RADIATION RESEARCH 9, 8 8 () -787/ $. y Rdition Reserch Society. All rights of reproduction in ny form reserved. Second Anlysis of Mortlity of Nucler Industry Workers in Jpn, 98 997 Tmiko Iwski,,e Motoi Murt,, Sumio Ohshim, Toshio Miyke, Shin-ichi Kudo, Ysushi Inoue, Minoru Nrit, Tkesumi Yoshimur, Suminori Aki, c Toshiro Tngo, d Ysuhiko Yoshimoto, e Yukiko Shimizu, f Tomotk Soue, g Shizuyo Kusumi, Chiko Ymgishi nd Hiromichi Mtsudir Rdition Effects Assocition, Kjicho, Chiyodku, Tokyo, -, Jpn; University of Occuptionl nd Environmentl Helth, - Iseigok, Yhtnishi-ku, Kitkyushu 87-8, Jpn; c Kgoshim University, 8--, Skurgok, Kgoshim, 89-8, Jpn; d Ntionl Institute of Pulic Helth, -, Minmi, Wko, -97, Jpn; e Ntionl Institute of Rdiologicl Sciences, -9-, Angw, Inge-ku, Chi, -8, Jpn; f Rdition Effects Reserch Foundtion, -, Hijiym Prk, Minmi-ku, Hiroshim, 7-8, Jpn; nd g Ntionl Cncer Center Reserch Institute, -- Tsukiji, Chuo-ku, Tokyo, -, Jpn Iwski, T., Murt, M., Ohshim, S., Miyke, T., Kudo, S., Inoue, Y., Nrit, M., Yoshimur, T., Aki, S., Tngo, T., Yoshimoto, Y., Shimizu, Y., Soue, T., Kusumi, S., Ymgishi, C. nd Mtsudir, H. Second Anlysis of Mortlity of Nucler Industry Workers in Jpn, 98 997. Rdit. Res. 9, 8 8 (). A cohort study of nucler industry workers ws initited in 99 to determine the possile helth effects of low-level rdition. A totl of,7 deths were scertined mong 7, mle workers who hd een retrospectively nd/or prospectively followed for n verge of 7.9 yers during the oservtion period 98 997. Sttisticl nlyses were mde minly on the prospective follow-up outcome of, workers followed for n verge of. yers. The stndrdized mortlity rtio (nd its 9% confidence intervl) ws.9 (.9,.97) for,9 cses of ll cuses comined nd.8 (.8,.9) for, cses of non-mlignnt diseses comined, which suggested helthy worker effect. For,9 cses of ll cncers comined, it ws.98 (.9,.), indicting no difference in mortlity from tht of the generl popultion. In tests for trend of deth rte with incresing rdition dose, no significnt correltion ws found for ll cncers comined. For site-specific cncers, most cncers including leukemi showed no positive correltion with dose, except for cncers of the esophgus, stomch nd rectum nd multiple myelom. Externl cuses showed significnt correltion with dose. A seprte questionnire study indicted tht these positive findings could e scried in prt to lifestyle chrcteristics of the workers. For leukemi only, we ttempted to estimte the excess reltive risk per unit dose of rdition, which, with reservtions ecuse of its wide confidence intervl, ws within the rnge of vrition of the risks reported in other rdition epidemiologicl studies. This popultion must e studied for longer time nd with considertion of the possile effects of confounding fctors. y Rdition Reserch Society INTRODUCTION Current rdition protection stndrds throughout the world re sed lrgely upon the recommendtions of the Address for correspondence: Institute of Rdition Epidemiology, Rdition Effects Assocition, -9-, Kji-cho, Chiyod-ku, Tokyo -, Jpn; e-mil: mmurt@re.or.jp. Interntionl Commission on Rdiologicl Protection (). The risk estimtes dopted in the recommendtions rely in prt on epidemiologicl studies of tomic om survivors in Hiroshim nd Ngski who were exposed to high doses of cute rdition (). It ws ssumed tht the findings on cncer mortlity from those studies could e extrpolted to low-dose nd low-dose-rte rdition exposures with some modifictions. However, considerle uncertinties re ssocited with tht extrpoltion process, ecuse the iologicl mechnisms of rdition crcinogenesis re still not well understood (). It is thus importnt to provide sis for the dequcy of current rdition protection mesures, t lest for rdition workers, in terms of the helth effects of low-level rdition. A numer of epidemiologicl studies of nucler industry workers hve een undertken to otin informtion on the helth effects of low-dose nd low-dose-rte rdition exposure ( ). In Jpn, the Institute of Rdition Epidemiology (IRE) of the Rdition Effects Assocition (REA) initited cohort study of nucler industry workers in 99. The Ministry of Eduction, Culture, Sports, Science nd Technology (MEXT) of Jpn hd entrusted this study to REA. The results of the first nlysis, deling with oservtions for the period 98 99, were reported previously (7). The present report presents the results of the second nlysis with extension of the oservtion period to 997 nd with lrger numer of study sujects. MATERIALS AND METHODS This study ws reviewed nd pproved y the Reserch Ethics Committee nd conformed to the Guidelines for Ethics of Epidemiologicl Study of the Jpnese government. Popultion nd Follow-up of Vitl Sttus In Jpn, registry system of rdition workers t nucler fcilities ws estlished in 977 tht is operted y the Rdition Dose Registrtion Center (RADREC) of REA. Its opertionl system nd dtse of dose records were descried in detil in the previous pper (7). The present follow-up study ws sed upon pproximtely, rdition 8
MORTALITY OF NUCLEAR INDUSTRY WORKERS IN JAPAN 9 workers, including, mles nd, femles, who were registered y RADREC s of the end of Mrch 99 nd who stisfied certin requirements such s hving record of exposure dose nd Jpnese ntionlity. Compred to our first report (7), which delt with 8, workers registered y RADREC s of Mrch 989, the present study ws supplemented y out, workers, consisting of those who worked only t fuel-processing plnts tht were excluded from the previous nlysis (out,) nd those who hd first engged in rdition work fter Mrch 989 (out,). The requirement of Jpnese ntionlity ws set to use residence registrtion crds (RRC) for follow-up studies. In Jpn, every municiplity mintins RRC of residents, including their nme, dte of irth, gender, ddress, first dte of residence, nd previous ddress. If ny resident dies or moves to nother municiplity, the RRC is updted with the dte of deth or reloction nd the next ddress in the cse of reloction nd is mintined for exctly yers, fter which it is deleted from the registrtion file. For the follow-up of study sujects, personl identifiction informtion ws first provided y RADREC, including their