Recent analyses of data from nuclear industry workers in the United Kingdom's National Registry for Radiation Workers (NRRW) [1, 2] and from nuclear industry workers from five sites in the United States of America [3] have shown raised risks for the grouping of all solid cancers combined following protracted low-dose exposure to low-LET radiation, such as x-rays and gamma radiation. However, the US study found that the risk estimates were significantly greater for those 56% of workers first hired in 1960 or later in comparison with the full cohort (1944–2016) of 101 363 workers, for all solid cancers combined and for certain individual site-specific solid cancers, such as lung, colon, and pleural cancer. A combined analysis of mortality data for 309 932 workers from the UK, the USA and France in the recent International Nuclear Workers Study (INWORKS) [4] also found higher risks for solid cancers among those 77% of workers first hired after 1957 and even more so among those 61% first hired after 1964.

Two letters to the editor by Wakeford [5, 6] discuss, among other things his concerns with this pattern of differing risks by calendar period of first hire.

We have now examined this issue among UK radiation workers using data from the recently published updated NRRW-3 study [1] which examined cancer incidence among about 173 000 radiation workers (90% men) from 15 different sites.

2.1. The NRRW cohort

The NRRW is a large cohort of occupationally exposed workers from the UK. The dataset from the NRRW analysed here comprised 172 452 workers with a total follow-up of about 5.25 million person-years (an average follow-up of approximately 30 years per worker). Mortality due to solid cancer was identified from 1955 and cancer registrations from 1971 through to the end of follow-up in 2011. Cancer incidence analysis in the updated NRRW-3 is based on a combination of registration and mortality data; comprising incidence data from 1971 and cancer deaths prior to that year. Detailed information about the definition of the study population, cohort design, data collection, the characteristics of the workers and follow-up procedures for this updated NRRW-3 cohort are given in an earlier study [1]. The NRRW includes workers from a wide range of employer organizations across the UK: British Nuclear Fuels Ltd (BNFL), Atomic Weapons Establishment, UK Atomic Energy Authority, British Energy Generation and Magnox Electric, the Ministry of Defence (MoD), GE Healthcare, Rolls-Royce, as well as many smaller organizations in the research and industrial sectors from all parts of the UK.

Poisson Regression is used to investigate the relationship between cumulative external radiation dose and solid cancer incidence rates (and mortality rates prior to 1971) and the main (linear) dose-response model is expressed as: predicted rate = Background rate *(1 + ERR), where background rate is based on a parametric model which includes attained age, sex, birth cohort, the first employer, the industrial status (industrial/non-industrial) and the duration of employment, where ERR is the excess relative rate. More details about the statistical models are given in the updated NRRW-3 study [1].

2.2. Dose data in the NRRW

The analysis in the NRRW focused on recorded whole-body dose from penetrating radiation (x-ray and gamma radiation) assessed by personal film dosemeters (latterly by thermoluminescent dosemeters) and of a generally smaller neutron component usually estimated in part by the same dosemeter worn for the photon component; all dose estimates are recorded in sievert (Sv) or expressed as millisieverts (mSv). Only a small number of radiation workers (12%) in the NRRW cohort are identified with exposure to neutrons [7]. However, the analysis in the NRRW does not take account of neutrons doses because of limited information provided by employers and they are subject to some uncertainties, which will be discussed later. Internal tritium exposure in the NRRW cohort was treated in terms of whole-body dose as if it were external dose and, in some records, this contribution has been added to the annual recorded dose, but only 3% of cohort members were exposed and information on dose estimates from tritium is also limited [8]. In addition, around a quarter of workers in the NRRW were also monitored for exposure to internal emitters, mainly plutonium and uranium, but estimates of doses from internal emitters were not available, only a yes/no indicator of whether a worker was ever monitored for internal exposures. A sensitivity analysis was conducted to assess the impact of those workers monitored for exposure to internal emitters on the main findings in relation to external dose in the previous NRRW incidence study [1].

Two-thirds of workers in the NRRW have cumulative external doses of less than 10 mSv. Workers' annual doses from external exposure in the early years of the UK nuclear programme were higher than in later years: for workers starting radiation work in the 1960s and 1970s, their mean lifetime cumulative doses were 28.6 mSv and 22.9 mSv, respectively, but mean annual dose had fallen to 0.6 mSv towards the end of the 1990s. Among the 14% of the NRRW workers first hired before 1960 the mean lifetime cumulative dose was 74.3 mSv compared to 17.0 mSv among those first hired later. About 64% of workers were first hired after 1969, and the cumulative dose reduced to 13.0 mSv, compared with 46.1 mSv among those first hired before 1970. Of those workers first hired before 1960, 26% stopped working before 1960 and their accumulated mean dose was 14.3 mSv, but the other 74% had worked after 1959 and received substantial exposure through the 1960s and 1970s and their mean cumulative dose was 95.5 mSv.

