,h is the excess risk per unit dose (Gy 1 ) for the cancer type of interest at high acute doses (H) of reference high-energy photons (γ); ρ is the effectiveness ratio for the lower-energy photons or electrons at low doses or low dose rates; and DDREF is the dose and dose-rate effectiveness factor, which takes into account that the risk per unit dose at low doses or low dose rates of photons or electrons of any energy may differ from the risk per unit dose at high acute doses. The cancer risk,, estimated using Equation (2.2) can be the excess absolute rate (EAR) or the excess relative risk (ERR) given by EAR/B, where B is the baseline rate of the cancer type of interest due to all other causes. Specification of the excess risk per unit dose for the reference radiation, R γ,h, as a value at high acute doses is a consequence of the use of data in Japanese atomic-bomb survivors to estimate those risk coefficients. Use of an effectiveness ratio ρ L then gives an estimated excess risk at low doses or low dose rates of the lower-energy photons or electrons of interest. Although a DDREF for photons or electrons could depend on energy (Trabalka and Kocher, 2007), an energy-dependent DDREF has not been used in cancer risk assessments. A linear-quadratic dose-response relationship for induction of leukemias is often assumed under conditions of acute exposure to low-let radiations (e.g., Land et al., 2003; NA/NRC, 2006). The excess risk of leukemias associated with a given absorbed dose of lower-energy photons or electrons then is estimated in accordance with one of the following equations: = α(ρ D) + β(ρ D) 2, acute exposure (2.3) = α ρ D, chronic exposure (2.4) where α (Gy 1 ) and β (Gy 2 ) are the coefficients of the linear and quadratic terms, respectively, in the assumed dose-response for leukemias under conditions of acute exposure to high-energy photons; as in Equation (2.2), the excess risk,, can be the EAR or ERR. Under conditions of 37

998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 chronic exposure, only the linear term in the dose-response for acute exposure is assumed to apply (Land et al., 2003). A DDREF is not explicit in Equations (2.3) and (2.4). However, the assumption of a linear-quadratic dose-response relationship under conditions of acute exposure incorporates a dose-dependent DDREF >1 implicitly. The models in Equations (2.2) through (2.4) incorporate an assumption that the effectiveness of lower-energy photons and electrons in inducing cancer in humans is independent of cancer type. For alpha particles and fission neutrons, which are high-let radiations, radiobiologic and epidemiologic data indicate that those radiations are substantially less effective in inducing leukemias than solid cancers, and some data suggest that their effectiveness in inducing solid cancers may depend on cancer type (NCRP, 1990; 2012; ICRP, 2003). However, such evidence is lacking for lower-energy photons or electrons, and it is assumed in this Report that any dependence of ρ for those radiations on cancer type is negligible compared with the uncertainty in estimating a ρ that would apply to all cancers. This should be a reasonable assumption when the biological effectiveness of lower-energy photons and electrons is expected to be much less than the effectiveness of alpha particles and fission neutrons (NCRP, 1990; 2012; ICRP, 2003). 2.6 Approach to Evaluation of Biological Effectiveness An evaluation of the effectiveness of lower-energy photons and electrons in inducing cancer in humans is challenging when definitive epidemiologic and radiobiologic data are not in evidence. Epidemiologic studies that could demonstrate, at a high level of confidence, that lower-energy photons and electrons are more effective than high-energy photons in inducing cancer are inherently difficult when very large study populations and highly accurate estimates of cancer risks would be required to observe a presumably small difference in effectiveness (NCRP, 2012). For example, an evaluation by Kocher et al. (2002; 2005) indicated that estimated risks of thyroid and other cancers in atomic-bomb survivors exposed to high-energy gamma rays and medical patients exposed to lower-energy x rays were much too uncertain to observe a difference in biological effectiveness of the two radiation types. The relevance of RBEs obtained from radiobiologic studies in cells and laboratory animals can be questioned on the grounds that 38

1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 many of the endpoints studied are not cancer or directly related to cancer or that RBEs for endpoints judged relevant to cancer may not apply to cancer induction in humans. The biological effectiveness of lower-energy photons and electrons also has been studied using microdosimetric calculations, which are based on models of scattering and absorption of radiation and certain assumptions about the dependence of biological effectiveness on patterns of energy deposition (track structure) in tissue. However, those assumptions may not adequately represent how the chemical and biological processes that lead to cancer or other health effects depend on the energy of photons or electrons (Nikjoo and Lindborg, 2010). Given the lack of definitive evidence from epidemiologic, radiobiologic, or microdosimetric studies, the approach in this Report is to evaluate all lines of scientific evidence that might bear on the question of the effectiveness of lower-energy photons and electrons in inducing cancer in humans. Sections 4 through 7 present evaluations of information derived from the following lines of evidence: Section 4: microdosimetric calculations; Section 5: studies of damage to DNA, including theory, calculations, and experimental data; Section 6: radiobiologic studies in cell and animal systems; and Section 7: human epidemiologic studies. Each section that considers a particular line of evidence presents a review and evaluation of available information. The evaluation of information for each line of evidence leads to the development of a subjective PDF to represent the state-of-knowledge of the biological effectiveness of photons or electrons of specified energies based solely on that line of evidence. The PDF for each line of evidence (e.g., cell studies) can incorporate judgments about the quality of data on RBE for particular endpoints (effects) and the relevance of available data for specific endpoints to induction of cancer in humans, as well as uncertainties in estimates of RBE. Subjective PDFs of biological effectiveness based on the data for the different lines of evidence are denoted by R i, where the subscript i denotes a particular line of evidence. The PDF of each R i 39

1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 is intended to represent the state-of-knowledge on the dose of a reference radiation relative to the dose of a radiation of interest that produce equal effects at low doses or low dose rates of interest in most assessments of cancer risks in humans. However, the quantity R i is not an RBE as that term is strictly defined (Section 2.3), because it is not the result of a specific measurement or calculation and it may be based to a significant degree on judgment. A subjective PDF to represent uncertainties in the effectiveness of photons or electrons of specified energies in inducing cancer in humans at low doses or low dose rates [an effectiveness ratio (ρ) introduced in Sections 2.3 and 2.5] then is developed from the PDFs of R i based on each line of evidence. The ρ values for each of five defined lower-energy groups of photons or electrons (now denoted as ρ L ) represent the ratio of the effect (human cancer risk) of a particular lower-energy group to the effect of a reference radiation at equal doses. The process of developing a PDF of ρ L for each lower-energy group assumes that both ρ L and R i are evaluated at doses sufficiently low that the dose-response relationships of low-let radiations are linear, and it includes an evaluation of the degree-of-belief that the PDF of each R i is indicative of the effectiveness of a lower-energy group in inducing cancer in humans. The approach to uncertainty analysis in this evaluation and the results of that evaluation, which involves Bayesian methods, is described in Section 8 and Appendix A. An important aim of the approach is to provide transparency in assumptions and their effects on results. It is important to emphasize that the PDFs of the quantity R i that represent information on biological effectiveness derived from each line of evidence, and the PDFs of the quantity ρ L that represent the effectiveness of lower-energy photons or electrons in inducing cancer in humans, are subjective representations of the current state-of-knowledge. The PDFs of the effectiveness ratios (ρ L s) developed in this Report clearly do not represent statistical distributions that would result from measurements, if such measurements were possible. In evaluating the effectiveness of photons or electrons of various energies in inducing cancer in humans, an important judgment involves assumptions about the energy range over which a particular PDF of the ρ L applies or, alternatively, assumptions about a functional relationship between an effectiveness ratio and energy in a manner analogous to assumptions about the energy dependence of w R for neutrons used in radiation protection (ICRP, 1991; 2007). Those 40

1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 judgments can be based on the extent to which the biological effectiveness of photons or electrons appears to vary with energy and practical considerations of implementation of assumptions about an energy dependence in cancer risk assessments. In this Report, subjective PDFs of ρ L are developed for the following five lower-energy groups of low-let radiations, which are defined by the type of radiation (photons or electrons) and the energies of the photons or electrons included in each group, where the subscript L refers to one or more of the lower-energy groups being discussed: photons of energy about 1.5 kev; photons of energy in the range of 15 to 30 kev; photons of energy in the range of 40 to 60 kev; photons of energy in the range of >60 to 150 kev; and electrons produced in beta-particle decay of tritium. The reference radiation, with a defined biological effectiveness of unity, is assumed to be photons of energy in the range of about 0.5 to 2 MeV. The lower portion of this range includes the energies of 60 Co and 137 Cs gamma rays that have been used as reference radiations in many radiobiologic studies and the upper portion of this range includes the predominate energies of gamma rays in exposures of Japanese atomic-bomb survivors (Section 2.4). High-energy electrons of equivalent biological effectiveness also are a suitable reference radiation [e.g., the use of 1 MeV electrons as the reference radiation in calculations by Bellamy and Eckerman (2013) and Bellamy et al. (2015) (Section 2.1.2)]. The five lower-energy groups considered in this Report were chosen on the basis of the availability of information on biological effectiveness from radiobiologic or epidemiologic studies or microdosimetric calculations and their importance in occupational or public exposures. The lowest photon energy of about 1.5 kev is representative of ultrasoft x rays of energy less than about 5 kev, which have been widely studied both experimentally and theoretically. Those studies suggest that the biological effectiveness of photons may be highest near 1.5 kev and that the dependence on energy may be most pronounced at the lowest energies (e.g., Nikjoo and 41

1123 1124 1125 1126 1127 1128 1129 Lindborg, 2010). The photon energy ranges of 15 to 30 kev and 40 to 60 kev are representative of spectra of x rays that are widely used in mammography and CT, respectively, and the highest energy range of >60 to 150 kev is representative of spectra of orthovoltage x rays that were used widely in conventional radiology and used as reference radiations in many radiobiologic studies. The spectrum of low-energy electrons from beta-particle decay of tritium has a maximum and average energy of 18.6 kev and 5.7 kev, respectively (ICRP, 2008). 42

1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 3. Spectral Characteristics of Representative Low-LET Radiations When photons (x or gamma rays) are incident on matter, the absorbed dose at a location of interest is mainly a consequence of two processes: 1. production of energetic secondary electrons by the incident photons, referred to as firstcollision electrons; and 2. production of higher generations of lower-energy electrons by interaction of the secondary electrons with atomic electrons in the irradiated material. Detailed descriptions of energy spectra of electrons at specific depths in tissue that are produced by spectra of incident photons are not generally available and may require extensive Monte Carlo calculations. For purposes of this Report, it is sufficient to discuss the two processes that produce energetic electrons. Section 3.1 describes the three processes (the photoelectric effect, Compton scattering, and positron-electron pair production) by which photons incident on matter produce energetic secondary electrons. The remainder of Section 3 discusses spectral characteristics of representative low-let radiations of interest to this Report, including spectra of incident photons or electrons and spectra of electrons produced by the incident photons. Section 3.2 describes spectra of photons for the following sources: 29 kv x rays, which are used in mammography and are assumed to represent photons at energies of 15 to 30 kev; 120 kv x rays, which are used in CT and are assumed to represent photons at energies of 40 to 60 kev; 250 kv x rays, which are representative of orthovoltage x rays that were frequently used in conventional radiology and as reference radiations in radiobiologic studies and are assumed to represent photons at energies of >60 to 150 kev; photons (gamma rays) from beta-particle decay of 60 Co, which are assumed in this Report to be an appropriate reference radiation with a defined biological effectiveness of unity; and 43

1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 photons (prompt and delayed, x and gamma rays) from detonation of the atomic bomb at Nagasaki, Japan, which are representative of spectra of photons in exposures of Japanese atomic-bomb survivors that have been analyzed to estimate risks of most cancer types in humans from exposure to ionizing radiation. Section 3.3 describes spectra of first-collision electrons that are produced by interactions of the photon spectra of interest in tissue. Consideration of those spectra is important when differences in the biological effectiveness of photons of different energies, if any, must arise from differences in the spectra of first-collision electrons. Section 3.3 also presents the spectrum of electrons from beta-particle decay of tritium, which is another low-let radiation of interest to this Report. Finally, Section 3.4 discusses spectra of higher generations of lower-energy electrons that are produced by interaction of first-collision electrons with atomic electrons in tissue. Those electrons are referred to as delta rays. 3.1 Production of Energetic Secondary Electrons by Incident Photons The photoelectric effect, Compton scattering, and positron-electron pair production are the three primary processes by which photons incident on matter produce energetic secondary electrons. The probability of occurrence of each of those processes depends on the photon energy and the atomic number of the irradiated material. In the photoelectric effect, a photon interacts with a bound atomic electron, thereby disappearing with the ejection of an electron, referred to as a photoelectron, of energy equal to E E b, where E is the incident photon energy and E b is the binding energy of the atomic electron. At photon energies above 10 kev, the energy of a photoelectron is much higher than the binding energy of the atomic electron and, thus, is nearly the same as the photon energy. 15 In the spectra of incident photons discussed in Section 3.2, there are few photons of energy below 15 In water (tissue), most photoelectrons are produced by interaction of an incident photon with a K-shell electron in an oxygen atom, which has a binding energy of about 0.53 kev (Nikjoo et al., 1989). Thus, at a photon energy of 10 kev, the energy of a photoelectron is about 95 % of the photon energy. Filling of the resulting vacancy in the oxygen K shell with an electron from an outer shell most often leads to emission of an Auger electron at an energy of about 0.5 kev (Nikjoo et al., 1989), which could have a biological effectiveness per unit absorbed dose greater than the biological effectiveness per unit absorbed dose of the photoelectron (Goodhead et al., 1979). Auger electrons are not included in spectra of first-collision electrons discussed in Section 3.3. 44