centrl registrtion numer, nme, gender nd dte of irth. Next, their residence ddress informtion ws otined with the coopertion of the nucler fcilities t which they worked. Then copies of RRC were cquired from the respective locl government offices to scertin the vitl sttus of the study sujects. Becuse of the limited period of mintennce of RRC for those who died or emigrted, follow-up checks must e continued t intervls of less thn yers, since those who hve died or emigrted would e lost from the study popultion otherwise. Using RRC, the cohort cn e followed up oth retrospectively nd prospectively. For the retrospective oservtions, the initil dte ws determined y tking the lter of either April of the first yer of enggement in rdition work or the dte yers preceding the dte when the RRC ws cquired t the first follow-up. Since this follow-up study strted in 99, the erliest yer of oservtion would e 98. The end of the retrospective oservtion period ws the dte when the RRC ws first cquired. Among the totl popultion of, workers, pproximtely 77, (7.%) were followed up successfully. The remining workers were lost due to either imprecise ddress informtion provided y the respective nucler fcilities or expirtion of the storge period of the RRC t the respective municipl offices for deths or reloctions tht hd occurred prior to the first follow-up. Prospective follow-up ws mde for out, workers who were confirmed to e live t time of the first follow-up. Thus the initil dte of the prospective oservtion period ws the dte when the RRC hd first een cquired. The terminl dte ws the dte when the RRC hd een otined in the most recent follow-up, the dte of deth or reloction, or the dte of dropout from the study due to filure of follow-up, whenever these events occurred erlier thn the closing dte of the oservtion period, i.e. Decemer, 997. Otherwise, this ltter dte ws the terminl dte. The dropout rte ws only.% of the popultion, nd the mjority of this resulted from those who hd emigrted from certin municiplity ut who for some unknown reson could not e confirmed to hve settled successfully in the municiplity given s the destintion of reloction. For those whose deth ws scertined from the RRC, the cuse of deth ws otined y record linkge with mgnetic tpe copies of vitl sttistics deth records (for the period 98 997) provided y the Ministry of Helth, Lor nd Welfre of Jpn. Indices used for record linkge were the dte of irth, dte of deth, sex nd municiplity of residence. Iwski et l. () reported successful determintion of cuses of deth y mens of this record linkge in 99.7% of sujects. Rdition Dose Records The effective dose dt registered t RADREC were used to determine the cumultive doses of suject workers. Dosimetry records of rdition workers re prepred y the respective nucler fcilities for the purpose of rdition protection mngement, nd exposure doses re reported to RADREC ech yer s the nnul (msv), externl nd internl doses comined. Internl doses were lmost negligile. Film dge dosimeters nd, especilly in recent yers, thermoluminescence dosimeters were used for determintion of externl rdition doses. Doses elow the detectle level, which ws round. msv for rys ut ws slightly vrile depending on the different types of dosimeters, were reported to RAD- REC s x vlues. Such doses were counted s msv in this study. For dose records missing for such resons s the loss of dosimeters, the respective fcility mde estimtes of dose sed on certin predetermined rules, such s use of mesurement y supplementry dosimeter or use of dosimeters of other workers who hd engged in work t the sme time. Although RADREC initited the registrtion in 978, erlier dosimetry records of workers hd een mintined t the respective nucler fcilities ck to 97, nd these were provided to RADREC. Consequently, dose dt ville for this study cover the period from 97 to 997; these dt were used to clculte the cumultive rdition doses for individul workers. The verge nnul dose per person hs grdully decresed over the yers, from out. msv efore 98 to out. msv fter 99, which is scrile minly to technicl improvements in procedures to reduce the rdition exposure of workers. Dose records filed in RADREC reflect chnges over time in the definition of rdition quntities nd units, dvnces in the methods of dose mesurement nd evlution of dose, nd methodologicl differences etween the respective nucler fcilities. The Rdition Dosimetry Committee ws orgnized to investigte nd exmine from n expert viewpoint the prolems involving rdition control prctices nd the methods of dose mesurement nd evlution. The investigtions nd ssessments were crried out sed on the results of questionnire survey nd onsite inspections of ll nucler fcilities, nd lso y tking into considertion vrious pertinent technicl mterils. The results of these exmintions indicted tht the qulity of the dose records ws dequte nd proper for use in this epidemiologicl study, s descried in the previous report (7). Chrcteristics of the Study Popultions Since the follow-up of the cohort popultion ws mde oth retrospectively nd prospectively, the following two study popultions were selected for the sttisticl nlysis. First ws the totl study popultion, consisting of pproximtely 7, mle workers whose vitl sttus hd een verified through either retrospective or prospective follow-up for the period 98 997. From this popultion, pproximtely 9, workers who hd een followed up prospectively for the period 99 997 were seprted nd designted s the prospective study popultion. All of the remining workers (out 7,) were those sujects who hd een dded to the cohort popultion susequent to the first nlysis. Femle workers were excluded from the study popultions ecuse they were too few in numer, i.e. out 9. Becuse follow-up ws egun retrospectively y use of RRC in this study, there is no fixed single strting point common for ll sujects in the popultion. Thus Tle shows the distriution of the irth yers of the study sujects insted of their ge distriution t the strt of the follow-up. The irth yer rnged from 9 to 977, thus covering more thn 7 yers. It ppers tht the distriution is skewed slightly to lter yers in the totl study popultion compred to the prospective study popultion, reflecting the fct tht the former popultion hd lrger proportion of those recently engged in work thn the ltter. Tle shows the distriution of cumultive doses y the end of 997 in oth the totl nd prospective study popultions. Among those clssified in the -msv dose clss, the dose vlues for % of the totl nd 8% of the prospective study popultions were ctully elow the detectle level ( msv). In 7% of those in the msv group, the dose exceeded msv ut ws less thn msv. For the totl study popultion, the men cumultive dose per person ws pproximtely. msv nd the popultion dose ws out,9 person-sv. The men cumultive dose per person in the prospective study popultion ws p-
