Document EdxZqrmNEdoOKDzp4Rz4Rp2YL

RADIATION RESEARCH 137, S68-S97 ( 1 9 4 ) Cancer Incidence in Atomic Bomb Survivors. t * L "Department of Statistics, Radiation Effecrs Research Foundation. Hiroshima, Japan: Deparrment of Clinical Stiidies. Radiation Effects bcr d Foundation, Hiroshima. Japan: 'Deparrment of .Mrdicine. A -bomb Institute of Yagasaki L'niversirv. .Vagasaki. Japan: 'Radiation Epi&m+, 3 Branch. .Varional CJncer Institute. National Institutes of Health. Bethesda. ,Mayland 20892: 'Department of Epidemiology. Radiation 8- S 7 Research Foundation. Hiroshima. Japan: 'Research Institiire for .Vuc!eur Medicine and Biology. Hiroshima University, Hiroshima Japan: D of Internal Medicine. Hiroshima A-bomb Hospital. Hiroshrnrn.Japan: 'Department 0%Intrrnal .Medic.inr. .Vagasuki City Hospital. .Vagasaki jaP... 4 'Departmentof Clinical Studies. Radiation Effects Research Foundation, .Vagasaki. Japan; 'BiostatisticsCenrer. George Washington Univc- ' Rockville. .Maryland 20852: unci 'Department of Epidemiologic Parholog!. Radiarion E f f r ~ Rt ~esrarch Foicndation. .Vagasaki,Japan _- Ife Preston, D. L., Kusumi, S., Tomonaga, ,M., Izumi, S., Ron, E., K u r a m o t o , A., K a m a d a , N., Dohy, H., Matsuo, T., Nonaka, H., Thompson, D. E., Soda, M. and Mabuchi, K. Cancer Incidence in Atomic Bomb Survivors. Part 111: Leukemia. Lymphoma and Multiple Myeloma. 1950-1987. Radiat. Res. W7, S68-S97(1994). INTRODUCTION By the late 1940s there were suggestions of an inam,( risk of leukemia among the survivors of the a t o m i c w r o of Hiroshima and Nagasaki. These early observatim kd !o the establishment of a registry of cases of leukemia and rclJI I ed disorders. including lymphomas. among atomic bomb \-: T h i s p a p e r presents a n analysis of d a t a on the incidence of leukemia, lymphoma and myeloma in the Life Span Study cohort vivors (I). In 1952 Folley et al. ( 2 )reported clear evident` ,,; of atomic bomb survivors during the period from late 1950 an excess risk of leukemia. making this disease oae of I-!. through the end of 1987 (93.6% survivors accounting for 2,778,000 first long-term health effects to be noted in this populat~n~ person-years). These analyses add 9 additional years of follow-up The Leukemia Registry data have been the basis of a serve. for leukemia and 12 for myeloma to that in the last comprehensive of reports (3-5) that have helped to clarify our understandlrlv reports o n these diseases. This is the first analysis of the lymphoma incidence data in the cohort. Using both the Leukemia Registry and the Hiroshima and Nagasaki tumor registries, a total of 290 leukemia, 229 lymphoma and 73 myeloma cases were identified. T h e primary analyses were restricted to first primary t u m o r s diagnosed among residents of the cities or surrounding of the risks of radiation-induced leukemia. Since their e s ~ h lishment in 1958. the Hiroshima and Nagasaki tumm IC.: istries have also collected information on hematopoietic and lymphatic malignancies in these cities. The most recent comprehensive reports on leukernia ri4\ areas with Dosimetry System 1986 dose estimates between 0 and 4 in &heatomic bomb survivor population appeared more than G y k e r m a (231 leukemias, 208 lymphomas a n d 62 myelomas). 10 years ago. These reports considered the nature of the dcw Analyses focused on time-dependent models for the excess absolute risk. Separate analyses were carried out for acute lymphocytic leukemia (ALL), acute myelogenous leukemia (AML), chronic myelocytic leukemia (CML) and adult T-cell leukemia (ATL). The:e were few cases of chronic lymphocytic leukemia in response ( 6 ) .general patterns of leukemia incidence in : n ~ Life Span Study (LSS') cohort from 1950 through 1975 ( - I the distribution of onset times for leukemia cases reponed 11' the Leukemia Registry between 1946 and 1375 (8). .I,,.' this population. There was strong evidence of radiation-induced risks for all subtypes except ATL, a n d there were significant sub- type differences with respect to the effects of age a t exposure and sex a n d in the temporal pattern of risk. T h e AML dose-response function was nonlinear, whereas there was no evidence against hearity for the other subtypes. When averaged over the follow-up period, the excess absolute risk (EAR) estimates (in cases per 10' PY Sv) for t h e leukemia subtypes were 0.6, 1.1 and 0.9 for ALL, 'Abbreviations used: AHS. Adult Health Study: ALL. acute lymp* cytic leukemia: A,MFIT. Additive Multiplicative Fitting Program ' $ I f analysis of data Cor cohort survival from Epicure User's Guide (sec :<! 26); XML. acute myelogenous leukemia: ATB. at time of bombin<* ATL. adult T-cell leukemia: DATAB. computer program from &pat .a'User's Guide (see ref. 26): DS86. Dosimetry System 86: CLL,chri*r:- AML a n d CML, respectively. The corresponding estimated aver- lymphocytic leukemia; CML. chronic myelocytic leukemia; EAR.e.rc.-* age excess relative risks a t 1 Sv are 9.1, 3.3 and 6.2, respectively. absolute risk: ERR. excess relative risk: HTLV. virus knowo to c a w T h e r e was some evidence of an increased risk of lymphoma in adult T-cell leukemia; FAB, French-Americ~n-British classifica~l*~~ us.males (EAR = 0.6 cases per 10' PY Sv) but no evidence of any ICD-0. International Classification of Diseases-Oncology: 1 e x c e s in females. There was no evidence of an excess risk for mul- Span Study: NHL. non-Hodgkin's leukemia: NIC. oot in city: REKI tiple myeloma in our standard analyses. Radiation Effects Research Foundation: T65D. tentative 1965 & J U ~ ~ try: T65DR. tentative 1965 dosimetry revised. 0033-7587/9455.00 a1994 by Radiation Research Society. All rightsof reproduction in any form reserved. S68 HEMATOPOIETIC TUMORS IN ATOMIC BOMB SURVIVORS S69 ' 'remia incidence among in uferosurvivors during the pen- 1945 through 1979 (9). Since the publication of these reports.leukemia incidence or mortality has been considered in several general LSS publications (10-14). In the mid-1980s over 60% of the cases in the Leukemia Registry were reclassified using modern diagnostic criteria and nomenclature. including the French-American-British (FAB) classification (15, 16). Recently. Tomonaga et al. (17) analyzed the leukemia subtype data using all (LSS and non- LSS) reclassified cases for the period from 1945 through 1980.They suggest that relative risks for acute lymphocytic leukemia (ALL) and chronic myelocytic leukemia (CML) are greater than those for acute myelogenous leukemia (AML) and that there is no evidence of an excess risk for adult T-cell leukemia (ATL) or chronic lymphocytic leukemia (CLL). Analyses of the A-bomb survivor population suggest that the risk of radiation-induced leukemia rises rapidly after exposure and then declines. In general. the initial peak in the excess (absolute o r relative) risk is greater. and the decrease in the risk with time is more rapid, for those exposed when younger: indeed. for survivors who were over age 40 years at exposure, the decrease in the excess risk with time appears to be small or even nonexistent (IO).Although leukemia is a rare disease. accounting for only 4% of the cancer deaths in the LSS. leukemias are estimated to have accounted for )re than 20% of the excess cancer deaths in this cohort *tween 1950and 1985. Over the years there has also been an ongoing interest in the risks of lymphoma and multiple myeloma. Early analyses :(18)of autopsy data suggest an increased prevalence of mul+ tiple myeloma. lymphosarcoma and Hodgkin's disease b o n g survivors exposed within 1400 m of the hypocenter. Since then, Nishiyama er af. (19) noted an increased preva- flence of lymphoma and multiple myeloma between 1945and 5 among survivors with Tentative 1965Dosimetry System 65D)dose estimates over 1Gy. lchimaru er al. (5)conclud- (e.d that between 1950and 1976there was a statistically sipif- icant increase in the incidence of multiple myeloma among Survivors with T65D dose estimates over 1 Gy that became ..apparent about 20 years after exposure. Both lymphoma and eloma are studied routinely as a part of the LSS data. From 1950through 1985(14) 110deaths were s malignant lymphoma and 36as multiple myeloma. no evidence of an excess risk of lymphoma, but ed excess relative risk (ERR) for multiple myelowas statistically significant. Indeed. the point estimate of is the largest for any of the cancers studied other is report presents comprehensive data on incidence of mias, lymphomas (including Hodgkin's and non- kin's) and multiple myelomas occurring among mem-of the LSS cohort during the period from 1October 1950, the date from which the LSS cohort was defined, through 31 December 1987. Unlike the companion paper on solid iumor which is based solely on data tumor registries. the pre- sent analyses also include cases ascertained by the Leukemia Reg-istry.. A-s was the case with the data for solid tumor inddence. analyses were limited to first primary cancers diag- nosed in the catchment qey,of#h&-Hkyhima and Nagasaki tumor registries. MATERIALS AND METHODS Stud.v Po.pulation These ana!yses are based upon a portion of the extended Life Span Study (LSS-ESS) cohort of A-bomb survivors in Hiroshima and Nagasaki. As currently defined. the cohort includes 120.321 people. of whom 93,741 were in either city at the time of the bombings (ATB). Additional details about the LSS cohort and its extensions are given elsewhere (11, 20). In this study. as in most recent analyses of the LSS data. the 26.580 LSS members who were not in either city (NIC) ATB and 45 survivors for whom information on vital status was unobtainable were excluded from all analyses. Thus the basic data set includes 93.696 survivors. Except where explicitly noted. data for 7103 survivors for whom Dosimetry System 1986 (DS86) doses were not available and 262 survivors with DS86 doses greater than 4 Gy were excluded. A n additional 38 people with cancer diagnoses prior to 1 October 1950 were also excluded. Thus the main analyses were based on the 86.293 LSS cohort members with DS86 kerma estimates less than 4 Gy. Figure 1 summarizes the composition of the cohort used for xhis study b! age at exposure. sex and vital status at the end of follow-up on -71 December 1987. As suggested by the data in this figure. the cohort included a relatively htph proportion of women (58.4%) due to the relatively small number of males between the ages of 15 and 50 years in Hiroshima and Nagasaki ATB. Roughly two-thirds of the survivors in the LSS were in Hiroshima at the time of exposure. The average age at exposure of Nagasaki survivors is slightly less than that of Hiroshima survivors. As noted above. a special Leukemia Registry was developed in the late 1940s and early 1950s (I).The Leukemia Registry includes data on cases of leukemia. lymphoma. multiple myeloma and related hematopoietic disorders diagnosed after 1945 among all survivors resident in Hiroshima and Nagasaki and surrounding areas. whether or not they are included in the LSS cohort. All cases of leukemia and multiple myeloma accepted by the Leukemia Registry are confirmed and classified by two or more experienced Leukemia Registry hematologists on the basis of relevant clinical records and histological materials. In the past. Leukemia Registry diagnoses were made using the nomenclature prevailing at the time of diagnosis and coded according to an ad hoc classification system. During the latter part of the 1980s. most Leukemia Registry leukemias were reclassified (15-17) using modern diagnostic criteria and disease classification systems. including the FAB system for acute leukemias (21).The reclassification also allowed identification of recently recog- nized disease entities. such as myelodysplastic syndrome and ATL. Sufficient material was available to reclassify about 60% of the Leukemia Registry leukemia cases. However. no additional efforts were made to reclassify Leukemia Registry cases for which diagnostic peripheral blood or bone marrow' slides were inadequate or could not be located. The reclassified diagnoses will be called Leukemia Registry-FAB diagnoses. Since 1957 in Nagasaki and 1958 in Hiroshima the tumor registries have been collecting cancer incidence data in the two cities. Details OD tumor registry procedures and data quality are contained in papers by Mabuchi et OL (22)and Thompson er al. (20). m S70 Tro'oa, In PRESTON ET AL. 1-nT'Om Women - C 0 -. $&- ,$m \ & t: a 1 ~ . n FIG. 1. LSS population by age at exposure. sex and vital status. I 9-59 bo A detailed review and a comparison of the data on LSS cases from the various registries were carried out to identify potential cases. Appendix 1 contains a description of this review. Cases were classified in terms of International Classification of Disease-Oncology (ICD-0) morphology codes (23).The leukemia cases used hers have ICD-0 morphology codes in the range 980 to 994 augmented w t h adult T-cell leukemias (ICD-0 morphology code 9702-3). Lymphoma cases have ICD-0 morphology codes associated with Hodgkin's and non-Hodgkin's lymphomas (959-963.969) or reticulosarcoma (961). Multiple myeloma cases have a three-digit ICD-0 morphology code of 973. A total of 339 leukemias. 304 lymphomas and 94 myelomas occurred among members of the LSS. Analyses were limited to cases among exposed members of the LSS diagnosed between October 1.1950 and the end of 1987. Following the procedures developed for the report on the incidence of solid tumors (20).the primary analyses were limited to initial primary cancers diagnosed among residents of the two cities. Table I presents summary information on the final diagnoses classified by subtype. ,Most analyses were based on the 181 cases with DS86 kerma estimates of 4 Gy or less. Dosimerry The latest version of the DS86 ( 2 4 , IS) was used for these analyses. This version of DS86 provides dose estimates for of the survivors in the LSS-E85 cohort. Although the basic tabulations (described M o w ) were made in terms of total bone marrow dose in grays. DS86 weighted bone marrow doses in sieverts were computed as the sum of the DS86 yray dose to the bone marrow and ten times the DS86 neutron dose to the bone marrow. This is equivalent to the assumption of a constant relative biological effectiveness (RBE) of ten for neutrons. Weighted doses were used for all analyses. Occasionally, risks for exposed and comparison groups are compared. The comparison group includes all survivors in the LSS-E85 cohort with total bone marrow doses below 0.01 Gy. Srarisrical dnal?;su Person-wur cornpurarion. A s in the companion report on sotid b, incidence (3)a.nalyses were based on a detailed tabulation of cay QU~,, and penon-years stratified by age at exposure (13 categories), D S 6 h. marrow dose (11 categories, including J dose-unknown category). alt.du time (10categories). sex and city. The cutpoints for dge at exposlae .C 10. 