1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 10 kev, and the energy of a photoelectron is assumed to be equal to the incident photon energy in the spectra of first-collision electrons presented in Section 3.3. In Compton scattering, a photon is deflected at an angle relative to its original direction by scattering from an atomic electron, resulting in a loss of energy that is transferred to the recoiling electron, referred to as a Compton electron. All angles of scattering of incident photons are possible, and the maximum transfer of energy to a Compton electron occurs at a scattering angle of 180. In contrast to the photoelectric effect, Compton scattering of photons of energy E produces a continuous spectrum of recoil electrons with kinetic energies that range from zero to a maximum equal to E [2α/(1 + 2α)], where α = E/m 0 c 2 and m 0 c 2 is the rest-mass energy of an electron (0.511 MeV). At photon energies of a few tens of kev or lower, relatively little energy is transferred to a Compton electron. As the photon energy increases, the maximum energy of a Compton electron approaches the photon energy. The average kinetic energy of recoil Compton electrons at incident photon energies of 0.01 to 20 MeV is given in Table 3.1. These data, which indicate the increase in energies of Compton electrons relative to the photon energy with increasing photon energy noted above, are useful in interpreting spectra of first-collision electrons produced by photon sources of interest. In positron-electron pair production, a photon of energy greater than 1.02 MeV (twice the electron rest-mass energy), when passing near an atomic nucleus, spontaneously disappears and its energy appears as a positron and an electron, which have a total kinetic energy equal to the photon energy minus 1.02 MeV. Spectra of positrons and electrons produced by photons of a given energy are continuous and range from zero to the total kinetic energy of the pair. Photon interaction cross sections for the three processes in water (a surrogate for tissue) are presented in Figure 3.1. The photoelectric effect is dominant at energies below about 20 kev, and Compton scattering is dominant at energies above about 40 kev. Pair production is of only minor importance at the highest photon energies of interest to this Report (about 15 MeV). 45

1221 1222 1223 1224 1225 1226 Table 3.1 Average kinetic energy of recoil electrons produced by Compton scattering of Photon Energy (MeV) incident photons of various energies a Average Energy of Compton Electrons (MeV) Photon Energy (MeV) Average Energy of Compton Electrons (MeV) 0.01 0.0002 0.80 0.327 0.02 0.0007 1.0 0.440 0.04 0.0027 2.0 1.06 0.06 0.0056 4.0 2.43 0.08 0.0094 6.0 3.86 0.10 0.0138 8.0 5.34 0.20 0.0432 10.0 6.84 0.40 0.124 20.0 14.5 0.60 0.221 a Source: Table 8.1 in Turner (1995). 1227 46

1228 1229 1230 1231 1232 1233 1234 Fig. 3.1. Photon interaction cross sections for photoelectric effect, Compton scattering, and positron-electron pair production in water. 16 16 Eckerman, K.F. (2014). Personal communication (Oak Ridge National Laboratory, Oak Ridge, Tennessee). 47

1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 3.2 Representative Spectra of Incident Photons Representative spectra of photons for sources of interest to this Report are presented in Figure 3.2 [(a) and (b)] and Figure 3.3 [(a) and (b)], where E is the photon energy and n(e) is the number distribution per photon. 17 When these spectra are plotted as E n(e) on a linear scale versus E on a log scale, the area under a curve between any two energies gives the number fraction of photons in that energy range. Not shown in Figures 3.2 and 3.3 are the energies of the two 60 Co gamma rays of equal intensity at 1.17 and 1.33 MeV. The spectra of x rays and Nagasaki photons in Figures 3.2 and 3.3 were calculated on the basis of the following assumptions: 29 kv x rays Mo target, 0.1 mm Al inherent filtration, 0.03 mm Mo filtration; 120 kv x rays W target, 0.5 mm Al inherent filtration, 3 mm Al filtration; 250 kv x rays W target, 1.2 mm Al plus 1 mm Cu inherent filtration; and Nagasaki photons Distance from hypocenter of 1,500 m and DS02 dosimetry. The x-ray spectra include discrete characteristic x rays produced by de-excitation of atomic electrons in the target material (molybdenum or tungsten) and a continuous spectrum of bremsstrahlung with a maximum energy equal to the peak tube potential and an average energy of about one-third of the maximum. The spectra of 120 kv and 250 kv x rays are presented in separate figures to allow contributions from the characteristic tungsten x rays in the two spectra to be shown separately. In Figures 3.2(b) and 3.3(b), peaks from the most intense characteristic x rays are truncated to show more details at lower intensities in all spectra. Average (mean) photon energies in the spectra of 29 kv, 120 kv, and 250 kv x rays are 17, 57, and 109 kev, respectively. Characteristic x rays are the most prominent in the spectrum of 29 kv x rays and contribute about 27 % of all photons, whereas the contributions from 17 In all spectra of incident photons and first-collision electrons presented in Sections 3.2 and 3.3, the values of E n(e) or E 2 n(e) on the vertical axis depend on the chosen size of an energy bin for normalizing the total area under a curve to unity and, thus, have no particular meaning. 48

6 x 10-3 5 x 10-3 29 kv x rays 120 kv x rays Nagasaki photons 4 x 10-3 E*n(E) 3 x 10-3 2 x 10-3 1 x 10-3 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 0 0.01 0.1 1 10 E(MeV) Fig. 3.2(a). Representative spectra of 29 kv and 120 kv x rays and photons at Nagasaki: E is photon energy, and n(e) is number distribution per photon. 18 Area under a curve between any two energies gives number fraction of photons in that energy range. Not shown are energies of 60 Co gamma rays at 1.17 and 1.33 MeV. NOTE: In Figure 3.2(a) and (b), Figure 3.3 (a) and (b) and Figure 3.4 (a) and (b), the spectra presented in each figure will be more clearly presented, perhaps in different colors. 18 Eckerman, K.F. (2014). Personal communication (Oak Ridge National Laboratory, Oak Ridge, Tennessee). 49

1 x 10-3 8 x 10-4 29 kv x rays 120 kv x rays Nagasaki photons 6 x 10-4 E*n(E) 4 x 10-4 2 x 10-4 1275 1276 1277 1278 0 0.01 0.1 1 10 E(MeV) Fig. 3.2(b). Same as Fig. 3.2(a) with peaks in x-ray spectra truncated. 50

6 x 10-3 5 x 10-3 29 kv x rays 250 kv x rays Nagasaki photons 4 x 10-3 E*n(E) 3 x 10-3 2 x 10-3 1 x 10-3 1279 1280 1281 1282 1283 1284 1285 0 0.01 0.1 1 10 E(MeV) Fig. 3.3(a). Representative spectra of 29 kv and 250 kv x rays and photons at Nagasaki: E is photon energy, and n(e) is number distribution per photon. 19 Area under a curve between any two energies gives number fraction of photons in that energy range. Not shown are energies of 60 Co gamma rays at 1.17 and 1.33 MeV. 19 Eckerman, K.F. (2014). Personal communication (Oak Ridge National Laboratory, Oak Ridge, Tennessee). 51

1 x 10-3 8 x 10-4 29 kv x rays 250 kv x rays Nagasaki photons 6 x 10-4 E*n(E) 4 x 10-4 2 x 10-4 1286 1287 1288 1289 0 0.01 0.1 1 10 E(MeV) Fig. 3.3(b). Same as Figure 3.3(a) with peaks in x-ray spectra truncated. 52

1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 characteristic x rays are <10 % in the spectra of 120 kv and 250 kv x rays. 20 The mean energy of the two 60 Co gamma rays is 1.25 MeV, and the mean energy in the spectrum of photons at Nagasaki is 3.57 MeV. 3.3 Spectra of First-Collision Electrons and Tritium Beta Particles Spectra of first-collision electrons produced by the photon spectra of interest in water (tissue) are presented in Figure 3.4 [(a) and (b)], where E is the electron energy and n(e) is the number distribution per electron. In Figure 3.4(b), peaks in the spectra of electrons produced by 29 kv x rays and 60 Co gamma rays are truncated to show more details at lower intensities in all spectra. When these spectra are plotted as E 2 n(e) on a linear scale versus E on a log scale, the area under a curve between any two energies gives the fraction of the total dose due to electrons in that energy range. In the spectrum of first-collision electrons produced by 29 kv x rays, about 25 % of the total dose is due to photoelectrons produced by the characteristic molybdenum x rays. In the spectra of first-collision electrons produced by 120 kv and 250 kv x rays, contributions to the total dose from photoelectrons produced by the characteristic tungsten x rays are much lower, due to the dominant importance of the Compton effect at those photon energies. Average (mean) energies of first-collision electrons produced by 29 kv, 120 kv, and 250 kvx rays are 14, 10, and 20 kev, respectively. The mean energy of the electrons produced by 29 kv x rays is close to the mean photon energy of 17 kev, due to the dominant importance of the photoelectric effect at photon energies below about 20 kev. In contrast, mean energies of the electrons produced by 120 kv and 250 kv x rays (10 and 20 kev, respectively) are much lower than the mean photon energies of 57 and 109 kev, respectively, due to the dominant importance of Compton scattering at photon energies above about 40 kev and the much lower energies of Compton electrons relative to the photon energies. The mean energy of the electrons 20 Some radiobiologic data suggest that the mean photon energy in a spectrum of x rays may not be a valid indicator of relative biological effectiveness. In studies of induction of dicentric chromosome aberrations in human blood lymphocytes discussed in Section 6.2, for example, an estimated RBE M for 29 kv x rays, with a mean energy of 17.4 kev, was about twice the RBE M for 17.4 kev photons (NCRP, 2012; Schmid et al., 2002; 2003) A similar result was obtained when estimates of RBE M for 60 kv x rays, with a mean energy of 48 kev, and 40 kev photons were compared. Differences in those estimates of RBE M were interpreted as indicating that a large fraction of the effects induced by the spectra of x rays should be attributed to photons of energy well below the mean. 53

1316 E 2 *n(e) 5 x 10-3 4 x 10-3 3 x 10-3 2 x 10-3 First-collision electrons 29 kv x rays 120 kv x rays 250 kv x rays Co-60 photons Nagasaki photons 1 x 10-3 1317 1318 1319 1320 1321 1322 1323 1324 1325 0 0.001 0.01 0.1 1 10 E(MeV) Fig. 3.4(a). Spectra of first-collision electrons in water (tissue) produced by representative spectra of 29 kv, 120 kv, and 250 kv x rays, 60 Co photons (gamma rays), and representative spectrum of photons at Nagasaki: E is electron energy, and n(e) is number distribution per electron. 21 Area under a curve between any two energies gives fraction of total dose due to first-collision electrons in that energy range. 21 Eckerman, K.F. (2014). Personal communication (Oak Ridge National Laboratory, Oak Ridge, Tennessee). 54

E 2 *n(e) 1 x 10-3 8 x 10-4 6 x 10-4 4 x 10-4 First-collision electrons 29 kv x rays 120 kv x rays 250 kv x rays Co-60 photons Nagasaki photons 2 x 10-4 1326 1327 1328 1329 1330 0 0.001 0.01 0.1 1 10 E(MeV) Fig. 3.4(b). Same as Figure 3.4(a) with peaks in spectra of first-collision electrons produced by 29 kv x rays and 60 Co photons (gamma rays) truncated. 55

