## Abstract

Background: The risk of developing nervous system tumors following exposure to ionizing radiation is not well quantified. We characterized the incidence of nervous system tumors among atomic bomb survivors as a function of radiation dose. Methods: Tumors of the nervous system and pituitary gland diagnosed between 1958 and 1995 among 80 160 atomic bomb survivors were ascertained using the Hiroshima and Nagasaki tumor registries, medical records, and death certificates. Pathologists reviewed slides and medical records to provide histologic diagnoses. Poisson regression analyses were used to characterize radiation effects on tumor incidence, which are expressed as excess relative risk per sievert (ERRSv). All statistical tests were two-sided. Results: A statistically significant dose-related excess of nervous system tumors was observed in the cohort (ERRSv = 1.2, 95% confidence interval [CI] = 0.6 to 2.1). The highest ERRSv was seen for schwannoma (4.5, 95% CI = 1.9 to 9.2). The risk for all other nervous system tumors as a group is also statistically significantly elevated (ERRSv = 0.6, 95% CI = 0.1 to 1.3). Risk increases, although not statistically significant, were seen for meningiomas (ERRSv = 0.6, 95% CI = –0.01 to 1.8), gliomas (ERRSv = 0.6, 95% CI = −0.2 to 2.0), other nervous system tumors (ERRSv = 0.5, 95% CI = <−0.2 to 2.2), and pituitary tumors (ERRSv = 1.0, 95% CI = <−0.2 to 3.5). The dose–response relationships were linear. For nervous system tumors other than schwannoma, excess risks were higher for men than for women and for those exposed during childhood than for those exposed during adulthood. Conclusions: A statistically significant dose response was observed for all nervous system tumors combined and for schwannoma considered separately, indicating that exposure to even moderate doses (i.e., <1 Sv) of radiation is associated with an elevated incidence of nervous system tumors.

Tumors of the brain and other parts of the nervous system are a heterogeneous group of neoplasms (1). Many adult neuroepithelial tissue tumors—gliomas including astrocytomas—arise from the glial cells that surround and support neurons. The most common form of malignant tumor, glioblastoma multiforme, is an extremely aggressive and often fatal cancer (2). The most common forms of benign nervous system tumors are meningiomas and schwannomas (sometimes called neurilemomas). Meningiomas (tumors of the meninges) usually arise from arachnoidal cells, whereas schwannomas arise from Schwann cells in the peripheral nerves. Schwann cells are the peripheral nervous system analogues of glial cells. Less common among adults are malignant lymphomas of the central nervous system and germ cell tumors involving regions near the pineal and pituitary glands. Tumors of the pituitary and pineal glands are also intracranial tumors, but the epidemiology and etiology of these endocrine tumors are substantially different than those for nervous system tumors (3,4).

Although the etiology of nervous system tumors is not well described, several epidemiologic studies (512) have reported increased risks of meningiomas, nerve-sheath tumors, and gliomas after high-dose medical treatment, especially for those exposed during infancy or childhood. Less epidemiologic data are available regarding the risk of nervous system tumors for those exposed as adults. After reviewing five epidemiologic studies, Preston-Martin and White (13) concluded that increased risks of meningioma and schwannoma were associated with a history of diagnostic x-ray exposures (both dental and medical) performed many years ago when the level of radiation exposure may have been substantial. However, radiation associated with dental x-rays has not been related to an excess risk of nervous system tumors in other studies (14,15). Previous radiotherapy to the head has been related to excess risks of meningiomas and schwannomas (16). Information on glioma risks after low-dose radiation exposure or exposure as an adult is limited. Mortality studies (1719) of workers exposed to low radiation doses did not find statistically significant effects, but mortality data are a poor measure of risk for nervous system tumors with relatively good survival, such as meningiomas and schwannomas.

Using data for the period from 1961 through 1975, Seyama et al. (20) reported a fivefold increase in brain tumor incidence among male atomic bomb survivors exposed to radiation levels of 1 Gy or more. In the most recent general analysis of Life Span Study (LSS) cancer incidence data (21), an increased risk of extracranial neural tissue tumors was suggested, but there was little evidence of a radiation-associated increase in brain tumor risk. More recently, increased risks of meningioma associated with atomic bomb radiation exposure or distance from the bombs were reported in Hiroshima (22) and in Nagasaki (23).

To further understand the role of radiation in the etiology of brain and other neural tumors, we conducted a detailed incidence study, including a pathology review, in the LSS cohort of atomic bomb survivors. Specifically, we examined the dose–response relationship, quantified radiation risks for specific histologic types of malignant and benign tumors, and evaluated the role of modifying factors on the dose response.

## Materials and Methods

Gy (gray units) are used to refer to doses of radiation in which no allowance is made for the biologic effectiveness of different types of radiation. If allowance is made for the different effectiveness of various types of radiation, then the resulting dose equivalent is expressed using sieverts (Sv). An absorbed dose of 1 Gy is equal to 100 rad, and 1 Sv is equal to 100 rem. A given dose of neutrons is believed to have greater biologic effectiveness than the same dose of γ-rays or x-rays.

### Life Span Study Population

Case ascertainment was carried out for the full LSS cohort, which includes 93 000 atomic bomb survivors from Hiroshima and Nagasaki and 27 000 people who were not in the cities at the time of exposure. The population used in these analyses includes 80 160 members of the LSS cohort for whom organ dose estimates can be computed, who were in Hiroshima or Nagasaki at the time of the atomic bombings, and who were alive and not known to have had cancer at the time of the establishment of the Hiroshima and Nagasaki Tumor Registries (January 1958). The population differs from that described in (21) because of the inclusion of 234 people with dose estimates of 4 Gy or more and the exclusion of 46 people on the basis of new follow-up data. At the end of follow-up, in December 1995, slightly more than 50% of the cohort members were still alive. There are more women (60%) than men in the cohort (Table 1), particularly in the group of people who were 20–39 years old at the time of the atomic bombings. Almost 68% of the study group was exposed to radiation from the bomb in Hiroshima. Individual weighted brain doses were calculated as the sum of the γ-ray dose plus 10 times the neutron dose by using the DS86 system (24,25). Weighted dose was used to allow for the greater biologic effectiveness of neutron radiation doses. The dose estimates incorporate a correction for bias arising as a result of random errors in individual dose estimates (26). Approximately 40% of the cohort had a weighted brain dose of less than 0.005 Sv. Only 2811 survivors (3.5%) had dose estimates greater than 1 Sv. Mabuchi et al. (27) and Thompson et al. (21) have described the characteristics of the LSS cohort in detail.