15.20.25.30.35.40.15.50.55 and 60 yean. The cutpoints for tkbm marrow dose (in Gy) are 0.01.0.10.0.20.0.50. 1.00. 1.50.2.00.3.00ad~m with a categoy for suMvors without DS86 estimates. Follow-upa m h d from 1 October 1950 until the earliest of (a) the date of diagnosisoftbc (b.r primary cancer (including in sinc and occult cancer). (b)the date of death M (c) 31 December 1987. The calendar time cutpoints were 1 January IY" 193.1958. 1961.1966.1971. 1976.1981 and 1986. The tables contain a row fa: each combination of the stratificalw- variables in which at least one person is a t risk. In addition to case cow:: and person-year totals. each row in these tables includes person-verr weighted averages of age at exposure. attained age. time since elpowrrc and bone marrow dose ( y rays. neutrons and total). T'hese tables YCIC made using DATAB ( 2 6 ) . There were a total of 2.778.000 PY among exposed members d Ibc LSS-E85 cohort. This total includes 2.551.OOO PY for the surviv0n .?lh DS86 bone marrow doses between 0 and 4 Gy. 216.000 PY aaymuhlnl by people for whom DS86 estimates Are not available. and 7000 PY fa survivors with DS86 kerma estimates ober 4 Gy. Migrarion adjusrrnenr. An adjustment for the effects of d p b0 the completeness of ascertainment similar to that used in the comprdos e*..report on solid tumor incidence (3w)as made using methods by Sposto and Preston ( 2 7 ) .Cases diagnosed outside the tumoT catchment areas were excluded and person-years were scaled sex-, age- and time-specific weights based upon an analysis OfWf'Jw (Em)rates in the Radiation Effects Research Foundation Adon Health Study (AHS)cohort for the period 1958 through 1987. HEMATOPOIETIC TUMORS IN ATOMIC BOMB SURVIVORS S71 TABLE I Leukemia, Lymphoma in Myeloma in the LSS. Summary of Final Diagnosis Eligible for analysis Exclusions MGy >4Gy unknown dose Total Nonresident Second primary c1950or >I987 NIC Total Leukemia 231 6 24 ALL 32 2 4 AML 103 2 12 CML 57 2 3 Orher 39 0 5 Lymphoma 191 0 19 Non-Hodgkin's 170 0 18 Hodgkin's 21 0 I Myeloma 59 1 5 Total 48 1 7 48 261 38 117 62 44 210 188 22 65 536 21 8 17 32 339 6 I 2 4 51 8 3 6 I8 152 4 2 9 2 79 3 2 0 8 57 12 7 10 65 304 8 6 10 55 -167 4 1 0 10 37 2 6 i 20 94 35 21 28 117 737 "Summarizes the final diagnoses for all cases accepted during the review. The main analyses were base on the 481 first-primary cases in the leftmost column of the table. Survivorswith Dosimetry System 1986 (DS86) kerma greater than 4 Gy or unknown were used in analyses of effect modificatiq. Cases diagnosed among persons who were not resident in the cities (NIC) at the time of diagnosis were excluded as were all second primary cases among members of the Life Span Study ( U S )and cases with diagnosis dates outside the follow-up period for the present study. follow-up for this study began in 1950. a slight modification to the person- year adjustment procedure was necessar).. This was done b) linear extrapolation of the catchment-area residency probabilities from a value of one on 1 October 1950 to the 1958 fitted values estimated from the A H S . After the residency adjustment. the numbers of leukemia. lymphoma and multiple myeloma cases were reduced by 21. 12 and 2. respectively (Table I). The number of person-years for survivors with DS86 kerma nates between 0 and 4 Gy was reduced 12% to 2.243.000. Table I1 ides a summary of the effects of the person-year adjustment by city. sex and age at exposure. There has been a substantial amount of mipration among the youngest survivors. especially in Nagasaki. the log rates among survivors with doses less than 0.1 Gy were described as logquadratic functions of attained age. We then examined sex. city and cohort effects. choosing parameters with statistically significant ( P< 0.05) effects for inclusion in the final model. After developing a model for the radiation effect. we re-examined the background rate model. We routinely considered linear. linear-quadratic and linear-spline models for the dose-response function. The linear-quadratic doseresponse function can be written as p(d) = yld + y2d2. whereas the linear-spline model is Statistical Methodr Modeling methods based on Poisson regression were used to study the variation in risk as a function of weighted dose (d), sex (s).city (c), attained age (a), birthcohort (6). age at exposure (g) and time since exposure (0. Two classes of models were considered (1)ERR models of the form A(cse.b)[l + p(d)~,(cs,g~e)], where A(.) is a model for the background rate. Le.. the incidence function among people with zero dose. p(d) is a dose-response function, and er(.) describes effect modification: and (2) excess absolute risk (EAR) models of the form A(csm,b) + p(d)e,(csg.f.n). Recently. radiation effects on cancer morbidity have usually been described in terms of time-constant ERR models. In the companion paper on solid tumor incidence in the LSS (20) we have taken that ..'hpproach. However, in these analyses EAR models are emphasized :*&. cause (1)the timedependent E R R models needed to model the risk ?.Of radiation-induced leukemias are just as complex as the EAR models, .+.>"$ad (2) presentation of the results in terms of excess cases and time;&pendent excess rates provides a clearer picture of the risks,especially ifd<c ifd2c. When y1is fixed at 0 this model becomes a linear-threshold model with a threshold at dose c. We Tied c at 0.5 Gy and did not attempt to find a "best"estimate for the threshold. Log-linear functions. i.e., E ( Z ) = e'', were used to describe effect modification. When looking at temporal effects on the excess risk. both t and log(r) were considered. When r is used as a covariate in e the temporal trend will be called a log-linear model since log(ee') = pr. whereas when lo&) is used the model will be called a power function since eeloz")= t'. Analyses of the LSS data commonly have excluded data on survivors for whom dose estimates are unavailable. However, the data for these individuals can be used in inference about effect modification under the assumptions that: (1) among persons with unknown dose the distribution of the true (but unknown) dose is independent of the covariates in 2. and (2) we can classify people with unknown dose as exposed or unexposed. In this report we occasionally have uscd models in which the effect-modification term is multiplied by a dose-responsefunction of the form known DS86 unknown DS86. - I many of the malignanciesconsidered here, we tried to find the most h i m o n i o u s model for each disease. We started with a model in which Analyses based on this extended model were used only to supplement the standard analyses. Results based on these models are reported when they modify or clarify the findingsof the standard analyses. Parameter estimates were computed using maximum-likelihood methods for groupedsurvival data (Poisson regression). Hypothesis tests were based on likelihood-ratio tests. However, when implicit restrictions on the parameter space precluded computation of the likelihood-ratio PRESTON E T A L TABLE II 0-19 20-39 a0 PY Adjusted PY Percentage change PY Adjusted PY Percentage change PY Adjusted PY Percentage change Males 367,414 305.692 168% 141359 134.902 4.6% 158,022 151.758 4.0% Females 413.955 355,671 14.1% 394,855 377.105 45"o u7.973 229.792 3.4% Males 227334 163.141 282% 53,172 48.675 8 50'9 59286 SjL? 6396 251574 190.414 24396 159,605 146.125 8.4Ya 89.731 84.418 5.996 1,014,918 '-195% ' 748991 7oa8a7 5.6% 545M2 521,197 4.4% statistic. score tests were used. Ninety-five percent confidence intervals (95YaCI) were computed from the profile likelihood function (-78).Parameter estimation and testing were carried out using AMFIT (26). The Results section contains short, nonmathematical descriptions of the final models. supplemented by plots of the risk functions for selected covariate values. Because it is not possible to show data points on such plots. approximate 9506 confidence bands are used to dustrate the uncertainty in these fits. These bands were obtained from pointwise protile likelihood bounds. Tables of the estimated number of background and excess cases are also provided for most analyses. These estimates are sums of cell-specificvalues computed from the final risk model for the outcome of interest. The final model for each type of malignancy considered is pre- -sented in Appendix 3. For an EAR model of the form A(.) p ( d ) ~ & . )t.he estimated numbers of background and excess cases in a cell with P,per- son-years at risk are P , h ( . )and f ,p ( d ) E , ( . ) , respectively. Model-based. time-averaged summary estimates of the ERR, sv, EAR 3nd the attributable risk (AR) among survivors with doses over 0.01 Gg (A&, GY) are derived from the values in these tables. The computation of this ERR estimate uses the mean weighted dose and. for nonlinear dose-response models. the mean of the squared weighted dose for survivors in the exposed (bone marrow,dose greater than 0.01 Gy) group. These values are 0.26 Sv and 0.29 Sv-. respectively. The total number of person-year sieverts (after adjustment for migration) in the cohort to date is 173.860. the value used in the computation of the summary EAR estimate. RESULTS Table 111 presents a summary of the case counts. crude rates. person-years and mean weighted dose values for the three major classes of malignancies considered for this report. The data are classified by age at exposure, period and dose category. Table IV provides a similar summary of the case counts and crude rates for the leukemia s u b t p s considered in these analyses. Many of the patterns described on the basis of the models considered below can be seen in these data. For example, the leukemia rates suggest a strong dose response, the magnitude of which depends on time and age at exposure. Leukemia:All Tvpes As shown in Table I, 261 leukemia cases met the basic criteria for use in these analyses. Among these cases. 231 had DSS6 kerma estimates between 0 and 4 Gy and 6 had estimated kenna over 4 Gy; for 24 cases DS86 estimates could not be computed. Expected rates in the absence of exposure (background rares). The background rates for all types of leukemia a m bined can be described by a model in whch the log of the cate is a quadratic function of the log of attained age. The age*cific rates for women are about half of those seen in men, but the rate of increase with age does not appear to depend on sex. There is a sigmficant secular trend with age-specificrata increasing by 2.2% for each year increase in the year ofb&. The data do not suggest a City difference in the background rates when all types are combined. Parameter e s t i a e s for the leukemia backgound rates and excess risks are presented tosether with those for other tumor types in Appendix 2. Figure 2 contains plots of the fitted background rate as a fundon of aze for men and women in several birth cohorts. E X C ~ SriSsks. There is a statistically significant dose response ( P < 0.001) that appears to be nonlinear in dose ( P = 0.008). As indicated in Fig. 3. the dose response is ancave upward. Although the linear-quadratic and spline models provided comparable fits for these data. the hypothesis of a 0.5-Sv threshold could be rejected (P< 0.001). The EAR was best described by a time-dependent model in which the temporal pattern depends on both age at exposure and sex. If the effectsof aze at exposure and sex were ignored. the EAR was seen to decrease significantly with time ( P < 0.001) at a rate of about 6.5% per year. After allowance is made for this temporal trend, there is no evidence for an independent effect of age at exposure ( P = 0.5); however. the addition of both an effectof age at exposure and an interaction between age at exposure and time, in which risks for those exposed earlier in life decreased more rapidly than did risks for older survivors. improved the fit significantly ( P = 0.02. 2 df). Similarly. sex alone did not appear to have an effect ( P = 0.08). whereas the addition of an interaction between sex and time did lead to a better fit ( P = 0.01.2 dn. With this model, risks for women decreased less rapidly with time than did those for men. The effects of sex and age at exposure were largely independent. The EAR did not depend on city ( P > 0.5). Inclusion of survivors with unknown dose did not change the results significantly. HEMATOPOIETIC TUMORS IN ATOMIC BOMB SURVIVORS 573 . Tw Lcukenua Lymphoma Myeloma - Pcmnycan 4 TABLE I11 LSS Leukemia, Lymphoma and Myeloma Data Case Counts and Crude Rates' - AF at erporwe Dose 1950-1952 (year1 (Gv) 40.01 0.01-0.99 &I9 -3-39 a40 0-19 XI-39 a10 0-19 33-39 &I9 XI-39 a40 1 (03) (0.8) 2 10.6) 0 (01 3 11.21 1 10.3) 0 (01 0 (0) 0 (0) 38.126 24.786 30.893 n.00 0.00 n.00 4 (1.3) 1 (0.4) 5 (1.8) 0 10) 0 (0) 0 IO) 0 (0) 0 (0) 0 (0) 31534 22.626 27309 0.16 0.17 0.17 1953-1957 1958-1969 1970-1987 Toul 21 3 (15.41 3 (21.8) 2 (15.6) 0 (0) 0 (0) 0 (0) 0 (01 0 (0) 0 (01 1.954 1379 1286 1.83 1.7 1.76 CO.01 4 (0.51 I (0.2) 4 (0.6) 0 (01 2 (0.41 6 (1.0) 0 I01 0 (0) 0 IO) 79380 54.903 61.659 0.00 0.00 0.00 0.014.99 6 (0 9) (0.41 (1.3) 3 (05) 1 (0.2) 5 10.9) 0 IO) I (0.2) 0 IO) 66225 50.036 54.966 0.16 0.1' 0.17 -21 co.01 56 (12.0) (031 37 (9.9) 2 (051 12 (7.7) 0 (1.1) 9 (01 (021 0 IO (01 (0.71 0 17 (01 10.9) 00 IO) (0) 0I (0) 0 I01 4.151 (0.1) 1 10.1 I 19o.W 3.035 13332 395 1 1 4 2 I.e 0.00 1.78 0.00 1.75 0.00 0.014.99 9 10.6) 8 (0.7) 14 (1.4) 1 (0.1) 9 IO.?) If. 11.9) 0 IO) 1 10.1) 5 10.51 lMJ.494 121.61? 