2.3. Summary results of the updated NRRW incidence study

A total of 18 310 cases of solid cancers based on a 10 year lag following the start of employment, were registered in the updated NRRW-3 study [1]. The study demonstrated that the risk increased approximately linearly at low doses but flattened and started to reduce above a cumulative external dose of 400 mSv for solid cancers, which indicated evidence of nonlinearity based on a linear-exponential (LE) dose-response. However, only a small proportion (1.6%) of the total number of solid cancers cases occurred above 400 mSv, for which 84% of cases occurred among workers who were monitored for internal emitters and most of them (71%) were BNFL Sellafield workers who had long employments of 20 years or more. This nonlinearity appeared to be driven by the workers who were monitored for potential exposure to internal emitters and who had also received relatively high external doses. Among cohort members only monitored for exposure to external radiation, a strong linear association was found between external dose and solid cancers.

A similar LE pattern is also seen for lung cancer. Excluding lung cancer from the grouping of all solid cancers resulted in evidence of a linear association with external radiation dose, so suggesting some degree of confounding by smoking.

The study also reported that there were no modifying effects of sex, age at exposure or time since first radiation exposure on the risk model for males of working age [1]. Female workers form only small numbers (10%) of the NRRW population. For many specific cancer types, the incidence data in females are too sparse, and the low statistical power of the solid cancer analysis is due mainly to the relatively small numbers of females and the very low (5.6 mSv) mean lifetime dose for female workers. It should be stressed that much of the information on radiation and cancer from the NRRW relates to males; it is not surprising that the overall estimates in the NRRW are similar to those for males alone.

2.4. The US cohort and INWORKS

INWORKS has carried out a combined analysis of solid cancer mortality among nuclear industry workforces in France, the UK and the USA [4]. The cohort comprised about 309 932 workers (mostly men) with 10.72 million person-years of follow-up. The NRRW contributes data from five major employers with follow up from 1955 to 2012 that comprise about 86% of the workers reported on in the updated third analysis of the NRRW [1]. NRRW workers make up roughly half of the INWORKS cohort. The US study includes 101 363 workers from five nuclear facilities with the calendar year of follow up from 1944 to 2016 and the total number of person-years of follow-up was near 4 million.

Both for the INWORKS and the US workers cohort, analyses assume that the vast majority of external recorded whole-body dose was due to x-rays and gamma radiation; both studies also have limited dose information on neutrons, tritium, and intakes of other radionuclides. Doses for workers in the US study are based on recorded whole-body dose, while in the INWORKS for the purpose of the solid cancer analysis, recorded doses are converted into colon absorbed doses from photons expressed as in gray (Gy) or milligray (mGy). The mean estimated dose to the colon is reported to be lower than the mean recorded whole-body dose by 31% [8], which resulted in an increase in the solid cancer ERR/Gy by about 41% [4]. The analysis in both INWORKS and the US study include an adjustment for neutron monitoring status. However, deviations from the linear model were not examined in detail in the US study [3] or in the INWORKS [4], although there was some suggestion of downward curvature in the INWORKS solid cancers dose-response [4].

Table 1 shows the estimated ERR/Sv based on a linear model for the cancer groupings all solid cancers, lung cancer, and all solid cancers excluding lung cancer, using the full NRRW cohort and the sub-cohort of workers monitored for external exposure only (NRRW-External workers). Also shown are results for the full cohort and the external exposure only workers sub-cohort separately by period of first hire before and after the beginning of the years 1960, 1965 and 1970. To compare, table 1 also presents the estimates of the linear ERR from the US study [3] for the full cohort and for workers first hired after the beginning of 1960, and also from the INWORKS cohort for workers first hired from the start of years 1958 and 1965, and before 1958 [4]. It is unfortunate that the results were not presented for the pre-1960 hire sub-cohort in the US study and for the pre-1965 hire sub-cohort in the INWORKS.

Table 1. Comparison of estimates of the ERR/Sv (based on linear models, and whole-body doses unless indicated otherwise) of external radiation exposure for all solid cancers combined, lung cancer and all solid cancers excluding lung cancer, for incidence among NRRW for the full cohort and restricted sub-cohorts, and for mortality among the US workers and INWORKS for the full cohorts and various sub-cohorts (all with doses lagged by 10 years).