### Tumor Ascertainment

Tumors of the brain, cranial and spinal nerves, pituitary gland, and pineal gland were ascertained through the population-based Hiroshima and Nagasaki Tumor Registries, which were established in 1958 (27). To improve ascertainment for benign tumors, the tumor registry data were augmented with information obtained from the Hiroshima and Nagasaki Tissue Registries (established in 1973); autopsy, surgical pathology, and clinical records from the Radiation Effects Research Foundation (RERF); and from major medical institutions in Hiroshima and Nagasaki.

A broad range of reported tumor diagnoses were considered in the initial stages of the ascertainment process. The 9th revision of the International Classification of Diseases (ICD-9) topography codes (28) considered in the initial screening were codes 191 (brain), 192 (other and unspecified parts of the nervous system), 194.3 (pituitary gland), 194.4 (pineal gland), 171.9 (connective tissue including peripheral and sympathetic nerves), 225 (benign neoplasms of brain and other parts of nervous system), 227.3 (benign pituitary gland neoplasms), 227.4 (benign pineal gland tumors), 237 (neoplasms of uncertain behavior of endocrine glands and nervous system), 239.6 (brain neoplasms of unspecified nature), and 239.7 (neoplasms of unspecified nature of endocrine glands and other parts of nervous system). The study pathologists also reviewed the records of tumors occurring in neighboring anatomic locations that might include misclassified neural tumors.

The four study pathologists (S. Yonehara, T. Kobuke, H. Fujii, and M. Kishikawa) independently reviewed pathology slides, pathology reports, and clinical records and classified tumors by anatomic site (topography), histologic type (morphology), and tumor behavior, according to World Health Organization (WHO) criteria (29). When diagnoses differed, the pathology panel met to develop a consensus diagnosis.

### Data Organization and Statistical Methods

Incidence for the various tumor types was cross-classified into 5-year age-at-exposure groups; 5-year attained-age groups; calendar time periods with an initial 3-year interval (from January 1, 1958, through December 31, 1960); and 5-year intervals for the period from January 1, 1961, through December 31, 1995; weighted brain dose estimates with cut points at 0, 0.005, 0.05, 0.1, 0.2, 0.5, 1, 2, 3, and 4 Sv; sex; city (Hiroshima or Nagasaki); and membership in the Adult Health Study (AHS) cohort (30). Rates were computed with and without inclusion of tumors detected only at autopsy. Members of the AHS receive biennial clinical examinations at RERF and, therefore, the likelihood of early tumor detection is increased, particularly for tumors with only minor clinical symptoms. Autopsies were performed on large numbers of deceased members of the LSS in the course of a major autopsy program carried out between 1960 and the mid-1970s. Because benign and malignant tumors were a major focus of this program, the number of autopsies conducted could have influenced tumor incidence rates.

For each stratum, person-years, tumor counts, person-year-weighted average values for radiation exposure dose, attained age, age at radiation exposure, and time since exposure were computed. Person-years of observation were calculated from January 1, 1958, until the earliest of a) the date of diagnosis of the first primary tumor, b) the date of death or last known vital status, or c) the end of follow-up (December 31, 1995). Because of the completeness of the Japanese family registration system, less than 1% of the LSS cohort members have been lost to follow-up; about half the cohort members who were lost to follow-up could not be traced at the time the cohort was defined; the remaining half consisted of people who emigrated from Japan (emigration date is known). As with most previous studies of tumor incidence in the LSS [including (21,31)], we excluded tumors diagnosed outside the Hiroshima and Nagasaki tumor registry catchment area and adjusted the person-years on the basis of immigration and emigration information obtained from AHS cohort records (21,32). This adjustment, which depends on city, sex, age, and time period, reduces the effective number of person-years by about 14%.

Imprecision in LSS survivor dose estimates results in an underestimation of radiation risk and some distortion in the shape of the radiation dose–response curve. To adjust for the impact of radiation dose errors, extremely large shielded body surface dose estimates were truncated to 4 Gy, and DS86 estimates were replaced with expected survivor dose estimates computed using the method developed by Pierce et al. (26) to produce bias-corrected risk estimates. These adjustments increase estimates of the slope of the dose response curve by about 10% in linear models.

Poisson regression methods (33) were used to compute maximum likelihood estimates for both excess relative risk (ERR) and excess absolute rate (EAR) models (34). Parameter estimates, likelihood-ratio tests, and likelihood-based confidence intervals (CIs) (33) were computed with the AMFIT computer program (34). We analyzed the data using general ERR models (the background rate times 1 plus the ERR) written as

${\lambda}(\mathit{c,g,p,a,m})[1+{\rho}(\mathit{d}){\varepsilon}(\mathit{c,g,a,e,t,m})]$
and general EAR models (the background rate plus the EAR) written as
${\lambda}(\mathit{c,g,p,a,m})+{\rho}(\mathit{d}){\varepsilon}*(\mathit{c,g,a,e,t,m})$

In these models, λ(•) describes background nervous system tumor rates as a function of city (c), sex (g), time period (p), attained age (a), and membership in the AHS (m). These fitted background rates are estimates of the rates for an unexposed population. The function ρ(•) describes the dose–response shape. The functions ε(•) and ε*(•) describe effect modification in the ERR and EAR models, respectively. Potential effect modifiers included the covariates c, g, a, and m, as well as age at exposure (e) and time since exposure (t). We generally present excess risk (ERR or EAR) estimates corresponding to specified values of any effect-modifying factors in the model. For example, if a model includes age at exposure or attained-age effects, we present the excess risk estimates at attained age 60 for a person exposed at age 30. The ERR, a dimensionless ratio, is positive if radiation increases risk, zero if there is no radiation effect, and negative when radiation exposure reduces risk. EARs have units of excess cases per 10 000 person-years per Sv and are positive when radiation increases tumor risk.