IWASAKI ET AL. Birth yers 99 9 99 9 99 9 99 9 99 9 99 97 Totl TABLE Distriution of Birth Yers in the Study Popultions Totl study popultion Numer (%), (.7) 9,78 (.), (.8),77 (.),9 (.7),7 (.9),97 (7.) 7,99 () Prospective study popultion Numer (%), (.9) 7,9 (.), (7.),7 (7.8),7 (9.9),8 (7.) 7 (.) 9,8 () Four percent of totl nd % of prospective study popultions were orn erlier thn 9 nd the erliest irth yer ws 9. The ltest irth yer ws 977. proximtely. msv nd the popultion dose ws out,8 person- Sv. The difference in men cumultive dose etween the totl nd prospective study popultions reflects the lrger proportion of persons in the former group who hd recently ecome engged in rdition work. The distriutions of clendr yer of first enggement in rdition work nd length of employment in rdition work in the different dose groups s of the end of 997 re shown in Tles nd, respectively, for only the prospective study popultion. It ppers tht the higher the cumultive dose, the lrger is the percentge who strted rdition work in erlier yers s well s the percentge who hd een working for longer times. The ltest plce of residence of study sujects during the oservtion period ws clssified into eight mjor districts from north to south; their distriution in the individul dose groups is shown in Tle only for the prospective study popultion. A lrge difference in the residence re ws oserved mong the indicted dose groups with sttisticl significnce (P. y test). Among the eight districts, Knto ws the most common re of residence in lower-dose clsses, wheres in the higher-dose clsses, residents of Hokkido nd Tohoku, nd Hokuriku res were more common. The verge oservtion period ws 7.9 yers per person for the 7,99 workers in the totl study popultion (retrospective nd prospective follow-up sujects) nd thus generted.9 million personyers. In the prospective study popultion, the verge oservtion period for the 9,8 workers ws out. yers per person nd thus generted. million person-yers. The men ge t the end of oservtion period ws.7 nd 9. yers for the totl nd prospective study popultions, respectively. Sttisticl Anlysis Both externl nd internl comprisons were mde using methods similr to those of other rdition epidemiology studies ( ). The totl Cumultive dose groups (msv) Totl TABLE Distriution of Cumultive Doses to the End of 997 in the Study Popultions. Totl study popultion Numer (%),89 (7.9),9 (9.),7 (9.) 7,9 (.), (.) 7,99 () Men cumultive dose (msv)... 9.8.. Prospective study popultion Numer (%) 8, (9.), (.),97 (.), (.), (.) 9,8 () Men cumultive dose (msv).7..9 7..7. Of this clss, % in the totl nd 8% in the prospective study popultions hd msv exposure. Of this clss, 7% exceeded msv, ut received less thn msv. oservtion period ws divided into three clendr-yer periods (98 989, 99 99 nd 99 997), nd the totl ge rnge ( to 8 yers) ws strtified into -yer ge clsses (, 9,..., 8 8). Oserved person-yers were clculted seprtely for ech ge clss y ech oservtion period. Person-yers for those over 8 yers old were excluded ecuse deth certificte dignosis of cuses of deth is generlly less relile t older ges. Cuses of deths of interest in this study were primrily ll cncers nd site-specific cncers, ut they lso included ll cuses, non-neoplstic diseses nd externl cuses. In the externl comprison, the stndrdized mortlity rtio (SMR) nd its 9% confidence intervl (CI) were clculted compred to the generl Jpnese mle popultion. The SMR is the rtio of oserved to expected numer of deths (). The expected numer of deths ws clculted y pplying the mortlity rtes of the generl Jpnese mle popultion in individul ge clsses nd clendr-yer periods s descried ove to the corresponding person-yers of the study popultion. In the internl comprison, one-sided trend nlysis with z vlues ws performed (). Nmely, the study popultion ws divided into five groups y cumultive dose levels, i.e. less thn,,, nd msv or more, nd the dt were tested for whether the rtio of oserved (O) to expected (E) numer of deths (O/E rtio) in individul dose groups ws incresed significntly with incresing cumultive dose. Men cumultive doses over the respective person-yers were used s the scores of individul dose groups. When clculting the expected numer of deths, it ws stndrdized to the mortlity of the totl popultion, in ddition to the ove-mentioned ge clsses nd clendr-yer periods. The residence re ws lso djusted for y the clssifiction shown in Tle. As ws seen in the previous section, this popultion ws heterogeneous Yers 97 97 98 98 Totl TABLE Distriution of Clendr Yer of First Enggement in Rdition Work in Different Dose Groups s of the End of 997 in the Prospective Study Popultion Dose group (msv) Totl 9,7 (.7), (.8) 9,8 (.),9 (8.) 8, (.) Percentge in prentheses.,7 (.),8 (.), (.7), (8.9), (.), (.7), (.),7 (.), (.),97 (.),9 (9.7), (.), (.8) 787 (.9), (.), (.),8 (9.), (8.) (.7), (.),87 (.),7 (7.),9 (.) 9,7 (.) 9,8 (.)