1c.094 0.16 0.10.1: 51 tO.O1 5 11 (5.11 4 (OS1 22 (5.6) (IS) 18 (14.3) (2.9) 1 I1 (1.0) (0.5) 1 25 (1.4) (1.7) 21 (4.11 (3.4) 02 10) (0.1) 19 (1.4) 0 (0.6) 15 (0) (2.4) 9.835 228.453 7.178 146.959 4.882 62517 1.82 0.00 1.77 0.00 1.74 0.00 0.014.99 =I 20.01 0.014.99 11 (0.61 P (1.6) 8 (1.4) 12 (0.6) 11 (1.7) 10 (1.8) Z (0.2) 6 (0.4) 11 (2.01 191386 133553 56-31 0.16 0.10.17 32 (2.7) (4) 3 32 (4.1) (0.9) 4 36 116.1) (13) 0 20 (0) (0.4) 240 (2.7) (1.1) 1 14 14.0) 0 (1.6) 2 (0) (0.0) 0 10 10) (0.3) 0 16 (0) (0.61 11306 536.832 7393 359.991 2.W 269291 1.R: 0.00 1.75 0.u) 1.71 0.00 30 (0.7) 33 (1.01 34 (1.4) 16 (0.4) 33 11.0) M 11.2) (0.11 8 10.2) 16 (0.7) 450.839 327832 240.661 0.16 0.1' n.1- >1 It! (5.91 13 (6.81 1s (133) 1 (0.4) '1 (1.61 (2.7) 0 (0) 1 (0.5) 0 (01 27216 18.W 11247 1.82 1.76 1.74 'Cases are classified hr age ai exposure. time and DS86 bone mamou dose Rates p r 10OOO PY arc shown In parentheses Summan ~nfonnationon m~$ration-adlustedPi and average DSM wngbted tone marrou dose IS also shown As can be seen in Fig. 4, young men had hgh EAFb in the 2 n d from 5 to 10yean after exposure. but these risks have decreased rap- idl-y with time. The EARSfor older men are :not as high in the early years and decline more slowly than those for younger men. Women tended to have lower excess risks than men until roughly 20 years after exposure. The data for older women sui-g-est-that the risks have not decreased with time. TABLE IV Toul 4 . 0 1 0.01499 a 397 (0.1) (021 (2.6) 3 12 roo (0) 0 (0) 0 (0) 0 (0) 0 (0) 1 (01 10.1) 1z (0) -! (1.4) (0.1) (0.1) (1.4) (0.1) (0.0) (1.1) 0 1 00331 (0) (0) (0) IO) (02) (3.9) (021 (0.2) (01 (0.21 (01 (0) (0.1) (0.1) (0.9) I 7, - Males - - - - PRESTON E T A L 10 20 30 40 50 60 70 80 10 20 30 40 50 60 70 Attained age (years) -y$g FIG.2. Leuktma. all types-background rates. Fitted background rates for men and women UI selected buth cohorts. The curves arc based0 model given in Appendix I Table V presents a summary of the observed and fitted numbers of leukemia casesby sex. age at exposure. period and exposure statusderived from the final model developed for this report. If the data in Table V are used. a model-based estimate of the time-averaged ERRl sv adjusted for the nonlinearity in the dose response is 3.9. Model-based estimates of the EAR and &.o, are 2.7 cases per la' PY Sv and 5096, respective- ly. Despite the tendency for leukemia risks to decrease with time, these ERRl sv and Gv values are larger than the correspondingestimates for any of h e solid tumors. LI 01 234 Dose (Sv) FIG. 3. The solid curve is the fitted leukemia dose response based on a quadratic excess absolute risk (EAR) model without effect modification. The shaded area is a 95% confidence region for the fitted curve. The vertical lines are 95% confidence intervals for dose-category-specificrisks. Point estimates of the risk for each category are indicated by short horizontal lines. Acute Lymphocytic Leukemia auOf the 38 ALL cases that met the basic criteria for sion in the study. 32 had DS86 kerma estimates between and 4 Gv (Table I). Because of the small number of caw, and the ielatively high percentage of cases with unknown dose. the analyses of effect-modifying factors were supple mented by analyses that used suMvors with unknown D!% Background rates. Although the small number of ALL cases precluded precise estimation. patterns are described based on the limited data available. There was an increase in the ALL background rates that was roughly log-linear in attained age (Fig. 5 ) and was similar for males and female. (P > 0.5). There was no indication that the age-spec& rater changed with time (P = 0.4). Excess risks. The test for a dose response was hiwy si? nificant (P < 0.001). with a weak suggestion of upward cur vature in both the quadratic (P = 0.16) and linear-spline (i' = 0.13) models. There was no significant difference betwecri the fit of the linear-spline and threshold models (P> 0.5 Analyses of effectmodification described below are based oil the linear model; virtually identical results were obtained for the quadratic model. The EAR has decreased with time since exposure (P 0.001) and the time-averaged EAR decreased with increas. ing age at exposure ( P = 0.01). After allowing for the temporal trend, the evidence for an effect of age at exposure on the risk was increased (P = 0.002). When sex was considered as an effect modifier in the standard analyses. i.e. excluding survivors with unknown DS86 dose estimate 01 with estimates over 4 Gy,the EAR for women was eStbM[ed to be about 40% of that for men. but this difference Way Males \ HEMATOPOIETIC TUMORS IN ATOMIC BOMB SURVIVORS -Age ATB 10 years --- AgeATB25y-n - - - - -AgeATB40yean r*O Females t I10 l5 s75 10 20 30 40 10 20 30 40 Time sinc exposure (years) FIG.4. Leukemia. all types-excess absolute nsks. Fitted EAR estimates for survicors receiving 1-Svbone marrow equivalent doses Curves are for male and female survivors at selected ages at exposure. The curves are based on the model presented in Appendlx 1 >I 'noi statistically signficant (P = 0.09: 95% CI = 13-113%). and women exposed to 1 Sv at the ages of 10, 25 and 40 Hawever, when the effect-modification analyses were years. Confidence bands are shown only for males. 'repeated with the addition of survivors with unknown doses. Over the current follow-up the standard summary esti- 4 the sex difference was found to be statistically significant (P mate of the ERR, s\. for ALL is 10.3 (95% CI = 4.3-25). = 0.03.95% CI = 12-91Yo). The EAR did not appear to dif- EAR and AR, o1 Gv estimates derived from this value are [fer between the two cities (P> 0.5). 0.57 cases per 10' PY Sv (95% CI = 0.35-0.78) and 72% Our final model for the ALL risk is a linear dose- (95% CI = 51.5436.5%), respectively. The observed and fit- se model in which the EAR has decreased rapidly ted values in Table VI provide a more detailed picture of ime. about 14% per year. Chiidren exposed under age these data. The model-based summary risk estimates derived 1 ad the highest excess risks. The EAR was estimated to from this table are ERR, Sr = 9.1, EAR = 0.62 per lo' PY 1 ease by about 5% for each year's increase in age at expo- Sv and A&,ol Gy= 70%. 1 > .+or any age at exposure or time since exposure, the estimates for women were estimated to be less than Acute Myelogerwus Leukemia .i those for men. Figure 6 shows plots of the EAR for men There were 103eligible cases among survivors with DS86 estimatesless than 4Gy. In addition there were twocases whose DS86 dose estimateswere greater than 4 Gy and 12 cases for I, i1 / whom DS86 estimatescould not be computed (Table I). Background rates. AML background rates in this cohort could be described by a model in which the log rate varied as a sex-dependent quadratic function of log attained age with n an effect of birth cohort. The sexdependent difference could )I not be described fully in t e r n of a simple ratio since the pat- tern of the increase in risk with attained age also differed by . sex (P= 0.02, after albwing for a difference in the level of .- risk for men and women). Age-specific rates for women 1- under the age of 40were slightly higher than the correspond- n ing rates for men, whereas thispattern was reversed at later 1- K)L,, ,., ,I . ,8 I. I.I ages (Fig.7). Age-specific rates for both sexes increased by 20 3 0 4 0 50 60 m 0 about 2.5% for each year increase in the year of birth (P= ,I AttehSdagsW) 0.01). There was no signtfimnt city difference (P= 0.13). .I- Is Acute lymphocytic leukemia-background rates. The fitted d rate is baredon the model given in AppeDdix 1. This model Excess risk There was strong evidence of a dose response (P < 0.001) and a suggestion of nonlinearity,as indicated by depend on sex. city or birth cohort. testsbased on either quadratic (P=0.05) or spline models (P TABLE V Leukemia, All Types, Distribution of Observed and Fitted Cases by Sex, Time, Exposure Status and Age at Ex Age at exposure (ycan) 0-19 20-39 =a All ages Dose category (Gy) caries Fitted background Fitted excess c- _. Fitted background Fitted excess cases Fitted background Fitted excess cases Fitted background Fitted excess 1950-1965 <MI MdeS 3 291 0.06 0.014 19 262 16.16 3. 8 2.68 249 0.02 5.98 8 17 839 7.99 0.03 10.71 14 13.99 0.11 44 13.10 32.86 1966-1987 e0.01 0.014 98 10.85 9.91 0.00 1.08 13 852 0.00 12 7.98 0.74 11 7 8.73 8.44 0.01 1.69 33 28.10 0.01 27 2632 350 1950-1987 <a01 0.01-4 12 13.76 0.07 n 1253 1724 -'3%y" - -r 1 . 16 1121 0.02 20 10.47 6.n 19 17.12 0.04 24 16.43 12.40 47 4209 0.12 71 39.43 36.36 0-19 20-39 =40 All ages C3ses Fitted background Fitted excess C3ses Fitted background Fitted exces cases Fitted background Fitted excess CaSeS Fitted background Fitted excess Females 4 I1 1.73 t32 0.04 9.56 36 3.24 3.19 0.03 7.68 5 13 5.90 538 0.03 6.42 12 10.87 0.10 30 10.09 23.66 6 6.74 0.01 13 1234 0.02 12 855 0.03 31 27.62 0.07 8 5.90 3.20 20 11.96 5.03 12 7.85 6.64 40 25.71 14.88 10 8.46 0.05 16 15.58 0.05 17 14.45 0.06 43 \ 38.49 0.16 .r 19 7.43 12.76 26 15.15 12.72 25 133 13.06 70 35.81 38.54 All ages cases Fitted background Fitted excess Both sexes 26 74 24.85 23.20 0.21 56.52 64 55.73 0.08 67 52.04 1838 90 80.58 0.29 141 75.24 74.90 "The fitted value for the background cases is the number of cases predicted by the final model assuming no radiation exposure. The fitted value for the excess cases is computed as the difference between the estimated number of background cases and the total number of cases predicted by the model. The specific model used is presented in Appendix 1. The only constraint in these fitted values is that the total number of fitted (background plus excess) cases equal the total number of observed cases. = 0.04). After allowance was made for the effectsof age at exposure, nonlinearity of the dose response became more apparent (P= 0.02). The 0.5-Gy threshold model did not fit the data as well as the spline model (P= 0.05). The AML dose-response function is contrasted with those or other leukemia subtypesin a later section. Age at exposure had a significant effect on the EAR (P = OM), with the youngest survivors having the highest average absolute excess risks. Evaluation of other potential effectmodifying factors individually revealed no sigrdicant effectsof time since exposure (P= 0.4), sex (P= 0.4) or city (P= 0.2). However, after allowing for the effect of age at exposure, the HEMATOPOIETICTUMORS IN ATOMIC BOMB SURVIVORS s77 5 10 15 20 25 30 35 40 45 Years since exposure " I 5 II I II I I I 10 15 20 25 30 35 40 45 Years since exposure 10 20 30 40 10 20 30 Years since exposure 40 10 20 30 40 FIG. 6. Acute lymphocytic leukemia-xcess absolute risks. Fitted EAR estimates for survivors receiving 1 -Sv bone marrow equivalent doses. Curves are shown for male and female survivorsat selected ages at exposure. The curves are based on the model presented in Appendix 1. The shaded areas in the lower panels indicate 95% confidence regions for the curves for males. a addition of both log time and an interaction between age at exposure and log time leads to a marginal improvement in the fit of the model (P= 0.15,2 df). Inclusion of the data for sur- vivors with unknowndoses (14 cases and 186300PY) strengthened the evidence for temporal trends dependent on age at exposure in the EAR (P= 0.005.2 df). To reduce the likelihood of oversmoothing due to the use of a product of two continuous covariates to describe the interaction between age at exposure and time, the final model allows for separate trends for each of three age-at-exposure groups (0-19, 20-39 and ->a)As.can be seen from Fig. 8, the fitted EARSdecreased aithtime for survivorswho were under 20 years ATE3 but were amstant or increasingwith time for survivorsolder ATB. VI1 summarizes the observed and fitted values for AML model Based on these data the model-based averaged ERRl sv is estimated to be 3.3. The comeding EAR estimate is 1.1cases per lo" PY Sv, and the p o 0 *oy is46%. ~nicMyelocyriCLeukemia f Of the 62 cases that were eligible for inclusion in these two had kerma estimates greater than 4 Gy (Table I). There were three additional first primary cases for whom a DS86 dose estimate could not be computed. Background rates. The increase in the background CML rates for this cohort is roughly proportional to attained age squared. Background rates differed significantly by sex (P= 0.008) with age-specific rates for women about 60% of those for men. The differencebetween the cities was even more p r e nounced (P= 0.003) with the rates in Hiroshima for either sex being more than three times those in Nagasaki (Fig. 9). There was no indication of a secular trend in the CML background rates, nor did the data suggest that the slope of the increase in the rateswith attained age depended on either sex or city. Excess risks. The CML data are consistent with a linear nonthreshold dose-response model. In particular,the excess risk exhibited a significant linear increase with dose (Pe 0.001) that was not improved by the addition of either quad- ratic (P> 0 5 )or spline (P> 05) terms in the weighted dose. The hypothesis of a 0.5-Gy threshold could be rejected for these data (P= 0.003). The CML EARShave decreased rapidly with the passage of time (PC 0.001). A log-linear (es') model described this - .- ... _~___ ~ PRESTON ETAL. TABLE VI Acute Lymphocytic Leukemia, Distribution ofObserved and Fitted Cases by Sex, Time, Exposure Status and Age at Exposure' Age at exposure (yean) 0-19 %39 -- ,a All ages Dose category (Gy) cases Fitted background Fitted excess cases Fitted background Fitted excess CaSeS Fitted background Fitted excess CaSeS Fitted background Fitted excess 1950-1965 <Offl 0.014 Males 09 0.26 023 0.06 7.84 00 027 0.25 0.01 1.22 12 0.68 0.64 0.00 0.73 12 1M 1.13 0.07 9.78 1964-1987 <om 0.014 01 022 0.6 0.01 0.90 21 0.46 0.43 0.00 0.14 10 0.41 0.40 0.00 0.05 31 159 1.49 0.01 1.09 0 0.98 0.07 2 0.73 0.01 2 1.09 0.00 4 2.79 0.08 - v. 1 0.68 135 2 1.04 0.78 3 261 , 10.87 0-19 20-39 All ages cases Fitted background Fitted excess cases Fitted background Fitted excess CaSeS Fitted background Fitted excess CW Fitted background Fitted excess Females 053 1 032 O X 0.89 0.78 0.03 3.73 0.00 0.44 01 1 1 0.66 0.65 133 1.29 0.01 137 0.00 0.16 1 200 0.94 0.86 0.79 0.73 0.W 035 0.00 0.03 1842 1.92 1.79 3.01 2.80 0.M 5.44 0.00 0.63 3 69 121 1.06 0.03 4.17 12 1.99 1.94 0.01 153 12 1.73 159 0.00 038 5 10 4.93 459 0.04 6.07 All ages cases Fitted background Fitted excess Both sexes 2 10 7 3 3.12 2.92 4.60 423 0.11 15.23 0.01 1.72 9~ 13 7.73 7.20 0.12 16.94 "The fitted value for the background cases is the number of cases predicted by the final model assuming no radiation exposure. The fitted value for the excess cases is computed as the difference between the estimated number of background cases and the total number of cases predicted by the model. The specific model used is presented in Appendix 1. The only constraint in these fitted values is that the total number of fitted (background plus excess) cases equal the total number of observed cases. temporal trend somewhat better than did a power-function model (ts). Although the temporal trend was not found to depend on age at exposure (P> OS), there were signrficant sex differences. Averaged over the follow-up period, men had greater excess risks (P= 0.03), and, after allowing for this difference, the temporal pattern of the excess risk was found to differ by sex (P = 0.01). As shown in Fig. 10, the fitted EAR ormen had a high initial peak and has subsequently decreased rapidly with time, whereas the fitted EAR for women has stayed roughly constant over the years. These data also show a si@cant difference between the EARSfor Hiroshima and Nagasaki (P= 0.005). The magnitude of the city effect was roughly proportional to that seen in the background risk, which suggests that differences in the number of excess cases for the twocities can be explained by differences in the Hiroshimaand Nagasaki background rates. Tables Wr and IXpresent expected numbers of casesbased upon the final fitted model stratified by sex and city, respective- ly. The city difference is particularly strikingsince only 6 of the 57 CML cases were diagnosed in Nagasaki. On the basis of the HEMATOPOIETIC TUMORS IN ATOMIC BOMB SURVIVORS s79 f 2.0 It 1.5 P L1.0 1 " ;0 0.5 0.0 - 1 I,,.,.,., 10 20 30 40 50 Attained age (years) .. 