The log of the background rates was modeled as a sex-specific linear function of log-attained age with additional effects for time period and birth cohort. Categorical time period effects (1958–1971, 1972–1981, 1982–1995) were included to allow for the impact of the autopsy program. Neither city nor AHS membership had appreciable effects on the background rates. In some analyses, we excluded cases diagnosed only at autopsy. All P values are based on two-sided statistical tests.

## Results

In this study, we examined the nature of radiation effects on nervous system tumor incidence in a large, fixed cohort of atomic bomb survivors. A total of 2655 tumors was screened to identify primary nervous system, pituitary gland, and pineal gland tumors among cohort members. From this initial screening, 467 primary nervous system and pituitary gland tumors were identified. The consensus diagnoses were made on the basis of a review of pathologic specimens (including autopsy specimens) for almost 80% of the tumors, on clinical records for 10%, and solely on death certificate information for 10%.

Analyses were limited to tumors diagnosed in the tumor registry catchment area between January 1, 1958, and December 31, 1995, among persons in Hiroshima or Nagasaki at the time of the atomic bombings for whom DS86 dose estimates can be computed. With the use of these criteria, 146 tumors were excluded because 27 cases occurred in cohort members outside the study period, 73 tumors occurred in cohort members not in Hiroshima or Nagasaki at the time of the bombings, 11 tumors occurred in cohort members who lived outside the tumor registry catchment area at the time of diagnosis, and 35 tumors occurred in cohort members who do not have DS86 dose estimates. Because treatment for the first primary tumor could cause a subsequent tumor and because close medical surveillance might increase the chance of diagnosing a subsequent tumor, we excluded 58 nervous system tumors diagnosed after one or more earlier primary tumors. Seven of the exclusions involved simultaneous nervous system tumors, including three people with two primary meningiomas, one person with two schwannomas, two people with a meningioma (taken as the first primary) and a schwannoma, and one person with both an astrocytoma (taken as the first primary) and a cavernous hemangioma.

After these restrictions and exclusions, 228 first primary tumors of the brain or other parts of the nervous system, 35 tumors of the pituitary gland and adjacent areas (sellar region), and no tumors of the pineal gland were identified among the LSS members (Tables 1 and 2). Of the 263 tumors, 169 occurred among women and 94 occurred among men. The most frequent tumor types were meningioma (88 tumors) and schwannoma (55 tumors). There were also 43 gliomas (including astrocytomas), 15 other nervous system tumors of known type, 27 other nervous system tumors for which the available information was not adequate to histologically classify the tumor, and 35 pituitary gland tumors. The meningiomas were largely calvarial (69 cases, 78%). There were only three malignant meningiomas, all of which were calvarial. More than half the schwannomas were cranial (33 cases, 60%), generally occurring in the acoustic nerve (27 cases). The majority of gliomas were glioblastomatous (24 cases, 56%) or astrocytic tumors (14 cases, 33%). Tumors of the sellar region included 34 pituitary adenomas and one craniopharyngioma.

Many of the tumors diagnosed before 1978 were identified through the RERF autopsy program, which was active from the late 1950s through 1977. Indeed, 55% (70/128) of the tumors were diagnosed on the basis of autopsies before 1978 compared with only 4% (6/135) of the tumors diagnosed after 1977. Approximately 20% of the malignant tumors and 35% of the benign tumors were diagnosed at autopsy. In addition, 36% of the meningiomas and 49% of the pituitary gland tumors were diagnosed at autopsy. Variation in the proportion of autopsy-detected tumors with radiation dose was not statistically significant (P = .3). The number of clinically unapparent tumors was greatly reduced after the end of the autopsy program.

We investigated the adequacy of the linear dose–response model by considering various alternative descriptions of the dose response. Fig. 1 presents linear (solid line), nonparametric (points), and smoothed nonparametric (thick dashed line) (35) dose–response functions for schwannoma and for nervous system tumors other than schwannoma. The thin dashed lines indicate the uncertainty (± one standard error) in the nonparametric smoothed dose response. There is very good agreement between the linear fit and the smoothed data for nervous system tumors other than schwannoma, suggesting that the linear dose– response model describes these data quite well. For schwannoma, the relatively low but poorly estimated risk at very high doses tended to reduce the apparent linear low-dose slope and had an even stronger influence on the smoothed dose–response function. More formally, consideration of linear quadratic dose–response models provided little evidence of a statistically significant lack of fit of the linear dose–response model for schwannoma (P = .09) and no indication of a significant lack of fit for other nervous system (P>.5) tumors. For both groups, the linear model appears to provide a good description of the low-dose (e.g., <1 Sv) risks. For nervous system tumors as a group (P = .01) and for schwannoma (P<.001), the dose response was statistically significant when the analyses were carried out using only cohort members with doses of less than 1 Sv. In neither of these groups, nor for nervous system tumors other than schwannoma, was there any evidence of statistically significant differences in the slope over this low-dose range and that over the full-dose range. Taken together, these findings suggest that the LSS data are consistent with a linear dose response for doses ranging from zero to two or more Sv.

With only four excess pituitary tumors, it is not possible to make meaningful inferences about variations in the excess pituitary tumor risks with age at exposure or time. Evaluation of the effect of sex also is limited by the number of cases. However, the estimated ERRSv for women was considerably greater than that for males (P = .13). For women, the pituitary ERRSv estimate was 2.0 (95% CI = –0.1 to 8.9). For men, the best estimate of the ERRSv was negative, resulting from a smaller-than-expected number of cases in the highest dose group. However, this apparent reduced risk is not statistically significant (P>.5, with an upper 95% confidence bound of 2.2).

Because inclusion of the autopsy-only cases could bias estimates of effect modification, particularly in terms of age and time, we also assessed effect modification excluding these cases. There were no marked changes in the parameter estimates or conclusions about the statistical significance of these effects.

## Discussion

Although our data show a statistically significant increased risk of all nervous system tumors combined, the risk was substantially higher for schwannoma than for any other nervous system tumor type. When considered individually, risks for glioma, meningioma, or other non-schwannoma nervous system tumors were not found to exhibit statistically significant radiation-related increases in risk. However, because the ERR estimates were elevated for each histologic type, analyses of all non-schwannoma nervous system tumors combined demonstrated a statistically significant radiation effect.