1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361 produced by 120 kv x rays is the lowest, and mean energies of the electrons produced by 29 kv and 250 kv x rays do not differ greatly. To illustrate the differences described above, the energy spectra of 29 kv and 120 kv x rays and first-collision electrons produced by those x rays, plotted as E n(e) versus E on linear scales, are compared in Figures 3.5 and 3.6, respectively. In the spectra of first-collision electrons, the area under a curve between any two energies gives the fraction of the total dose due to electrons in that energy range, whereas in the x-ray spectra, the area under any portion of a curve gives the fraction of the total energy carried by photons in that energy range. The similarity of the two spectra in Figure 3.5, which is a consequence of the dominant importance of the photoelectric effect for 29 kv x rays, and the pronounced difference in the two spectra in Figure 3.6, which is a consequence of the dominant importance of Compton scattering for 120 kv x rays, are apparent. In Figure 3.6, the broad peaks at low energies in the spectrum of first-collision electrons are Compton electrons produced by the two groups of characteristic tungsten x rays centered at about 59 and 68 kev. The mean of the dose-weighted distribution E n(e) (as plotted in Figures 3.5, 3.6 and 3.7) is the dose mean energy, which is the energy at the mean dose due to all first-collision electrons. Dose mean energies in the spectra of first-collision electrons produced by 29 kv, 120 kv, and 250 kv x rays are 15, 20, and 37 kev, respectively. In contrast to the mean electron energies noted above, the dose mean energy increases with increasing mean photon energy in the spectra of incident x rays. This increase from mean electron energy to dose mean energy provides an indication of the relative dose contribution from higher-energy electrons. The increase is much greater for 120 kv and 250 kv x rays than for 29 kv x rays. If it is assumed that the biological effectiveness of photons would increase with decreasing dose mean energy of first-collision electrons at photon energies of importance in the three x-ray spectra, the biological effectiveness of 29 kv x rays might be somewhat greater than the biological effectiveness of 120 kv x rays, which in turn might be somewhat greater than the biological effectiveness of 250 kv x rays. Whether these statements are true depends on whether differences in the spectra of first-collision electrons produced by the three x-ray spectra are 56

3.5 x 10-1 3 x 10-1 2.5 x 10-1 29 kv x rays X rays First-collision electrons E*n(E) 2 x 10-1 1.5 x 10-1 1 x 10-1 5 x 10-2 1362 1363 1364 1365 1366 1367 1368 1369 1370 0 0 0.005 0.01 0.015 0.02 0.025 0.03 E(MeV) Fig. 3.5. Energy spectra of 29 kv x rays and first-collision electrons produced by those x rays: E is photon or electron energy, and n(e) is number distribution per photon or electron. 22 In x- ray spectrum, area under curve between any two energies gives fraction of total energy carried by photons due to photons in that energy range. In spectrum of first-collision electrons, area under curve between any two energies gives fraction of total dose due to electrons in that energy range. 22 Eckerman, K.F. (2014). Personal communication (Oak Ridge National Laboratory, Oak Ridge, Tennessee). 57

6 x 10-2 5 x 10-2 120 kv x rays X rays First-collision electrons 4 x 10-2 E*n(E) 3 x 10-2 2 x 10-2 1 x 10-2 1371 1372 1373 1374 1375 1376 1377 1378 1379 0 0 0.02 0.04 0.06 0.08 0.1 0.12 E(MeV) Fig. 3.6. Energy spectra of 120 kv x rays and first-collision electrons produced by those x rays: E is photon or electron energy, and n(e) is number distribution per photon or electron. 23 In x-ray spectrum, area under curve between any two energies gives fraction of total energy carried by photons due to photons in that energy range. In spectrum of first-collision electrons, area under curve between any two energies gives fraction of total dose due to electrons in that energy range. 23 Eckerman, K.F. (2014). Personal communication (Oak Ridge National Laboratory, Oak Ridge, Tennessee). 58

1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 preserved to a sufficient degree when higher generations of lower-energy electrons that deliver most of the dose are produced (Section 3.4). In Figure 3.4, a difference in the spectra of first-collision electrons produced by 60 Co gamma rays and Nagasaki photons, which have mean energies of 0.65 and 2.12 MeV and dose mean energies of 0.83 and 2.26 MeV, respectively, is apparent. However, this difference might be largely damped out and lost as the energies of first-collision electrons in those spectra are degraded over many generations of interaction to the much lower energies of electrons that deliver most of the dose (see Section 3.4). If so, the biological effectiveness of the two radiations might be about the same, as assumed in this Report. Figure 3.7 compares the spectrum of first-collision electrons produced by 29 kv x rays (mean and dose mean energies of 14 and 15 kev, respectively) with the spectrum of electrons from beta-particle decay of tritium, which has an endpoint energy of 18.6 kev, a mean energy of 5.7 kev, and a dose mean energy of 5.9 kev. When E n(e) and E are plotted on linear scales, the area under a curve between any two energies gives the fraction of the total dose due to electrons in that energy range. If it is assumed that the biological effectiveness of electrons would increase with decreasing energy below about 20 kev, these data suggest that the biological effectiveness of tritium beta particles might be greater than the biological effectiveness of 29 kv x rays. 3.4 Spectra of Lower-Energy Electrons Produced by First-Collision Electrons First-collision electrons deposit their energy by interacting with atomic electrons of the irradiated material. Since all electrons have the same rest mass, a first-collision electron can transfer up to its entire energy to another, secondary, electron. Some of these secondary electrons are given enough energy that they produce recognizable tracks of additional ionizations, and are referred to as delta rays. These delta rays also interact with atomic electrons and produce higher generation electrons, some of which have sufficient energy to produce additional delta-ray tracks. The energy-dependence of the spectrum of electrons produced by interaction of first-collision electrons is approximately 1/E 2, up to the maximum of the energy of the incident electron. As a result, the spectrum of electrons which produce ionization and 59

3 x 10-1 2.5 x 10-1 29 kv first-collision electrons H-3 beta particles 2 x 10-1 E*n(E) 1.5 x 10-1 1 x 10-1 5 x 10-2 1412 1413 1414 1415 1416 1417 1418 0 0 0.005 0.01 0.015 0.02 0.025 0.03 E(MeV) Fig. 3.7. Energy spectra of first-collision electrons produced by 29 kv x rays and electrons from beta-particle decay of tririum: E is electron energy, and n(e) is number distribution per electron. 24 Area under a curve between any two energies gives fraction of total dose due to electrons in that energy range. 24 Eckerman, K.F. (2014). Personal communication (Oak Ridge National Laboratory, Oak Ridge, Tennessee). 60

1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 excitation in irradiated material is dominated by low-energy secondary and higher-generation electrons. The result is illustrated in Figure 3.8. Figure 3.8 gives the mean number of electrons with starting energy greater than E for incident electrons with initial energies E 0 of 0.1, 1, 10 and 100 kev. The curves are normalized by E 0, so, for example, each 1 kev first-collision electron results, on average, in about 8 electrons with energy >20 ev [Figure 3.8 gives N(E)/E 0 = 8 x 10-3 ev -1, so N(E) = 8 x 10-3 ev -1 x 1 kev = 8], including two electrons of energy >100 ev. On average, each 100 kev electron results in ~150 electrons of energy >100 ev, including ~15 of energy >1 kev and two of energy >10 kev. As a result, a large fraction of the energy initially carried by first-collision electrons is deposited in clusters at the ends of a large number of delta-ray tracks. The markers in Figure 3.8 show that 50 % of the ions produced when a 100 kev electron interacts with tissue are actually produced by electrons (delta rays) of energy <800 ev. As a result, differences in the spectra of electrons depositing energy are substantially less than the differences in the first-collision electron spectra produced by incident photons. Assuming that biological effectiveness is related in some way to the characteristics of energy deposition, it is likely that differences in firstcollision electron spectra will overestimate differences in biological effectiveness. The dashed line in Figure 3.8, the mean free path for ionization, indicates that electrons of ~100 ev result in the minimum average spacing between ionizations, resulting in clusters of ~5 ions. 61

1439 1440 1441 1442 1443 1444 1445 1446 1447 Fig. 3.8. Mean numbers of secondary and higher generation electrons with starting energies greater than E produced in water by electrons with initial energies E 0 of 0.1, 1, 10 and 100 kev. Values of N(E) have been divided by E 0. Arrows indicate median electron energies of the slowing-down spectra produced by the indicated primary electron energy. Dashed line indicates the mean distance between ionizations as a function of electron energy (Paretzke, 1987). NOTE: The slanted figure will be fixed. 62

1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 4. Line of Evidence: Microdosimetry The basic assumption underlying the study of the effects of ionizing radiation on biological systems, and therefore the health risks resulting from radiation exposure, is that ionizing radiation initiates chemical changes which start a sequence of biochemical and biological changes eventually leading to the endpoint being studied. This is the basis for developing mathematical models to predict biological response based on the physical characteristics of the energy deposition by different radiations. Both ionization and excitation of atoms or molecules can initiate chemical changes, but ionization can initiate a wider range of changes and is usually assumed to be the dominant source of DNA damage. Because it is not known which specific chemical changes initiate a specific biochemical process, and because the probability of producing a specific chemical change is related to the amount of energy deposited by ionizing radiation, irradiation is typically described in terms of the energy deposited per mass of target material, the absorbed dose. In the case of high-energy photon (x or gamma ray) irradiation essentially all of the energy is actually deposited by the energetic electrons produced when the photons interact with the irradiated material. Photons generally transfer much of their energy to a single electron during Compton scattering, all of their energy to a single electron during the photoelectric interaction, or all of their energy to an electron and a positron in pair production. In biological materials the photoelectric effect dominates for low-energy photons, and Compton scattering dominates for higher-energy photons, so only a few ionized molecules are produced by the typical photon itself. However, the energetic electrons that are produced in those interactions transfer some energy to many of the molecules and electrons that they pass, and deposit nearly all of the photon s original energy. The high-energy electrons interact with the electrons of the irradiated material via Coulomb forces, transferring any amount of energy up to the total kinetic energy of the electron. These interactions are random in nature, with the PDF of energy transferred to a target electron depending on the velocity of the incident electron, the distance between the path of the incident electron and the target electron, and the binding of the target electron to its atom or molecule. Most interactions are at large impact parameters and result in very small energy depositions, typically enough to excite the target molecule, but not enough to actually knock an electron out of it. A small fraction of the interactions result in ionization of the 63

1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512 target molecule, and give the freed electron a small kinetic energy. A small fraction of those interactions ionize the target and give the freed electron (then known as a delta ray) sufficient energy for it to ionize other molecules. The collection of the ionizations and excitations produced by an electron is referred to as its track (ICRU, 2011). Since the cross sections for nearly all of the energy transfer interactions between electrons and biological materials are known with reasonable accuracy, it is possible to use Monte Carlo calculations to simulate individual electron tracks, as illustrated in Figure 4.1. The results of these simulations illustrate several characteristics of electron tracks. First, there are more excitations than ionizations along the track. Second, the average distance between ionizations or excitations decreases as the incident electron energy (or velocity) decreases, reflecting the fact that cross sections for most types of interaction increase as the electron velocity decreases, down to about 100 ev. At lower energies the cross sections decrease and the distance between the last few interactions increases (Figure 3.8). Third, the tracks show substantial random changes in direction, and fourth, the distance between successive ionizations along an individual electron track varies widely, reflecting the random nature of the various interactions. The result of this randomness is that there are significant clusters of ions as well as random gaps in individual electron tracks. A direct consequence of our initial assumption, that biological consequences of irradiation are initiated by the chemical changes produced by electron interactions, is that any differences in the biological effects of the same dose of different radiations must be related to the differences in the track structures of those radiations. The differences can be expressed in many ways. Perhaps the most commonly used characterization of a charged-particle track is the electron (nonradiative) stopping power, also known at the unrestricted LET, the expectation value of the rate of energy loss at a point along the track. Other ways of describing the differences in the way a fixed amount of energy is deposited by electrons of different energies include the total electron track length and the probability density of energy deposited in specified volumes of approximately DNA feature or cellular dimensions, the lineal energy (y). There are two quantities traditionally used to describe the interaction of photons with the irradiated material. The kerma is the expectation value of energy lost by photons per mass at a point. The absorbed dose is the expectation value of the energy absorbed per mass at a point. 64

1513 1514 1515 1516 1517 1518 1519 1520 1521 1522 1523 Fig. 4.1. Monte Carlo simulation of electron tracks in water vapor calculated with a Monte Carlo simulation program that follows particles down to 10 ev; red points represent ionization, and green points represent excitation (ICRU, 2011). 65