MORTALITY OF NUCLEAR INDUSTRY WORKERS IN JAPAN Period (yers) 9 Totl TABLE Distriution of Durtion of Employment in Rdition Work in Different Dose Groups s of the End of 997 Dose group (msv) Totl 8,8 (.) 8,7 (.),7 (.8), (9.) 8, (.) Percentge in prentheses.,8 (.8),78 (.),9 (.),87 (7.), (.) 89 (.),9 (.), (8.) 8,7 (.7),97 (.) (.) 78 (.7) 99 (.), (8.), (.) (.) (.) (.),8 (9.), (.),9 (.9), (9.7),9 (.) 8, (.) 9,8 (.) with respect to vrious chrcteristics, such s strting ge nd strting yer of rdition work, durtion of enggement in work, residence re nd so forth (Tles ), nd some of these fctors were potentil confounding fctors in the present study. Among these chrcteristics, t lest residence re would proly e ssocited with mortlity of the study sujects, since geogrphic vrition in mortlity is well known for vrious kinds of cncer in Jpn (). Although it would e etter to choose the plce of residence in the erlier or min prt of life of individuls s confounder, such informtion ws not ville ecuse of the short length of the present follow-up period. We therefore decided to include the ltest residence s potentil confounder in the internl comprison. The other fctors, though they my lso e ssocited with the mortlity of the study sujects, were not dopted s controlling fctors simply ecuse the oservtion period ws still too short to djust for these timedependent fctors in the nlysis. The internl comprison ws performed in two wys, either with or without considertion of the ltent period for rdition-induced cncer. In deling with the ltent period, we dopted the method used in the study of NRPB in the United Kingdom (). Nmely, yers for leukemi nd yers for other cncers were set s the minimum ltent periods, nd rdition doses were summed from the eginning of rdition work up to these numers of yers efore the terminl dte of the oservtion. Thus some sujects might move from certin dose clss to lower dose clss. Similrly, the first yers of follow-up for leukemi nd yers for other cncers, respectively, fter first employment to rdition work were excluded from the oservtion period. Accordingly, the numers of deths during these periods were lso excluded from the nlysis. If the oserved numer of deths ws less thn nd the estimted trend P vlue ws lower thn., the results were confirmed y the simultion method (). Furthermore, to void inflted detection of sttisticl significnce during repeted multiple tests, supplementry multiple comprison nlysis (Bonferroni s method) ws performed (). Externl Comprison RESULTS The totl numer of deths ws,7, including,8 cncer deths, during the oservtion period 98 through 997 in the totl study popultion (Tle ). The SMR (nd its 9% CI) for deths due to ll cuses, non-neoplstic diseses, nd ll cncers ws.9 (.87.9),.8 (.77.8) nd.9 (.9.98), respectively. They were ll significntly less thn. The totl numer of deths ws,9, including,9 cncer deths in the prospective study popultion during the oservtion period 99 through 997 (Tle 7). The SMR (nd its 9% CI) for deths due to ll cuses, non-neoplstic diseses, nd ll cncers ws.9 (.9.97),.8 (.8.9), nd.98 (.9.), respectively. The former two were oth significntly less thn, wheres tht of ll cncers ws not significntly different from. In ddition, ll of the SMRs for these cuses of deth were lower in the totl study popultion thn in the prospective study popultion, proly due to certin mount of incomplete retrospective follow-up (see the Discussion). The SMR for site-specific cncers, including leukemis, ws not significntly different from in the prospective study popultion. Incidentlly, no cse mong the leukemi ws certified s chronic lymphocytic leukemi (CLL). TABLE Distriution of the Finl Residence Are of Study Sujects in the Prospective Study Popultion y the 8 Mjor Districts in the Entire Country y Dose Group District A B C D E F G H Totl Dose group (msv),9 (.),8 (.), (7.),7 (7.),79 (.), (.), (.),8 (7.) 8, (.),9 (8.),88 (.9),9 (.) (.),9 (.) 9 (7.) (.), (9.), (.), (.7),8 (.),7 (.) 788 (.),9 (.) 7 (.) 78 (.), (7.),97 (.),9 (.7), (.) 89 (.) 8 (.8), (.9) (.9) 88 (.) (.), (.) From north to south, A: Hokkido nd Tohoku, B: Knto, C: Hokuriku, D: Chuu, E: Kinki, F: Chugoku, G: Sikoku, H: Kyusyu. Percentge in prentheses.,8 (9.) 7 (.9) 8 (.) (.) (.) 9 (.7) (.8) (.), (.)
IWASAKI ET AL. TABLE SMR y Cuse of Deth during the Oservtion Period 98 997 in the Totl Study Popultion Cuse of deth All cuses Externl cuses Non-neoplstic disese All neoplsms Mlignnt neoplsms All sites Orl, phrynx Esophgus Stomch Colon Rectum Liver Gllldder Pncres Lung Prostte Bldder Kidney nd other nd unspecified urinry orgns Brin nd centrl nervous system Non-Hodgkin s lymphom Multiple myelom Leukemi All sites except leukemi Neoplsm of mlignnt, enign nd unspecified nture. Numer of cses of chronic lymphocytic leukemi ws zero. Oserved numer,7 9,,8,8 8 8 9 7 7 97 7,78 Expected numer,8.8 9.7,9.8,.,7.9 8. 9. 8.9. 9. 9. 7. 9..9.8. 7. 7.7 7. 7.8 7.,. SMR.9..8.9.9.79.8.89..87..9.98.97.9..8.9.8..89.9 9% confidence intervl (.87.9) (.99.) (.77.8) (.9.98) (.9.98) (..8) (.8.) (.8.98) (.8.) (.7.) (.9.) (.7.) (.8.7) (.87.7) (..) (.7.8) (.8.) (..) (.9.7) (.9.7) (.8.) (.9.98) P vlue for two-sided test.8...9.8..87.8.7.7.8..88...7..8.9.7 Internl Comprison The results of the internl comprison to e presented will e limited to the prospective study popultion, ecuse, s will e discussed lter, retrospective follow-up of the totl study popultion suffered from selection is. The result of the nlysis without considertion of the ltent period re shown in Tle 8 nd those including correction for the ltent period for cncers re presented in Tle 9. The totl numer of deths with cncer ws,9 without considertion of the ltent period, ut it ws reduced to,7 with inclusion of the ltency. The results of nlyses were lmost comprle etween these two groups. Although the totl mortlity rte incresed significntly with incresing cumultive dose (P.7), it ws minly scried to deths due to externl cuses; the ltter showed clerly nd significntly elevted mortlity in higher compred to lower dose clsses (P.), wheres nonneoplstic diseses showed no pprent trend of increse with dose (P.7). For cncers of ll sites, though the O/E rtio exhiited slight tendency of n increse from the lowest to the highest dose clsses, the trend ws not sttisticlly significnt, either with (P.99) or without (P.8) considertion of the ltent period. Even fter the exclusion of leukemi, the results remined lmost the sme. For leukemi, too, no trend of n increse ws found (P. with nd P.8 without considertion of the ltent period). For site-specific cncers, n increse in the mortlity rte with incresing dose ws not pprent for most cncers except for cncers of the esophgus (P. nd P.), stomch (P. nd P.) nd rectum (P. nd P.7), nd multiple myelom (P.7 nd P.7), where the two P vlues in prentheses re with nd without considertion of the ltent period, respectively. For multiple myelom, P vlues were otined y the simultion method (). When the multiple comprison test (Bonferroni s method) ws pplied to the results for these cncers, the trends for cncer of the stomch (P.8 nd P.) nd rectum (P.7 nd P.8) nd multiple myelom (P.78 nd P.7) were found to e non-significnt, while significnt trend remined for cncer of the esophgus (P. nd P.). DISCUSSION Results of the externl comprison reveled tht, for mjor cuses of deth, the clculted SMR vlues were generlly lower in the totl study popultion thn in the prospective one. This my e due in prt to the fct tht the retrospective follow-up period covered more young nd thus newly employed workers thn the prospective one, so tht the helthy worker effect should e mnifested more strongly (7). However, it seems prole tht the lower
MORTALITY OF NUCLEAR INDUSTRY WORKERS IN JAPAN TABLE 7 SMR y Cuse of Deth during the Oservtion Period 99 997 in the Prospective Study Popultion Cuse of deth All cuses Externl cuses Non-neoplstic diseses All neoplsms Mlignnt neoplsms All sites Orl, phrynx Esophgus Stomch Colon Rectum Liver Gllldder Pncres Lung Prostte Bldder Kidney nd other nd unspecified urinry orgns Brin nd centrl nervous system Non-Hodgkin s lymphom Multiple myelom Leukemi All sites except leukemi Neoplsm of mlignnt, enign nd unspecified nture. Numer of cses of chronic lymphocytic leukemi ws zero. Oserved numer,9 97,,,9 9 7 8 8 8, Expected numer,7.7 8.7,.7,.,..7.. 77. 8... 9.9.... 7. 9...9,8.7 SMR.9..8.98.98.9.97.9.8.8...9..97..97.8.7.79.9.98 9% confidence intervl (.9.97) (.9.) (.8.9) (.9.) (.9.) (.8.) (.7.) (.8.7) (..) (..) (.99.8) (.8.) (.7.) (.9.7) (..9) (.7.7) (.9.) (..9) (.7.) (..7) (..) (.9.) P vlue for two-sided test.8...7.8....7.9.779..98.987.98..9...97 SMR in the totl study popultion ws cused minly y the fct tht, t the strt of the cohort study, more precise informtion on residence ddress could e otined for workers who were live nd thus currently working compred to those who were decesed. Moreover, since currently employed workers were included more frequently in the higher-dose group thn in the lower-dose group, differentil follow-up rtes were introduced into the retrospective study (dt not shown). The prospective follow-up is less ffected y this selection is. In the prospective study popultion, externl comprisons showed significntly lower totl mortlity rte thn Jpnese mles in generl, possily due to the helthy worker effect (7). The totl cncer mortlity rte, for which the influence of this effect is considered to e smll (8 ), ws nerly the sme s tht of the generl Jpnese mle popultion. For ll site-specific cncers, including leukemi, the mortlity rte ws not significntly different from tht of the generl popultion. The results of the internl comprisons did not demonstrte ny significnt trend of n increse in totl cncer mortlity with rdition dose, either including or excluding leukemi, oth with nd without considertion of the ltent period. No significnt trend ws found for n increse in leukemi mortlity, wheres significnt trend ws detected for certin gstrointestinl trct cncers nd for multiple myelom. It should e emphsized tht sttisticlly significnt result might e otined y chnce even when lowlevel rdition my not hve rel helth effect. Thus multiple comprison nlysis ws performed for the results of internl comprisons for individul site-specific cncers. With this nlysis significnt trend remined only for cncer of the esophgus. The results of similr studies reported from other countries re quite vrile with respect to cncer sites, with significnt ssocition of mortlity with rdition dose eing found for ll sites (, 7, 9), esophgus (, ), lung (, 7, 9), skin (), ldder (), rin nd centrl nervous system (), multiple myelom (,,, ), leukemi (, 8,,,, ). Although these results, including those of the present study, re ssocited with lrge vritions, t lest the significnt findings for cncer of the esophgus nd multiple myelom, which re in greement with the present study, deserve ttention. For multiple myelom, however, specil emphsis should not e plced on the present results ecuse the oserved numer of cses ws only eight, six of which were in the less thn msv dose clss (Tle 8). It should e noted tht cncer of the pncres, which showed positive trend in the first nlysis (7), demonstrted no increse of mortlity in the present nlysis. For cncer of the esophgus, on the other hnd, the significnt trend with dose oserved in the present study is notle ecuse it ws confirmed y multiple comprison
IWASAKI ET AL. TABLE 8 Trend Anlysis y Cuses of Deth during the Oservtion Period 99 997, fter Adjustment for Age, Clendr Yer nd Residence Are, in the Prospective Study Popultion, without Considertion of the Ltent Periods for Cncers Cuses of deth All cuses Externl cuses Non-neoplstic diseses Mlignnt neoplsms All sites Orl, phrynx Esophgus Stomch Colon Rectum Liver Gllldder Pncres Lung Prostte Bldder Kidney nd other urinry Brin nd CNS Non-Hodgkin s lymphom Multiple myelom c Leukemi d All site except leukemi Cumultive dose group (msv).97(.9.).88 (.77.) 9.98 (.9.) 8.98 (.9.) 8. (..) 8.8 (.9.).9 (.7.) 8. (.7.).9 (..).99 (.8.) 9.8 (.8.).97 (.7.9) 7. (.8.) 8. (.8.8). (.7.).8 (..).9 (.9.8) 7.9 (..7). (.7.8) 9.99 (..) 8.98 (.9.). (.98.).8 (.79.). (.9.).7 (.89.8). (..9).8 (.8.7). (.97.).7 (..8).8 (..8). (..).7 (..9). (.8.9). (.88.8). (..). (..).98 (..). (..9). (..). (..7).7 (.9.).8 (.9.9) 9. (..). (.9.7). (.9.) 8. (.9.). (..) 8. (.7.) 8. (.7.). (.7.9) 7. (..77) 9.9 (.8.7).8 (..). (..). (.7.). (..89).8 (..).8 (..). (..7). (.7.8). (..9).7 (..). (.9.) 8.9 (.7.7).8 (.87.8) 7.88 (..).8 (..8). (..).8 (.7.).8 (.9.). (.9.7).7 (.9.) 8.8 (.9.). (..). (..8). (..).7 (..9). (..8).8 (..7). (..). (..). (.9.).8 (..) 9.8 (..). (.99.) 8.8 (..8). (.7.). (.8.). (..78). (..).8 (.99.). (..).8 (..9) 7.9 (.8.9). (..7).9 (..).7 (..9).99 (.8.). (. 9.9). (..). (. 8.). (. 