An enhanced risk of glioma has been seen following moderate-dose radiotherapy for tinea capitis (5,6), and large risks of gliomas have been reported following very high radiation doses, e.g., after prophylactic radiation therapy in childhood for acute lymphoblastic leukemia (11,12) or after radiation therapy for pituitary tumors (36). In some studies, excess relative risks were greater for schwannoma than for other nervous system tumors (5,6,12). Our results are generally consistent with these observations.

To date, most reports of radiation-associated nervous system tumors have been based on childhood exposures. The LSS is unique in that it permits direct assessment of how radiation effects on nervous system tumor incidence vary over a large range of ages at exposure. This study has several additional strengths. LSS cohort members have individually estimated organ doses. Tumor incidence could be ascertained on the basis of data from population-based tumor and tissue registries, with final pathology review carried out on the basis of consistently applied, modern diagnostic criteria. The fact that 20% of the benign tumors in this study were identified at autopsy is also a potential strength of the study because inclusion of these cases increases the ability of the study to characterize radiation effects. Although the use of these cases raises concerns about potential biases in the radiation risk estimates resulting from dose-related ascertainment rate differences, our analyses suggested that after allowing for temporal variation in autopsy rates, there are no indications of bias in the radiation risk estimates.

Our results for nervous system tumors other than schwannoma and for meningioma considered separately are consistent with a marked decrease in the ERRSv with increasing age at exposure, but the effect is not statistically significant, possibly because of the relatively small number of radiation-associated cases in this population. The lack of a statistically significant attained-age effect on the radiation risks for nervous system tumor incidence may also reflect these limitations of the data. However, the age-group-specific point estimates of the ERRSv do not suggest a consistent pattern of change with increasing age, and statistically significant attained-age effects have not been reported in other studies.

Although the evidence for sex differences in excess relative risks was not strong, men generally had higher excess relative risks than women, even for meningioma, for which background rates are considerably greater for women than for men. In one population-based incidence study, meningioma rates increase with age, are about two times higher among women than among men, and typically account for about 20% of intracranial tumors (37). In the LSS data, background meningioma incidence rates increased with age, women had about three times the incidence of men, and meningioma comprised almost 50% of the nervous system tumors, mostly because of the large number of meningiomas detected at autopsy. The background sex difference was reduced, but not eliminated, when analyses were restricted to meningiomas not discovered at autopsy. Although the sex difference in the ERRSv for meningioma in this cohort was not statistically significant (P = .4), the point estimate of the sex ratio (female-to-male ratio = 0.3) of the excess relative risks is consistent with observations from another study (38), in which it was noted that men have somewhat greater age-specific rates of meningioma than women.

The mechanisms of tumorigenesis in the human brain are believed to be different than those in other tissues and organs because the brain is well protected, not exposed to many exogenous agents, and only glial cells (probably astrocytes) proliferate after puberty (39). Inskip et al. (3) have suggested that traumatic injury to the brain increases cell proliferation or breakdown of the blood–brain barrier, thus increasing the risk of brain tumor development.

This study indicates that relatively low-dose radiation exposure plays a role in the etiology of nervous system and pituitary tumors, with a weak indication of somewhat higher relative risks for those exposed during childhood. Although most of the tumors were benign, radiation-induced meningiomas reportedly have more atypical or anaplastic histology than spontaneous meningiomas and have a high rate of recurrence (4143). Among the LSS cases, there were three anaplastic and one atypical meningioma (two in cohort members with zero dose and two in cohort members with doses of about 1 Sv), making it difficult to draw any firm conclusion regarding radiation-induced meningioma.

Table 1.

Selected characteristics of atomic bomb survivors in the Life Span Study: cancer incidence follow-up, 1958–1995

No. of subjects with nervous system tumors* No. of subjects with pituitary tumors* Total No. of subjects Person-years†
*First primary tumors only. — = not applicable.
‡The clinical cohort is a subset of the Life Span Study whose members were invited to participate in biennial medical examinations at the Radiation Effects Research Foundation.
§Weighted doses were used to allow for greater biologic effectiveness of neutron doses. Individual weighted brain doses were calculated as the sum of the γ-ray dose plus 10 times the neutron dose using the DS86 system (24,25) and are expressed in sieverts (Sv).
Sex
Female 150 19 47 755 1 244 140
Male 78 16 32 405 745 157
City of exposure
Hiroshima 171 28 54 153 1 376 381
Nagasaki 57 26 007 612 916
Clinical cohort membership‡
Yes 68 14 325 366 500
No 160 28 65 835 1 622 797
Age at exposure, y
0–9 36 17 681 483 820
10–19 49 11 17 180 491 553
20–39 68 22 369 645 557
≥40 75 12 22 930 358 367
Attained age, y
<35 — 361 520
35–49 44 — 529 660
50–64 80 — 599 261
65–79 73 11 — 400 106
≥80 — 98 750
Weighted brain dose, Sv§
<0.005 86 32 559 805 018
0.005–0.099 70 19 30 275 755 742
0.10–0.49 37 11 237 280 956
0.50–0.99 15 3 278 80 761
≥1.00 20 2 811 66 820
Total 228 35 80 160 1 989 297
No. of subjects with nervous system tumors* No. of subjects with pituitary tumors* Total No. of subjects Person-years†
*First primary tumors only. — = not applicable.
‡The clinical cohort is a subset of the Life Span Study whose members were invited to participate in biennial medical examinations at the Radiation Effects Research Foundation.
§Weighted doses were used to allow for greater biologic effectiveness of neutron doses. Individual weighted brain doses were calculated as the sum of the γ-ray dose plus 10 times the neutron dose using the DS86 system (24,25) and are expressed in sieverts (Sv).
Sex
Female 150 19 47 755 1 244 140
Male 78 16 32 405 745 157
City of exposure
Hiroshima 171 28 54 153 1 376 381
Nagasaki 57 26 007 612 916
Clinical cohort membership‡
Yes 68 14 325 366 500
No 160 28 65 835 1 622 797
Age at exposure, y
0–9 36 17 681 483 820
10–19 49 11 17 180 491 553
20–39 68 22 369 645 557
≥40 75 12 22 930 358 367
Attained age, y
<35 — 361 520
35–49 44 — 529 660
50–64 80 — 599 261
65–79 73 11 — 400 106
≥80 — 98 750
Weighted brain dose, Sv§
<0.005 86 32 559 805 018
0.005–0.099 70 19 30 275 755 742
0.10–0.49 37 11 237 280 956
0.50–0.99 15 3 278 80 761
≥1.00 20 2 811 66 820
Total 228 35 80 160 1 989 297
Table 2.