1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549 1550 1551 1552 1553 1554 1555 1556 Both of these quantities are specified in joules per kilogram (gray) but their values are not identical except in cases where secondary particle equilibrium occurs. That is, when the expectation value of energy carried by secondary electrons leaving the volume is exactly matched by the expectation value of energy carried into the volume by secondary electrons formed elsewhere. When the absorbed dose is large the average distance between electron tracks is small compared to the length of those tracks so several tracks will intersect each small volume within the irradiated material and the relative variance in the amount of energy deposited in adjacent small volumes is small. In such situations the absorbed dose is a realistic representation of the amount of damage. However, when the average distance between tracks is large compared to the track length many small volumes will not be intersected by any track and the variation in the energy deposited in small volumes will be much larger. The energy deposited in an individual small volume (site) during a given time interval is often zero, but ranges up to many times the expectation value. In that situation, which is often the case when the absorbed dose is low, the average quantities such as absorbed dose and LET are inadequate descriptions of the energy deposited in or near individual biomolecules. As a result absorbed dose and kerma are not useful indicators of biological response. In order to characterize the radiation it is necessary to consider the probability density of either the energy of the electrons at the point of interest, the energy deposited in a small volume representative of the volume where the chemistry which initiates the radiation response occurs, or some other measure of the clustering of energy deposition along electron tracks. The probability density of the electron energy, the particle radiance, at the point of interest has the advantage that any desired description of the clustering of energy deposition can be calculated from it and a description of the atomic composition and density of the target, and has been recommended as the preferred way to characterize heterogeneous radiation exposures (ICRU, 2011). However, very little data on biological damage as a function of spectral radiance is available, and the spectral radiance can be difficult to ascertain in some radiation exposure situations. It may be easier to measure the probability density of the lineal energy [f(y)] of the energy deposited by individual electron events where y is the energy imparted in a defined 66

1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 volume divided by the mean chord length of that volume. Lineal energy is the stochastic analog of linear energy transfer (LET). However, it differs from LET because it includes the effect of energy loss straggling and energy transport by delta rays. Although f(y) is relatively easily obtained for most radiations it has the disadvantage that it is a function of the size and geometry of the site used in the determination, and measurements for several different site sizes, or spectral radiance data, would be needed to determine f(y) for other site sizes. Attempts to describe clustering along electron tracks, for example the proximity function 25 (Kellerer and Chmelevsky, 1975), as a way of characterizing radiation exposures have resulted in descriptions which are generally too complex to be widely used. Although the spectral radiance of secondary electrons has not been experimentally determined or evaluated through Monte Carlo calculations for most photon irradiations, there are a number of features of these distributions that can be predicted through consideration of the fundamental energy-deposition processes and have been described in Section 3. Computed and experimental data on f(y), along with their likely impact on biological effectiveness as represented by R i, will be described in the subsections of Section 4. 4.1 Lineal Energy Distributions Produced by Photons On the assumption that energy absorbed in some volume initiates the chemical reactions that eventually lead to a biological consequence, the lineal energy, y ε l, where ε is the energy imparted and l is the mean chord length of the unit density volume of interest, may be the quantity most directly related to biological damage. Alternativly, on the assumption that the probability of a biological consequence depends on the spatial distribution of chemical changes within a particle track, y may be a useful measure of that spatial distribution. However, ε l is dependent on the size and geometry of the volume in which ε is evaluated. Four characteristics of charged particle tracks are primarily responsible for this variation; the dependence of stopping power on particle velocity, the production of energetic delta rays, energy loss straggling, and the limited range of low-energy particles. If the diameter of the site is larger than the particle range, the particle will lose all of its energy in the site and ε will equal that energy. The value of y will 25 The proximity function is essentially the probability density of the distance between ionizations, for each pairing of ions produced by a charged-particle track. 67

1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 then be lower than it would be in a smaller site since ε will be the same but l will be larger. Thus, only if the range is sufficient for the particle to cross the site will y be indicative of the ionization density along the track. Independent of the site size, y is indicative of ϵ, the energy deposited in the site, and the related quantity, specific energy, z = ϵ/m, where m is the mass of the site, may be useful. Similarly, if the site is small the energy loss straggling will result in substantial differences in the energy deposited by independent tracks, thus resulting in larger relative variance in f(y) in smaller sites. Delta rays carry more energy out of small sites than out of large sites resulting in lower values of average y in small sites and in the delta-ray effect being more significant for higher-energy electrons since their average delta-ray range is longer. Finally, since the stopping power reaches its maximum value for only a very narrow range of energies, the stopping power averaged over the path through a site, and therefore y, decreases as the path length increases. Depending on the chemical or biological process being initiated, the energy deposition in volumes ranging from a few nanometers in diameter to larger than the cell nucleus, corresponding to the rate of energy deposition averaged over some small distance, may determine the biological effectiveness. In a very small volume containing the end of an electron track, y may be as large as 100 kev µm 1, while in a sphere a micrometer or more in diameter, crossed by a high-energy electron track, y may be less than 0.5 kev µm 1, but no more than about 15 kev µm 1. This relatively low value of y occurs even though that volume contains several electron track ends which result in high values of y in very small regions within the larger volume. Unfortunately there is no clear indication of the relevant site size for specific biological endpoints. For example, for DNA strand breaks the relevant site size is on the order of 10 nm or less (Goodhead, 1994a), while for cell lethality at low doses some studies have suggested similarly small site sizes (Brenner and Zaider, 1984; Goodhead and Brenner, 1983) but others have suggested that the relevant site size is on the order of 0.2 µm (Neary et al., 1967), or 1 µm (Kellerer and Rossi, 1972) or assumed even larger site sizes similar to a cell nucleus (Bond and Varma, 1983). A corresponding variety of site sizes have been suggested for other cellular effects relevant to carcinogenesis. For some processes which are related to the bystander effect the overall target size is likely to be much larger than a single cell (>10 µm) (Morgan, 2003a), but the nature of the sensitive sites for the initiating radiation damage is less clear. 68

1619 1620 1621 1622 1623 1624 1625 1626 1627 1628 1629 1630 1631 1632 1633 1634 1635 1636 1637 1638 1639 1640 1641 1642 1643 1644 Lineal energy distributions can be obtained experimentally or through Monte Carlo calculations, but experimental values for sites less than about 0.5 µm in diameter are hard to obtain because of limitations on the gas gain of detectors simulating such small sites. The variance method (Bengtsson, 1970) can be used to determine the dose mean lineal energy, y d 26, for smaller site sizes, but few measurements are available. Furthermore, data for detector volumes in a uniform homogeneous medium, typically requiring a wall less detector, are needed in order to avoid distortions in the measured distributions due to wall effects. As a result there is relatively little published data that is directly relevant to the RBE for photons of different energies. Figure 4.2 gives examples of the probability density, f(y), and the cumulative probability F(y) for 60 Co gamma rays and three typical x-ray spectra as measured by a wall-less detector inside a gas volume surrounded by a tissue equivalent plastic wall (Ellettt and Braby, 1972). As expected, as the average photon energy increases, resulting in the average electron energy increasing, the average lineal energy decreases. The maximum value of y for any of the photon spectra in this size site is about 15 kev µm 1, which is determined primarily by the maximum average stopping power and scattering of low-energy electrons crossing the site at the maximum chord length. Very small energy depositions, corresponding to single ionizations, are relatively common, but difficult to measure due to electronic noise. Consequently, f(y) measured by low pressure wall-less proportional counters becomes progressively more uncertain at lower values of y. Since it is often inconvenient to compare detailed f(y) distributions it is common to compare radiations in terms of the mean (or expectation) values. Because f(y) is likely to have large numbers of small events, where electronic noise makes it difficult to determine f(y), the frequency mean, y y y y, which is strongly influenced by the total number of events, has relatively large uncertainty while the mean value of y weighted by the contribution of each value of y to the absorbed dose, the dose mean, y y 2 y y y y y, is much less uncertain. d 26 The dose mean is the average of the quantity in terms of its contribution to the dose, whereas the frequency mean is the average value of the quantity itself. 69

1645 0.08 0.92 µm diameter 0.07 0.06 0.05 yf(y) 0.04 0.03 0.02 0.01 1646 1647 0 0.01 0.1 1 10 100 y, kev/µm 1 0.1 0.01 F(y) 0.001 0.0001 0.00001 0.000001 0.01 0.1 1 10 100 y, kev/µm 60Co 250 kv, HVL= 1.88 mm Cu 250 kv, HVL=0.44 mm Cu 65 kv, HVL=1.9 mm Al 1648 1649 1650 1651 1652 1653 tritium Fig. 4.2. Probability density, f(y) and cumulative probability, F(y) in a 0.92 µm diameter, spherical, tissue equivalent site in a uniform homogeneous tissue medium (Ellett and Braby, 1972). 70

1654 1655 1656 1657 1658 1659 1660 1661 1662 1663 1664 1665 1666 1667 1668 1669 1670 1671 1672 1673 1674 1675 1676 1677 1678 1679 1680 1681 It should be noted that, although the definition of y does not exclude any value of the mean chord length, f(y) for site sizes that are comparable to or larger than the range of the electrons produced by incident photons is of little use for evaluating track structure. If the site size is large the energy imparted, ε, will be the energy of the electron and if the site size increases further ε remains constant but l increases so y decreases. Consequently use of lineal energy to evaluate R i will be limited to situations where the site diameter is smaller than the range of most of the incident electrons or to estimates derived from models that are based on the total energy imparted rather than the track structure. Experimental results for dose distribution, d(y) = y f(y) 27, in wall-less detectors simulating a range of site sizes are illustrated in Figure 4.3. The data in Figures 4.2 and 4.3, as is customary in microdosimetry (ICRU, 1986), are plotted as y f(y) or y d(y) versus log y so that equal areas under the curve represent equal numbers of events or equal doses, respectively. Figure 4.3, for 25.3 kev photons, producing electrons with range of about 12 µm, illustrates the effect of the range of the electron relative to the diameter of the sensitive site. For small sites the particle can cross the diameter and the energy deposited shows the variation due to the distribution of path lengths, changing stopping power,and energy loss straggling. For larger sites the electron range is not much greater than the site diameter, and the energy deposited is primarily the energy of the electron, modified by wall effects. The change in the shapes of these curves, as well as the mean values, illustrates the significance of the site size when using microdosimetry data to estimate the biological consequences of irradiation. Experimentally evaluated f(y) and d(y) distributions have been obtained using various detector designs, but in all cases the detector volume is located in a gas-filled volume large enough to minimize wall effects and surrounded by an electrically conducting plastic which is intended to be tissue equivalent. Consequently these data are specific to the secondary electron spectrum produced in the plastic wall and filling gas. The tissue equivalent plastic most commonly used (A150; Shonka, 1958) and the gasses used (typically mixtures of a hydrocarbon 27 d(y) and f(y) are probability densities and are, by definition, normalized and unitless. d(y) is the density of dose, which is proportional to the energy per event times the number of events at each value of y. Normalization, dividing by the sum of y f(y) from 0 to infinity, results in the unitless d(y). 71

1682 1683 1684 1685 1686 1687 1688 1689 1690 1691 1692 1693 1694 1695 Fig. 4.3. Dose distributions for 25.3 kev photons in tissue equivalent sites from 0.25 to 8.0 µm in diameter obtained with a wall less tissue equivalent proportional counter (Kliauga and Dvorak, 1978). NOTE: Before publication, the values on Figure 4.3 given in ug/cm 2 (bottom right within the graph) will be converted to um (e.g., 28 ug/cm 2 will become 0.28 um) for consistency with the rest of this Report. NOTE: The slanted figure will be fixed. Fig. 4.3. Dose distributions for 25.3 kev photons in tissue equivalent sites from 0.25 to in diameter obtained with a wall less tissue equivalent proportional counter. (Kliauga and Dvora1978) 72

1696 1697 1698 1699 1700 1701 1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 1720 1721 1722 1723 1724 with nitrogen and CO 2 ) do not contain phosphorus and sulfur, and so do not produce exactly the same secondary electron spectrum found, for example, in a phosphorus rich cell nucleus. Although there is some variation in experimental details the available data are typically for a depth of approximately 6 mm in a 12 mm diameter tissue sphere. Monte Carlo calculations of the energy deposited in volumes of specified size in a uniformly irradiated homogeneous medium can also be used to evaluate f(y). Monte Carlo calculations are not subject to electronic noise or gas avalanche based restrictions on the site size, but are subject to errors due to uncertainty in cross sections and are typically limited to evaluation of energy deposition in small sites (less than 1 µm) due to the large amounts of computer time required to simulate long segments of electron tracks. These calculations are complicated by low probability that delta rays will reach the volume of interest if the primary track does not cross that volume. Since the Monte Carlo calculation is inherently stochastic, and available computer time limits the number of tracks that can be simulated, there is some uncertainty in each calculation of average quantities such as y. Energy deposition in small cylindrical volumes, with length equal to the diameter and suggestive of the sizes of features of DNA and other components of chromatin, was initially calculated using water vapor cross-section data (Nikjoo et al., 1994a; 1994b). Subsequent comparisons have been made using cross-section data for liquid water (Liamsuwan et al., 2012; Nikjoo et al., 1994c). The high stopping power of the low-energy electrons results in large values of y, while short segments of higher-energy electron tracks produce only a few ionizations and correspondingly low values of y. Thus y will decrease slowly as the electron energy d increases. Frequency mean and dose mean values of y for different photon energies and site sizes are given in Table 4.1. The origins of these values, experimental or calculated, are listed in the footnotes. 73