60 ,, 70 FIG.7. Acute myelogenous leukemia-background rates. The fitted background rate for men and women in selected birth cohorts based on the model given in Appendix 1. values in these tables we computed the time-averaged. model- based ERR, s" as 6.2. The estimated EAR is 0.9 cases per lo' PY Sv, and the Gv is estimated to be 62%. Adult T-cell Leukemia-Nagasaki As indicated in Table X, 25 of the 43 cases not classified in one of the three subtypes considered thus far were diagnosed as ATL, and all except one of these cases occurred among Nagasaki survivors. In view of the differencesbetween the two cities and because it is felt that radiation effects on ATL risk might differ from those on the risk of other types of leukemia, we analyzed the Nagasaki ATL data separately from the Hiroshima data for leukemias not classified as ALL, AML or CML.The ATL analyses were based on the 22 Nagasaki cases with DS86 dose estimates between 0 and 4 Gy. It is well known that, as indicated by these data, ATL is endemic to Nagasaki and extremely rare in Hiroshima (29). However, it does seem surprising that ATL is virtually the only leukemia subtype other than ALL. AML or CML diag- nosed among Nagasaki survivors. In fact only 2 of the 71 Nagasaki cases were other non-ATL leukemias. Data from the "open-city population" suggest that a slightly higher per- centage, about 6%. of the Nag& cases with FAB diagnoses could not be classified as CML. A M L , ALL or ATL (17). Background rares. The background rates for ATL in Nagasaki rose rapidly with increasing attained age, but these rates did not appear to depend on either sex (P > 0.5) or brth cohort (P= 0.15) (Fig. 11). Excess risks. There was no indication of a dose response for the Nagasaki ATL data. The point estimate of the EAR was negative with an estimated upper 95% confidence limit of 0.41 cases per la' PY Sv. There was no statistically sipnificant variation in ATL excess risk by sex or time since exposure. -Age ATE 0-19 - - AgeATB20-39par~ .----Age ATB a40yean 0 L, 0 I I I I .1 10 20 30 40 Age A'IB 0-19 yenm I1: ! L AQOATB24OyWS ~ G. 8. Acute myelogenous leukemia-cxcess absolute risks. Fitted E A R estimates for survivorsreceiving 1-Sv bone mamow equivalent doses. res are shown for survivOn at selected ages at exposure. The Bnal model, shown in Appendix 1. did not depend on sex or city. The shaded areas in ower panels indicate 95%confidence regions for the curves for males. S80 PRESTON ET AL. TABLE M .. -- Acute Myelogenous Leukemia, Distribution of Observed and Fitted by Sex, Time, Exposure Status and Age at Exposure" Cases,-.-2.--q ~ 1m1965 1966-1987 Age at exposure (year) Dose category (Gy) 0-19 20-39 cases Fitted background Fitted excess CaSeS Fitted background Fitted excess <0.01 0.01-4 Male 0 0.16 0.01 4 0.15 3.14 02 0.96 0.88 0.0 0.94 co.01 5 426 0.00 6 s.27 0.00 0.01011 5 3% 129 8 1.92 0.95 ,.'&$-. 5I 4.42 0.01 i 6 10 623 sa 0.00 240 cases 17 6 6 10 Fitted backgound 426 4.06 5.34 5.16 9.60 Fitted excess 0.01 357 0.00 2.43 0.01 All ages Cases Fitted backgound Fitted excess 4 13 538 5.09 0.01 7.65 17 11.88 0.01 19 14.M 1.73 21 20.26 0.02 0-19 20-39 sa All ages cases Fitted background Fitted cxcess cases Fitted background Fitted excess cases Fitted background Fitted excess cases Fitted background Fitted excess Females 3 '5 1.63 1.4s 0.01 3.71; 23 155 1.52 0.0 2.26 33 2.03 1.86 0.01 3.69 a 11 521 4.83 0.02 9.67 2 3.E 0.00 6 533 0.01 6 3.02 0.01 14 12.27 0.02 1 3.45 1.55 6 5.15 252 7 2.78 3.75 17 1138 7.81 5 555 0.01 8 6% 0.01 9 5.05 0.02 22 17.48 0.04 526 9 667 1.723 10 4.64 7.45 28 1621 17.48 All ages Cases Fitted background Fitted excess Both sexes 12 24 10.60 9.93 0.04 17.32 31 27.15 0.03 36 25.40 1254 43 37.74 0.06 60 35.33 29.80 "The fitted value for the background cases is the number of cases predicted by the final model assuming no radiation exposure. The fitted valuc ! the excess cases is computed as the difference between the estimated number of background cases and the total number of cases predicted by I ) model. The speafic model used is presented in Appendix 1. The only constraint in these fitted values is that the total number of fitted (backgour : . plus excess) cases equal the total number of observed cases. Other Leukemias-Hiroshima Of the 19 leukemia cases not classified as ALL, AML, CML or ATL. 17 were diagnosed in Hiroshima. Because of the small number of Nagasaki cases. the following analyses were limited to the 15 Hiroshima cases with DS86 dose esti- mates. These cases included 7acute leukemias of unspecified type, 4 chronic lymphocytic leukemias, 2 myelodysplastic syndromes and 2 hairy-cell leukemias. Background rates. Background rates or "other" leukemias in Hiroshima showed no significant dependence on attained age (P > 0.5) or birth cohort (P = 0.14). However, rates for males were estimated to be almost five times tbav for females (P = 0.03) (Fig. 11). xcess rids. There was a sipruficantlinear dose respomc m the data or"other" leukemias in Hiroshima (P = 0.004),bn there was no evidence of nonlinearity (P > 0.5). The data not suggest that the risk varied with either time Since-e (P>05) or age at exposure (P = 0.19).However, the nspaaC was signrficantly differentfor men and women (P = 0.a) particular, the estimated time-constant E A R for WomcO 0.44(9% CI = 0.14-0.%)excess cases per 104PYSv. the point estimate for men was negative with an e s h d upper 95% confidence limit of 0.20 casesper la'PY SV. '~ FK "un. HEMATOPOIETIC TUMORS IN ATOMIC BOMB SURVIVORS S81 I i 0.4 bF ----- -----Males Hiroshima NapaJaki Females / .-- /--- _____--_------ 10 20 30 40 50 60 70 80 Attained age (years) FIG.9. Chronic myelogenous leukemia-background rates. The cityspecific fitted background rate for men and women based on the model +givenin Appendix 1. There is no effect of birth cohort in this model. Table XI presents observed and fitted values for the data for "other" leukemias for Hiroshima by sex. Using these data, the model-based sex- and time-averaged summ&y estimate of the ERRl sv is 3.6, whereas the EAR and &.ol Gy are estimated to be 0.21 cases per lo4 PY Sv and 51%. respectively. Leukemia Joint Analysis The results of the type-specific leukemia analyses described above suggest that there are differences in the nature of the dose response for the leukemia types.These differences can be seen in Fig. 12. which compares the dose-response curves for the four diagnostic categories considered above, 5 and 35 years after exposure for three ages at exposure. To examine this issue. we carried out a joint analysis of the data for AML, ALL and CML using the methods described by Pierce and Preston (30).The joint analysis SUggests that there are significant intertype differences in both - 15 rz h r 10 b 8 I: 5 10 15 20 25 30 35 Years since exposure si01520253035 5 10 1 5 2 0 2 5 3 0 3 5 Years slnce exposure 10. ChroNc myelogenous leukemia-cxass absolute risks. Fitted EAR estimates for survivors receiving 1-Svbone marrow equivalent doses. shown for male and female swivors in each City. In the final model, shown in Appendix 1, the EAR does not depend on age at exposure. er panels display 95% likelihood-basedconfidence regions for the curves for Hiroshima. s82 PRESTON E T A L TABLE VIlI 0-19 cases Males Fitted background Fitted excess t 20-39 cases Fitted background Fitted excess 2 40 CaseS 0.01 3.m 0.00 0.11 0.01 .0 6 2 1 2 7 -*.-- Fitted background Fitted excess 2.15 237 135 1.74 3.71 4.11 0.03 536 0.00 0.11 0.03 5.46 Q -eAll ages CaseS a- 2 18 7 4 9 22 Fitted background Fitted excess 3.1 1 338 451 4.88 7.62 826 0.08 1459 0.00 0.45 0.08 15.04 * 0-19 20-39 CaSeS Fitted backmound Fitted exce& cases Fitted background Fitted excess Females 1 00 3 13 0.2 0.22 1.01 1.00 123 121 i- 0.01 2 3 0.01 1.79 0.03 4.02 0 -7 3 4 3 6 0.79 0.89 2.16 239 2.% 3.3 0.01 238 0.01 1.82 0.02 4.20 240 All ages cases Fitted background Fitted excess cases Fitted background Fitted excess 0 613 4 9 159 1.67 138 1.67 3.17 3 3 0.01 1.74 0.01 0:78 0.02 252 1 8 7 10 8 18 2.60 Ln 4.76 5.06 736 7.83 0.04 635 0.m 439 0.07 10.74 All ages CaSeS Fitted background Fitted excess Both sexes 3 26 5.71 6.15 0.12 20.94 14 9n 0.03 14 9.94 4.54 17 14.98 0.15 40 16.09 25.78 - T h e fitted value for the background cases is the number of cases predicted by the final model assuming no radiation exposure. The fitted value f t u the excess cases is computed as the difference between the estimated number of background cases and the total number of cases predicted bv rhc model. The specific model used is presented in Appendix I.ll,: onty constraint in these fitted values is that the total number of fitted (backgroum plus excess) cases equal the total number of observed cases. the temporal pattern (P = 0.02.2 df) and the nature of the variation of the risk with age at exposure (P= 0.002). After allowing for these differences a test for nonlinearity in the dose-response function. with a common amount of curvature for ail three types assumed, was statistically significant (P = 0.008),whereas the test for type-specific differences in the shape of the dose-response function was not sigmficant (P = 0.35.2 df). Although the latter test is not powerful. this result suggests that the differences in the shape of the dose- response function described in the eartier sections should be interpreted with caution. There was strong evidence against 0.5-Gy threshold ( P < 0.001). A joint analysis of the data for ATL and "other" leukemias for the two cities suggested significant city d8erenm in both the background rates and the EARS (P = 0.05. 5 df). The details of this analysis will not be presented here. Lymphoma A total of 210 lymphoma cases, including 22 Hodgkh'. and 188non-Hodgkin's cases, met the basic criteria for indu- HEMATOPOIETIC TUMORS IN ATOMIC BOMB SURVIVORS S83 TABLE IX Chronic Myelocytic Leukemia, Distribution of Observed and Fitted Cases by City, Time, Exposure Status and Age at Exposure" Age at exposure (year) Dose category (Gy) 0-19 20-39 240 All ages cases Fitted background Fitted excess cases Fitted background Fitted excess cases Fitted background Fitted excess CaSeS Fitted background Fitted excess 0-19 20-39 m AII ages cases Fitted background Fitted excess cases Fitted background Fitted excess cases Fitted background Fitted excess cases Fitted background Fitted excess AI1 ages cases Fitted background Fitted excess 1950-1965 4.01 0.01-4 Hiroshima 14 0.42 0.49 0.05 7.47 18 1.21 1.44 0.02 4.92 0 11 3.13 3.76 0.03 6.69 2 23 4.76 5.70 0.10 19.09 Nasasaki 12 0.11 0.06 0.01 0.97 00 0.22 0.12 0.00 0.48 01 0.61 0.28 0.01 0.41 13 0.94 0.46 . 0.02 1.85 Both cities 3 26 5.71 6.15 0.12 m.94 1966-1987 - 4.01 0.014 2 2.01 0.01 5 3.05 0.01 6 2.66 0.00 13 7.73 0.03 3 232 1.80 6 3.64 1.79 4 3.20 0.84 13 9.16 4.43 01 0 3 0.27 0.00 0.22 10 056 029 0.00 0.14 00 0.48 021 0.00 0.04 11 154 0.78 0.00 0.41 14 14 927 9.94 0.03 4.84 1950-1987 4.01 0.014 3 2.44 0.06 6 4.26 0.03 6 5.80 0.04 15 12.50 0.13 7 2.81 927 14 5.09 6.71 15 6.% 7.54 36 14.86 2352 13 0.61 0-34 0.01 1.19 10 0.79 0.40 0.01 0.62 01 1.09 0.49 0.01 0.45 24 2.48 1.23 0.M 226 17 14.98 0.15 40 16.09 25.78 h e fitted value for the background cases is the number of cases predicted by the h a 1 model assuming no radiation exposure. The fitted value for the ex& cases is computed as the difference between the estimated number of background cases and the total number of cases predicted by the model. The specific model used is presented in Appendix 1. The only constraint in these fitted values is that the total number of fitted (background plus excess) cases equal the total number of observed cases. sion in the analyses. The primary analyses were based on the 170cases of non-Hodgkin's lymphoma (NHL) among survivors with DS86 kerma below 4 Gy. There were an additional 21 cases of Hodgkin's lymphoma in this group. Because of the small number of cases, detailed analyses of the Hodgkin'scases were not carried out. Background rtes. The increase in the NHL background rates was roughly proportional to the fourth power of attained age. Rates for women were about 60% of those for m `P= 0.002).The data suggested (P= 0.01) that age`SaC,.ic background NHL rates have been increasing by about 3% per year over the course of this study (Fig. 13). Neither the rate of increase with attained age (P= 0.4)nor the secular trend (P> 05) appeared to depend on sex. Background rates for the twocities did not differ(P= 0.4). Excess risks. The test for a linear dose response in the EAR model was not significant (P= 0.09),nor was the fit improved by the addition of a quadratic term in the dose (P > 0.21).An unconstrained sexeffect model could not be fit- ted because of the negative value of the excess risk for females. However, a model in which the EAR for females was constrained to be equal to zero fit significantly better PRESTON E T A L TABLE X Distribution of ATL and Other Leukemia Types by City and Sex' Type ATL Non-ATL Total Hiroshima M e Female 01 7 8 (1) 7 9 (1) Nagasaki Male Female 9 (1) 13 (1) 1(1) 0 (1) 9 (2) 14 (2) 'Numbers of additional cases with unknown DS86 dose or DS86 kerma estimates greater than 1 Gy are given in parentheses. 