Sex-specific crude incidence rates* by dose group for various tumor types in Life Span Study subjects

Weighted brain dose (Sv)†
<0.005 0.005–0.099 0.10–0.49 0.50–0.99 ≥1.00 Total No. of cases No. of autopsy cases‡
*Cases per 100,000 person-years. NOS = not otherwise specified.
†Weighted doses were used to allow for greater biologic effectiveness of neutron doses. Individual weighted brain doses were calculated as the sum of the γ-ray dose plus 10 times the neutron dose using the DS86 system (24,25) and are expressed in sieverts (Sv).
‡Cases identified only at autopsy.
Male
Nervous system
Glioma 2.9 1.8 2.0 3.4 10.9 20
Meningioma 1.6 1.8 1.0 3.4 7.2 14
Schwannoma 2.3 0.7 4.0 13.6 21.7 23
Other and NOS 2.6 2.5 2.0 3.4 10.9 21
Pituitary 1.0 3.5 1.0 6.8 16 11
Person-years§ 306 178 281 956 100 016 29 378 27 629 745 157
Female
Nervous system
Glioma 2.2 1.7 1.7 1.9 23
Meningioma 6.0 5.1 7.2 7.8 7.7 74 28
Schwannoma 2.2 1.7 3.9 5.8 7.7 32
Other and NOS 1.0 2.3 2.8 21
Pituitary 0.6 1.9 2.2 7.7 19
Person-years§ 498 840 473 786 180 940 51 383 39 191 1 244 140
Weighted brain dose (Sv)†
<0.005 0.005–0.099 0.10–0.49 0.50–0.99 ≥1.00 Total No. of cases No. of autopsy cases‡
*Cases per 100,000 person-years. NOS = not otherwise specified.
†Weighted doses were used to allow for greater biologic effectiveness of neutron doses. Individual weighted brain doses were calculated as the sum of the γ-ray dose plus 10 times the neutron dose using the DS86 system (24,25) and are expressed in sieverts (Sv).
‡Cases identified only at autopsy.
Male
Nervous system
Glioma 2.9 1.8 2.0 3.4 10.9 20
Meningioma 1.6 1.8 1.0 3.4 7.2 14
Schwannoma 2.3 0.7 4.0 13.6 21.7 23
Other and NOS 2.6 2.5 2.0 3.4 10.9 21
Pituitary 1.0 3.5 1.0 6.8 16 11
Person-years§ 306 178 281 956 100 016 29 378 27 629 745 157
Female
Nervous system
Glioma 2.2 1.7 1.7 1.9 23
Meningioma 6.0 5.1 7.2 7.8 7.7 74 28
Schwannoma 2.2 1.7 3.9 5.8 7.7 32
Other and NOS 1.0 2.3 2.8 21
Pituitary 0.6 1.9 2.2 7.7 19
Person-years§ 498 840 473 786 180 940 51 383 39 191 1 244 140
Table 3.

Fitted linear excess relative risk estimates by tumor type for all cases (autopsy and nonautopsy) and excluding autopsy-only cases* in Life Span Study subjects

Nervous system Glioma and astrocytoma Meningioma Schwannoma Other nervous system† Pituitary
*Maximum likelihood estimates from a linear dose–response model with adjustment for age, birth corhort, period, and sex effects on the background rates; CI = confidence interval; RR = relative risk; Sv = sievert.
†This category includes 27 tumors of unknown histology and 15 of known histology.
All cases
Excess RR per Sv (95% CI) 1.2 (0.6 to 2.1) 0.56 (−0.2 to 2.0)< 0.64 (−0.01 to 1.8) 4.5 (1.9 to 9.2) 0.51 (<−0.2 to 2.2) 0.98 (<−0.2 to 3.5)
No. of cases 228 43 88 55 42 35
No. of excess cases 31.7 3.1 6.8 20.3 2.8 3.7
Excluding autopsy-only cases
Excess RR per Sv (95% CI) 1.4 (0.6 to 2.5) 0.60 (<−0.2 to 2.4) 0.82 (−0.1 to 2.6)0 4.6 (1.5 to 10). 0.75 (<−0.2 to 2.8) 01.5 (<−0.2 to 6.7)
No. of cases 169 34 56 42 37 18
No. of excess cases 26.9 2.6 5.9 15.7 3.6 3.1
Nervous system Glioma and astrocytoma Meningioma Schwannoma Other nervous system† Pituitary
*Maximum likelihood estimates from a linear dose–response model with adjustment for age, birth corhort, period, and sex effects on the background rates; CI = confidence interval; RR = relative risk; Sv = sievert.
†This category includes 27 tumors of unknown histology and 15 of known histology.
All cases
Excess RR per Sv (95% CI) 1.2 (0.6 to 2.1) 0.56 (−0.2 to 2.0)< 0.64 (−0.01 to 1.8) 4.5 (1.9 to 9.2) 0.51 (<−0.2 to 2.2) 0.98 (<−0.2 to 3.5)
No. of cases 228 43 88 55 42 35
No. of excess cases 31.7 3.1 6.8 20.3 2.8 3.7
Excluding autopsy-only cases
Excess RR per Sv (95% CI) 1.4 (0.6 to 2.5) 0.60 (<−0.2 to 2.4) 0.82 (−0.1 to 2.6)0 4.6 (1.5 to 10). 0.75 (<−0.2 to 2.8) 01.5 (<−0.2 to 6.7)
No. of cases 169 34 56 42 37 18
No. of excess cases 26.9 2.6 5.9 15.7 3.6 3.1
Table 4.