Table 4.1 Frequency mean and dose mean lineal energy, y and y in kev µm 1, as a function of site size and photon energy. d Simulated Diameter (µm) Tritium Beta Particles 1.5 kev 15 to 30 kev 40 to 60 kev >60 to 150 kev 60 Co 0.0023 a y = 15 b d c y = 16 y = 6.9 y = 6.6 y = 6.7 y = 25 b y = 28 c y = 14.1 d y = 12.8 d d d d d d d d y = 12.4 d no data for y y = 30 k d 0.01 a no data for y or y d no data for y or y d no data for y or y d no data for y or y d no data for y or y d no data for y y = 15 k d 0.03 a y = 8 b y = 14 y = 4.9 y = 19 b y = 24 c d d c d d y = 18.8 d y = 2.1 d y = 10.5 n d y = 1.6 d y = 6.3 d d no data for y y = 10 k d y = 11 n d y = 5.4 d d y = 10 n d y = 8.3 n d 0.1 a y = 7 b y = 10 y = 4.1 y = 14 b y = 15 c d d c d d y = 13.4 y = 8 n d no data for y y = 0.72 d y = 7.7 n y = 3.5 d d d y = 7.2 n d no data for y y = 5.2 k d y = 8.0 m d y = 5.5 n d 0.5 y = 3.75 e y = 6.88 e d no data for y or y d y = 2.56 f y = 1.99 g y = 1.22 o y = 4.8 n d y = 1.36 h y = 4.45 h d y = 0.46 i y = 0.36 j y = 6.42 f d y = 4.42 o d y = 4.1 n d y = 2.16 i d y = 4.91 g d y = 2.1 j d y = 4.6 n d y = 2.0 k d y = 2.6 n d 74

1 y = 3.06 e y = 5.39 e d no data for y or y d y = 2.13 f y = 1.88 g y = 1.14 o y = 3.8 n d y = 1.19 h y = 3.80 h d y = 0.38 i y = 0.28 j y = 5.14 f d y = 4.09 g d y = 3.5 o d y = 3.4 n d y = 1.65 i d y = 1.59 j d y = 3.5 n d y = 1.66 k d y = 1.4 m d y = 2.0 n d 2 y = 2.39 e y = 3.92 e d no data for y or y d y = 2.0 f y = 1.18 g y = 1.04 o y = 2.77 o y = 3.16 h y = 0.26 j y = 1.15 y = 0.33 d d h i y = 4.07 f d y = 1.42 i d y = 3.47 g d y = 1.22 j d 3 y = 1.84 e y = 2.91 e d no data for y or y d y = 1.84 f y = 1.03 y = 1.10 y = 0.29 y = 3.4 f y = 2.14 o y = 2.61 h d o d d h y = 0.99 i d i 1725 1726 1727 1728 1729 1730 1731 1732 1733 1734 1735 1736 1737 1738 1739 5 y = 1.38 e y = 2.08 e d no data for y or a Cylindrical sites with length equal to diameter. y d b Liamsuwan et al. (2012), 5,000 ev data (calculation). y = 1.55 f c Liamsuwan et al. (2012), average of 1 and 2 kev data (calculation). d Nikjoo et al. (1994a) (calculation). e Ellett and Braby (1972) (measurement). f Ellett and Braby (1972), 65 kv x rays (measurement). g Kliauga and Dvorak (1978), Sn K α x ray (measurement). h Ellett and Braby (1972), 250 kv x ray (measurement). i Ellertt and Braby (1972), 60 Co (measurement). j Kliauga and Dvorak (1978), 60 Co (measurement). k Grindborg and Olko (1997) (calculation). m Lindborg et al. (2013) (calculation). n Chen et al. (2006) (calculation, spherical volumes). o Kliauga and Dvorak (1978), Np L γ x ray (measurement). y = 1.0 y = 1.02 y = 0.27 y = 2.68 f y = 1.8 o y = 2.05 h d d o d h y = 0.79 i d i 75

1740 1741 1742 1743 1744 1745 1746 1747 1748 1749 1750 1751 1752 1753 1754 1755 1756 1757 1758 1759 1760 1761 1762 1763 1764 1765 1766 1767 1768 1769 1770 4.2 Prediction of R i Based on f(y) No description of the track structure of a charged particle that correlates directly with R i for that radiation has been found. R i appears to depend on the interaction between the characteristics of the energy deposition and the structure and biochemical processes functioning in the irradiated system. Various conceptual and mathematical models have been formulated to attempt to describe this interaction between the energy deposited and the biological system. These models can be used to formulate different predictions of R i for radiations using appropriate quantities based on the track structure of the radiation. Many models assume that the linear/quadratic behavior of typical survival curves for low LET radiation reflects a fundamental characteristic of radiation damage that single energy depositions can lead to the endpoint (linear component) but that interaction of damage produced by two independent energydeposition events can also lead to the endpoint (quadratic component). However, many other mechanisms have been proposed as possible causes for the observed shapes of dose-response curves for low LET radiation. Some rely on spontaneous and radiation-induced misrepair causing the increase in effectiveness with increasing dose. Another mechanism which would lead to increased effectiveness with increased dose is depletion of repair capacity. Other models assume that some minimum amount of damage must be done in individual targets in order to produce lethal damage. The following explores the implications of several of these models for R i of photon irradiation as a function of photon energy. Most models of biological response give the ratios of doses that are expected to give the same biological effect. These values are not actually RBEs since they are not derived from observed values of the dose to produce a specified effect. Section 4 refers to these estimates as R i, sometimes with a subscript to indicate the model that produced the specific value, for example R yd for the ratio of dose mean lineal energies. At low doses the ratio of doses for the same effect and the ratio of effects for equal dose are equal. Values for R i as predicted using several models and a variety of assumed relevant site sizes are tabulated in Table 4.2. Based on the hypothesis (Neary, 1965) that lethal lesions can be formed directly by a single energy-deposition event or by the interaction of sublesions formed in separate 76

1771 1772 y d(test) /y d(1250) for 1 µm site y d(test) /y d(1250) for 5 µm site y d(test ) /y d(100 kev ) for 2.3 nm site y d(test ) /y d(100 kev ) for 30 nm site y d(test ) /y d(100 kev ) for 100 nm site Table 4.2 -- Estimates of R i relative to 60 Co gamma rays. a 1.5 kev Electrons 10 to 30 kev Electrons 40 to 60 kev Electrons 100 to 150 kev Electrons Tritium Beta Particles 2.65 2.35 2.32 3.37 3.39 2.27 2.59 2.63 2.25 1.14 1.03 2.01 3.0 1.85 0.95 2.3 2.8 2.0 1.4 2.6 Hit size effectiveness 1.2 1.2 1.2 Ratio of doses to produce same number of events larger than 3 kev µm 1 in 1 µm sites Ratio of doses to produce same number of events larger than 5 kev µm 1 in 5.5 µm diameter sites 4.0 2.8 4.2 3.3 1773 1774 Ratio of doses to produce same number of events larger than 100 ev in 3 nm sites (relative to 100 kev) 3 1.14 a Unless otherwise indicated in the reference radiation subscript. 77

1775 1776 1777 1778 1779 1780 1781 1782 1783 1784 1785 1786 1787 1788 1789 1790 1791 1792 1793 1794 1795 1796 1797 1798 1799 1800 1801 1802 1803 1804 1805 1806 energy-deposition events, Kellerer and Rossi (1972) developed the dual radiation action model and concluded that R i = y d(test) /y d(reference). Based on comparison of f(y) distributions with experimental data for biological effects such as chromosome aberrations and mutation as a function of radiation type they concluded that damage produced by separate radiation events can interact over distances comparable to the cell nuclear diameter. However, there is no clearly defined reason for using any specific site size or geometry for estimating the R i for the various biological endpoints and cell types. Fortunately, for most of the photon energies of interest here, y d changes relatively slowly with site size. Table 4.2 includes estimates of R i based on experimental measurements of y d in 1 µm and 5.5 µm diameter sites. There is some uncertainty in these values due to: uncertainty in experimental measurements of y; imprecise reproduction of the atomic composition of cellular components by tissue equivalent plastic; and use of an arbitrary depth in simulated tissue (the tissue-equivalent wall of the vacuum chamber housing the wall-less detector used to make the measurements). However, it is evident that if this model is correct and the relevant site size is 1 to 5.5 µm, the value of of R i, for 10 to 30 kev electrons, relative to 60 Co gamma rays is in the range from a little less than 2.6 to a little over 3.4. If the interaction of sublesions produced by a single charged particle were responsible for the increase in biological effectiveness with increasing LET it would be possible to determine the distance over which interactions occur by irradiating cells with pairs of particles with known spacing between the particle tracks. The molecular ion experiment (Bird, 1979; Rossi, 1979) was designed to do this. Single charged hydrogen molecules, often D + 2, were accelerated to energies of a few million electron volts per nucleon. When these molecules interact with matter (the surface of a plastic film supporting mammalian cells in vitro) the electron is stripped off and the two hydrogen ions become independent of each other. Random scattering causes the average distance between the two ions to increase with depth in the material. Cells can then be irradiated with pairs of ions at essentially the same time but with different average spatial separation, depending on the thickness of the plastic film the cells are attached to. 78

1807 1808 1809 1810 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820 1821 1822 1823 1824 1825 1826 1827 1828 1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 These experiments showed essentially no difference in biological effect over the range of ion separations that could be produced by available accelerator energies and plastic film thicknesses. Analysis of the results in terms of the later distance-dependent form of the theory of dual radiation action (Kellerer and Rossi, 1978) led to the conclusion (Kellerer et al., 1980; Zaider and Brenner, 1984) that the interactions of products of a single charged particle track are limited to very short distances, probably less than 10 nm, and that these interactions are biochemically different than those that occur between the products of separate tracks and occur over distances larger than could be investigated with the dual ion experiment, probably several micrometers. Thus the ratios of dose mean lineal energy for nanometer scale sites should also be considered when estimating R i. The ratios of dose mean lineal energies for 10 to 30 kev electrons relative to 60 Co gamma rays in 100 nm diameter sites is in the range of 1 to 2.5, depending on the gamma-ray data used. It is about 1 for a 2.3 nm diameter region but 1.3 to 1.8 for a 30 nm site. At low doses, where the occurence of two or more electron tracks in a cell nucleus is rare, R i is probably influenced primarily by the energy deposited by single particle tracks in nm scale sites. It also appears that the structure of the cell influences the interaction of radiation-induced biochemical changes. For example, in an attempt to separate the effects of cell structure from the effects of genetic differences between cell types, Carpenter et al. (1989) irradiated 5 cell lines with similar sensitivity to high-energy electrons, but different nuclear shapes, from nearly spherical to relatively thin disks, with the same volume. In studies of cell survival, Carpenter et al. (1989) found that cells were more sensitive to radiation damage and their RBEs for lowenergy electrons were lower if their nuclei were flatter. Although dose-response data for most biological systems are consistent with a linear/quadratic model over the range of doses that can be investigated experimentally, there is no evidence that the linear/quadratic model is any more than an adequate fit to results of a complex combination of biological processes. A wide variety of experiments, many of them based on the response of cultured cells in vitro, suggest that extensive repair of most types of radiation-induced damage occurs and that both the amount of damage and the complexity of the damage impact the probability of repair success. Furthermore, studies at low doses or with 79