10.75 025 I than the simple linear dose-response model ( P = 0.04). In view of the lack of significance of the test for a dose response it is better to consider a comparison with the no dose- response model based on two degrees of freedom. By this more stringent criterion these data still suggest a significant excess risk for males ( P = 0.03, 2 df). The EAR did not appear to depend on time since exposure ( P > 0.5). attained age (P = 0.25) or age at exposure (P> 0.5). Analyses of the lymphoma ERR failed to demonstrate a significant overall dose response ( P > 0.5) or to provide evidence of a sex dif- ference in the risks ( P = 0.3). In the final model the time- constant EAR for men is estimated to be 0.56 cases per 10' PY Sv (95% CI = 0.08-1.39). whereas that for women is 0 with an upper 95% bound of 0.28 cases per 10' PY Sv. Based on the observed and fitted values summarized in Table XI1 the sex- and time-averaged model-based estimates of the ERR, sv, EAR and are 0.31.0.32 cases per 10' PY Sv and 7.6%. respectively. If only data for males are used. the model-based estimates of the ERRl sv, EAR and A&,o, Gv are 0.62.0.56 cases per 10' PY Sv and 13%, respectively. EAR analyses of the Hodgkin's cases failed to demon- strate a significant dose response ( P > 0.5). The sign of the dose coefficient was negative. When separate dose-response functions were allowed for men and women. the coefficient for men was positive whereas that for women was negative. However, there was no evidence of a significant dose response for either sex. An analysis of all lymphoma cases as a group led to essentially the samc results as the NHL analy- ses described above. Multiple Myeloma As indicated in Table I. 73 cases of multiple myeloma were diagnosed among A-bomb survivors in the LSS between 1 October 1950 and the end of 1987. A relatively high proportion of these cases (8.6%) were classified as second primaries, compared to leukemia (2.8'6) or lymphoma (3.196). Among the 65 cases that met the basic inclusion criteria 59 had DS86 kerma estimates below 4 Gy. Background rares. The increase in multiple myeloma background rates is proportional to the eighth power of attained age. This rapid increase is consistent with multiple 1.1 1 5 15 25 35 45 55 65 75 Attamed age (years) FIG. 11. Other leukemia types-background rates. Backgroua rates for other types of leukemia. excluding ATL. for Hiroshima malesad females and background ATL rates for both sexes in Nagasaki. n e shaded areas in the lower panels indicate %YO confidence regions for the curves for Hiroshima. myeloma being primanly a disease of old age. The LSS data also suggest that age-specific background rates have increased by about 5% per year over the course of this stud\ ( P = 0.04) (Fig. 14). Rates did not appear to differ by sex ( b = 0.4) or city ( P > 0.5). Excess risks. When analyses were limited to first-primary multiple myeloma cases with DS86 estimates below 4 Gy kerma there was no evidence of a significant dose response ( P = 0.12). nor was there any evidence of nonlinearity in the dose response ( P > 0.5). The point estimate of the time- averaged EAR for this model was 0.08 cases per la' PY Sv (9596 CI < 0-0.3). The EAR did not appear to vary with sex ( P = 0.4). time since exposure ( P = 0.4); age at exposure ( P = 0.4) or city ( P > 0.5). Because of the contrast between this tinding and findings of recent mortality analyses based on death certificates (13) and an earlier incidence study (9,both of which reported evidence of a radiation effect,differences between the earlier data sets and the present data were examined in detail. Four of the 36 cases in the data set in the LSS mortality report were excluded in the present analysis: three. with DS86 dose estimates of 0.15.0.21 and 1.5 Gy, had been rejected by the leukemia registry, and one high-dose case (5.6 Gy) was a sec- ond primary. Nine cases were excluded from the mortality analysis because of unknown DS86 estimates. Recent extensions of the DS86 system made it possible to compute dose estimates for 6 of the 9 unknown-dose cases excluded from the mortality analyses, however: one of these cases (0.2 Gy) was a second primary. The present series includes 22 first- and 3 second-primarycases with DS86 estimates and 3 cases with unknown dose not included in the mortality series. Among the 22 new cases used in these analyses. 4 were diag- HEMATOPOIETICTUMORS IN ATOMIC BOMB SURVIVORS S85 TABLE XI Other Leukemias-Hiroshima Only, Distribution of Observed and Fitted Cases by Sex, Time, Exposure Status and Age at Exposure' Age at exposure (year) Dose category (Gy) 195C!-1%5 4.01 0.014 1-1987 4.01 0.014 1950-1987 4.01 0.01 4 C19 20-39 MO All ages C-!19 20-39 240 All ages AU ages cases Fitted background Fitted excess cases Fitted background Fitted excess Cases Fitted background Fitted excess Cases Fitted background Fitted excess Cases Fitted background Fitted excess cases Fitted background Fitted excess Cases Fitted background Fitted excess cases Fitted background Fitted excess cases Fitted background Fitted excess Males 1 0.73 0.00 0 0.86 0.00 10 035 0.41 0.00 0.00 20 0.54 0.66 0.00 0.00 40 1.61 1.93 0.00 0.00 Females 01 0.18 021 0.01 0.91 10 0.19 0 3 0.01 1.a2 01 0.16 0.19 0.00 0.76 12 053 0.63 0.M 268 Both sexes 52 2.15 256 0.02 2.68 01 0.93 1.10 0.00 0.00 11 0.38 0.45 0.00 0.00 00 0.26 0.33 0.00 0.00 12 158 1.88 0.00 0.00 00 0.24 0.21 0.01 1.14 13 0.24 0.29 0.01 1.21 01 0.11 0.13 0.00 051 14 059 0.68 0.M 2.86 26 2.16 256 0.02 2.86 1 1.65 0.00 0.00 21 0.73 0.86 0.00 0.00 20 0.80 0.99 0.00 0.00 52 3.19 3.81 0.00 0.00 01 0.42 0.48 0.01 2.05 23 0.43 032 0.01 2.22 02 0.27 031 0.01 126 26 1.12 131 0.03 554 78 431 5.12 0.03 5 5 4 ' 'Because of the small number of cases in Nagasaki these analvses were restncted to Hiroshima. The fitted value for the background cases is the number of cases predicted by the final model assuming no radiation exposure. The fitted value for the excess cases IS computed as the difference between the estimated number of background cases and the total number of cases predicted by the model. The specific model used is presented in A I p n d i x 1. The only Constraint in these fitted values is that the total number of fitted (background plus excess) cases equal the total number of observed cases. &. I 34 low-dose cases and only one new highdose case diag- ukemia Registry (DS86dose estimates of 1.53 Gy, 0 kerma estimates below 1 Gy. and unknown), an additionalcase for which a DS86 dose To obtain a better understandingof the hpact of exclud- Id not be computed, and one case with a DS86 kerma ing the highdose and second-primarycases,the EAR model PRESTON ET A& Age at exposure - 1950 10 8 0 4 2 0.0 0.5 1.o 1.5 2.0 0 0.0 0.5 1.o 1.5 2.0 ;6 lo > F -ii 6 - ET a 4- , ,, , Age at expos1u0re-25 years 8- - 6- 4- - I 2- - &- -*--e.-- ---..=S.S c*- I 0.0 0.5 1.o 1.5 2.0 0 - ; F-P -, 0.0 0.5 1.o 1.5 2.0 Age at exposure 40 years 10 , 0.0 0.5 1.o 1.5 2.0 0.0 0.5 1.o 1.5 2.0 Dose (Sv) FIG.12. Type-specific leukemia dose-response functions. The panels in this figure compare the dose-response functions for the major leukemia subtypes in 1950 (column 1) and 1970 (column 2) for survivors aged 10 (row 1). 25 (row 2) and 40 (row 3 ) at the time of exposure. The Curves computed from the models described in Appendiv 1 by setting the time-since-exposureand age-at-exposure variables to the appropriatevalues. I HEMATOPOIETIC TUMORS IN ATOMIC BOMB SURVIVORS S87 I I , , I ,,8 0 I myeloma in the U S between 1 October 1950 and the end of 1987, adding 9 years of follow-up for leukemia and 12 years for multiple myeloma since the last LSS incidence reports of these diseases (5. 7). This is the first analysis of the U S inci- dence data for malignant lymphoma. The unified approach to identification and classification of cases using data from all relevant sources and the use of new statistical methods (IO) have made possible more detailed and quantitative analyses than were possible in the past. The principal finding of the leukemia analyses was that the risk associated with radiation exposure vaned sigmficant- ly among the various types of leukemia. The magnitude of risk exhibited statisticall! si-pificant and often complex vari- OL, I Iability with time since exposure. aye 3: .r':psure and. to a 40 50 60 70 80 somewhat lesser exterit, sex. Tlxse fiiv.:?~c.le-srly demon- Attained age (years) strate that leukemia risk estimates tiszci ior risk assessment FIG. 13. Non-Hodgkin`s lymphoma-background rates. The fitted background rate for men and women in selected birth cohorts based on the model given in Appendix 1. and the comparison of various populations should allow for the effects of age. time and sex on the excess risk. Summary risk estimators that do not allow for these important factors can be misleading. was refitted using several expanded data sets. Including sur- The present analyses emphasize the importance of allow- vivors with DS86 estimates over 4 Gy (one additional case ing for background incidence rates when assessing radiation- , and 6150 PY) had virtually no effect on the EAR or the sig- induced leukemia risks. In particular. it is essential to recog- / nilicance level of the test (P = 0.10)..4ddition of the five sec- nize the differences between Hiroshima and Nagasaki CML I ond-primary cases with known dose also had little effect on and ATL basehe rates when interpreting the data on excess I the fit or the test for a dose response (P = 0.09). When both risks for these subtypes. Earlier misunderstandings of the the second primaries and the case in the >4 Gy kerma group nature of the excess risks for the two cities arose because 'wereainndclbuadreedly, the point reached estimate statistical of the EAR significance increased to ( P = 0.05). intercity differences in background rates were not considered. Although the LSS data are of limited use for detailed S m e the U S mortality analysis was based on ERR models, analyses of background rates because of the nature of the we fitted ERR models to these data. Using the basic data set cohort. it is noteworthy that these data do not provide evi- (first primaries with kerma under 4 Gy). the ERR, point dence of an elevated background rate for ALL for the estimate of 0.25 was not significantly different from 0 (P> youngest survivors. This finding is consistent with other 05). However, addition of both the case in the >4 Gy kerma reports on the incidence of ALL in Japan (31,32).Recently. `goup and the second-primary cases increased the estimated Bessho (33)has suggested that this observation may be a ERR,sv to 0.9 and was statistically significant (P= 0.02). result of misclassification of ALL in Japanese children. How- On the basis of these analyses, we feel that the current LSS ever, in view of the generally high diagnostic standards of the bpdence data provide little evidence for an increased risk of Leukemia Registry, which has included reviews by Japanese myeloma. Thus our final model for myeloma incidence and U.S. hematologists and the recent Leukemia assumes no effect of radiation exposure. Table XI11 summa- Registry-FAB review of the Leukemia Registry materials, rizes the observed and fitted numbers of cases using this misclassification seems less likely for these data. model. Earlier results based on the mortality data appear to Two questions regarding the temporal pattern of be h e a d y dependent on the inclusion of questionable diag- leukemia risks after radiation exposure are of interest. The Doses and of both highdose cases and second primaries that first concerns estimates of the excess leukemia risk among have routinely been excluded in the current analyses of the the A-bomb survivors during the 5 years before the initiation incidence data A detailed review of all of the lymphoma and of follow-up of the LSS cohort. The second concerns the evi- -#yeloma cases is being planned. Perhaps this review, along dence for an excess risk of leukemia in recent years. Because `hth the extension of the follow-up perioct, w i a~llow a more of problems with case ascertainment and determination of &f%tive statement about myeloma risks based on the US. the population size it is not possible to compute risk estiI mates for the years before 1948, but Folley et d (2) evaluat- DISCUSSION ed leukemia incidence in Hiroshima and Nagasaki between 1948and 1950in relation to distance from the hypocenter. In This report provides a comprehensive analysis of the inci- their study, cases were obtained from the Leukemia Registry -e of leukemia, malignant lymphoma and multiple but not classified by type. Population estimates were based PRESrON ET AL. TABLE XII Non-Hodgkin's Lymphoma, Distribution of Observed and Fitted Cases by Sex, Time, Exposure Status and Age at Exposure" 1950-1965 19661987 1950-1987 -- 'tI i4 0-19 20-39 cases Fitted background Fitted excess cases Fitted background Fitted excess 35 OX7 0.82 0.01 138 42 w 2.16 0.00 0.67 6 10.05 0.01 9 9.19 0.00 8 9.23 1.71 10 8.60 0.71 9 10.93 0.02 13 11.54 0.01 I3 10.05 3.10 12 1an 138 * rm AU ages cases Fitted background Fitted excess Cases Fitted background Fitted excess 7 5.03 0.01 14 11.24 0.02 3 7.65 0.98 10 10.63 3.03 12 8.69 0.00 27 27.94 0.02 13 839 0.47 31 2623 2.89 19 16.71 0.01 41 39.18 0.04 16 16.04 1.45 41 36.86 5.93 0-19 cases Females 20 8 4 10 4 Fitted background 0.72 0.63 7.83 6.83 835 7.46 Fitted excess 0.00 0.00 0.00 0.00 0.00 0.00 20-39 240 AH ages CaSeS Fitted background Fitted eycess CaSeS Fitted background Fitted excess cases Fitted background Fitted excess 3 3 .