Sex-Specific Lung Cancer Risk among Radiation Workers in the Million Person Study and TB-Fluoroscopy Patients

Abstract

Background:

The study of Japanese atomic bomb survivors, exposed briefly to radiation, finds the risk of radiation-induced lung cancer to be nearly three times greater for women than for men. Because protection standards for astronauts are based on individual lifetime risk projections, this sex-specific difference limits the time women can spend in space. Populations exposed to chronic or fractionated radiation were evaluated to learn whether similar differences exist when exposures occur gradually and over years, and not acutely.

Methods and Materials:

Five occupational cohorts within the Million Person Study of Low-Dose Health Effects (MPS) and a Canadian Fluoroscopy Cohort Study (CFCS) of tuberculosis patients who underwent frequent chest fluoroscopic examinations are evaluated. Included are male and female workers at the Mound nuclear facility, nuclear power plants, and industrial radiographers. Workers at the Mallinckrodt Chemical Works and military participants at aboveground nuclear weapons tests provide information on the risk among males. Cox proportional hazards and Poisson regression models were used to estimate sex-specific radiation risks for lung cancer and to compare any differences.

Results:

Overall, 453,746 subjects were studied: 50,679 women and 403,067 men. The mean radiation absorbed cumulative dose to the lung was 166.3 mGy with maximum values greater than >1,000 mGy. Of the 12 estimates of radiation-related risk, only one, for male industrial radiographers, showed a significant elevation (ERR 0.13; 95% CI 0.02-0.23, at 100 mGy). In contrast, the dose response for the male nuclear power plant workers was negative and nearly significant (ERR −0.06; 95% CI −0.11, 0.01, at 100 mGy). Combined, these two similar cohorts provided little evidence for a radiation effect among males (ERR 0.001; 95% CI −0.04, 0.05, at 100 mGy). There was no significant evidence for a dose response among females within any cohort. There was no difference in the sex-specific estimates of lung cancer risk.

Conclusions:

There was little evidence that chronic or fractionated exposures increased the risk of lung cancer. There were no differences in the risks of lung cancer between men and women. However, the sex-specific analyses are limited because of small numbers of women and relatively low doses, and await confirmation from the ongoing MPS study of medical radiation workers which includes 65,000 women and 65,000 men. Additional understanding will come as well from the ongoing follow-up of the Canadian Fluoroscopy Cohort Study.

Introduction

The evidence that low-LET radiation causes lung cancer comes mainly from acute high-dose rate exposures and at moderate radiation doses (UNSCEAR 2008). The Japanese atomic bomb survivor study provides evidence for an increased radiation risk of lung cancer but also indicates that the risk of radiation-induced lung cancer is nearly three times greater for women than for men (Ozasa et al. 2012; NCRP 2014; Cahoon et al. 2018). In contrast, studies of tuberculosis patients undergoing frequent chest fluoroscopic examination to monitor lung collapse therapy in Canada and in Massachusetts have failed to find an increased risk of lung cancer despite lung doses that exceed those of the Japanese atomic bomb survivors (Davis et al. 1989; Howe 1995). It has not been clear why such differences exist. One possibility is that fractionation of dose over time results in a lower risk than when the exposure is received in a short period of time. This difference in sex-specific radiation lung cancer risk is of special interest to NASA which bases its protection standards on lifetime risk estimates for individuals based on the atomic bomb survivor study. Since women have a much higher risk of lung cancer on a relative scale than men, their lifetime risk estimates per unit dose are much higher and thus their time in space for long-term missions is limited.

Several radiation cohorts within the MPS have sufficient numbers of female workers to be able to contrast the risks among males with those among females. The studies evaluated here include the Mound facility workers in Ohio (Boice et al. 2014), nuclear power plant workers (Boice et al. 2019b), and industrial radiographers (Boice et al. 2019c). These studies were analyzed in similar fashion and allowed for estimates to be made for both male and female lung cancer risk. Male workers at the Mallinckrodt Chemical Works (MCW) in Tennessee and participants at above ground nuclear weapons tests (Boice et al. 2019a) were also included to provide additional estimates of lung cancer risk among males, and because a component of the lung dose among MCW workers comes from intakes of high-LET emitters (uranium and radium) (Ellis et al. 2018; Golden et al. 2019).

Materials and Methods

Statistical Methods

For the MPS cohorts, Cox proportional hazards models were used to compute hazards ratios (HR) and corresponding 95% confidence intervals (CI) for a continuous measure of radiation dose to the lung (Cox 1972). Age was used as the underlying timescale for the hazard function. To allow for a possible latent period between radiation exposure and any effect consequent to it, doses were lagged 10 years. All Cox models were conducted using SAS/STAT software (SAS/STAT software, Version 9.4 of the SAS System for Windows, SAS Institute Inc., Cary, NC, USA). Analogous Excess Relative Risks per Gy (ERR Gy⁻¹) along with corresponding CIs were estimated using the Peanuts program in Epicure (Preston et al. 2018). All models were adjusted for year of birth and a measure of socioeconomic status which varied by cohort but was generally obtained from work histories and was either education level or pay type, such as hourly versus salary.