1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866 1867 1868 particle microbeams show that signals passed from irradiated cells to neighboring cells result in DNA changes (Nagasawa and Little, 1992) and changes in DNA repair associated proteins (Morgan, 2003a; 2003b) in cells which did not receive energy deposited by ionizing radiation. The initiation of many of these processes appears to require a minimum energy deposition which would be expected to result in a threshold for either the energy deposited in a single event (y) or the total energy deposited in a small volume during an irradiation (z). At higher values of y or z additional biochemical processes may be activated and make significant contributions to the total effect of a specific radiation exposure. In the simplest case, a threshold in y, y, which must be exceeded in order to produce a given biological effect, the R i would be the ratio of the doses needed to produce equal numbers of cells with events larger than y in identical populations of cells. The value of R i would then depend on the value of y and on the shape of the f(y) curves for the two radiations being compared. The shape of the f(y) will depend on the size and shape of the volume where it is evaluated as well as the track structure of the radiation. Thus there is a great deal of uncertainty associated with predicting R i assuming a simple threshold model. The fraction of the events that are larger than y can be read from a plot of Φ(y), the frequency of events per dose, Figures 4.4 and 4.5. Thresholds may occur for different processes in different size sites and at different energies. It has been proposed (Goodhead and Brenner, 1983) that the threshold for damage leading to cell lethality is 100 ev in a 3 nm diameter volume, probably relating to DNA. This leads to R i from 1.1 to 2.25 for 20 and 1.5 kev electrons, respectively (Figure 4.4). Thresholds for processes such as repair depletion and misrepair may involve sites on the order of the diameter of the cell nucleus and energy depositions from a few hundred ev to several kev. These processes would result in R i up to perhaps 4 or 5 for 10 to 30 kev photons (Figure 4.5). A simple threshold is probably an overly simplistic model for biological response. A more realistic model would allow for a combination of independent processes leading to the same biological effect. The hit size effectiveness approach (Bond and Varma, 1983; Bond et al., 1985) attempts to address this type of situation. In this approach it is assumed that the probability of the effect, E(y) is a function of the magnitude of a physical measure of radiation 80

10 3 nm 1 0.1 frequency x 10 8 /cgy 0.01 0.001 0.0001 1.5 kev 20 kev 50 kev 100 kev 1869 0.00001 0 200 400 ε/ev 10000 25 nm 100000 100 nm 1000 10000 100 1000 10 100 1 10 0.1 1 1870 1871 1872 1873 1874 1875 1876 1877 1878 1879 0.01 0.1 0 2000 4000 6000 0 5000 10000 Fig. 4.4. Calculated event frequencies in nm scale cylindrical sites (height equal to diameter) in water. The uncharacteristic shape of the line for 20 kev electrons in 25 and 100 nm sites may be the result of random variations when a small number of simulated tracks were sampled (Nikjoo et al., 1994a). The statistical reliability of the frequency distributions for 20 kev electrons in 25 and 100 nm cylinders is particularly limited by the small number (20) of Monte Carlo tracks generated for these two computations. NOTE: In Figure 4.4, the legend on the right in the top graph applies to all 3 graphs. It will be relocated appropriately later. 81

1880 1881 0.58 µm diameter 100 10 1 Event Frequency, Φ(y) 0.1 0.01 0.001 0.0001 60Co 250kV 1.8mmCu 65kV 1.9mmAl 0.00001 0.000001 0.01 0.1 1 10 1882 1883 1884 1885 1886 Fig. 4.5. Measured values of event frequencies for photon irradiation of micrometer scale sites (Ellett and Braby, 1972). y/kevµm 1 82

1887 100 0.92 µm diameter 10 1 0.1 0.01 0.001 0.0001 0.00001 1888 1889 0.000001 0.01 0.1 1 10 100 1.85 µm diameter 100 5.55 µm diameter 10 10 1 1 0.1 0.1 0.01 0.01 0.001 0.001 0.0001 0.0001 0.00001 0.00001 1890 1891 1892 1893 0.000001 0.000001 0.01 0.1 1 10 0.01 0.1 1 10 Fig. 4.5. (Continued). Measured values of event frequencies for photon irradiation of micrometer scale sites (Ellett and Braby, 1972). 83

1894 1895 1896 1897 1898 1899 1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 quality, which is usually considered to be y, in a site of specified size, Figure 4.6. The effect of a specific irradiation is assumed to be the sum of the effects produced by each increment of the f(y) distribution. The R i is then the ratio of the doses required to produce equal effects. The E(y) distributions are determined from experimental data for biological effects in individual biological systems. Morstin et al., (1989) developed a probabilistic approach to obtaining E(y) from limited experimental data, and their results for the RBE of low-energy x rays, derived from data on chromatid exchanges in CH2B2 cells, mutation in V79 cells, mutation of human fibroblasts, and inactivation of human T-1 kidney cells, range from 1.0 to 1.2. Their approach has the advantage of being independent of specific assumptions about radiation damage mechanisms, but does assume a single biological target size for each biological endpoint, and is limited by the lack of data for different photon spectra. 4.3 Evaluation of the PDF of R i There is no fixed way to construct a PDF of R i from the data in Table 4.2. The selection of models and site sizes clearly influences the value of R i, and the selection of values to be entered in Table 4.2 was driven by the availability of data more than by any consideration of relevance to biological processes. Consequently the significance of each value of R i, relative to all the other values of R i for the specific radiation, must be carefully considered before determining the PDFs of R i for each lower-energy group listed in Section 2.6, Figure 4.7. Each PDF represents the credibility of alternative values to represent the state-of-knowledge of R i and is based on critical evaluation of relevant findings and supporting evidence for that lower-energy group. It is essentially the result of comparing all of the strengths and weaknesses of the individual models described in the Section 4.2, and avoiding the bias which might result from an abundance of data for one model or a lack of data for another. Some models, for which there are essentially no data, or apparent large uncertainties in the data, may be more useful in establishing maximum and minimum values for R i than for determining the probability of intermediate values. For example, there is essentially no data available to establish the shape of the hit size effectiveness relationship for any endpoint for lowenergy photons. All of the available data are for higher values of y, primarily from neutron and 84

1925 1926 1927 1928 1929 1930 1931 1932 1933 1934 Fig. 4.6. Typical hit size effectiveness functions. These examples are for chromatid exchanges (solid line) and abnormal metaphase in CH2B2 hamster cells (dashed line) (Morstine et al., 1989). 85

1935 1936 f(r i ) 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 about 1.5 kev Quartiles 1st = 2.25 2nd =2.9 3rd = 3.4 0 1 2 3 4 5 6 R i 1937 1938 1939 1940 1941 1942 1943 1944 1945 1946 1947 1948 1949 1950 f(r i ) 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 10 to 30 kev 0 1 2 3 4 5 6 Fig. 4.7. PDFs of R i relative to 60 Co, estimated based on the energy-deposition characteristics. R i Quartiles 1st = 2.66 2nd = 3.36 3rd = 3.81 86

1951 1952 f(r i ) 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 40 to 60 kev Quartiles 1st = 2.08 2nd = 2.5 3rd = 2.92 0 1 2 3 4 5 6 R i 1953 1954 f(r i ) 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 100 to 150 kev Quartiles 1st = 2.25 2nd = 2.667 3rd = 3.25 0 1 2 3 4 5 6 R i 1955 1956 1957 Fig. 4.7 (continued). (Continued). PDFs of R i relative to 60 Co, estimated based on the energy- deposition characteristics. 87

1958 1959 f(r i ) 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 tritium beta particles Quartiles 1st = 2 2nd = 2.5 3rd = 3 0 1 2 3 4 5 6 R i 1960 1961 1962 Fig. 4.7 (continued). (Continued). PDFs of R i relative to 60 Co, estimated based on the energy- deposition characteristics. 88

1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 charged-particle irradiation. The hit size effectiveness function, Figure 4.6, can be extrapolated to very low values of y, but there is clearly a large uncertainty involved in that extrapolation. Furthermore R i based on hit size effectiveness function is only available for two biological endpoints, which are only indirectly related to the endpoints of primary concern for this study. Consequently, considering the uncertainty in the extrapolation, the hit size effectiveness model serves primarily to suggest that the lower bound for the PDF for each radiation should be 1.0. R yd, the ratio of y for the test radiation to the value for the reference radiation was evaluated at d a range of site sizes, based on the data in Table 4.1. The result for 1 µm diameter sites was given slightly more weight than data for other site sizes since radiobiology data seems to indicate that micrometer site data correlates best with mutation induction and other endpoints which seem to be related to the late effects of radiation. Similarly, the ratio of doses to give equal frequencies of events larger than 100 ev in 3 nm diameter sites was given more weight than data for other threshold values and site sizes because of its consistency with experimental data for very soft x- ray experiments (Goodhead and Brenner, 1983). The predictions of R i based on other models were given less weight due to the lack of extensive data on R i prediction. In all cases the PDF was broadened by the uncertainty in the measured or calculated values of the energy imparted. There is significantly more data for 15 to 30 kev photons than for other lower-energy groups. Although the same process was followed for constructing PDFs of R i for the other lower-energy groups, the lack of data resulted in greater broadening of the PDF relative to the values from the models. Since the area a PDF is 1.0, the first quartile is the value of R i which is the upper edge of the segment of the curve with the area equal to 0.25. The second quartile bounds the portion of the curve with half of the total area and the third quartile bounds the part with area equal to 0.75. For very low absorbed doses the probability of more than one charged particle track crossing a typical cell nucleus becomes very small and only single event processes are likely to contribute to the biological effectiveness. However, for doses of only 3 mgy, the average number of tracks crossing cell nuclei is approximately 1 and roughly 1/3 of the nuclei have 2 or more track crossings, resulting in a significant chance that interactions of the products of two 89

1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 tracks contribute to the biological effect. Thus R i values from models involving interactions in the nucleus are included in evaluation of the PDFs. R i determined by any of these models, reflects only the initial stage of the sequence of processes leading to the health effect, the deposition of energy which initiates biochemical changes. This initial stage is indirectly linked to ρ L, the effectiveness ratio for human cancer, as it is to the RBE for a specific endpoint in a particular animal. However R i is expected to be relevant to ρ L because: R i is an indicator of the energy-deposition pattern, which presumably is important to induction of cancer; the initial types of radiation damage, such as DSB and free-radical production, are generally independent of the biological system or endpoint (effect) under study; and DNA damage and repair mechanisms may differ quantitatively but probably are similar qualitatively in in vitro and in vivo experimental systems and in humans. However, the PDFs of R i based on energy-deposition patterns are relatively broad because of uncertainty in the: relevant site size for energy deposition; appropriate model for the specific health effect; and uncertainty in experimental and calculated values of f(y). 90

2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041 2042 2043 2044 2045 2046 5. Line of Evidence: Deoxyribonucleic Acid Damage 5.1 Deoxyribonucleic Acid Damage from Ionizing Radiation It is generally recognised that damage to cellular DNA can lead to a variety of long-term cellular and health effects. Ionizing radiation is an effective agent at creating DNA damage, both via direct interaction of the ionizing particle with the DNA and via production of reactive radicals in the immediate neighborhood of the DNA. A wide variety of DNA damage is produced. The highest yields are for single-component damages such as modified bases and single-strand breaks (SSB) in the sugar-phosphate backbone of the DNA, but a range of more complex DNA damage also occurs due to the spatial clustering of interactions within individual radiation tracks. Table 5.1 illustrates the yields and types of damage within a cell after 1 Gy of gamma-ray irradiation (Goodhead, 1994a). Most of the DNA damage is rapidly repaired by the very efficient and specialized biochemical repair systems in the cell, but a small proportion may be misrepaired or remain unrepaired for longer times and lead to permanent cellular consequences (Jeggo et al., 2011; Mladenov and Iliakis, 2011). DNA double-strand breaks (DSB) are of particular concern as a pathway to permanent alterations to the DNA sequence, with potential long-term cellular and health consequences. Although the majority of DSB are correctly rejoined in normal cells, and even in most DNArepair deficient cells, a significant proportion may be misrepaired or remain unrejoined. Factors that influence the probability of failure include the competitive biochemistry of perfect repair versus alternative error-prone reactions, the phase of the cell in its cell cycle, the location of the particular DSB in a region of greater (heterochromatic) or lesser (euchromatic) chromatin condensation, and the complexity of the particular DSB. Unrepaired or misrepaired DSB are responsible for loss of, and/or changes to DNA sequence. The changes may be lethal to the cell or its progeny but in many cases they are transmissible as viable mutations. Although ionizing radiation is particularly effective at inactivating cells, it is also efficient at producing viable mutations in cells that survive, including large DNA deletions and chromosome rearrangements visible under the light microscope. A variety of lines of evidence have led to the conclusion that it is DSB that are the initial damage predominantly responsible for the deletion 91