7.8). (.8.).9 (..). (.8.) P vlue..7.7..8.7.8.8.8.89.89......9.7.7..9.9.9.7.97.99..9.7.9...8.97.9.9.7.9.8.8.7 Oserved numer of deths re shown in the upper row, nd O/E rtio (9% confidence intervl) re shown in the lower row, respectively. In the rightmost column, trend z vlue nd P vlue in the upper row nd lower rows, respectively. Mlignnt, enign nd unspecified neoplsms were included. c P vlue ws estimted using the simultion method. d Numer of cses of chronic lymphocytic leukemi ws zero. tests. However, exmintion of the O/E rtio for this cncer, s seen in Tle 8 for exmple, showed the difference to e s lrge s out threefold etween the lowest dose clss ( msv) nd the two highest dose clsses comined ( msv). This mgnitude of difference could not e explined s eing entirely due to the effect of rdition in view of recent risk estimtes otined from the study of A- om survivors (). This difference my hve een cused y the influence of confounding fctors, such s lifestyle, which were not tken into ccount in the present study. One reson to suspect the ssocition of confounding fctors with the positive trend oserved for certin cncers is the finding for externl cuses. At present, no explntion cn e offered for externl cuses of deth, which showed
MORTALITY OF NUCLEAR INDUSTRY WORKERS IN JAPAN All sites Orl, phrynx Esophgus Stomch Colon Rectum Liver Gllldder Pncres Lung Prostte Bldder TABLE 9 Trend Anlysis y Cuses of Deth during the Oservtion Period 99 997, fter Adjustment for Age, Clendr Yer nd Residence, in the Prospective Study Popultion for Cncers with Ltent Periods of Yers for Leukemi nd Yers for Solid Cncers, Assumed in the Anlysis Cuses of deth Kidney nd other urinry Brin nd CNS Non-Hodgkin s lymphom Multiple myelom c Leukemi d All site except leukemi Cumultive dose group (msv) 77.97 (.9.) 7. (..9) 8.7 (..8).89 (.7.) 7. (.7.7) 9.88 (.9.7) 9.98 (.8.) 8.8(.8.). (.7.). (.8.) 8. (.7.79). (.8.9).8 (..) 7. (.9.).97 (..7). (..) 9.98 (.9.) 7.97 (.9.).9 (.9.).9 (..).7 (.7.77) 8. (.9.97). (..).8 (..98). (..).7 (..) 8. (.8.). (.78.79). (..97). (..7). (..).9 (..9).8 (..). (..9).7 (.9.).9 (.9.). (.9.). (.8.9). (..). (.8-.8) 7. (..) 7.9 (..7).9 (.77.7). (..).9 (..7). (.7.7). (..9). (..9).99 (.-.). (..7).8 (..9). (..7).8 (..8). (.9.).89 (..). (..) 8.77 (. 7.). (.7.).8 (..8).88 (..8).7 (..). (..89).7 (..). (.7.9). (.7 7.98). (..9). (..9). (..). (. 7.). (. 8.).8 (..).89 (..).7 (.8.88). (. 7.8). (. 9.8). (..). (..88). (. 7.).7 (..77). (..8).97 (..).7 (..). (.7.). (..). (. 9.99). (. 8.9). (..9).8 (. 87.9). (..7). (.8.87) P vlue.88.99.99.8.7..9..8.7.98..8..8.97.9.7.98.88.7.9..99....8.8..7.9..97. Oserved numer of deths re shown in the upper row, nd O/E rtio (9% confidence intervl) re shown in the lower row, respectively. In the rightmost column, trend z vlue nd P vlue in the upper row nd lower rows, respectively. Mlignnt, enign nd unspecified neoplsms were included. c P vlue ws estimted using the simultion method. d Numer of cses for chronic lymphocytic leukemi ws zero. highly significnt trend with rdition dose. Externl cuses consist minly of suicide nd ccidents. Almost ll previous rdition epidemiology studies (, 8,, ) hve filed to show ny increse in these cuses of deth in reltion to rdition exposure, except those of Ashmore et l. (9), who found significnt positive ssocition of deth from ccidents with rdition exposure, nd those of Gilert et l. (), who found non-significnt ssocitions for externl cuses. They did not interpret these findings. Since no cusl reltionship of rdition exposure, especilly to such low doses s eing considered in this study, could e expected for externl cuses of deth, the present findings were proly influenced y some confounders. Fctors ssocited with these cuses of deth include the personlity, lifestyle, occuption nd socioeconomic sttus of the individuls nd the nture of the surrounding society, ccording to studies in other countries ( ). Although epidemiologicl studies of these cuses of deth re limited in Jpn, the vitl sttistics collected y the government () showed tht the mortlity rte from externl cuses of deth is chrcterized y gret vriility mong different industries. The possile confounding fctors must e tken into considertion since the effect of low-level rdition exposure on cncer mortlity is expected to e very smll. Since the
IWASAKI ET AL. present study did not include informtion on such fctors s the lifestyle of the rdition workers, which might e confounding the risk ssessment, questionnire survey on lifestyle ws susequently performed seprtely in 997 through 999 (7). The sujects of tht survey were,9 mles nd 7 femles currently engged in rdition work. A self-dministered questionnire form ws distriuted to nd collected from the sujects y the mngement of nucler fcilities. Vlid nswers were otined from 8,8 mle nd 8 femle sujects. Exmintion of whether the lifestyle differed mong different dose groups ws mde y the Mntel extension sttisticl test (8) only for the mle respondents, with cumultive rdition doses strtified into five clsses similrly to those used in this study. The results, in reltion to the cumultive rdition dose, could e summrized s follows.. A significntly lrger percentge of tocco smokers ws found with incresing dose. Furthermore, the numer of cigrettes smoked per dy ws lrger nd the ge of commencement of tocco smoking ws younger in the higher-dose groups.. Although the percentge of lcohol drinkers did not differ mong the dose groups, those in the higher-dose groups tended to e hevier lcohol drinkers nd were younger when they egn to drink.. Regrding the hit of drinking te nd suchlike, lower percentge of lck te drinkers ws found with incresing cumultive dose.. Pst history of occuptionl contct with certin specific hrmful mterils such s sestos, orgnic solvents, etc. ws more frequent in the higher-dose groups.. Sujects in higher-dose groups underwent X-ry exmintions of the upper digestive trct, s well s other kinds of rdiologicl exmintions, less frequently. These results indicted tht, mong these rdition workers, different dose groups hd distinct chrcteristics in their lifestyles. Though only out 8% (,7) of the prospective study popultion exmined in the present nlysis ws lso included in tht study, the proportion included incresed with incresing dose, up to % in the msv