Further details of the analytic methods for the Canadian Fluoroscopy Cohort Study (CFCS) are available in Howe (Howe 1995). Briefly, standard Poisson regression models (Breslow 1987) were used to compute relative risks and corresponding CIs for a continuous dose to the lung with a 10-year lag applied to account for a latent period. The Epicure program (Preston et al. 2017) was used to compute linear excess relative risks. All models included adjustment for age at risk, calendar year at risk, and sex by stratification.

For all cohorts, analyses were conducted within cohort except for the NPP and IR cohorts which were examined both individually as well as combined. This could be done with minimal bias because both populations were identified from the same databases, had the same co-variates, and used the same procedures for obtaining and adjusting for socioeconomic status. Models were first stratified by sex and then combined over sex with further adjustment for male versus female. To test whether sex was an effect modifier, nested models with and without an interaction term between sex and radiation dose were compared using the likelihood ratio test (LRT).

Results and Study Descriptions

The occupational studies within the MPS and the CFCS studies are described below and in Tables 1–3. Other studies with estimates of sex-specific lung cancer risks are also briefly summarized: Massachusetts TB-fluoroscopy patients (Boice et al. 1978; Davis et al. 1989; Brenner et al. 2007); scoliosis patients (Ronckers et al. 2010); Hodgkin lymphoma patients (Gilbert et al. 2003); and workers in the Mayak (Gilbert et al. 2004, 2013; Gillies et al. 2017); INWORKS (Richardson et al. 2018), and Rocketdyne (Boice et al. 2011) studies.

Table 1.

Subject counts and descriptive measures of dose, by cohort and sex

	FEMALES					MALES
Cohort	N subjects	N cases lung cancer	Person-Years	Mean [Median] Dose, mGy	95th percentile [Maximum] Dose, mGy	N subjects	N cases lung cancer	Person-Years	Mean [Median] Dose, mGy	95th percentile [Maximum] Dose, mGy
Mallinckrodt (Golden et al. 2019)	na	na	na	na	na	2,514	157	107,920	67.3 [30.7]	273 [829]
Mound (Boice et al. 2014)	971	21	41,750	24.9 [1.6]	132 [1,460]	3,983	182	159,771	112.9 [14.7]	555 [17,478]
Atomic Veterans (Boice et a. 2019a)	na	na	na	na	na	113,806	8,027	5,350,400	6.21 [2.28]	25 [972]
Nuclear Power Plant (NPP) (Boice et al. 2019b)	4,975	57	147,896	16.2 [3.1]	71.6 [466]	140,234	3,633	4,221,255	39.1 [19.6]	147 [1,085]
Industrial Radiographers (IR) (Boice et al. 2019c)	12,934	55	210,930	2.0 [0.02]	8.3 [302]	110,580	2,061	2,427,392	12.0 [0.4]	67 [1,435]
Combined NPP and IR	17,909	112	358,826	6.0 [0.08]	34 [466]	250,814	5,694	6,648,647	27.1 [9.7]	121 [1435]
Canadian Fluoroscopy Cohort Study1 (Howe 1995)	31,917	266	883,172	1072 [622]	4115 [9640]	32,255	912	725,318	1038 [617]	4210 [9280]

¹Excludes person-years for age at risk <20 years and deaths and person-years at risk within 10 years of exposure.

Table 3.

Sex-adjusted lung cancer ERR and HR at 100 mGy by cohort, combined over sex

Cohort	Total Subjects both sexes	Total Lung cancers both sexes	Mean [Median] dose, mGy	95th percentile [Maximum] dose, mGy	Sex-adjusted ERR at 100 mGy (95%CI)	Sex-adjusted HR at 100 mGy (95%CI)	p 1
Mallinckrodt	2,514	157	67.3 [30.7]	273 [829]	−0.06 (−0.18, 0.06)	0.95 (0.81, 1.12)	na
Mound	4,954	203	95.6 [9.1]	442 [17,478]	0.01 (−0.03, 0.05)	1.01 (0.99, 1.03)	0.72
Atomic Veterans	113,806	8,027	6.21 [2.28]	25 [972]	0.08 (−0.06, 0.22)	1.04 (0.90-1.21)	na
Nuclear Power Plant (NPP)	145,209	3,690	38.3 [19.1]	145 [1,085]	−0.06 (−0.11, 0.001)	0.95 (0.90, 1.02)	0.35
Industrial Radiographers (IR)	123,556	2,116	10.9 [0.3]	61 [1,435]	0.13 (0.02, 0.23)	1.09 (1.02, 1.17)	0.43
Combined NPP and IR	268,765	5,806	25.7 [8.5]	116 [1,435]	0.01 (−0.04, 0.06)	1.01 (0.96-1.06)	0.84
Canadian Fluoroscopy Cohort Study	63,707	1,178	1055 [620]	4146 [9640]	0.00 (−0.005, 0.005)	0.96 (0.88, 1.05)	0.097

¹p-value from likelihood ratio test (LRT) for interaction between sex and dose in a Cox proportional hazards model with males and females