%J 0.00 6 6.90 0.00 11 10.97 0.00 4 331 000 4 630 000 5 1024 0.00 19 16.34 0.00 15 10.39 0.00 42 3.56 0.00 18 15.84 0.00 5 9.55 0.00 27 32.23 0.00 22 190' 000 21 17 29 0.00 53 4534 0.00 32 19 15 0.00 9 15.85 0.00 35 42.46 0.00 All ages CaSeS Fitted background Fitted excess Bothsexes 15 18 2.21 20.87 0.02 3.03 69 6250 0.m 58 58.45 2.89 94 84.71 0.04 6' 7932 593 T h e fitted value for the background cases is the number of cases predicted by the final model assuming no radiation exposure. The fitted value for the excess cases is computed as the difference between the estimated number of background cases and the total number of cases predicted by the model. The specific model used is presented in Appendix 1. The only constraint in these fitted values is that the total number offitted (background plus excess) cases equal the total number of observed cases. on the 1950 census of atomic bomb survivors subsequently used in the definition of the LSS cohort. The results of Folley et aL provide clear evidence of an increased risk of leukemia among proximally exposed (less than 2OOO m from the hypocenter) survivors. To compare the risk estimates of Folley et al. with those seen in the first few years of follow-up of the LSS. we constructed a special data set. Cases and person-years for the period from 1October 1950through 31 December 1953were cross-classified by sex, city and distance with the categories used by Folley er al. Table XIV compares the populations and crude estimates of the risk derived from the data of Folley et aL with estimates based upon the LSS data. The results of Folley et al. appear to underestimate the incidence for distal survivors. especially in Nagasaki. There may also be some under-ascertainment of cases among the Hiroshima proximal suMvors. Despite the problems with case ascertainment and population definition. leukemia risks averaged over the 3 years before the start of the U S appear to have been elevat- ed. The report of Folley et af. suggests that the largest excess risks are seen in survivors under the age of 45 at the time of exposure and that CML accounts for a substantial fraction of ' the early excess risk. These findings are consistent with the results of our analyses of the first years of follow-up in the LSS.This comparison highlights an important limitation of the LSS leukemia data: HEMATOPOIETIC TUMORS IN ATOMIC BOMB SURVIVORS S89 "II*.O h' -Ban 1935 ..-..---Ban1920 Brm 1 8 M Indeed the point estimate of the Nagasaki EAR for this penod was negative, though not significantly so. Because of differences in the types of leukemia seen in the two cities, particularly the relatively large number of ATL cases in Nagasaki, and the differences in temporal patterns and levels of 8 a -c '.O .,... _ _ - - -O.{/: __.- ,_.....'. o,o . . . . . ( '?.?.-.-.:: "' L-2 40 50 i:;,:.:! y. /..' ,...(..., I 60 70 2 80 Attained age (years) FIG.14. Multiple myeloma-background rates. The fitted background rate far selected birth cohorts based on the model given in Appendix 1. excess risk for the leukemia types. the city difference should be interpreted with caution. Among the three major tvpes of leukemia associated with radiation. ALL and CML (at least for males) share similar responses characterized by a linear, nonthreshold dose response and an initial h g h risk (in 1950) followed by a gradual decline. However. as noted in the presentation of the results of the joint analysis. there is no strong evidence against a null hypothesis of equal curTmturc for the ALL. AML and CML dose responses. Our results also suggest that there are detectable excess risks for leukemia at doses below 0.5 Gy. Based upon a proportional incidence analysis of the Leukemia Registry data. Tomonaga er af.(17) suggested pre- viously that the CML dose response is well described by a data for the first 5 years of follow-up are not incuded. linear. nonthreshold function. As noted above. our results Assuming the 1948-1950 risks are of the size suggested by are not inconsistent with theirs. However. the risks at v e p the data of Folley et ai.. it is clear that a signrficant number of low doses predicted by the models considered here are much excess cases are not included in the U S .Crude calculations lower than those suggested in the crude incidence analysis of suggest that inclusion of the early years of follow-up would the earlier paper. increase the average leukemia risk estimates (all types) @veri The present analysis also clearly indicates that the EAR earlier in this report by about 10to 15%. If the risk models in for CML is lugher in H i r o s h a than in Nagasaki. but this dif- . 'F report were to be used for other populations. bounds on ference appears to be proportional to the differences in the ,risk could be obtained by allowing the EAR to increase background rates. Lnlike ALL and CML. the risk for AML linearly from zero at 2 years after exposure to the model esti- risk is better described by a nonlinear response with a tenden- mate at 5 years after exposure or by considering the risk to cy to decrease with time for younger survivors but remain sta- $I"abfeteerqeuxaplotsoutrhe.e 1950 value for the period from 2 to 5 years ble .or possibly increase for older survivors. The temporal pattern for AML among survivors exposed as adults resembles The issue of whether or not an increased risk of leukemia that which has generally been found for solid tumor incidence S is seen 30 or more years after exposure has received much (20). The various temporal patterns found for the three attention. To address this question, supplementary analyses leukemia types were generally consistent with the schematic $of the pooled data for the period from 1976 through 1987 summary of leukemia risks presented in previous reports (4, were carried out. A total of 72 cases occurred: 39 among sur- 7). Our analyses found a sigruficant difference in the temporal evivors exposed to less than 0.01 Gy and 33 among those with pattern of the CML excess risks for men and women. espe- higher doses. A significant linear dose response was apparent cially in the early years of follow-up.But. because of the small (P= 0.005). and addltion of a quadratic term in the weighted number of cases, it is possible that this is a chance finding. ose did not improve the fit ( P = 0.3). The model for the Since the RERF data cannot shed additional light on this led leukemia data predicts about 7 excess cases among question, data hom other populations should be examined to 33 cases that occurred among survivors with doses over determine if there is evidence for a similar interaction ere was a weak suggestion of heterogeneity in the between sex and time in the excess risks for CML. h age ATB (P = 0.07). The largest excess risks Further studies are necessary to investigate possible bio- to have occurred among survivors who were under logical and epidemiological bases for age, sex and time differ- or over 40 years old ATB with little, if any, excess risk ences in the type-specific patterns of leukemia EARS.A link arent for survivors who were 20 to 39 years old ATEL between CML and ALL is suggested by the Philadelphia ese findings indicate the importance of continued follow- chromosome, or variants of translocation involving the bcr during the last 12 years of follow-up did not and ab1 genes, present in almost all CML cases and some with time since exposure ( P > 0.5)or sex (P ALL cases (34).Ionizing radiation was shown to produce a ever, there was a si@cant city difference ( P = number of chromosomal abnormalities, including changes ki risks being less than those in Hiroshima morphologically similar to the Philadelphia chromosome, in I S90 PRESTON ETAL. 3. TABLEwr :w Multiple Myeloma,Distribution of Observed and Fitted Casesby Sex, Time, Exposure Status and Age-__ .--1 ' C . < wL L 1966-1987 _ _Age at exposure (year) Dose category (Gy) <0.01 0.014 co.01 0.014 <O.Ol Om-4 0-19 cases Males 00 1 2 2 2 2,: Fitted background Fitted excess 0.02 0.02 0.00 . 0.00 1.46 136 1.48 138 .i: om 40.00 0.00 0.00 -1 2S39 CaSeS Fitted background 00 0 3 0.18 4 2% 4 2.75 4 3.16 4 293 T .4 I 0 Fitted excess 0.00 0.00 0.00 0.00 0.00 om -$& 8 240 CaSeS 02 .. 4 4 46 b Fitted background 1.47 139 3.64 332 5.11 4.91 4 Fitted excess 0.00 0.00 0.00 0.00 0.00 0.00 Q AU ages cases 02 10 10 12 12 *A it Fitted background Fitted excess 1.68 139 0.00 0.00 8.06 7.63 9.74 922 * t - 3 9 0.00 0.00 0.00 0.00 Females 0-19 cases Fitted background 00 0.03 0.02 0 2.01 2 1.74 0 2.03 21.76 IT& Fitted excess 0.00 0.00 0.00 0.00 0.00 0.00 E 339 cases Fitted background 13 0.44 0.44 6 3 78 8.57 8.32 9.01 8.76 l Fitted excess 0.00 0.00 0.00 0.00 0.00 0.00 240 cases 0 2 12 8 12 10 Fitted background 2.1; 1.93 732 6.90 9.65 8.83 Fitted excess 0.00 0.00 0.00 0.00 0.00 0.00 All ages cases Fitted background Fitted excess 15 259 239 0.00 0.00 18 18.10 0.00 13 16.96 0.00 19 20.69 0.00 18 1935 0.00 All ages cases Fitted background Fitted excess Bothsexes 17 423 3.98 0.00 0.00 28 26.16 0.00 23 2459 0.00 29 30.43 0.00 30 2857 0.00 T h e fitted value for the background cases is the number of cases predicted by the final model assuming no radiation exposure. The fitted vaiue for the excess cases is computed as the difference between the estimated number of background cases and the total number of cases predicted by the model. The specific model used is presented in Appendix 1. The only constraint in these fitted values is that the total number of fitted (background plus excess) cases equal the total number of observed cascs. apparently healthy subjects (35).and X irradiation in vitro was found to induce bcr-ab1fusion genes. including the types specific to CML (Ito, personal communication). Studies are under way to investigate molecular changes. including bcrab1 fusion genes in exposed and nonexposed leukemia cases. The translocation involving the bcr and ab1 genes has rarely been found in AML cases, but epidemiological studies, including a case-control study of the A-bomb survivors (361, demonstrated that exposure to benzene increases the risk of leukemia, especially AML (37). In trying to learn about mechanisms of radiation leukemogenesis. especially for AML, it would be important to consider the possible effect of other leukemogenic factors. such as benzene and other chemical agents. Although certain cytogenetic effects caused by radiation may be sufficient for the induction of some types of leukemia such as CML and ALL. additional changes might be necessary for other types such as AIfL in adults. Nagasaki is an endemic area of HTLV-1. an endogenous virus known to cause ATL. Roughly 30% (22 out of 58) of the Nagasaki leukemia cases were diagnosed as having ATL. whereas less than 1% (1 out of 173) of the leukemias in Hiroshima were ATL. The failure to find a sigdicant radiation effect for ATL suggests that radiation exposure does not interact with HTLV-1 in inducing ATL. Recent data ah0 I HEMATOPOIETIC TUMORS IN ATOMIC BOMB SURVIVORS S91 $ Distance TABLE X N Early Leukemia Deaths, Comparison of the 1948-1950 Data of Folley ef al. and 1950-1953 Data of the Life Sp_a_n Study (LSS)" - Population LSS population. 1950-195-3 - Open population (Folley et ai.). 1948-1950 __ Crude rate Crude Cruderate Crude per rate per per rate per Person 1.OoO.Mx) 1o.Ooo Person years 1.o0o.ooo 1o.ooo years Cases people pk' Population (approximate) Cases people IT Proximal Hiroshima 0-1999m Nagasaki Total 36.163 7.618 33.811 83.723 27 1.032 3 2 30.998 24.485 7 915 2.86 8.39 108208 34 1.006 3.14 3 9 3 7 %.?I4 25.768 122.482 15 484 155 7 84s 2.72 22 560 1.a0 Distal Hiroshima 22OOOm Nagasaki Total 35.805 24.079 59.w 114297 76.828 191.125 5 140 0.44 67267 3 125 0.39 88.703 8 134 0.42 155.970 209.873 276.753 186.626 4 59 0.19 3 34 0.11 7 45 0.14 Total Hiroshima Nagasaki Total 61.968 31.727 93.695 1Y8.020 52 516 101313 10 315 299333 42 348 1.62 93.265 0.99 %.%2 1.40 195.227 306557 302.521 609.108 19 193 10 103 29 149 - 0.62 033 0.38 T h e data for cases and persons at risk in the open population for the period from 1 January 1948 through 51 December 1950 were taken from Folley er al. ( 2 ) .The population size was determined from the 1950 census of A-bomb survivors. Person-years were estimated as 3.12 times the population size. The 4% correction was used to allow for deaths during the 3-year follow-up period. T h e magnitude of this correction was estimated using the U S data. The LSS data cover the period from 1 October 1950 through 31 December 1953. show that HTLV-I infection rates are not associated with exposure to A-bomb radiation (Matsuo, personal communication). As a result of the Leukemia Registry-FAB reclassification, most of the cases previously classified as CLL or "lymphosarcoma leukemia." including many of the cases msidered in the earlier report on CLL by Finch and Hoshino (38),are now classified as ATL. After the Leukemia Reg* ktry-FAB reclassification there were only four CLL cases in LSS (17).The small number of CLL cases is consistent the rarity of this type of leukemia in the general Japse population (39.40). After exposure to ionizing radiation. significant excesses all types of leukemia. excluding CLL, have been reported nsistently for almost 50 years (4146).In the aggregate. of the association between leukemia and external es to low-LET radiation delivered at hgh dose rates nstrate that leukemia should be considered separately lid cancers because both the magnitudes of the risk risk patterns differ. gh many studies have demonstrated an excess of nic leukemia, few have sufficient data to evaluate the f the dose-raponse curve or provide detailed informabout temporal trends or effect modification (47-52).The appears to flatten at high doses, presumably due to g effects, but the small number of cases suggests cauinterpretation of the findings.Unlike solid tumors,radileukemias start to OCCUT about 2 years after exposure. absolute and relative risks generally appear to reach a ak soon thereafter and then gradually decline. Among ults exposed to radiation a positive association between ukemia risk and age at irradiation has been reported among the patients with anklosing spondylitis (53)and amon%women treated for benign gynecolo@caldiseases (16,54)T.he results of our stud! are generally in line with other studies. A statistically significant increased risk of multiple myeloma associated with exposure to A-bomb radiation has been reported in previous analyses of LSS mortality and incidence data (5.13.55).The magnitude of risk estimated was quite large. exceeded only by that for leukemia. The present study did not provide evidence of such an association. As described in the Results, the change is due primarily to increased follow-up. but differences in diagnostic criteria and the decision to consider only first-primary cancers also affected the risk estimates. The decision to limit analyses to cohort members with DSS6 kerma estimates below 4 Gy did not have a major impact on the findings. It is noteworthy that no new cases of multiple myeloma were identified in the high-dose group during the last 12 years of follow-up. One might argue that misclassification or underdiagnosis of multiple myeloma has contributed to the above observation. but this seems unlikely in view of the general improvement in diagnostic procedures and inclusion of cases from all available sources. Earlier results based-on the mortality data are dependent on the inclusion of misclassified cases, second primaries and cases with dose estimates over 4 Gy. A