Parameter estimates (excess relative risk [RR] and 95% confidence intervals [CIs]) and hypothesis tests for the effect modification* in Life Span Study subjects

Schwannoma Nervous system tumors excluding schwannoma Meningioma
*Maximum likelihood estimates in an excess relative risk model with adjustment for city, sex, age, birth cohort, and period effects on background rates. Two-sided P values and confidence intervals are based on likelihood ratio tests (33,34). CI = confidence interval.
†Two-sided P value for testing the hypothesis that the excess RR exhibits a log-linear trend with age at exposure. The excess RR is given for a reference age at exposure of 30. The effect of interest is expressed as the percentage change in the excess RR per decade increase in age at exposure. Thus, for schwannoma, the estimated risks for people exposed at ages 20 and 40 are 4.7 (= 4.2 × [1 + 0.12]) and 3.7 (= 4.2 × [1 − 0.12]), respectively.
‡Two-sided P value for testing the hypothesis that the excess RR varies with age at exposure. The parameters are excess RR per sievert estimates for the indicated age at exposure groups.
§Two-sided P value for testing the hypothesis that the excess RR exhibits a log-linear trend with attained age (age for diagnosis). The excess RR is given for a reference age of 60. The effect of interest is expressed as the percentage change in the excess RR per decade increase in attained age. Thus, for schwannoma, the estimated risks for ages 50 and 70 (assuming exposure prior to the age of interest) are 4.9 (= 3.6 × [1 + 0.35]) and 2.3 (= 3.6 × [1 − 0.35]), respectively.
∥Two-sided P value for testing the hypothesis that the excess RR varies with attained age. The parameters are excess RR per sievert estimates for the indicated age groups.
Sex
Male 8.0 (2.7 to 21) 1.4 (0.4 to 3.3) 1.6 (−0.04 to 7.1)
Female 2.3 (0.3 to 7.0) 0.1 (−0.2 to 0.9) 0.4 (−0.2 to 1.7)
P for difference .12 .05 .4
Age at exposure, y†
Excess RR per sievert for survivors exposed at 30 years of age 4.2 (0.9 to 8.9) 0.15 (0.0 to 1.0) 0.5 (<0 to 1.8)
Change per decade, % −12 (−60 to 32) −62 (−95 to 4) −38 (−95 to 60)
Ptrend >.5 .06 .3
<20 6.0 (2.1 to 14) 1.2 (0.3 to 2.9) 1.3 (0.01 to 4.5)
20–39 2.6 (<−0.2 to 10) 0.3 (<−0.2 to 1.6) 0.5 (−0.05 to 2.8)
≥40 3.3 (0.33 to 11) 0.1 (<−0.2 to 1.2) 0.3 (<−0.1 to 2.0)
P for heterogeneity‡ .3 .25 >.5
Attained age, y§
Excess RR per sievert at age 60 3.6 (0.8 to 8.2) 0.6 (0.08 to 1.3) 0.6 (<0 to 2.3)
Change per decade, % −35 (−74 to 9) −28 (−95 to 100) −54 (−95 to 130)
Ptrend .11 .5 .12
<50 8.4 (2.7 to 22) 0.6 (<−0.1 to 1.3) 2.0 (−0.1 to 12)
50–69 3.0 (0.4 to 8.7) 0.6 (−0.2 to 1.8) 0.5 (<−0.1 to 2.2)
≥70 3.0 (0.2 to 10) 0.8 (−0.1 to 2.7) 0.7 (<−0.1 to 2.4)
P for heterogeneity∥ .3 .5 >.5
Schwannoma Nervous system tumors excluding schwannoma Meningioma
*Maximum likelihood estimates in an excess relative risk model with adjustment for city, sex, age, birth cohort, and period effects on background rates. Two-sided P values and confidence intervals are based on likelihood ratio tests (33,34). CI = confidence interval.
†Two-sided P value for testing the hypothesis that the excess RR exhibits a log-linear trend with age at exposure. The excess RR is given for a reference age at exposure of 30. The effect of interest is expressed as the percentage change in the excess RR per decade increase in age at exposure. Thus, for schwannoma, the estimated risks for people exposed at ages 20 and 40 are 4.7 (= 4.2 × [1 + 0.12]) and 3.7 (= 4.2 × [1 − 0.12]), respectively.
‡Two-sided P value for testing the hypothesis that the excess RR varies with age at exposure. The parameters are excess RR per sievert estimates for the indicated age at exposure groups.
§Two-sided P value for testing the hypothesis that the excess RR exhibits a log-linear trend with attained age (age for diagnosis). The excess RR is given for a reference age of 60. The effect of interest is expressed as the percentage change in the excess RR per decade increase in attained age. Thus, for schwannoma, the estimated risks for ages 50 and 70 (assuming exposure prior to the age of interest) are 4.9 (= 3.6 × [1 + 0.35]) and 2.3 (= 3.6 × [1 − 0.35]), respectively.
∥Two-sided P value for testing the hypothesis that the excess RR varies with attained age. The parameters are excess RR per sievert estimates for the indicated age groups.
Sex
Male 8.0 (2.7 to 21) 1.4 (0.4 to 3.3) 1.6 (−0.04 to 7.1)
Female 2.3 (0.3 to 7.0) 0.1 (−0.2 to 0.9) 0.4 (−0.2 to 1.7)
P for difference .12 .05 .4
Age at exposure, y†
Excess RR per sievert for survivors exposed at 30 years of age 4.2 (0.9 to 8.9) 0.15 (0.0 to 1.0) 0.5 (<0 to 1.8)
Change per decade, % −12 (−60 to 32) −62 (−95 to 4) −38 (−95 to 60)
Ptrend >.5 .06 .3
<20 6.0 (2.1 to 14) 1.2 (0.3 to 2.9) 1.3 (0.01 to 4.5)
20–39 2.6 (<−0.2 to 10) 0.3 (<−0.2 to 1.6) 0.5 (−0.05 to 2.8)
≥40 3.3 (0.33 to 11) 0.1 (<−0.2 to 1.2) 0.3 (<−0.1 to 2.0)
P for heterogeneity‡ .3 .25 >.5
Attained age, y§
Excess RR per sievert at age 60 3.6 (0.8 to 8.2) 0.6 (0.08 to 1.3) 0.6 (<0 to 2.3)
Change per decade, % −35 (−74 to 9) −28 (−95 to 100) −54 (−95 to 130)
Ptrend .11 .5 .12
<50 8.4 (2.7 to 22) 0.6 (<−0.1 to 1.3) 2.0 (−0.1 to 12)
50–69 3.0 (0.4 to 8.7) 0.6 (−0.2 to 1.8) 0.5 (<−0.1 to 2.2)
≥70 3.0 (0.2 to 10) 0.8 (−0.1 to 2.7) 0.7 (<−0.1 to 2.4)
P for heterogeneity∥ .3 .5 >.5
Table 5.