Mound nuclear facility (Tables 1–3)

The Mound Nuclear Facility near Dayton, OH, employed 7,270 workers during the years 1944-1971, including 4,977 workers monitored for radiation exposure (Boice et al. 2014). The mean lung dose was 132 mGy and the maximum was nearly 1.5 Gy. However, there was no evidence for an increase in lung cancer related to radiation (Tables 2–3). There was no increased risk in lung cancer among the 971 female workers (ERR at 100 mGy = −0.01; 95% CI −0.07, 0.07) nor among the 3,983 male workers (ERR at 100 mGy = 0.01; 95% CI −0.02, 0.07) (Table 2), and accordingly no difference in the sex-specific lung cancer risk estimates (LRT p=0.72).

Table 2.

Sex-specific lung cancer excess relative risks (ERR) and hazard ratios (HR) at 100 mGy, by cohort and sex

	FEMALES		MALES
Cohort	ERR at 100 mGy (95% CI)	HR at 100 mGy (95% CI)	ERR at 100 mGy (95% CI)	HR at 100 mGy (95% CI)
Mallinckrodt	na	na	−0.06 (−0.18, 0.06)	0.95 (0.81, 1.12)
Mound	−0.01 (−0.07, 0.07)	0.86 (0.48-1.55)	0.01 (−0.02, 0.04)	1.01 (0.99, 1.03)
Atomic Veterans	na	na	0.08 (−0.06, 0.22)	1.04 (0.90-1.21)
Nuclear Power Plant (NPP)	0.63 (−0.91, 2.17)	1.61 (0.81, 3.21)	−0.06 (−0.11, 0.01)	0.95 (0.89, 1.01)
Industrial Radiographers (IR)	−0.33 (−0.54, 0.12)	0.69 (0.14, 3.49)	0.13 (0.02, 0.23)	1.09 (1.02, 1.17)
Combined NPP and IR	0.20 (−0.69, 1.09)	1.17 (0.61-2.23)	0.001 (−0.04, 0.05)	1.01 (0.96-1.06)
Canadian Fluoroscopy Cohort Study	−0.007 (−0.015, 0.002)	0.97 (0.81, 1.15)	0.002 (−0.003; 0.008)	0.96 (0.87, 1.057)

Nuclear power plant workers (Tables 1–3)

Over 145,000 workers first monitored at nuclear power plants in the United States from 1957-1984 were evaluated through 2011 (Boice et al. 2019a). There was no evidence for an increase in radiation-related lung cancer, and the sex-adjusted dose response was negative and of borderline significance (ERR at 100 mGy = −0.06; 95% CI −0.11, 0.01). The mean dose to the lung was 38 mGy and ranged up to 1,085 mGy. Overall, there were 4,975 female workers and 140,234 male workers (Table 1). A radiation-related dose response was not seen for the 4,975 female workers (ERR at 100 mGy = 0.63; 95% CI −0.91, 2.17) nor among the 140,234 male workers (ERR at 100 mGy = −0.06; 95% CI −0.11, 0.01). The estimates of radiation risk were non-significantly different between females and males (LRT p=0.35).

Industrial radiographers (Tables 1–3)

Nearly 130,000 industrial radiographers in the United States who were first monitored 1939-2011 were studied (Boice et al. 2019b). The workers over the years were exposed only to external radiation and mainly iridium-192 and cobalt-60. The mean lung dose was 10.9 mGy and ranged up to 1,435 mGy. The sex-averaged lung cancer risk estimate was significantly elevated (ERR at 100 mGy = 0.13; 95% CI 0.02, 0.23). There were 12,946 female industrial radiographers and 110,610 male radiographers (Table 1). The mean dose to the lung was 2.0 mGy for females and 12.0 mGy for males. The radiation risk estimates were non-significantly negative for females (ERR at 100 mGy = −0.33; 95% CI −0.54, 0.12) and significantly positive for males (ERR at 100 mGy = 0.13; 95% CI 0.02, 0.23). While a significant dose response was observed for radiation-induced lung cancer among male radiographers, the difference in risk estimates between female and male workers was not significant (LRT p=0.43).

The nuclear power plant worker and the industrial radiography worker studies were constructed using the same methodologies. The NRC REIRS database (NRC 2018) and the Landauer dosimetry records were used to select the study cohorts and the analytical and dosimetry approaches were exactly the same, including the adjustments for socioeconomic status based on areawide educational categories. A pooled analysis of these two cohorts was conducted to increase statistical power. The pooled analysis included 17,921 females and 250,844 males (Tables 1–3). There was no significant increase in the sex-adjusted radiation risk (ERR at 100 mGy = 0.01; 95% CI −0.04, 0.06), nor among females (ERR at 100 mGy = 0.20; 95% CI −0.69, 0.09) or males (ERR at 100 mGy = 0.001; 95% CI −0.04, 0.05). The pooled analysis did not support a radiation-related risk for lung cancer nor a statistically significant difference in sex-specific radiation lung cancer risk estimates (LRT p=0.83).