detailed review of the cases of multiple myeloma wiU be undertaken soon. Together with the extension of the follow-upperiod, this review should help clarify the nature of the risk of multiple myeloma in the US. Like CLL, multiple myeloma is a malignant proliferative disorder of B-cell lineage and is considered to originate in terminally differentiated B lymphocytes. In view of the lack of evidence for a relationship between radiation exposure 1 S92 PRESTON ET At. and CLL it may not be surprisingthat radiation is not related changes of specific diagnoses within the to multiple myeloma. Epidemiologicalstudies of populations types considered here. with protracted low-level radiation exposures have suggested Although we reviewed the records an increased risk of multiple myeloma (56.57). It k possible, of multiple myeloma and lymphoma in though unlikely, that protracted exposures produce multiple based on the kind of definitive consider myeloma but not leukemia (45) or that the reported effects materials involved in the Leukemia Re reflect confounding or other mechanisms. The most recent Therefore, we chose to place less emp analysis of mortality of Hanford workers provides less evi- ic analyses of these cancers in thisr e p dence than previously for increased risk of multiple myeloma nation of the material available for these sites associated with radiation dose (32?,39). under way, and detailed reports are forthcoming. . *.I, For malignant lymphoma, limited evidence suggesting an detailed analyses should provide additional insightsm t 0 - h excess risk associated with A-bomb radiation was available ation effects for these cancers.but it is unlikely that the g w . from early prevalence studies (It?. 19,60),but mortality data al conclusions regarding multiple myeloma and lymphoma have not demonstrated an association (14). In the literature, this paper will be affected in any substantive way. there is reasonable agreement that Hodgkin's disease is not Although much has been learned from these data, &e increased in irradiated populations (41). but findings regard- next 10 to 20 years will be important in understanding &e ing NHL are variable. Increased risks were found in women risks of radiation-induced hematopoietic tumors and lym- treated with radiation for ankylosing spondylitis (44) or for phomas. The temporal pattern of leukemia risks, cervical cancer (61).Increased mortality from lymphosarco- AML, among young survivors remains to be studied in ma was found in early cohorts. but not in recent cohorts, of detail. Also of interest are whether the relatively constant American radiologists (62). No increased risk for hHL was excess absolute risk of lymphoma in males persistS:&~d reported from a recently updated and expanded study of whether there are detectable differences in the risk of women treated with radiation for various gynecological dis- Hodgkin's disease and NHL. orders (54) or of Chinese radiation workers (631. The pre- sent finding of an increased risk of NHL. apparently limited APPENDIX 1 to males. warrants further careful study. It might be argued that this finding reflects a bias arising from the inclusion of malignant lymphoma cases identified at autopsy. However, Ascertainment of LSS Leukemia, Lymphoma and Myeloma Cases this seems unlikely since there is no ekidence for time trends A master list of potential cases was created by merging (1) in the excess risk. Since Hodgkin's disease is relatively infre- Leukemia Registry records for members of the U S ,(2) rele- quent in the LSS. it is difficult to reach any conclusions vant tumor registry records for LSS cohort members, (3)infor- about the effects of radiation on Hodgkin's disease. More mation on deceased persons in the LSS whose cause of death definitive evidence on this issue is expected from a study as coded on the death certificate was a diagnosis of interest. now in progress involving standardized reviews of the LSS and (4)information obtained from recent special reviews of lymphoma cases. In speculating on mechanisms for a weak myeloma cases. An initial review revealed that all death oertifi- association between radiation and NHL. as suggested by cate cases not included in either the Leukemia Registry or the present data. it is of interest to note recent findings sug- tumor registries had previously been reviewed and rejected by gesting immune disturbance, as indicated by an increased either the Leukemia Registry or tumor registries. The master prevalence of Epstein-Barr virus reactivation. associated list contained 1233individuals (including 197people who were with A-bomb radiation (64). NIC A m ) . Table AI summarizes the cases in the master list A major difficulty in long-tern follow-up studies such as by ascertainment source and diagnosticcategory. this one involves consistency in the classification of out- The 4 5 Leukemia Registry noncancer cases that were not comes in the face of changes in diagnostic procedures. crite- included in the tumor registries as hematopoietic cancers or ria and nomenclature. The leukemia analyses in this study lymphomas were excluded from additional review. The benefited greatly from the recently completed Leukemia remaining 788 cases were each broadly classified as a Registry-FAB reclassification. Leukemia Registry-FXB leukemia, lymphoma or myeloma, and the diagnoses were diagnoses were wadable for more than 606 of the leukemia compared. As the data in Table AI1 indicate, there were some cases. whereas diagnoses in the remaining cases were based discrepancies between the two sources, due primarily to the on supplementary reviews of available records by study differing coverage periods for the two registries and a lag in hematologists. Although differences in diagnostic criteria the processing of Leukemia Registry cases accessed during the may introduce some biases into type-specific analyses. these 1980s. The data for discordant pairs were reviewed by the effects are not large since the availability of Leukemia Reg- authors (S. K. and K. M.) to determine the final diagnosk A istry-FAB diagnoses appears to be independent of dose special review of myeloma diagnoses found to differ from (17).Furthermore, the changes in diagnosis resulting from those used in earlier analyses was carried out by hematolokFj the Leukemia Registry-FAB reclassification involved affiliated with the Leukemia Registry. ' HEMATOPOIETIC TUMORS IN ATOMIC BOMB SURVIVORS s93 TABLE AI Comparison of Tumor Registry and Leukemia Registry Cases" TABLE An Comparison of Tumor Registry and Leukemia Registry Diagnoses" Tumor registry Tumor registry diagnosis Leukemia registry Cancer Noncancer Not included Total Lymphoma or hematopoietic cancer 541 19 112 672 Other cancer 12 1-04 116 Not included 104 341 - 445 Total 657 464 112 1233 "This table contains information on potential cases of leukemia. Ivmphoma or multiple myeloma identified from tumor registry ( T R ) . leukemia registry (LR) or mortality surveillance records. Data on 197 NIC members of the LSS are included in the table. The 445 :xes ilot classified as a disease of interest in any registry were excluded from further consideration. This group includes five death-certificate-only cases that were rejected in earlier LR reviews. To compare the Leukemia Registry and tumor registry diagnoses at a more detailed level, cases were classified using ICD-0 morphology codes (23).Table AI11 shows the Leukemia regisuy diagnosis Other or not Leukemia Lymphoma Myeloma included Total Leukemia with FAB Leukemia without FAB Lymphoma Myeloma Otherhot included Total 136 % 0 0 18 250 - 12 0 39 187 11 0 26 133 202 0 31 233 0 84 20 101 93 20 0 131 318 104 116 788 T h e comparison is based on broad diagnostic categories defined at an earl!- stage in the review process. Leukemia cases from the leukemia registry with and without Leukemia Registry-FAB classification are t r e a t d separately. Ttus table contains data on 126 NIC members of the LSS. correspondence between FAB classifications. the traditional Leukemia Registry leukemia diagnostic categories. and the ICD-0 morphology codes used in this study. When there was a discrepancy. and a Leukemia Registry-FAB diagnosis was available. it was chosen as the final diagnosis. Diaposuc caicgor) Acute IynphoeytK leukcma (ALL) chronic myelogenousleukemia ( a L ) TABLE AI11 Comparison of Leukemia Diagnostic~Ca_tego~rie~s" _ _ _ - _ _ _ _ _ _ - 100-M 98213 Subacute lymphoid leukemia 98403 Eqthroleutcnua 98613 Acute myeloid leukenua 98623 Subacute mveloid leukemia 98643 Aleukcnuc myeloidleukcma 98663 Acute pmmveloMlc leukeuua 98913 Acute monOCVtK leukcma 98633 chronic mveloid leukemia 96653 N e u u o p b ~ l ~Iecukenua Leukenua repsm rcclasbificationcalegon (Leulenua Rrgctn -F4B) .4cute 1:mphocvtic leukemia Lynphgnic leukenua Npes I and If . Mvclold lcukcrma type VI Acute myeloid leukenua Myeloid leukcma types I. 11and IV Acuw leukemia NOS Refractor)memu with excess blasts (R4EB) RAEB-Tcell Hgopkutic leukemia Myelad kukcnua typc 111 Myelotd leukemia type V an Chronic mvelornonocvticleukenua Onpnal leukemia repst? tvpe Acute Ivmphcatic leukemia Erythroleukemu Acute p a a u l o q i c leukemia Acute myelomoaocylic leukemta ANIC r n o n w i c leukemia Chronic ganulocytic leuliema 97022 ATL An 983(u Pl85nmall Icukenu. 98233 (II Phsma-all leukemia CZL 98)13 Acute leukcmu NOS 98603 MrC)dd kukemuNOS 98au L c v t e m ~NOS 99y13 H q c r U kukcmu ACUWkukcmu NOS LeukemiaNOS 99641 MDS -s p s94 PRESTON ET AL. i. TABLEAIV Summary of Final Diagnoses by Source and Perioda TABLE AV .? Reclassified (Leukemia Registry-FAB) Diagnoses i d i Final Leukemia Diagnoses in This Study" Both tP leukemia Leukemia registryand Tumor Leukemia Registry-FAB diagnosis reetry tumor registry Diagnosis Period only registry only Total Leukemia subtype Yes No Tatal Leukemia 1946-1958 1958-1980 1981-1989 Total 49 12 -1. 65 37 0 86 173 0 185 19 15 68 259 15 339 Lymphoma 1946-1958 15 18 0 33 1958-1980 1 154 23 187 `i 1981-1989 Total 1 17 37 46 84 209 78 304 4 Myeloma 19461958 00 1 1958-1980 50 2 52 1981-1989 31 9 41 Total 81 11 94 ALL ALL CML An. Other Total 59 (9%) 22 (69%) 43 (75%) 12 (520/,) 1 (2506) 140 (61%) 44 (43Yo) 10 (3lwo) 14 (370) 11 (4896) 12 (75O6) 91 (3996) 103 32 6 57 23 16 231 I "The 737 cases identified in the review are classified by the date of diagnosis and the availabilityof tumor registry or leukemia registry data. This table includes informationon 117 SIC membersof the U S cohort. "The distribution of Leukemia Registry-FAB diagnoses for the leukemia subtypes is described for the 231 cases used in the primary analyses for this paper. h tion is used: r, time since exposure: a. attained age: g, age at Discrepant cases with no Leukemia Registry-FAB diagno- exposure: s. sex: d . weighted dose in sieverts. sis were reviewed individually. and a consensus diagnosis was As indicated in the expressions below. time since sxpo- made. For 34 cases the tumor regsty and Leukemia Registry sure or, where appropriate. its logarithm IS centered at 25 diagnosis dates differed by more than 1year. These cases were years after exposure. Sirmlarly. attained ase is centercd at 50 all reviewed. For cases with smaller discrepancies. preference years. Thus the leading term in the background-rate models was given to the Lzukemia Registry date. A diagnosticcertain- refers to the risk for a jO-year-old, whereas the leading term ty code was assigned to each case on the basis of the Leukemia in the excess-risk models is an estimate of the risk coefficient Registry certainty code or noncomputerized information when in August 1970. i.e., 25 years after exposure. When age at such records could be found. A total of 51 cases were rejected exposure is used as a continuous variable it is centered at age as it result of the case review. leaving 737 cases for further con- 25. In these cases the leading coefficient describes the risk in sideration. Tumor registry data and coding rules (22) were 1970 or a person who was 25 years old in 1945. Secular used to assign a sequence number to each case. trends in background rate models are written in terms of (g - Table AIV compares the Leukemia Registry and tumor 25); this is possible because all members of the cohort were registry data by period. Agreement between the two registries is excellent from 1958 through 1980. whereas the Leukemia Registry is the primary source of cases before 1 9 3 , as.! :It- m n c r :egistries are more important after 1980. Table 1il: the main text summarizes the results of the diagnostic review and case selection.Table AV provides informa- exposed at (essentially)the same time. Thus the age at exposure covariate is equivalent to 1970minus the year of birth. Background rates are per la` PY, whereas the units for the EAR estimates are given in terms of excess cases per la`W .;: . Leukemia: All Types tion on the use of the Leukema Registry-FhB in the final leukemia diagnoses by subtype for the 231 cases with DS86 estimates between 0 and 4 Gy that were used in the main analyses. There is no significant difference between the cities with respect to the proportion of cases with new diagnoses. nor are there sigmficant differences for any subtype. Seventeen of Background rate i0.91e4`M2(g-3)+3DB In(dYJ)+l22In'(a5Ol X( s.a .g) = 0.45e4)M2(g-Y)+-708In(u50)+112 In'(u50) Excess risk male female the cases without Leukemia Registry- F A B diagnoses were Nagasaki cases (5 AMI-.1ALL, 10ATL, and 1other type). APPENDIX 2 Fitted Models P(LI)E(g.LS) 0.33(d+0.?9d2)e41'(`-3) 0.66(d+0.79d2)e4""-L5' I0.Jg(d+0.79d2)e4'3``-3) = I0.97(d +0.79d2)e403(r-") males age ATB 0-19 females age ATB 0-19 males age ATB 20-39 females age ATB 20-3' This appendix contains the final fitted models for the major analyses described in this paper. The following nota- I1.31(d+0.79d2)e4m(r-3) 2.64(d+0.79d2)e003"-?5) males males age ATB age ATB I HEMATOPOIETIC TUMORS IN ATOMIC BOMB SURVIVORS s95 Acute Lymphocytic Leukemia Rackground rate A(C. s. a, e )= 0.07e1Yb'"a) Excess risk I . e-0 d e@ ~ ~de. I) = q~-O.Q7@-25)-014(I-rr' '- ~ ) - ~ I ~ I I - B ) mfeamleasles Acute Myelogenous Leukemia I -Background rate o.41ea.w-~315 i ~ l (aow,-a73Irzr.