Study cohortMean dose (mSv)All solid cancerLung cancerAll solid cancers excluding lungNo. of cases/deathsERR/Sv (95% CI)p*No. of cases/deathsERR/Sv (95% CI)p*ERR/Sv (95% CI)p* NRRW-Full data (1955–2011)24.918 3100.20 (−0.001; 0.43)a 32750.16 (−0.27; 0.70)a 0.24 (0.04; 0.45) Hired <196074.351530.14 (−0.08; 0.38)0.201138−0.006 (−0.39; 0.52)0.060.20 (−0.05; 0.48)>0.5Hired 1960+17.013 1570.39 (0.04; 0.76)21370.915 (−0.006; 2.00)a0.33 (−0.04; 0.74) NRRW-External workers (1955–2011)13.113 1990.52 (0.11; 0.96) 21980.76 (−0.31; 2.06)a 0.53 (0.09; 1.02) Hired <196037.323370.39 (−0.13; 1.01)>0.54930.59 (−0.84; 2.45)>0.50.42 (−0.15; 1.10)>0.5Hired 1960+10.910 8620.63 (0.09; 1.22)17050.91 (−0.48; 2.59)a0.63 (0.05; 1.28) US- Full (1944–2016)26.512 0690.19 (−0.10; 0.52)b 38710.65 (0.09; 1.30)b −0.01 (−0.34; 0.36)b Hired 1960+14.537122.23 (1.13; 3.49)b NP2.90 (1.00; 5.26)b 1.88 (0.59; 3.41)b  INWORKS-Full (1944–2016)17.728 0890.52 (0.27; 0.77)b,c,d  NP 0.46 (0.18; 0.76)b, c, d Hired <1958NP13 2210.20 (−0.07; 0.49)b,c,d  NP NP Hired 1958+NP14 8681.22 (0.74; 1.72)b,c,d  NP 1.20 (0.65; 1.78)b, c, d Hired 1965+NP81191.44 (0.65; 2.32)a,b,c,d  NP 1.38 (0.47; 2.39)b, c, d  NRRW-Full data Hired <196559.485190.21 (−0.01; 0.45)a>0.51792−0.06 (−0.42; 0.43)0.0020.32 (0.065; 0.61)a0.11Hired 1965+14.697910.18 (−0.24; 0.64)14831.92 (0.64, 3.44)a−0.09 (−0.51; 0.39) NRRW-External workers Hired <196530.948800.63 (0.14; 1.21)0.419800.35 (−0.87; 1.86)0.310.77 (0.21; 1.41)0.15Hired 1965+9.383190.30 (−0.31; 0.99)12181.49 (−0.24, 3.64)0.13 (−0.51; 0.86) NRRW-Full data Hired <197046.111 5870.21 (0.004; 0.44)a0.523610.09 (0.04; 0.62)0.130.28 (0.04; 0.54)0.23Hired 1970+13.067230.07 (−0.52; 0.74)9141.45 (−0.25; 3.59)−0.16 (−0.77; 0.55) NRRW-External workers Hired <197023.775850.53 (0.09; 1.02)>0.514720.89 (−0.26; 2.31)>0.50.53 (0.06; 1.06)>0.5Hired 1970+7.956140.47 (−0.42; 1.48)7260.007 (−2.44; 3.01)0.53 (−0.41; 1.63)

* p-value: test for heterogeneity between two sub-cohorts; a: The LE model is best fitted (p < 0.05); b: additional adjustment for neutron monitoring status; c: colon dose; d: 90% CI; NP: not provided.

3.1. All solid cancers combined

For all solid cancers combined there is reasonably good agreement, in that the confidence intervals overlap, between the risk estimates based on the NRRW-full data, the NRRW-external workers, the full US study cohort and the full INWORKS cohort, although given that the INWORKS cohort contains a significant part of the updated NRRW cohort and the full US data considerable similarity would be expected. It should be noted that the INWORKS confidence intervals are 90% CIs, which are narrower than the 95% CIs used in other studies.

In the NRRW cohort, the risk for all solid cancers using both the NRRW-full and NRRW-external workers is greater among workers first hired in 1960 or later than among those who started work earlier, but not significantly so (p = 0.20 and p > 0.5 respectively) and all estimates are consistent with the main analysis of the NRRW-full and NRRW-external workers sub-cohort.