Mallinckrodt Chemical Works (Tables 1–3)

The Mallinckrodt Chemical Works (MCW) study evaluated 2,514 white male workers employed 1942-1966 (Boice et al. 2018; Ellis et al. 2018, Golden et al. 2019). The workers over the years were exposed to external gamma radiation plus internal radiation from inhalation and ingestion of pitchblende dust (uranium, radium, and silica). Lung dose from occupational medical x-rays were included in the individual dose reconstructions as were ambient levels of radon and radon progeny. The mean dose to the lung was 67.3 mGy and ranged up to 829 mGy (Table 3). There was no evidence for a radiation dose response (ERR at 100 mGy = −0.06; 95% CI −0.18, 0.06) and the negative trend was nearly significant. Because only males were included, the MCW study could not evaluate sex-specific radiation risks. It did suggest that chronic radiation exposures, including that from intakes of radionuclides that provide high-LET dose to lung tissue, is not associated with a detectable increase of lung cancer risk at the radiation doses received. A negative radiation dose response for non-malignant lung disease (Golden et al. 2019) provided some additional evidence that low-dose radiation exposures to the lung experienced over a period of years may be much less carcinogenic than when the doses are received all at once and at a high level. The small numbers in this study, however, temper the strength of conclusions that might be drawn.

Nuclear weapons test participants (atomic veterans) (Tables 1–3)

The United States conducted over 200 above-ground atmospheric weapons tests during the Cold War and many involved military personnel at the Nevada Test Site (NTS) and the Pacific Proving Grounds (e.g., Bikini Atolls) (Till et al. 2014, Caldwell et al. 2016, Beck et al. 2017).

The epidemiologic study of above-ground nuclear weapons test participants (atomic veterans) included 113,806 male military personnel who participated in weapons testing between 1945 and 1962. The atomic veterans are included as a component of the MPS that provides information on the radiation risk of lung cancer among males. Vital status was obtained for nearly 97% of the veterans and over 70% had died. There were 8,512 deaths due to lung cancer. Overall, radiation doses to the lung were small, mean 6.21 mGy with a maximum of 972 mGy (Table 1). There was no statistical evidence for a radiation dose response over the very narrow range of lung doses (ERR 0.08; 95% CI −0.11, 0.01; at 100 mGy).

Canadian Fluoroscopy Cohort Study (Tables 1–3)

A lung cancer mortality study was conducted between 1950-1987 among 64,172 Canadian tuberculosis patients who received multiple chest fluoroscopic examinations during the monitoring of lung collapse therapy (Sherman et al. 1978; Howe and Yaffe 1992; Howe 1995). The current analyses are based on the previously published data (Howe 1995) using Poisson and Cox regression models. There were 31,917 female patients and 32,255 male patients (Table 1). The mean dose to the lung was 1,055 mGy and the maximum was 9,640 mGy (Table 3). Overall there was no evidence for a radiation-related increase in lung cancer among either females (ERR at 100 mGy = −0.007; 95% CI −0.015, 0.002) or males (ERR at 100 mGy = 0.002; 95% CI −0.003, 0.008) (Table 2). There was no difference between the sexes in their risk of radiation-induced lung cancer (LRT p=0.097). These data are consistent with the Massachusetts TB-fluoroscopy study which also did not find a radiation-related risk for lung cancer (Davis et al. 1989; Brenner et al. 2007).

Summary of other published studies

TB- Fluoroscopy Massachusetts Study

The Massachusetts TB-fluoroscopy study included 13,385 patients treated between 1925-1954 (Boice et al. 1978; Davis et al. 1989; Brenner et al. 2007; Little 2018). The patients received an average of 77 x-ray fluoroscopic examinations, and the mean dose to lung tissue was 840 mGy. Risks were not reported for women and men separately. Lung cancer (SMR 0.8, n = 69) was not elevated despite a range of doses to the lung that reached over 8,000 mGy; there was no evidence of a dose response. Adjustments for smoking and the amount of lung tissue at risk did not appreciably modify these observations. The authors concluded that frequent exposures to low doses of radiation over a period of several years did not increase the risk of lung cancer and that fractionated exposure appeared less effective in causing lung cancer than a single exposure of the same total dose.

U.S. Scoliosis Study

The U.S. Scoliosis Study includes 5,573 female patients monitored with frequent spinal x-rays for curvature changes during the adolescent growth spurt and before 20 years of age, 1912–1965 (Hoffman et al. 1989; Ronckers et al. 2010). The average number of x-rays was 22.9 and ranged up to 553. The average cumulative lung dose was 41 mGy and the maximum was 676 mGy. Breast cancer was significantly elevated. Lung cancer (SMR 0.8, n = 57), however, was not elevated; and the dose response was negative (ERR Gy⁻¹-1.4; 95% CI −7.1, 3.1). These results are identical to the Canadian and Massachusetts TB-fluoroscopy studies where significant positive dose-response relationships for breast cancer were observed (Boice et al. 1979) but evidence for increased lung cancer risk was absent (Davis et al. 1989; Howe 1995). These three studies of patients monitored with diagnostic radiation examinations for many years, however, are not consistent with the Japanese study of atomic bomb survivors exposed acutely to radiation where strong relationships where found for both breast and lung cancer (Preston et al. 2002; Cahoon et al. 2017). Possible explanations raised by the authors for these discrepancies with the lung cancer findings included a low-dose threshold or nonlinear radiation dose response when doses are fractionated, or failure to adequately control for the carcinogenic effects of smoking (Ronckers et al. 2010).