~wi A( c. s. a g ) = 0.45~4CEe- YI * L76 In(o,(01- IsOl&u~XJ) Excess risk males females Chronic Myelogenous Leukemia Background rate h(c.s,a)= 0,23eZ.2b InjadO) Hiroshima males 0.08e2.26 h{d50) Nagasaki males 0.12el.'6h(a/SO) Hiroshima females 0.04eZ.26 h(o:MI Nagasaki females .xcessrisk 1 OtherLeukemia-Hiroshima ?Background rate f k:A(c.s. a)= iExcess risk &(S. 1) = k44 Hiroshima males Nagasaki males Hiroshima females Nagasaki females males females males females Excess risk Multiple Myeloma Background rate males females Excess risk None ACKNOWLEDGMENTS The authors wish to acknowledge the major contributions of Drs. Michito Ichiman: and Stuart C. Finch. without whose long-term involvemefit i n a,.ii suppm: of the Leukemia Registry this report would not have been possible. In addition. we want to thank Mr. Morito Dote. Ms. Sachiyo Funamoto and Ms. Ritsuko Sat0 for technical support. The Hiroshima and Nagasaki tumor registries are partially supported by NCI contract number NCI-4893-8-001.This publication is based on researc h protocol R P 29-60 and manuscript TR 24-92 concerning research performed at the Radiation Effects Research Foundation (RERF). Hiroshima and Nagasaki. Japan. RERF is a private nonprofit foundation funded equally by the Japanese Ministp of Health and Welfare and the United States Department of Energy through the National Academy of Sciences. REFERENCES 1. S.C. Finch. Z. Hrubec. M.D. Nefzger. T. Hoshino and J. Iroga. Derecrion of Leukemia and Relaled Disorders, Hiroshima and "Vugasaki.Research Plan. TR 5-65. Atomic Bomb Casualty Commission. Hiroshima. 1965. 2. J . H. Folley. W.Borges and T. Yamawaki. Incidence of leukemia in survivors of the atomic bomb in Hiroshima and Nagasaki. Japan. Am. J. Med 13. -311-321 (1952).[ABCC TR 30-591 3. A. B. Brill. M. Tomonaga and R. M. Heyssel. Leukemia in man following exposure to ionizing radiation. A summary of findings in Hiroshima and Nagasaki and comparison with other human experi- ence. Ann. Intent Med 56,59040!9 (1%2). [ABCCT R 15-59] 4. M. Ichimaru. T. lsfumaru and J. L. Belsky. Incidence of leukemia in atomic bomb survivors belonging IOa fixed cohort in Hiroshima and Nagasaki, 1950-71;radiation dose, years after exposure, age at exposure, and type of leukemia. 1. Radial. Res. 19,262-282 (1978). [FERFTR 10-761 5 M. Ichimaru. T. Ishimaru. M. Mikami and M. Matsunasa. Multiple myeloma among atomic bomb survivors. Hiroshima and Nagasaki. 1950-76. "V. Engl. 1. Med. 301,439440 (1979).[ R E W TR 9-79] 6. T. Ishimam. M. Otake, M. Ichimaru and M. Mikami. Dose-Respome Relationship of Leukemio Incidence among Atomic Bomb Survivors and Their Controls by Absorbed Marrow Dose and Two Types of Leukemia. Hiroshima and Nagasaki, October 1950-December1978. T R 1@811Radiation EffectsResearch Foundation. Hiroshima 1981. 7. M.Ichimaru, T.Ishimaru, M.Mikami, Y.Yamada and T. Ohkita, Incidence of Leukemia in a Fixed Cohon ofAtomic Bomb Survivors and Controls, Hiroshima and Nagasaki. October 1950-December 1978. TR 13-81. Radiation Effects Research Fouadati~m,Hiroshima, 1981. 8. T. Ishimaru,M. Ichimaru. M.Mikami. Y. Yamada and Y.Tomona- ga. Disrribution of Onset of Leukemio among Atomic Bomb Survivors in the Leukmia Registry by Dose, Hiroshima ond Nagasaki, 2946-75. TR 12-81, Radiation Effects Researcb Foundation, Hiroshima,1981. 9. T. Ishimaru. M. Ichimaru and M. Mikami, Leukemio Incidence among Individuals Exposed In Utero, Children of Atomic Bomb Survivors and Thcir Controk Hiroshima and Nogmaki, 1954-79. TR 11-81, Radiation Effects Research Foundation,Hirodtima. 1981. II . S% PRESTON ET AL. IO. D. A. Pierce, D.L. Preston and T. Ishimaru, A Method of Anulvsis of 26. D. L. Preston, J. H Cancer Incidence in Japanese Atomic Bomb Survivors with Applica- HiroSoft International Cop., Seattle, WA, . rion toAcute Leukemia TR 15-83, Radiation Effects Research Foun- .*, dation, Hiroshma. 1983. I I . D.L.Preston. H. Kato. K. J. Kopecky and S. Fujita. Studies of the mortality of A-bomb S U M V O ~ . 8. Cancer mortality. 1950-82. R a d k Res. Ill. 151-178 (1987). [RERFTR 1-86, Report 10) 12. D. L.Preston and D. A. Pierce, The effect of changes in dosimetry o n cancer mortality nsk estimates in the atomic bomb survivors. Radiat. R a . 1 1 4 , 4 3 7 4 (1988). [RE- TR 9-87] 28. Hall.London, 1974. B. J 13. Y.Shimizu, H. Kato, W J. Schull, D. L. Preston, S. Fujita and D. A. Pierce, Life Span Study Report 11. Part 1. Companson of nsk coefficients for site-specific cancer mortality based on the DS86 and u). J s T65DR shielded kerma and organ doses. Rodiat. Res. 118,502-524 (1989). [RERF TR 12-87] TR 17-911 0 t I 4 Y.Shimizu. H . Kat0 and W J Schull. Life Span Study Report 11. Part 2. Cancer mortality in the years 1950-85 based on the recently 31. H. Nishiyama. Relative frequency and mortality rate of of 111 Japan. Gann 60. 7141 (1%9). revtsed doses (DSS6) Radrut Res. U1.120-141 (1990) [ R E M TR 32. I. F Fraumeni and R. Miller. Epidemiology of leukemwc rrccat 5-yl)l observations f Natl. Cancer Inst. 38.593-605 (1%7). 15. A. V. Pisciotta. Irf. Ichimaru and %I. Tomonaga. Morphological 33. F. Bessho. Acute non-lymphocytic leukemia is not a major of a.Review of Peripheral Blood and Bone Marrow of Acute and Chronic childhood leukemia in Japan. Eur. J . Cancer. Clin. Omo& 1 Leukemia in Atomic Bomb Survivors, Hiroshima und Nagusaki. RP 4-84. Radiation Effects Research Foundation. Hiroshima. 1984. 729-732 (1989). 'r .*t:. 34. B. Clarkson and A. Strife, Discordant maturation in cluo&ipye. 16. T. Matsuo, M. Tomonaga. J. M. Bennett, K. Kuriyama. F. Imanaka. logenous leukemia. In Chronic Myelogenous Leukemia. M&l A. Kuramoto. N. Kamada. M. Ichimaru. S. C. Finch and A. V. Pis- Dekkrr. New York. 1991. ciotta. Reclassification of leukemia among A-bomb survivors by 35. K - 0 Goh. Total-body irradiation and human chromosomes; e French-American-British (FAB) classification. 1. Concordance of diagnosis in Nagasaki cases by RERF members and a member of FAB cooperative goup. Jpn f. Clin. Oncol. 18.91-96(1988). [RERF TRI-871 netic studies of the peripheral blood and bone marrow l e ~ a seven years after total-body irradiation. Rodiat. Res. 35, ~ S ~ I M (1968). 0 36. T. Ishimaru. H. Okada. T. Tomiyasu. T. Tsuchimoto. T.Hoshioo a d 17. M.Tomonaga. T. Irlatsuo, R. L. Carter. J. M. Bennett. K. Kuriyama. M. Ichimaru. Occupational factors in the epidemiology of leukemia F. Imanaka. S. Kusumi. K. .Mabuchi. A. Kuramoto. N. Kdmata. W. in Hiroshima and Nagasaki. Am. J . Epidemiol. 93. 157-165 (1971). Ichimaru. A. V. Pisciotta and S . C. Finch. Differential Effecrs of Atomic Bomb Irradiution in Inducing .Major Leukemia Types IA.ML. j i . WHO. Evaluation of the carcinogenic risk of chemicals to humam. A L L and C.MLI:.inalvses of Open-Cirv Cases Inchding Life Span [ARC Monogr. 29.93-148 (1982). Sfudy Cohon Based upon Up-dated Diagnostic Systems and :he .Vew 38. S. C. Finch. T.Hoshino. T. [toga. M.Ichimaru and R. H. Ingram Jr.. Dosimerrv System I DSY6). TR 9-91, Radiation Effects Research Chronic lymphocytic leukemia in Hiroshima and Nagasaki, Japan. Foundation. Hiroshima. 1991. Blood 3.79-86 (1969). 18. R E. Anderson and K. Ishida. Malignant lymphoma in atomic bomb 39. The T- and B-Cell Malignancy Study Group. The third nation-wide survivors. Hiroshima. Ann. Intern. .Wed. 61.853-862 (1964). [REM study on adult T-cell leukemiailymphoma (ATL) in Japan. cfiarac- TR 2-61] teristic patterns of HLA antigen and HTLV-I infection in ATL 19. H. Nishiyama. R.E. Anderson. T.Ishimaru. K. Ishida. Y. ii and N. patients and their relatives. Int. J. Cancer 2 (19881. Okabe. Malignant lymphoma and multiple myeloma in atomic bomb 40. C. Muir. J. Waterhouse. T. Mack. J. Powell and S . Whelan, Concer survivors. Hiroshima-Nagasaki 1945-65. Cancer 32. 1301-1309 Incidence in Five Continents. Vol. 5. Scientific Publication No. 88. (1973). [REWTR 4-71] International Agency for Research on Cancer, Lyon. 1987. 20. D. E. Thompson. K. Mabuchi. E. Ron. M. Soda. M. Tokunaga. S. Ocfukubo. S. Su-@moto. T. Iksda. M. Terasaki. S. [zumi and D. L. Reston. Cancer incidence in atomic bomb survivors. Part 11: Solid tumor incidence. 1955-1987. Raiiiut. Res. U7.S17-S67 (1%). [RERFTR5-92] 41. Yational Research Council. Committee on the Biological Effects of Ionizing Radiation. Health Effects of Exposure to Low Lrvcls yf lonizing Radiation (BEIR V). National Academy Press. Washington. DC. 1990. 2 i . J. Bennett. E. Catovsky. M. Daniel. G . Flandrin, D.Galton. H. Gral- 42. UNSCEAR. Sources, Effects. drid Risk2 ,j ioni:i..;.g Rudiation. nick and C. Sultan. Proposals for the classification of the myelodys- Report to the General Assembly. with Anne.ses. L'riited ?;ations. plastic syndromes. Br. J . Harmatol. 51. 189-199 (1982). New York. 1988. 22. K. Mabuchi. M. Soda. E. Ron. M. Tokunaga. S. Ochikubo, S. Sugimoto. T. Ikeda. M.Terasaki. D. L. Preston and D. E. Thompson. Cancer incidence in atomic bomb suMvors. Part I: Use of tumor registries in Hiroshima and Vagasaki for incidence studies. Radiat Res. l37.S1-S16 (1993). [RERFCR 3-91] 23. Knternutional Classificution of Disrases for Oncology (ICD-0). 9th ed. World Health Organization. Geneva. 1977. 24. W. C. Roesch. Ed.. US-Japan Joint Reassessment ofAtomic Bomb Radiation Dosimrtp in Hiroshimu and .Vugusuki. Radiation Effects Research Foundation. Hiroshima. Japan. 1987. 25. S. Fujita. Versions of DS86. RERF Lpdure 12 (1989). 43. R. W. Miller and G. W. Beebe. Leukemia. lymphoma and multiple myeloma. In Radiation Corcinogenrsis (A. C. Upton. R. E. Albert. F. J. Bums and R. E. Shore. Eds.). pp. 245-260. Elsevier. New York. 1986. 44. S. C. Darby. R. Doll. S. K. Gill and P. G. Smith. Long term m o d t y after a single treatment course with X-rays in patients treated for ankylosing spondylitis. Br. 1. Cuncrr 55. 179-190 (1987). 45. J. D. Boice. Jr.. M. Blettner, R. A. Kleinerman. M.Stovdl, W. c. Moloney. G. Engholm. D. Austin. A. Bosch. D. Cookfair. E. Kre- mentz. H. Latourette. L. Peters, M. Schulz. M.Lundell. F. Peners- son. H. Storm. C. Bell. M. Coleman. P. Fraser. M. Palmer. P. Prior. , HEMATOPOIETIC TUMORS IN ATOMIC BOMB SURVIVORS s97 N. Choi. T. Hislop. M. Koch. D. Robb. D. Robson. R. Spengler. D. V.Fournier. R. Frischkorn. H. Lochmuller. V.Pompe-Kirn. A. Rimpela. K. Kjorstad. M. Pejovic. K. Sigurdsson. P. Pisani. H. Kucera and G. Hutchison, Radiation dose and leukemia risk in patients treated for cancer of the cervix. 1. Narl. Cancer Insr. 79. 1295-131 1 (1987). 6. P. D. Inskip. R. R. Monson. J. K. Wagoner. M. Stovall. F. G. Daws. R. A. Kleinennan and J. D. Boice. Leukemia foUowin_eradiotherapy for uterine bleeding. Rudiur. Res. 122.107-119 (1990). 7. R. E. Shore. R. E. Alben and B. S. Pasternack. Follow-up study of patients treated by X-ray epilation for tinea capitis: resurvey of posttreatment illness and mortality experience. Arch. nviron. Healfh 31. 17-24 (1976). 8. N. G. Hildreth. R. E. Shore. L. H. Hempelmann and M. Rosenstein. Risk of extra thyroid tumors following radiation treatment in infancy for thymic enlargement. Radiar. Res. 102.378-391 (1985). 9. E. Ron. B. J. Modan and J. D. Boice. Jr.. Mortality after radiotherapy for ringworm of the scalp. Am. J. Epideminl. U7.713-725 (1988). 50. J. M. Kalder. N. E. Day. A. Clarke. F. E. Van Leeuwen. hi. HenryAmar. M. V. Fiorentino. J. Bell. D. Pedersen. P. Band. D. Assouline. M. Kogh. W. Choi. P. Prior. V. Blair. F. Langmark. V. P. Kirn. F. ,.: Neal. D. Peters. R. Pfeiffer. S.Kajalainen. J. Cuzick. S. B. Sutcliffe. R. Somers. B. Pellae-Cosset. G. L. Pappagallo. P. Fraser. H. Stome 1 and M. Stovall. Leukemia following Hodgkin's disease. N. Engl. 1. J Med. 322.7-13 (1990). I, M. Greene.R. Young. J. Memll and V. De Vita. Evidence of a treat- ment dose response in acute nonlymphocytic leukemias which occur j after therapy of non-Hodgkin's lymphoma. Cancer Res. 43. 1891- and Nugusuki. TR 9-84.Radiation Effects Research Foundation. Hiroshima. 1984. 56. J. Cuzick. Radiation-induced myelomatosis. A'. Engl. J. Med. 304. 204-210 (1981). 57. G. M. Kendall, C. R. Muirhead. B. H. Macgibbon. J. A. O'Hagan. A. A. Goodill. B. K. Butland. T. P. Fell. D. A. Jackson, M. A. Webb. R. G. E. Haylock. J. M. Thoman and T.J. Silk, Mortality and OCCUpational exposure to radiation: first analysis of the national registry orradiation workers. Br. Med 1.304.220-25 (1992). 58. E. S. Gilbert. G. Petersen and J. Buchanan. Mortality for workers at the Hanford site: 1945-1981.Healrh fhys. 56.1-25 (1989). 59. E. S. Gilbert. E. Omohundro. J. A. Buchanan and N. A. Holter. Mortality of workers at the Hanford site: 1945-1986.Healrh Phys. 64. 577-590 (1993). 60. M. Ichimaru. T. Ohkita and T. Ishimaru. Leukemia. multiple myeloma. and malignani lymphoma. S a n n Monogr. Cancer Res. 32. 113-127 (1986). 61. J. D. Boice. Jr.. G. Eqholm. R. A. Xleinerman. M.Blettner. M.Sto- vall. H. Lisco. w . C. Moloney D. F. Austin. A. Bosch. D. L. Cookfair. E. T.Krementz, H. B. Latourette. J. A. Merrill. L. J. Peters. M. D. Schulz. H. H. Storm. E. Bjorkholm. F. Pettersson. C. M. Janine Bell. M. P. Coleman. P. Fraser. F. E. Neal. P. Prior. N. W o n Choi. T.G. Hislop. M. Koch. N. Kreiger. D. Robb. D. Robson. D. H. Thomson. H. Lochmuller. D. Von Fournier. R. Frischkorn. K. E. Kj0rstad. .4. Rimpela, M-H Pejovic. V. Pompe K m . H. Stankusova. F. Berrino. K. Sigurdsson. G. B. Hutchison and B. MacMahon. Radiation dose and second cancer risk in patients treated for cancer of the cervix. Radial. Res. 116.3-55(19881. .. 52. M. Tucker. C. Coleman. R. Cox. A. Varghese and S. Rosenberg. 1 Risk of second malignancies after treatment for Hodgkin's disease. A'. Engl. 1. Med. 318.76-81 (1988). 53. P. G. Smith and R. Doll. Mortality among patients with ankylosing & spondylitis after a single treatment. Br. Med. J. 284.449460(1982). 54. P. D. Inskip. R. A. Kleinerman. M. Stovall. D. L. Cookfair. 0.Had- -#- jimichael. W.C. Moloney. R. R. Monson. W.D. Thompson. 1. Wactawski-Wende. J. K. Wagoner and J. D. Boice. Jr.. Leukemia lymphoma. and multiple myeloma following pelvic radiotherapy for benign desease. Radial. Res. US.108-124 (1993). . D. G. Hoe1 and R. I. Jennrich. Life Table Anafysis wirh Small Numbers of Cares: A n Exumple-Multiple Myeloma in Hiroshima 62. G. M. Matanoski. P. Sartwell. E. Elliott. J. Tonascia and A. Sternberg. Cancer risks in radiologists and radiation workers. In Radiation Carcinogenesis: Epidemiology and Biological Significance (J. D. Boice. Jr. and J. F. Fraumeni. Jr.. Eds.). pp. 83-96.Raven Press. %ew York. 1984. 63. J-X W'ang. J. D. Boice. Jr.. B-X Li. J-Y Zhang and J. F. Fraumeni. Jr.. Cancer among medical diagnostic x-ray workers in China. J. .Varf. Cancer Insr. 80.344-350 (1988). 64. M. Akiyama. Y. Kusunoki. S. Kyoizumi. K. M. Ozaki. S. Fujita and . J. B. Cologne. Study on the titers of anti-Epstein-Barr virus antibod- ies in the sera of atomic bomb survivors. Radiar. Res. 133,297-302 (1993).[REWTR 14-92) -