clss. Thus it is felt tht these results represent the sitution mong memers of the present study popultion firly well, nd the differences mong them in lifestyle my hve confounded the results of the present study, prticulrly with respect to rdition dose. For instnce, tocco smoking nd lcohol drinking re well known to influence the incidence of gstrointestinl trct cncers, especilly cncers of the esophgus (9). The present finding regrding cncer of the esophgus my e explined in prt y this effect. The positive trend of stomch cncer mortlity with incresing dose, which cnnot e ignored ecuse of the lrge numer of cses detected in the present popultion, lthough the trend ws not significnt y multiple comprison nlysis, could lso hve een rought out y some influence of confounding fctors. For stomch cncer, however, tocco smoking nd lcohol drinking re not known to e strongly ssocited (). Dietry fctors such s slt-rich food s well s occuptionl fctors () should e tken into considertion for this. Furthermore, with respect to the increse in gstrointestinl cncers in higher-dose groups, the lst finding () in the ove-mentioned lifestyle study is especilly noteworthy. In Jpn, popultion X-ry exmintions of upper gstrointestinl trct cncers hve een promoted from the 97s s ntionl helth policy (). The finding tht those in higher-dose clsses received exmintions less frequently thn those in lower-dose clsses indictes tht the former workers tended to hve less recognition of the need for their own helth cre. Such lifestyle chrcteristics my increse their mortlity rte. At present, however, this is only specultion. The influence of confounding fctors in the dose response reltionship etween rdition dose nd mortlity must e exmined in more detil. The excess reltive risk of cncer mortlity per unit dose of rdition exposure (ERR/Sv) hs een estimted in mny rdition epidemiology studies (,,, 8 ). This ws not done in the present study ecuse the influence of confounding fctors ws strongly suspected, especilly for certin solid cncers. On the other hnd, there is no definite evidence of n ssocition of these lifestyle fctors with leukemi ccording to present epidemiologicl knowledge (). The only fctor whose possile confounding influence cnnot e totlly disregrded is occuptionl exposure to certin leukemogenic chemicl sustnces, s ws suggested from the ove-mentioned finding () from our questionnire study (7). At present, no dt re ville on the level of exposure to specific chemicl sustnces mong individul rdition workers. Another fctor strongly ssocited with leukemi, especilly in Jpn, is infection y humn T-cell leukemi virus (). However, the min endemic re for this virus is Kyushu islnd. Only very smll proportion of the present study popultion resided in this re (Tle ). Thus it seems unlikely tht this fctor hs influenced the present results. An ttempt to estimte the excess reltive risk ws mde only for leukemi mortlity y fitting the dt to liner reltive risk regression model () with the AMFIT progrm (). The ERR/Sv estimtes (9% CI) for leukemi otined in nlyses without nd with the correction for the ltent period were. ( 9.7, 8.9) nd. (.,.), respectively. Becuse of the wide confidence intervl, s ws expected from the smll numers of oserved cses (Tles 8 nd 9), their quntittive evlution is rther limited. Still, it cn e stted t lest tht the vlues re within the rnge (. 7.) oserved in other pulished studies (8, 9,,,, 8 ), some of which (8,,, ) hve estimted ERR only for leukemi other thn CLL. Furthermore, compred to the ERR estimte (.) from the study of the A-om survivors (), the centrl estimte of the ERR/Sv in our study ppers to e very smll, even though the difference ws not sttisticlly significnt.
MORTALITY OF NUCLEAR INDUSTRY WORKERS IN JAPAN 7 In conclusion, the present results hve not yielded ny definite evidence s to whether exposure to occuptionl low-level rdition increses cncer mortlity. The oservtion period in this study is short, nd the verge ge of the suject popultion is rther young. Efforts re eing mde to continue this study over longer period nd to consider the possile effects of confounding fctors to otin more relile informtion out the helth effects of low-dose nd low-dose-rte rdition. ACKNOWLEDGMENTS We thnk ll memers of the following committees for their collortion in this study: the Study Evlution Committee, the Study Opertion Committee, the Reserch Ethics Committee, nd the Rdition Dosimetry Committee, which were estlished under the egis of the REA Bord Chirmn to ensure tht the study would e conducted properly. Thnks re lso due to ll stff memers of IRE nd the Director of RADREC, who provided personl dosimetry records. Received: Septemer 7, ; ccepted: Septemer, REFERENCES. ICRP, 99 Recommendtions of the Interntionl Commission on Rdiologicl Protection. Puliction, Annls of the ICRP, Vol., No., Pergmon Press, New York, 99.. Y. Shimizu, H. Kto nd W. J. Schull, Studies of the mortlity of A- om survivors. 9. Mortlity, 9 98: Cncer mortlity sed on the recently revised doses (DS8). Rdit. Res., (99).. UNSCEAR, Sources nd Effects of Ionizing Rdition. United Ntions, New York,.. V. Berl, P. Frser, L. Crpenter, M. Booth, A. Brown nd G. Rose. Mortlity of employees of the Atomic Wepons Estlishment, 9 98. Br. Med. J. 97, 77 77 (988).. P. G. Smith nd A. J. Dougls, Mortlity of workers t the Sellfield plnt of British nucler fuels. Br. Med. J. 9, 8 8 (98).. E. S. Gilert, G. R. Petersen nd J. A. Buchnn, Mortlity of workers t the Hnford site: 9 98. Helth Phys., (989). 7. S. Wing, C. M. Shy, J. L. Wood, S. Wolf, D. L. Crgle nd E. L. Frome, Mortlity mong workers of Ok Ridge Ntionl Lortories Evidence of rdition effects in follow-up through 98. J. Am. Med. Assoc., 97 (99). 8. G. M. Kendll, C. R. Muirhed, B. H. McGion, J. A. O Hgn, A. J. Conquest, A. A. Goodill, B. K. Butlnd, T. P. Fell, D. A. Jckson nd T. J. Silk, Mortlity nd occuptionl exposure to rdition: First nlysis of the Ntionl Registry for Rdition Workers. Br. Med. J., (99). 