Rocketdyne workers

The plutonium contribution to lung dose for the Mayak workers (Gilbert et al. 2004; 2013; Gillies et al. 2017) was substantial and much larger than for the previously discussed studies of Mallinckrodt and Mound which also incorporated dose from internal intakes (see section below on Mayak nuclear facilities workers). The Rocketdyne study also incorporated intakes of radionuclides for lung dose, but again was small in comparison with the external dose contribution. The Rocketdyne (Atomics International) study involved 46,970 workers employed 1948-1999 (Boice et al. 2011). Overall, 5,801 workers were monitored for radiation activities (Boice et al. 2006). The mean dose to lung from external and internal radiation combined was 19 mSv (maximum 3,600 mSv). Overall, there was no evidence for radiation-associated risk for lung cancer. There was no evidence for a dose response among the radiation-exposed workers and the trend for lung cancer risk over categories of cumulative dose to the lung was negative but not significant. The RR at 100 mSv for lung cancer was estimated to be 1.01 (95% CI 0.89, 1.16). There was limited ability to evaluate sex-specific lung cancer radiation risk because only 8% of the population was female. Overall, the Rocketdyne study provides little evidence that fractionated exposures to the lung experienced over years is related to lung cancer risk. The study is limited by small numbers but the dose range was broad and much of the higher lung doses came from intakes of radionuclides emitting high-LET alpha particles.

Hodgkin lymphoma patients

Studies of patients with Hodgkin lymphoma, treated with chest radiotherapy (Travis et al. 2002, Dores et al. 2002, Gilbert et al. 2003), provide evidence that high-dose radiotherapy causes lung cancer but there was no evidence that this risk differed by sex. The data for lung cancer were obtained within a population-based cohort of 19,046 Hodgkin lymphoma patients diagnosed from 1965-1994. Radiation dose was determined to the specific location in the lung where the cancer developed, and tobacco use was adjusted for in the analyses. The mean dose to the lung was 36 Gy from the mantle radiotherapy received. Practically all radiotherapy doses to the lung were >5 Gy. Seventy-five percent of the patients with lung cancer within this nested case-control study were male and 25% were female. The radiation-related risk of lung cancer for the patients receiving >5 Gy was four times higher for males (RR 7.2; 95% CI 3.0, 18.6) than for females (RR 2.1; 95% CI 0.6, 11.5) but this difference was not statistically significant. This comprehensive study of lung cancer risk following radiotherapy at high doses provides no evidence for a difference in the sex-specific risk for radiation-induced lung cancer. The actuarial risk of lung cancer at 25 years after treatment was significantly higher in males (6.2%; 95% CI 5%, 7.4%) than females (3.2%; 95% CI 2.2%, 4.2%), although the radiation component was not separated from the entire cohort (Dores et al. 2002). The relevance of such very high doses to low-dose risk assessment is unclear.

INWORKS

The INWORKS occupational study consists of 308,297 nuclear workers in France, the United Kingdom, and the United States (Richardson et al. 2018). Overall there were 268,262 male workers and 40,035 female workers. The mean (and 95th percentile) cumulative lung doses were 22.8 mGy (106.3 mGy) for men and 4.8 mGy (18.8 mGy) for women. There were 5,802 lung cancer deaths. The maximum likelihood estimate of excess relative risk per Gy was 0.51 (90% CI 0.00, 1.09). A 10-year lag was applied, and the risk was sex-adjusted but sex-specific estimates were not presented. A hierarchical Bayesian analytical method was also applied to estimate ERR/Gy with slightly higher risk estimates for lung cancer than provided by the maximum likelihood approach. The individual site-specific estimates in INWORKS are difficult to interpret because very high and statistically significant risks were seen for sites not found or rarely found to be increased following radiation exposures such as cancer of the rectum, peritoneum, pleura and mesothelioma, testis, larynx, and skin (e.g. melanoma), whereas sites that are frequently increased following radiation exposure were not, such as cancers of the colon, liver, bladder, and kidney (UNSCEAR 2008). In comparison, the current study involves nearly 50% more subjects (453,746), twice as many men (403,067) and 25% more women (50,679). The current study has 3 times as many lung cancer deaths (16,049), and the mean cumulative lung dose (166.3 mGy) was seven times higher overall, and similar if the TB-fluoroscopy patients are excluded (21.2 mGy). Excluding the TB-fluoroscopy patients, the mean (and 95th percentile) cumulative lung doses were 21.9 mGy (101.1 mGy) for men and 6.9 mGy (36.9 mGy) for women. For all cohorts combined the mean cumulative lung doses were 102.0 mGy for men and 677.2 mGy for women.