Excess absolute rate estimates by nervous system tumor type in subjects from the Life Span Study

Tumor type Excess cases per 10 000 person-year Sv* (95% CI) Attained-age Ptrend (% increase per decade)
*In age-dependent models, these estimates are roughly equal to the (sex-averaged) excess absolute rate estimates at age 60. Sv = sievert; CI = confidence interval.
Schwannoma 0.67 (0.3 to 1.1) >.5 (3)
Nervous system, excluding schwannoma 0.28 (−0.03 to 0.7) .3 (30)
Meningioma 0.14 (0.00 to 0.45) .5 (20)
Tumor type Excess cases per 10 000 person-year Sv* (95% CI) Attained-age Ptrend (% increase per decade)
*In age-dependent models, these estimates are roughly equal to the (sex-averaged) excess absolute rate estimates at age 60. Sv = sievert; CI = confidence interval.
Schwannoma 0.67 (0.3 to 1.1) >.5 (3)
Nervous system, excluding schwannoma 0.28 (−0.03 to 0.7) .3 (30)
Meningioma 0.14 (0.00 to 0.45) .5 (20)
Fig. 1.

Nervous system tumor dose–response functions. The plots present smoothed and nonparametric estimates of the dose–response functions for schwannoma and other nervous system tumors. The points are nonparametric estimates of the excess relative risk (ERR) in various dose categories. The solid lines show the fitted linear dose response (sex-averaged for schwannoma). The thick dashed line is a simple smoothed version of the nonparametric estimates (32). The thin dashed lines are one-standard-error bounds for the nonparametric smoothed curves.

Fig. 1.

Nervous system tumor dose–response functions. The plots present smoothed and nonparametric estimates of the dose–response functions for schwannoma and other nervous system tumors. The points are nonparametric estimates of the excess relative risk (ERR) in various dose categories. The solid lines show the fitted linear dose response (sex-averaged for schwannoma). The thick dashed line is a simple smoothed version of the nonparametric estimates (32). The thin dashed lines are one-standard-error bounds for the nonparametric smoothed curves.

The operation of the Hiroshima and Nagasaki Tumor Registries is supported in part by contract numbers NO1-CP-21028 and NO1-CP-71015 from the National Cancer Institute, National Institutes of Health, Department of Health and Human Services. We thank S. Funamoto for extensive help in organizing and managing the data for this study. The Radiation Effects Research Foundation is a private nonprofit foundation funded equally by the Japanese Ministry of Health, Labor and Welfare and the U.S. Department of Energy through the National Academy of Sciences.
Present address: M. Kishikawa, Department of Morphology, Institute for Developmental Research, Aichi Human Service Center, Aichi, Japan.
Present address: M. Tokunaga, Kodama Hospital, Kagoshima, Japan.

## References

1
Kleihues P, Cavenee W. Pathology and genetics of tumours of the nervous system. 2nd ed. Lyon (France): IARC Press; 2000.
2
Prados MD, Berger MS, Wilson CB. Primary central nervous system tumors: advances in knowledge and treatment.
CA Cancer J Clin

1998
;
48
:
331
–60.
3
Inskip PD, Linet MS, Heineman EF. Etiology of brain tumors in adults.
Epidemiol Rev

1995
;
17
:
382
–414.
4
Davis FG, Preston-Martin S. Epidemiology: incidence and survival. In: Bigner DD, McLendon RE, Bruner JM, editors. Russell and Rubinstein’s pathology of the nervous system. Vol 1. 6th ed. New York (NY): Oxford University Press; 1998. p. 5–45.
5
Ron E, Modan B, Boice JD Jr, Alfandary E, Stovall M, Chetrit A, et al. Tumors of the brain and nervous system after radiotherapy in childhood.
N Engl J Med

1988
;
319
:
1033
–9.
6
Shore RE, Albert RE, Pasternack BS. Follow-up study of patients treated by X-ray epilation for Tinea capitis; resurvey of post-treatment illness and mortality experience.
Arch Environ Health

1976
;
31
:
21
–8.
7
Karlsson P, Holmberg E, Lundell M, Mattsson A, Holm LE, Wallgren A. Intracranial tumors after exposure to ionizing radiation during infancy: a pooled analysis of two Swedish cohorts of 28,008 infants with skin hemangioma.

1998
;
150
:
357
–64.
8
Little MP, de Vathaire F, Shamsaldin A, Oberlin O, Campbell S, Grimaud E, et al. Risks of brain tumour following treatment for cancer in childhood: modification by genetic factors, radiotherapy and chemotherapy.
Int J Cancer

1998
;
78
:
269
–75.
9
Sznajder L, Abrahams C, Parry DM, Gierlowski TC, Shore-Freedman E, Schneider AB. Multiple schwannomas and meningiomas associated with irradiation in childhood.
Arch Intern Med

1996
;
156
:
1873
–8.
10
Hawkins MM, Draper GJ, Kingston JE. Incidence of second primary tumours among childhood cancer survivors.
Br J Cancer

1987
;
56
:
339
–47.
11
Neglia JP, Meadows AT, Robison LL, Kim TH, Newton WA, Ruymann FB, et al. Second neoplasms after acute lymphoblastic leukemia in childhood.
N Engl J Med

1991
;
325
:
1330
–6.
12
Walter AW, Hancock ML, Pui CH, Hudson MM, Ochs JS, Rivera GK, et al. Secondary brain tumors in children treated for acute lymphoblastic leukemia at St Jude Children’s Research Hospital.
J Clin Oncol

1998
;
16
:
3761
–7.
13
Preston-Martin S, White SC. Brain and salivary gland tumors related to prior dental radiography: implications for current practice.
J Am Dent Assoc

1990
;
120
:
151
–8.
14
Ryan P, Lee MW, North B, McMichael AJ. Amalgam fillings, diagnostic dental x-rays and tumours of the brain and meninges.
Eur J Cancer B Oral Oncol

1992
;
28B
:
91
–5.
15
McCredie M, Maisonneuve P, Boyle P. Perinatal and early postnatal risk factors for malignant brain tumours in New South Wales children.
Int J Cancer

1994
;
56
:
11
–5.
16
Preston-Martin S, Pogoda JM, Schlehofer B, Blettner M, Howe GR, Ryan P, et al. An international case-control study of adult glioma and meningioma: the role of head trauma.
Int J Epidemiol

1998
;
27
:
579
–86.
17
Cardis E, Gilbert ES, Carpenter L, Howe G, Kato I, Armstrong BK, et al. Effects of low doses and low dose rates of external ionizing radiation: cancer mortality among nuclear industry workers in three countries.