2047 2048 2049 Table 5.1 Some of the damage in a mammalian cell nucleus from 1 Gy of low-let radiation. Initial physical damage Ionizations in cell nucleus Ionizations directly in DNA Excitations directly in DNA ~100,000 ~2,000 ~2,000 Selected biochemical damage (Ward, 1988) DNA single-strand breaks 8-hydroxyadenine T* (thymine damage) DNA double-strand breaks DNA protein cross links 1,000 700 a 250 a 40 150 2050 2051 2052 Selected cellular effects Lethal events Chromosome aberrations Hprt mutations ~0.2 0.8 ~1 ~10 5 a More recent estimates by Cadet et al. (2008) suggest replacement of these two entries with purine damage (~400) and pyrimidine damage (~890). 92

2053 2054 2055 2056 2057 2058 2059 2060 2061 2062 2063 2064 2065 2066 2067 2068 2069 2070 2071 2072 2073 2074 2075 2076 2077 2078 2079 2080 2081 2082 2083 2084 mutations, chromosome aberrations and other genetic changes induced by all qualities of ionizing radiation and that these are a major contributor to long-term health consequences, including cancer. Hence the RBEs for induction of DSB by photons of different energies are likely to be informative of the relative effectiveness of the photons for such long-term health consequences. Although the more minor classes of DNA damage (such as base damage and SSB) are produced in much greater abundance, they are much more efficiently repaired and less likely to result in permanent genetic changes. A large amount of experimental evidence has shown that DSB are produced linearly with absorbed dose of radiation, over an enormous dose range from tens of milligray to thousands of gray (Barnard et al., 2013; Durante et al., 2013; Frankenberg et al., 1999; Rothkamm and Löbrich 2003; Sutherland et al., 2002). These experimental observations are consistent with theoretical expectations based on the negligible probability of overlap of separate radiation tracks on the nanometer scale of the DNA molecule and the very short diffusion distances of reactive radiolysis products in the cellular environment (Goodhead and Nikjoo, 1989). From theoretical considerations and track structure simulations of electrons and heavier charged particles, as well as the observed linearity of the dose-response, it is apparent that each DSB results from the clustering properties of the ionizations within a single radiation track on the scale of DNA and its immediate surroundings. Clustered ionizations directly within the DNA molecule over distances of a few nanometers can cause multiple breaks to the DNA. In addition, reactive radical species (particularly the OH radical from ionization of a water molecule) can be produced by the radiation track in the vicinity of the DNA, and these too have the ability to lead to DNA breaks if they are sufficiently close. Since the mean diffusion distance of these radicals in the highly reactive cellular environment is very small (~ 4 nm) (Chapman and Gillespie, 1981; Roots and Okada, 1975), their ability to cause multiple breaks to the DNA, either on their own or in conjunction with direct ionizations, is highly dependent on the ionization-clustering within the radiation track on the scale of a few nanometers. In this way the clustered ionizations within a track are capable of producing not only simple DSB, composed of a single break on each of the two strands of DNA in close proximity, 93

2085 2086 2087 2088 2089 2090 2091 2092 2093 2094 2095 2096 2097 2098 2099 2100 2101 2102 2103 2104 2105 2106 2107 2108 2109 2110 2111 2112 2113 2114 2115 2116 but also a spectrum of more complex DSB, which may have additional breaks on one or both strands and/or damaged bases, all within a short distance along the DNA molecule [illustrated, for example, by Goodhead (2006; 2009)]. Track structure simulations suggest that most of these complex DSB are confined to less than about 10 base pairs of DNA (Nikjoo et al., 1997). While current experimental methods are not capable of distinguishing directly between simple and complex DSB, it has been found that synthetically-produced complex DSB are less readily repaired by cells than are simple DSB and that they are more likely to result in mutations (Almohaini et al., 2016; Dobbs et al. 2008). Additionally, there is experimental association between radiations with greater clustering of ionizations and reduced repairability of the DSB that they produce (Goodhead and Nikjoo, 1989; Lomax et al., 2013). As well as complex DSB, ionization clustering can lead to non-dsb clustered damage in DNA, consisting of multiple base damages or one or more base damages adjacent to an SSB (Eccles et al., 2011; Sutherland et al., 2002). These also are likely to present a challenge to the cell s repair system, depending on the complexity of the non-dsb clustered damage. 5.2 Photon and Electron Damage to Deoxyribonucleic Acid In situations of photon irradiation of biological systems, the vast majority of the energytransfer interactions with the medium are due to the large numbers of interactions of the secondary electrons ejected by relatively rare photon interactions. For example, for a single photon interaction that produces a Compton electron of say 200 kev, nearly 10 4 further ionizations result from the subsequent interactions of the Compton electron and its secondary electrons. Hence, it is these electrons that are overwhelmingly responsible for the DNA damage. The yield of DNA damage and its spectrum of complexity depend substantially on the energy of the electrons. Low-energy electrons of a few hundred to a few thousand electron volts are particularly efficient because of their short mean free paths (Figure 3.8) and consequent ionization-clustering properties on the scale of the DNA molecule. The primary electrons produced by the photon interactions generate a series of higher-order electrons as they interact and lose energy. Thus, at any plane in the irradiated medium there will be a wide spectrum of electrons of energies ranging from zero to the full energy of the primary electrons generated directly by the photon interactions. A typical electron track is composed of a primary electron and an array of primary and higher-order electrons of decreasing energy until they all come to 94

2117 2118 2119 2120 2121 2122 2123 2124 2125 2126 2127 2128 2129 2130 2131 2132 2133 2134 2135 2136 2137 2138 2139 2140 2141 2142 2143 2144 2145 2146 2147 2148 rest. It is this overall electron spectrum that determines the spectrum of DNA damage. For more detailed discussion of of the spectra of secondary electrons ultimately produced by a primary electron, see Section 3.4. All photon irradiations produce an abundance of low-energy electrons due to the above slowing-down and degradation processes. The fraction of overall absorbed dose that is deposited by the low-energy electrons is larger for photons that produce a greater proportion of lowerenergy primary electrons. These are generally the lower-energy photons, although the relationship with photon energy is not monotonic due to differences in the primary electron spectra from Compton and photoelectric processes (Section 3.3). Figure 5.1 illustrates analytical evaluations of the fraction of total absorbed dose deposited locally by electrons of a given energy at a point in the irradiated medium (Burch, 1957). From the upper panel in Figure 5.1 it can be seen that for 60 Co gamma rays about 33 % of the dose is deposited by low-energy electrons of less than 5 kev and that this increases to about 50 % for 220 kv x rays. The corresponding fractions are about 27 % (for 60 Co gamma rays) and 32 % (for 220 kv x rays) of deposition by electrons of less than 1 kev. As will be seen from the experimental data to be discussed below, it is these low-energy electrons that are particularly responsible for the production of DNA DSB. The enhanced efficiency of DSB production by low-energy electrons can be explained by the short mean-free path for ionization by such electrons (Figure 3.8), resulting in clustering of ionizations on the nanometer scale of the DNA molecule. It has been suggested that nearly all the DSB from irradiation with x or gamma rays are produced by the low-energy electron component of the degraded radiation field. For 1.5 kev photons all but a few percent of the absorbed dose is deposited by low-energy electrons. Corresponding but more recent evaluations of local energy deposition, now using Monte Carlo track structure simulations of electrons in liquid water (Figure 5.2), produce generally similar results to those shown above, except for a tendency towards somewhat larger fractional contributions of local deposition by electrons of the order of kiloelectron-volt energies (Bellamy and Eckerman, 2013; Bellamy et al, 2015). As mentioned above and elaborated below, experimental results on DSB induction by low-energy photons show that secondary electrons of up to a few kiloelectron volts are particularly efficient at producing such DNA damage. 95

2149 2150 2151 2152 2153 2154 2155 Fig. 5.1. NOTE: Figure caption is on next page. 96

2156 2157 2158 2159 2160 2161 2162 2163 2164 2165 Fig. 5.1. Local energy deposition as a function of electron energy for irradiation of water with 220 kv x rays, 60 Co gamma rays, monoenergetic electrons or tritium beta particles, shown as either cumulative fraction (upper panel) or differential fraction (lower panel) of total absorbed dose. For these evaluations local" was taken to include all collisional energy transfers (excitation and ionization) up to 100 ev. Secondary and higher order electrons produced with energy greater than 100 ev were treated separately, that is, as being distinct from their parent track. These data were replotted by Nikjoo and Goodhead (1991) from analytical calculations of Burch (1957) and, for 100 kev electrons, R.J. Munson [personal communication in Nikjoo and Goodhead (1991)]. 97

2166 2167 2168 2169 2170 2171 2172 2173 2174 2175 2176 Fig. 5.2. Local energy deposition as a function of electron energy for irradiation of water with monoenergetic electrons of selected energies from 10 to 1,000 kev (reproduced from Bellamy and Eckerman, 2013). These results correspond directly to those of Figure 5.1, except that they were evaluated using Monte Carlo electron track-structure simulations. The data show the cumulative fraction of absorbed dose deposited by primary and secondary electrons as a function of electron kinetic energy. The fractional contribution to absorbed dose by interactions entered into at energies below a cutoff of 5 kev is projected by the horizontal lines on the y-axis. 98

2177 2178 2179 2180 2181 2182 2183 2184 2185 2186 2187 2188 2189 2190 2191 2192 2193 2194 2195 2196 2197 2198 2199 2200 2201 2202 2203 2204 2205 2206 2207 2208 2209 2210 5.3 Experimental Data on RBE for DNA Double-Strand Breaks A variety of experimental techniques have been developed and applied over the years to measure the yield of DSB induced by ionizing radiations of different quality and hence provide RBE values for DSB induction. Evolving technical development has been towards methods of greater sensitivity and/or greater convenience. Table 5.2 summarizes the published measurements of RBE (for DSB production within cells) for monoenergetic and broad spectrum photons of energies below 120 kev. The early technique of neutral sedimentation is of high accuracy, but low sensitivity, and it was applied mostly to DNA from irradiated yeast cells rather than mammalian cells. Hence, very high radiation doses were required. However, in view of the accepted (Durante et al., 2013) linear dose-response for DSB and the similarities between yeast and mammalian cells in respect of primary DNA conformation and local environment, these measurements provide RBEs of more general relevance. Table 5.2 shows the RBE values for DSB in yeast for various photon energies (monoenergetic and broad spectrum), using as reference radiation either very highenergy (30 MeV) electrons (Frankenberg and Binder 1985; Frankenberg et al., 1999) or 60 Co gamma rays (Frankenberg et al., 1986, Kuhne et al., 2005). The numerical results, with relatively small CI, show a general increase in RBE with decreasing photon energy, rising to a maximum at the lowest measured energy of 0.28 kev C K characteristic x rays. Values of particular interest for the Committee s assessments are RBE of 1.35 for 50 kv x rays (with respect to 30 MeV electrons) and 2.2 ± 0.1 for 1.5 kev x rays (with respect to 60 Co gamma rays). For mammalian cells, most measurements providing relevant RBEs for DSB induction have been for very low-energy x rays (with 60 Co gamma ray as reference radiation), using the more-sensitive pulsed-field gel electrophoresis technique (Table 5.2). The RBE values from de Lara et al. (2001), including revised results from Botchway et al. (1997), again show the systematic increase in RBE with decreasing photon energy, in this case rising to 2.7 ± 0.3 for 0.28 kev C K x rays. A somewhat lower value (2.0 ± 0.3) for 0.28 kev C K x rays was measured by Kuhne et al. (2005) using the same technique but a different mammalian cell type; for 29 kv x rays they obtained the RBE as 1.15 ± 0.05. They state no detectable difference at high energies (i.e., between 60 Co gamma rays and 15 MeV electrons). Of particular interest for the 99