Mayak nuclear facility workers

A mortality study of lung cancer was conducted among 14,621 Mayak workers who were hired between 1948-1982 (Gilbert et al. 2004, 2013; Gillies et al. 2017). Significant differences between the female and male ERR/Gy for lung cancer for intakes from plutonium were reported with females being 4 times higher for mortality (Gillies et al. 2017). For internal dose, the excess absolute risk (EAR) for females was less than half that for males (Gilbert et al. 2004). The lung cancer risk estimates from gamma exposure where, on the other hand, 2-3 times lower than comparable risks in the Japanese atomic bomb survivor study (Preston et al. 2017). Interpretation, however, is challenging for several reasons such as the differences in smoking prevalence: 74% of all male workers smoked and only 5% of all females smoked and ERR/Gy risks varied significantly by smoking status which affects the scale of comparison (ERR or EAR) (Gillies et al. 2017); the overall lost-to-follow-up rate was high at 22% (Preston et al. 2017); it is challenging to separate out a gamma ray component to risk for workers who also had significant plutonium intakes; and the estimates of plutonium dose were based on incomplete coverage of only 42% of the workers employed at the plutonium facilities (Gillies et al. 2017). Further, “A particularly important limitation of our lung cancer risk estimates relates to intrinsic difficulties in estimating doses from plutonium exposure. Despite extensive efforts by Russian and non-Russian dosimetrists, these doses are subject to many uncertainties including especially those resulting from imprecision in the urine measurements. Only 33% of the 6,540 workers with estimates of positive plutonium doses had more than two measurements, although this percentage was higher (51%) for workers with estimated plutonium doses exceeding 0.2 Gy. For most workers, these measurements were taken many years after the exposure occurred, since routine plutonium monitoring did not begin until about 1970. Such measurement errors, often referred to as classical errors, are known to bias the dose-response coefficient toward zero and might also distort the shape of the dose response” (Gilbert et al. 2013). In a comparable but much less powerful study of Sellafield workers, there was no difference in the male and female risk estimates (Gillies et al. 2017).

When the analysis of Mayak workers was restricted to workers who received <200 mGy, there was no difference in the male and female estimates of risk (Gillies et al. 2017). The mean dose to gamma-rays was 455 mGy (maximum 7,595 mGy). The mean dose from plutonium-239 was 129-176 mGy (maximum 16,500-19,700 mGy). The differences related to the plutonium doses depends on whether the biokinetic models assumed either a slow clearance or a fast clearance from the lung, that is related to solubility. When the ERR analyses were allowed to depend on smoking status, the plutonium ERR for non-smokers was significantly greater than that for smokers which suggested a sub-multiplicative relationship. While the ERR for non-smokers was estimated to be four times that for smokers, the female/male ratio was estimated to be 1.0 (95% CI 0.3, 3.4). That is, when the plutonium ERR/Gy was allowed to depend on smoking, the ratio of female and male risk estimates was reduced from 3.3 to 1.0. The study does raise the issue as to the most appropriate scale for comparisons: the excess relative risk which depends on the background rates or the absolute excess risk which depends less so. This is a remarkably comprehensive and important study regarding the risks of high-dose plutonium and high-dose gamma exposures, but there remains some uncertainty in the estimated radiation risks and in the risks of sex-specific radiation-induced lung cancer.

Japanese atomic bomb survivor study

The study of Japanese atomic bomb survivors provides strong evidence that acute exposures to ionizing radiation can lead to an increased risk of lung cancer. Females are reported to be at 2-3 times higher risk of radiation-induced lung cancer than males (Furukawa et al. 2010; Ozasa et al. 2012; Cahoon et al. 2017). The lung cancer incidence analyses showed that lung cancer risk strongly depended on smoking patterns, with radiation risk highest for those smoking 10 cigarettes per day and approaching no risk at 20 cigarettes per day. Most men smoked whereas most women did not. The female to male incidence ratio of excess risk decreased with increasing radiation dose. Application of these risk estimates to American workers, and to the NASA astronaut corps (NCRP 2014), is somewhat problematic because of the different population characteristics and circumstances: the Japanese population was exposed briefly in 1945 to the atomic bombings in Hiroshima and Nagasaki and lived in a war-torn country for many years after (NCRP 2014; Boice et al. 2017).

The most recent analyses evaluated cancer incidence from 1958-2009 (Cahoon et al. 2017). There were 105,444 survivors evaluated of whom a little over 86,000 had estimates of radiation dose. For non-smokers the ERR/Gy estimate for lung cancer was 0.81 (95% CI 0.15, 1.18) and the female to male ratio was 2.83. The estimate of risk was significantly high for low-to-moderate smokers and no radiation risk was apparent for heavy smokers. Smoking information was available for 64,465 subjects (81%). For those with information on smoking, 86% of men and 18% of women indicated that they had ever smoked. The authors mentioned that the results should be interpreted in the context of several limitations, e.g., smoking history was unknown for approximately 40% of the lung cancer cases and approximately 60% of the follow-up time. It was also peculiar that no or very low radiation risks were seen for those who smoked heavily whereas moderate smokers had the highest radiation risk for lung cancer, as previously reported (Furukawa et al. 2010, 2017).