1995
;
142
:
117
–32.
18
Muirhead CR, Goodill AA, Haylock RG, Vokes J, Little MP, Jackson DA, et al. Occupational radiation exposure and mortality: second analysis of the National Registry for Radiation Workers.

1999
;
19
:
3
–26.
19
Doody MM, Mandel JS, Lubin JH, Boice JD Jr. Mortality among United States radiologic technologists, 1926–90.
Cancer Causes Control

1998
;
9
:
67
–75.
20
Seyama S, Ishimaru T, Iijima S, Mori K. Primary intracranial tumors among atomic bomb survivors and controls, Hiroshima and Nagasaki, 1961–75.
J Hiroshima Med Assoc

1981
;
34
:
1056
–65.
21
Thompson DE, Mabuchi K, Ron E, Soda M, Tokunaga M, Ochikubo S, et al. Cancer incidence in atomic bomb survivors. Part II: Solid tumors, 1958–1987.

1994
;
137
(2 Suppl):
S17
–67.
22
Shintani T, Hayakawa N, Hoshi M, Sumida M, Kurisu K, Oki S, et al. High incidence of meningioma among Hiroshima atomic bomb survivors.

1999
;
40
:
49
–57.
23
Sadamori N, Shibata S, Mine M, Miyazaki H, Miyake H, Kurihara M, et al. Incidence of intracranial meningiomas in Nagasaki atomic-bomb survivors.
Int J Cancer

1996
;
67
:
318
–22.
24
Radiation Effects Research Foundation (RERF). U.S.–Japan joint reassessment of atomic bomb radiation dosimetry in Hiroshima and Nagasaki: final report. Hiroshima (Japan): RERF; 1987.
25
Fujita S. Versions of DS86.
RERF Update

1989
;
1
:
3
.
26
Pierce DA, Stram DO, Vaeth M. Allowing for random errors in radiation dose estimates for the atomic bomb survivor data.

1990
;
123
:
275
–84.
27
Mabuchi K, Soda M, Ron E, Tokunaga M, Ochikubo S, Sugimoto S, et al. Cancer incidence in atomic bomb survivors. Part I: Use of the tumor registries in Hiroshima and Nagasaki for incidence studies.

1994
;
137
:
S1
–16.
28
World Health Organization (WHO). International Classification of Diseases: Manual of Statistical Classification of Diseases, Injuries, and Causes of Death, 1975 Revision. Vol 1. Geneva (Switzerland): WHO; 1977.
29
Kleihues P, Burger PC, Scheithauer BW. Histolgical typing of tumours of the central nervous system. New York (NY): Springer-Verlag; 1993.
30
Wong FL, Yamada M, Sasaki H, Kodama K, Akiba S, Shimaoka K, et al. Noncancer disease incidence in the atomic-bomb survivors: 1958–1986.

1993
;
135
:
418
–30.
31
Ron E, Preston DL, Kishikawa M, Kobuke T, Iseki M, Tokuoka S, et al. Skin tumor risk among atomic-bomb survivors in Japan.
Cancer Causes Control

1998
;
9
:
393
–401.
32
Sposto R, Preston DL. Correcting for catchment area nonresidency in studies based on tumor registry data. CR1–92. Hiroshima (Japan); RERF: 1992.
33
McCullagh P, Nelder JA. Generalized linear models. London (U.K.); Chapman & Hall Ltd; 1999. p. 21–45.
34
Preston D, Lubin J, Pierce D, McConney M. Epicure users guide. Seattle (WA): Hirosoft International Corporation; 1993.
35
Pierce DA, Preston DL. Radiation-related cancer risks at low doses among atomic bomb survivors.

2000
;
154
:
178
–86.
36
Brada M, Ford D, Ashley S, Bliss JM, Crowley S, Mason M, et al. Risk of second brain tumour after conservative surgery and radiotherapy for pituitary adenoma.
BMJ

1992
;
304
:
1343
–6.
37
Longstreth WT, Dennis LK, McGuire VM, Drangsholt MT, Koepsell TD. Epidemiology of intracranial meningioma.
Cancer

1993
;
72
:
639
–48.
38
Rabin BM, Meyer JR, Berlin JW, Marymount MH, Palka PS, Russell EJ. Radiation-induced changes in the central nervous system and head and neck.

1996
;
16
:
1055
–72.
39
Bigner DD, McLendon RE, Bryner JM. Russell and Rubinstein’s pathology of tumors of the nervous system. Vols 1 and 2. 6th ed. New York (NY): Oxford University Press; 1998.
40
Preston-Martin S, Pogoda JM, Mueller BA, Lubin F, Modan B, Holly EA, et al. Results from an international case-control study of childhood brain tumors: the role of prenatal vitamin supplementation.
Environ Health Perspect

1998
;
106
Suppl 3:
887
–92.
41
Shoshan Y, Chernova O, Juen SS, Somerville RP, Israel Z, Barnett GH, et al. Radiation-induced meningioma: a distinct molecular genetic pattern?
J Neuropathol Exp Neurol

2000
;
59
:
614
–20.
42
Soffer D, Pittaluga S, Feiner M, Beller AJ. Intracranial meningiomas following low-dose irradiation to the head.
J Neurosurg

1983
;
59
:
1048
–53.
43
Rubinstein AB, Shalit MN, Cohen ML, Zandbank U, Reichenthal E. Radiation-induced cerebral meningioma: a recognizable entity.
J Neurosurg

1984
;
61
:
966
–71.
44
Heros RC, Korosue K. Radiation treatment of cerebral arteriovenous malformations.
N Engl J Med

1990
;
323
:
127
–9.
45
Mettler FA Jr, Wiest PW, Locken JA, Kelsey CA. CT scanning: patterns of use and dose.

2000
;
20
:
353
–9.
46
Brenner D, Elliston C, Hall E, Berdon W. Estimated risks of radiation-induced fatal cancer from pediatric CT.
AJR Am J Roentgenol

2001
;
176
:
289
–96.