9. P. Frser, L. Crpenter, N. Mconochie, C. Higgins, M. Booth nd V. Berl, Cncer mortlity nd moridity in employees of the United Kingdom Atomic Energy Authority, 9 8. Br. J. Cncer 7, (99).. M. A. Griin, J. L. Weeks nd G. R. Howe, Cncer mortlity (9 98) mong mle employees of Atomic Energy of Cnd Limited with respect to occuptionl exposure to externl low-liner-energytrnsfer ionizing rdition. Rdit. Res., 7 8 (99).. E. S. Gilert, E. Omohundoro, J. A. Buchnn nd N. A. Holter, Mortlity of workers t the Hnford site: 9 98. Helth Phys., 77 9 (99).. E. S. Gilert, D. L. Crgle nd L. D. Wiggs, Updted nlyses of comined mortlity dt for workers t the Hnford Site, Ok Ridge Ntionl Lortory, nd Rocky Flts Wepons Plnts. Rdit. Res., 8 (99).. A. J. Dougls, R. Z. Omr nd P. G. Smith, Cncer mortlity nd moridity mong workers t the Sellfield plnt of British Nucler Fuels. Br. J. Cncer 7, (99).. L. Crpenter, C. Higgins, A. Dougls, P. Frser, V. Berl nd P. Smith, Comined nlysis of mortlity in three United Kingdom nucler industry workforces, 9 988. Rdit. Res. 8, 8 (99).. L. D. Wiggs, E. R. Johnson, C. A. Cox-DeVore nd G. L. Voelz, Mortlity through 99 mong white mle workers t the Los Almos Ntionl Lortory: Considering exposures to plutonium nd externl ionizing rdition. Helth Phys. 7, 77 88 (99).. E. Crdis, E. S. Gilert, L. Crpenter, G. Howe, I. Kto, B. K. Armstrong, V. Berl, G. Cowper, A. Dougls nd L. D. Wiggs, Effects of low doses nd low dose rtes of externl ionizing rdition: Cncer mortlity mong nucler industry workers in three countries. Rdit. Res., 7 (99). 7. Y. Hosod nd Epidemiologicl Study Group of Nucler Workers (Jpn), First nlysis of mortlity of nucler industry workers in Jpn, 98 99. J. Helth Phys., 7 8 (997). 8. E. L. Frome, D. L. Crgle, J. P. Wtkins, S. Wing, C. M. Shy, W. G. Tnkersley nd C. M. West, A mortlity study of employees of the nucler industry in Ok Ridge, Tennessee. Rdit. Res. 8, 8 (997). 9. J. P. Ashmore, D. Krewski, J. M. Zielinski, H. Jing, R. Semenciew nd P. R. Bnd, First nlysis of mortlity nd occuptionl rdition exposure sed on the Ntionl Dose Registry of Cnd. Am. J. Epidemiol. 8, 7 (998).. C. R. Muirhed, A. A. Goodill, R. G. E. Hylock, J. Vokes, M. P. Little, D. A. Jckson, J. A. O Hgn, J. M. Thoms, G. M. Kendll nd G. L. C. Berridge, Occuptionl rdition exposure nd mortlity: Second nlysis of the Ntionl Registry for Rdition Workers. J. Rdiol. Prot. 9, (999).. W. N. Sont, J. M. Zielinski, J. P. Ashmore, H. Jing, D. Krewski, M. E. Fir, P. R. Bnd nd E. G. Letourneu, First nlysis of cncer incidence nd occuptionl rdition exposure sed on the Ntionl Dose Registry of Cnd. Am. J. Epidemiol., 9 8 ().. T. Iwski, T. Miyke, S. Ohshim, S. Kudo nd T. Yoshimur, A method of identifying underlying cuses of deth in epidemiologicl study. J. Epidemiol., ().. N. E. Breslow nd N. E. Dy, Sttisticl Methods in Cncer Reserch: Vol. II. The Design nd Anlysis of Cohort Studies. Interntionl Agency for Reserch on Cncer, Lyon, 987.. S. Mizuno, H. Arimoto, N. Ymguchi nd S. Wtne, Age-stndrdized cncer mortlity rtios y prefecture for yers, 98 to 99. Jpn. J. Clin. Oncol., 7 (99).. E. S. Gilert, Some Computer Simultions Bsed on the Liner Reltive Risk Model. Report No. PNL-787, Pcific Northwest Lortory, Richlnd, WA, 99.. K. J. Rothmn nd S. Greenlnd, Modern Epidemiology, nd ed. Lippincott Willims nd Wilkins, New York, 998. 7. A. J. McMichel, Stndrdized mortlity rtio nd helthy worker effect : Scrtching eneth the surfce. J. Occup. Med. 8, 8 (97). 8. K. Koym, The helthy worker effect in long-term follow-up popultion. Jpn. J. Cncer Clin., 7 (999). [in Jpnese with English strct] 9. B. C. K. Choi, Definition, sources, mgnitude, effect modifiers nd strtegies of reduction of the helthy worker effect. J. Occup. Med., 979 988 (99).. R. Chen nd A. Seton, The influence of study chrcteristics on helthy worker effect: A multiple regression nlysis. Occup. Med., (99).. D. A. Pierce, Y. Shimizu, D. L. Preston, M. Veth nd K. Muchi, Studies of the mortlity of tomic om survivors. Report, Prt. Cncer: 9 99. Rdit. Res., 7 (99).. M. Miller, D. Hemenwy nd E. Rimm, Cigrettes nd suicide: A prospective study of, men. Am. J. Pulic Helth 9, 78 77 ().. J. Angst nd P. J. Clyton, Personlity, smoking nd suicide: A prospective study. J. Affect. Disord., (998).
8 IWASAKI ET AL.. A. J. Kposw, Suicide mortlity in the United Sttes: Differentils y industril nd occuptionl groups. Am. J. Ind. Med., (999).. A. E. Kunst, F. Groenhof, J. P. Mckench nd E. W. Helth, Occuptionl clss nd cuse specific mortlity in middle ged men in Europen countries: Comprison of popultion sed studies. EU Working Group on Socioeconomic Inequlities in Helth. Br. Med. J., (998).. Sttistics nd Informtion Deprtment, Ministry of Helth nd Welfre, Specil Report of Vitl Sttistics in FY 99: Occuptionl nd Industril Aspects. Kousei-Toukei-Kyouki, Tokyo, 999. [in Jpnese with English summry] 7. M. Murt, T. Miyke, Y. Inoue, S. Ohshim, S. Kudo, T. Yoshimur, S. Aki, T. Tngo, Y. Yoshimoto nd H. Mtsudir, Life-styles of rdition workers t nucler fcilities in Jpn: Bse-line dt of questionnire survey. J. Epidemiol., in press. 8. N. Mntel, Chi-squre tests with one degree freedom; extension of the Mntel-Henszel procedure. J. Am. Stt. Assoc. 8, 9 7 (9). 9. F. Levi, Cncer prevention: Epidemiology nd perspectives. Eur. J. Cncer, 9 9 (999).. S. Kono nd T. Hiroht, A review on the epidemiology of stomch cncer. J. Epidemiol., (99).. P. Cocco, M. H. Wrd nd E. Buitti, Occuptionl risk fctors for gstric cncer: An overview. Epidemiol. Rev. 8, 8 (99).. S. In, H. Hirym, C. Ngt, Y. Kurisu, N. Tktsuk, N. Kwkmi nd H. Shimizu, Evlution of screening progrm on reduction of gstric cncer mortlity in Jpn: Preliminry results from cohort study. Prev. Med. 9, (999).. H. Zee nd M. Blettner, Adult leukemi: Wht is the role of currently known risk fctors? Rdit. Environ. Biophys., 7 8 (998).. K. Tjim, The th ntion-wide study of dult T-cell leukemi/lymphom (ATL) nd clinicl fetures. The T- nd B-cell Mlignncy Study Group. Int. J. Cncer, 7 (99).. E. S. Gilert, S. A. Fry, L. D. Wiggs, G. L. Voelz, D. L. Crgle nd G. R. Petersen, Methods for nlyzing comined dt from studies of workers exposed to low doses of rdition. Am. J. Epidemiol., 97 97 (99).. D. L. Preston, J. H. Luin nd D. A. Pierce, Epicure Risk Regression nd Dt Anlysis Softwre. HiroSoft Interntionl Corportion, Settle, WA, 99.