The mortality data (Ozasa 2012) are more comparable to our mortality studies of worker populations and TB-fluoroscopy patients, although not of the diagnostic quality of that for incidence data. The Japanese mortality data are more complete than the incidence data in the sense that the entire country of Japan after 1950 was covered for death determinations, whereas the two cancer registries in Hiroshima and Nagasaki began in 1958 and could not include the entire country for incidence determinations. The mortality analyses focused on the 86,611 lifespan study cohort members with dose estimates. The sex-averaged ERR/Gy was 0.75 (95% CI 0.51,1.03). The ERR/Gy was greater among females (1.1; 95% CI 0.68, 1.6) compared with males (0.40; 95% CI 0.17, 0.67). The male to female ratio was 2.7 (95% CI 1.3, 6.8). The excess absolute risk (EAR) ratio of females to males, however, was 0.98 (95% CI 0.40, 1.8), indicating the importance of background rates of disease when making inferences on “risk differences”.

Animal Experiments

There is clear evidence in experimental studies that low dose-rate exposures are less effective in causing lung cancer than are high dose-rate exposures (Ullrich and Storer 1979a, 1979b; Ullrich et al. 1987; Fry 1996; NRC 2006;Figure 10b-1). The evidence for high-LET exposures, such as neutrons, is not as apparent (Ullrich et al. 1976, 1977; Storer et al. 1979, 1983; Storer and Fry 1995; Heidenreich et al. 2006). Interestingly, the relative risk estimates for man and mouse for lung tumors were not significantly different (Storer et al. 1988). Apparently, for low-LET radiation, low-dose rates consistently result in dose-response relationships and risks significantly lower than that observed for acute or higher-dose rate exposures (Ullrich et al. 1976, 1987; Ullrich and Storer 1979a, 1979b, 1976; Ullrich et al. 1987). However, the evidence that male and female risk estimates differ is not clear and depended upon mouse strain, dose rate, and radiation quality. To evaluate more fully the information available from animal experiments as well as in humans, NCRP has assembled scientific committee (SC 1-27), entitled “Sex-Specific Differences in Lung Cancer Risks and Guidance on Modification of Lifetime Risk Projections.”

DISCUSSION

There are few studies, other than that of the Japanese atomic bomb survivors (Ozasa et al. 2012; NCRP 2014; Cahoon et al. 2017) and the Mayak workers (Gilbert et al. 2013; Gillies et al. 2017), that have focused on the sex-specific differences in lung cancer radiation risks. An informative summary and evaluation of several studies can be found in a Statement by the German Commission on Radiological Protection (Strahlenschutzkommission 2009). The current study computed sex-specific lung cancer risks for occupational cohorts within the MPS and for TB patients who received frequent chest fluoroscopic examinations. Insights and preliminary conclusions can be drawn as to sex-specific radiation-related lung cancer risks when exposure is protracted or fractionated over time.

Six cohort analyses of occupational and patient exposures showed little evidence that prolonged or fractionated exposures increase the risk of lung cancer. Further, there were no significant differences in the radiation risk estimates between men and women. These data for low-LET radiation are consistent with studies of Hodgkin lymphoma patients and Rocketdyne workers. They are also consistent with animal studies showing a marked reduction in risk when low-LET radiation dose is delivered at a low rate, i.e. spread over time. The studies providing evidence for a sex-specific difference in lung cancer risk are those of the Japanese atomic bomb survivors and the study of Mayak workers exposed to exceptionally high doses of plutonium and to relatively high doses from gamma-rays. Animal studies do not consistently find an ameliorating effect of protracted radiation on lung cancer risk when the exposures are to neutrons.

These data suggest that for low-LET radiation, chronic or fractionated exposures to radiation have little effect on increasing lung cancer risk among either men or women. Further, the new analyses of the occupational cohorts and the TB-fluoroscopy patients find no evidence for any differences in lung cancer risk between men and women. The reduction in lung cancer risk with protracted exposures is not seen in studies of radiation-induced breast cancer (Preston et al. 2002), and suggests that different tissues respond differently to the carcinogenic effects of chronic or fractionated exposures. Interestingly, the TB-fluoroscopy studies showing minimal radiation risk for lung cancer (Davis et al. 1989; Howe 1995) are the same studies showing, on an absolute scale, that breast cancer risk is not appreciable influenced by fractionation of dose (Miller et al. 1989; Boice et al. 1991). Similar observations of a significant breast cancer risk and the absence of a lung cancer risk are seen in the U.S. Scoliosis Study (Ronckers et al. 2010).