2211 2212 NCRP SC 1-20 Table 5.2 Experimental RBEs for induction of DNA double-strand breaks by x rays. Publication X Rays Reference Radiation Frankenberg 2 kv (Al anode) 30 MeV and Binder 4 kv (Ag anode) electrons (1985) 1.1 1.5 kev 1.5 2.0 kev 2.0 4.0 kev Frankenberg et al. (1986) Frankenberg et al. (1999) Prise et al. (1989) Botchway et al. (1997); O Neill et al. (1997) De Lara et al. (2001) Kuhne et al. (2005) Hamada et al. (2006) Kegel et al. (2007) Beyreuther et al. (2010) 4.0 7.0 kev C K (0.28 kev) Al K (1.5 kev) 50 kv (low filtration: Be window + 1 mm Al * ) Al K (1.5 kev) Cu L (0.96 kev) Al K (1.5 kev) C K (0.28 kev) Cu L (0.96 kev) Al K (1.5 kev) Ti K (4.5 kev) C K (0.28 kev) 29 kv (W target and 50 µm Rh filter) Al K (1.5 kev) 25 kv (W anode) 120 kv (W anode) 10 kv (W anode) 25 kv (W anode) 60 Co gamma rays 30 MeV electrons 250 kv x rays 60 Co gamma rays 60 Co gamma rays 60 Co gamma rays (same for 15 MeV electrons) 240 kv x rays 137 Cs gamma rays 200 kv x rays (W anode; various filters) RBE for DSB a Cell Type Technique Comment 1.93 ± 0.12 1.50 ± 0.06 2.0 1.7 1.7 1.7 3.8 ± 0.2 2.2 ± 0.1 Yeast Yeast Neutral Sedimentation Neutral Sedimentation 1.35 Yeast Neutral Sedimentation 1.64 V79 hamster cells 3.0±0.3 2.3# 2.5±0.2 1.9# 2.7 ± 0.3 2.3# 1.9# 1.4 ± 0.1 2.0 ± 0.3 1.15 ± 0.05 V79 hamster cells V79 hamster cells HSF2 primary human skin fibroblasts 2.2 HE49 human diploid fibroblasts ~0.97 ± ~0.1 ~0.95 ± ~0.1 ~ 5 (at 2 h) ~ 2 (at 2 h) MRC 5 primary human fibroblasts 184A1 human mammary epithelial Neutral filter elution Pulsed field gel electrophoresis Pulsed field gel electrophoresis Pulsed field gel electrophoresis γh2ax foci γh2ax foci γh2ax+53bp1 co localized foci Inferred from spectral fractions and additional data on survival of rad52 strain Table 2 as unpublished. Filtration in Kuhne et al. (2005) Underestimate due to cell thickness? (de Lara et al. 2001) # Values revised for reference yield (by de Lara et al., 2001) # Revised values from Botchway et al. (1997) Also reduced repair and reduced repair fidelity Not a measure of initial DSB 100

2213 2214 NCRP SC 1-20 Beels et al. 100 kv (2mm Al) (2010) Shridar et al. (2010) Depuydt et al. (2013) Mills et al. (2015) 8 kev Cu K X rays 30 kv (Mo/Mo) Mammography 29 kv (Rh/Rh) Mammography 60 Co gamma rays 60 Co gamma rays 60 Co gamma rays 137 Cs gamma rays 1.19 implied by slopes (>> at mgy doses?) Human lymphocytes in whole blood 0.9 ± 0.3 Human glioma cells 1.35 Human [95 %: 0.95, lymphocytes 1.98] in whole blood 1.1 ± 0.2 MCF 10A human breast cell line a Uncertainties given are as stated in the cited publications. γh2ax foci Reference radiation not contemporaneous γh2ax foci 8 kev foci are larger γh2ax foci Blood from 5 human volunteers 53BP1 foci Low doses: 3 30 mgy 101

2215 2216 2217 2218 2219 2220 2221 2222 2223 2224 2225 2226 2227 2228 2229 2230 2231 2232 2233 2234 2235 2236 2237 2238 2239 2240 2241 2242 2243 2244 2245 2246 2247 NCRP SC 1-20 Committee s assessment is the RBE of 1.15 ± 0.05 for filtered 29 kv x rays (Kuhne et al., 2005) and 2.2 ± 0.1 for 1.5 kev Al K x rays (de Lara et al., 2001), both with respect to 60 Co gamma rays. Prise et al. (1989) used the neutral filter elution method to measure an RBE of 1.64 for Al K mono-energetic x rays, with reference to 250 kv x rays. This value is somewhat smaller than the above value reported by de Lara et al. (2001) with 60 Co gamma rays as reference, but this is likely to be at least partly due to underestimation of the cell thickness and hence overestimation of the Al K x-ray doses in the earlier work. The RBE difference may also be suggestive of a difference in yield between the two reference radiations. It is unlikely, however, that a reliable RBE for 250 kv x rays, relative to 60 Co gamma rays, can be usefully estimated from these measurements using different assay techniques by different groups in different laboratories. Additionally, the RBE estimate of Prise et al. (1989) is likely to be an underestimate because the Al K x-ray dosimetry was based on an assumed (rather than measured) cell thickness that was likely to be too large. The alkaline comet assay was used by Gomolka et al. (2005) to measure all single-strand breaks in DNA after 27 kv mammography x rays, 220 kv x rays, 137 Cs gamma rays, and 60 Co gamma rays. No statistical differences were observed in the yields. These results are not informative, however, on the RBE for DSB since only about 2 % of the total measured breakage was contributed by DSB so these experiments would be insensitive to differences in DSB yield. A more recent, and now widely applied, method for detection of DSB in individual cells after irradiation is by fluorescent staining of the phosphorylated histone variant γh2ax to produce foci visible under the light microscope. This protein, as well as others in the relevant repair pathways, is believed to be involved in biochemical detection and repair of DNA DSB. In this way numbers of DSB at a given time can be estimated by counting the number of individual foci, on the assumption that there is a one-to-one correspondence between foci and DSB present at that time. The precision of this assumption is particularly questionable when studying repair kinetics (Markova et al., 2007) and also in relation to reports of supralinear dose-responses, and surprisingly high yields of foci at very low doses in some systems (Barnard et al., 2013; Beels et al., 2010; Colin et al., 2011), when questions on thresholds and timescales for appearance and disappearance of visible foci may be relevant. 102

2248 2249 2250 2251 2252 2253 2254 2255 2256 2257 2258 2259 2260 2261 2262 2263 2264 2265 2266 2267 2268 2269 2270 2271 2272 2273 2274 2275 2276 2277 2278 2279 2280 NCRP SC 1-20 In studies with human diploid fibroblasts, Hamada et al. (2006) measured the RBE of 1.5 kev Al K x rays for induction of γh2ax foci in human fibroblasts to be 2.2, relative to 240 kv x rays, over the dose range 0.01 to 0.5 Gy. Shridhar et al. (2010) observed γh2ax foci in U251 human glioma cells 1 h after irradiation with 8 kev Cu K characteristic x rays or 60 Co gamma rays and reported no significant difference in numbers per unit dose, but that the 8 kev foci were significantly larger in size. From the limited information in Shridhar et al. (2010), an RBE of 0.91 ± 0.33 can be calculated. Kegel et al. (2007) report no significant differences between the numbers of γh2ax foci in MRC-5 primary human fibroblasts, assayed 15 min after irradiation with 1 Gy of 25 kv x rays, 120 kv x rays, or 137 Cs gamma rays. Reading off from their Figure 2 (pooling the data from plastic foil on plastic and on water, and neglecting the data for non-equilibrium materials glass and aluminium), implies RBE 0.95 for 120 kv x rays and 0.97 0.97 for 25 kv x rays, both with respect to 137 Cs gamma rays. A rough estimate of standard deviations suggests ±0.1 to 0.15. With the aim of measuring the time course of residual DNA DSB, with reduced cellcycle-dependent background in the γh2ax foci assay, Beyreuther et al. (2009) counted γh2ax foci that were co-localized with staining of 53BP1(tumor protein 53 binding protein) as a second signalling molecule downstream in the DSB recognition/repair pathway overlapping with the γh2ax signal. They reported numbers of such double-staining foci per cell at 2, 24, and 48 h after irradiation of the 184A1human mammary gland epithelial cell line with graded doses of 10 kv and 25 kv x rays (from the same tungsten target x-ray tube), compared to 200 kv x rays. During this sampling period the numbers of double-stained foci per cell increased to a dosedependent maximum at 0.5 to 2 h and then declined, unlike single-stained γh2ax foci that show a maximum number at very short times (within a few minutes or less) and then decline. The analysis by Beyreuther et al. (2009) of the data for their double-stained foci concentrated on the 24 h time point, for which they concluded that For the mammographic x rays [i.e., 25 kv] RBE values higher than one were achieved only for doses higher than 2 [Gy]. However, this conclusion was based on the parameters of fitted linear-quadratic equations which appear poor representations of the experimental data points at high and/or low doses. Inspection of the published data shows that for every individual dose at 24 h, and also at 2 h, the number of foci induced by 25 kv x rays is notably greater than by 200 kv x rays. For the lower doses (0.25 to 2 103

2281 2282 2283 2284 2285 2286 2287 2288 2289 2290 2291 2292 2293 2294 2295 2296 2297 2298 2299 2300 2301 2302 2303 2304 2305 2306 2307 2308 2309 2310 2311 2312 2313 NCRP SC 1-20 Gy), at 24 h the ratios of these numbers ranged from 1.7 to 3 and at 2 h ranged from 1.4 to 4.1. On this basis it appears that the RBEs of the 25 kv x rays, relative to 200 kv x rays, are substantially greater than unity even at the lower doses and are probably greater than 2 at both sampling times. For 10 kv x rays, the authors deduced RBEs of 4 to 7 relative to 200 kv at 24 h; direct observation of the published data for doses 2 Gy suggests that the RBEs may be somewhat greater, in the range of 6 to 10 at 24 h and about 3 to 8 at 2 h. These results from Beytheuther et al. (2009) are now represented in Table 5.2 by RBEs (at 2 h) of ~5 and ~2 for 10 kv and 25 kv x rays, respectively, both with 200 kv x rays as reference. It should be noted, however, that these combined γh2ax+53bp1 foci do not provide a measure of initial DSB, so they have not been included in Figure 5.3. Also there are differences between the original authors interpretation of the effectiveness of 25 kv x rays and the interpretation suggested here. Increases in γh2ax assay foci have been reported in normal human cells irradiated in vitro with very low doses of x rays from clinical mammography tubes operated at 28-30 kv. Colin et al. (2011) reported significant increases in foci in breast glandular epithelial cells after doses of only 4 mgy, but no other radiations were studied for comparison. Depuydt et al. (2013) compared the effects of 5 to 400 mgy in vitro mammography x-ray irradiation with effects of 60 Co gamma rays for induction of γh2ax foci and micronuclei in human lymphocytes in blood from 5 patients. Significant increases in number of foci were observed at doses as low as 10 mgy of the 30 kv x rays. Dose-responses for foci were linear (except perhaps at the very lowest doses) and the RBE from relative slopes was 1.35, with 95 % CIs 0.95 and 1.98. At each of the 5 dose points the yield of foci was greater for 30 kv x rays than for 60 Co gamma rays. The results for micronuclei showed linear-quadratic dose-responses and had larger RBEs over the full dose range studied, suggesting that DSB from 30 kv x rays may have greater biological severity, as well as being produced in larger numbers per unit dose (see also Section 5.5). In a recent publication, Mills et al. (2015) report on the induction and resolution of 53BP1 foci in the MCF-10A human breast cell line after low-dose irradiation with 29 kv mammography x rays compared to 137 Cs gamma rays. Using a highly sensitive automated 53BP1 assay after 29 kv x-ray doses of 3 to 30 mgy, they found no significant difference between the radiations for induction and resolution of the foci. The RBE for the initial number of induced DSB was calculated to be 1.1 ± 0.2 (standard deviation). They did observe, however, subtle differences between the radiations in the distribution of foci throughout the nuclei. 104

Fig. 5.3. Experimentally-measured RBE values for initial DNA DSB produced by low-energy photons, plotted from the results listed in greater detail in Table 5.2. The plot is presented only to give a concise visual summary of the available data. It should be noted that, for the purpose of this plot, broad spectrum x rays have been represented at a single energy, without consideration of the wide distribution of photon (and secondary electron) energies within the spectrum: points have been plotted at 35 kev for the 50 kv filtered x rays of Frankenberg et al. (1999); 20 kev for the 29 kv filtered x rays of Kuhne et al. (2005); 3.0 kev and 1.5 kev for the 4 kv (Ag anode) and 2 kv (Al anode) x rays, respectively, of Frankenberg and Binder (1985); 17 kev and 80 kev for the 25 kv (W anode) and 120 kv (W anode) x rays, respectively, of Kegel et al. (2007); 67 kev for the 100 kv x rays of Beels et al. (2010); and at the mean of the stated energy range of the spectral fraction results of Frankenberg and Binder (1985). All other RBE data points are for nominally monoenergetic K- or L-shell x rays and are plotted accordingly. Additionally, a variety of different reference radiations for the RBE have been used in the experiments, namely 60 Co gamma rays in all cases except for Prise et al. (1989) (reference 250 kv x rays), Hamada et al. (2006) (240 kv X-rays), Frankenberg et al. (1999) (30 MeV electrons) and Kegel et al. (2007) ( 137 Cs gamma rays). 105