Strengths and weaknesses

The occupational studies included in these analyses have been comprehensively conducted, including the epidemiologic methods and the extensive programs for dose reconstruction (NCRP 2018a; NCRP 2018b). The follow-up rate was over 95% overall and the cause of death determination nearly complete. The causes of death were obtained in the same manner by the same investigators relying upon state and national death information in the United States (Mumma et al. 2019). The numbers are large, with 453,746 individuals and 16,049 lung cancer deaths. The dose distribution was broad with mean lung doses ranging from 6 mGy to >1,000 mGy. The dose reconstruction procedures for each cohort followed guidance found in NCRP Report 178 (NCRP 2018a) by essentially the same team of investigators. The TB-fluoroscopy studies have been ongoing for over 40 years with a focus on dose reconstruction, follow-up and death ascertainment. The findings of the TB-fluoroscopy studies with regard to breast cancer have been influential in understanding radiation risks and guiding protection standards.

Nonetheless, there are limitations that temper the strength of conclusions that can be drawn. First, despite the larger number of women (50,679) and their high mean cumulative dose to lung (677.1 mGy), the statistical power to discern sex-specific differences is low, in part because only one of the 12 sex-specific estimates of risks was significantly different from the null and tests to compare “no effect” to “no effect” are not very informative. Secondly, information on smoking histories was only available for a fraction of subjects in the TB-fluoroscopy studies. In the Canadian Fluoroscopy Cohort Study, there was little variation in the prevalence of smoking by dose category, and thus smoking was not a confounder of the radiation risks of lung cancer (Howe et al. 1995). For the occupational studies, education or job category were taken as measures of socioeconomic status and as an indirect way to account for lifestyle and behavior factors, but there remains some uncertainty as to the effectiveness of this adjustment to adequately control for smoking. A small interview study of Rocketdyne workers did confirm that hourly workers compared with salaried workers were significantly more likely to have smoked cigarettes, to start at a younger age, to quit at an older age, and to smoke for more years (Boice et al. 2006). Thirdly, because of the relatively small numbers exposed to intakes of radionuclides, the studies could not powerfully evaluate lung cancer risk following high-LET exposures except perhaps for the Mound study which involved intakes of polonium and relatively high doses to lung. Other studies initiated or planned within the MPS will be able to address high-LET exposures more powerfully, such as the plutonium workers at the Los Alamos National Laboratory (Wiggs et al. 1994), the Rocky Flats Nuclear Weapons Plant and the Hanford site (Gillbert et al. 1993), and the uranium workers at the Fernald Feed Materials Production Center (Silver et al. 2013) and the Linde Air Products Company Ceramics Plant (Dupree et al. 1987).

The study of Japanese atomic bomb survivors continues to show a risk of radiation-induced lung cancer and a substantial difference between male and female risks. These observations could be causally related to acute exposures or perhaps related to uncontrolled confounding or incomplete smoking ascertainment, perhaps because few Japanese women smoked whereas most men smoked. Consistent with animal experiments a very brief exposure, or high dose-rate exposure, appears to carry more risk than a similar dose to the lung that is received gradually over time, or at a low dose rate. The Mayak study is exceptionally powerful and comprehensive but interpretations are with cautions given the uncertainties in the dose reconstructions and to the possible influence of the very high lung doses that caused deterministic effects (tissue reactions), e.g., pulmonary fibrosis (termed plutonium pneumosclerosis) (Okladnikova et al. 1994). Conceivably the influence of such high doses might be different than that of lower doses as suggested in the companion study of the Sellafield workers who received very much lower doses and had little evidence of a lung cancer risk (Gillies et al. 2017).

More work could be done with regard to evaluating the excess absolute risks in addition to the excessive relative risks per unit radiation dose. When differences in radiation risks between men and women are reported, they usually are based on a relative risk scale and not an absolute risk scale.

CONCLUSION

These analyses and evaluations provide little evidence that chronic or fractionated exposures to radiation increase the risk of radiation-induced lung cancer. Further there was no evidence that radiation-related lung cancer risks differed between women and men. However, these conclusions are preliminary and limited by relatively small numbers of women who received relatively high doses. Lung cancer risk following high-LET exposures could not be effectively studied in the current MPS cohorts but should be in the near future. The ongoing study of medical radiation workers within the MPS including 65,000 women and 65,000 men will provide substantial information as to the effectiveness of gradual exposures over time to cause lung cancer and to determine whether such exposure results in different radiation lung cancer risks between men and women (Boice 2017; Yoder et al. 2019). Canadian Fluoroscopy Cohort Study of 64,000 patients includes equal numbers of men and women exposed at all ages. It is currently being updated for mortality and incidence and should provide essentially lifespan follow-up of the cohort.

Biographies

John Boice is President of the National Council on Radiation Protection and Measurements and Professor of Medicine at Vanderbilt University. He is an international authority on radiation effects and served on the Main Commission of the International Commission on Radiological Protection and on the United Nations Scientific Committee on the Effects of Atomic Radiation. He directs the Million Person Study of Low-Dose Health Effects.

Lydia Zablotska is a Professor in the Department of Epidemiology and Biostatistics in the School of Medicine at the University of California, San Francisco (UCSF), where she serves as the Leader of the Occupational and Environmental Epidemiology Area of Concentration. She is the Principal Investigator of the NCI-funded study to update dosimetry and conduct lifespan mortality and incidence follow-up of the Canadian Fluoroscopy Cohort Study .