This is not the case for the US cohort where the risk for all solid cancers among workers first hired in 1960 or later is higher by a factor of 12 compared to the estimate from the full cohort and also larger by factors of 6 and 4 compared to the post-1959 NRRW value using the NRRW-full data and the NRRW-external workers sub-cohort, respectively. It should be noted that unlike the US analysis no neutron adjustment was considered in the updated NRRW-3 analysis.

A similar pattern is also seen in the INWORKS cohort for all solid cancers where the excess risk among those first hired post-1957 is larger by a factor of about 6 compared with the pre-1958 value and is not consistent with the estimate based on the full INWORKS cohort. Furthermore, the risk estimate for the INWORKS workers hired post-1957 is generally consistent with the estimate for workers hired post-1959 in the US study and among those NRRW-external workers hired post-1959, but not for the NRRW-full data (table 1).

Further investigation of the NRRW data for all solid cancers for those first hired pre-1965 or pre-1970 showed that the risk estimates are larger than for the respective post-1964 or post-1969 sub-cohorts, but the differences in ERR/Sv between the earlier and later periods are not statistically significant for either the NRRW-full data or the NRRW-external workers sub-cohort (table 1). The risk estimates for the periods of first hire pre-1965 and post-1964 and pre-1970 and post-1969, are all consistent with the risk estimates in the main analysis of the NRRW-full data and the external workers of the NRRW sub-cohort; for the NRRW-full data, the dose-response for the sub-cohort of workers first hired before 1965 and before 1970 are better fitted by LE model than linear models (table 1). In the INWORKS cohort, the result for all solid cancers among those radiation workers first hired post-1964 is larger but consistent with the estimate for the full INWORKS cohort. However, the INWORKS risk estimates for those hired in 1965 or later is larger by a factor of about 8 compared with those workers hired after 1964 from the NRRW-full data and 5 times greater for workers hired after 1964 from among the NRRW-external worker sub-cohort, and the INWORKS dose-response for those hired in 1965 or later is better fitted by a LE model [4].

3.2. Lung cancer

The overall risk estimates for lung cancer in both NRRW-full and NRRW-external workers sub-cohort are consistent with the US estimate for the full cohort.

For lung cancer, the risk estimate in the NRRW was higher for those workers first employed in 1960 or later compared to workers first employed earlier, but the evidence for differences is weak (P = 0.06) for the NRRW-full data. The same pattern is also observed for the NRRW-external workers sub-cohort, but the difference is not statistically significant (P > 0.5). All these hire-period-specific estimates are consistent with those based on the NRRW-full cohort and the NRRW-external workers sub-cohort. As with the NRRW-full data for all hire dates, the lung cancer dose-response for all workers hired in 1960 or later is better fitted by a LE model. The NRRW estimate for those workers hired post-1959 is lower (by a factor of about 3) than that for workers in the US study hired in the same period, albeit that the confidence intervals overlap. However, the risk estimate is higher with very wide confidence interval for the post-1959 estimate for lung cancer in the US study and just overlaps with the confidence interval for the estimate based on the full US cohort.

The lung cancer risk estimate using the NRRW-full data is statistically significantly higher for the post-1964 workers compared to that for workers first employed earlier (P = 0.002), and a LE model provides a better fit to the dose-response than a linear model. For the NRRW-external workers sub-cohort the risk is also higher for the workers hired post-1964 compared that for workers hired earlier, but the difference is not statistically significant (P = 0.31). The risk estimate in the NRRW-external workers hired pre-1970 is larger than the post-1969 value, but the opposite pattern was observed using the NRRW-full data. In neither case was the difference statistically significant.

3.3. All cancers excluding lung cancer

For the grouping of all cancers excluding lung cancer, the full US cohort showed no evidence of excess risk but there was good evidence of an elevated risk for the post-1960 hired workers. In the NRRW both the full cohort and external workers sub-cohort also showed good evidence of excess risk in this disease grouping and the results are consistent with the results based on the full data of INWORKS and of the US study.

The NRRW-full cohort showed little evidence of excess risk among workers hired pre-1960 or post- 1959, but among the NRRW-external workers for those hired post-1959 there was raised risk. However, a test of heterogeneity did not find evidence for a difference between the risk estimates for the two hire periods and all these estimates were consistent with those from the NRRW-full cohort and the external workers sub-cohort.

The estimate of risk in the US cohort for workers hired after 1959 is larger compared with the estimates based on either the NRRW-full data (by a factor of about 6) or the NRRW-external workers sub-cohort (by a factor of about 3) but is consistent with the post-1957 estimate from INWORKS. The post-1957 INWORKS estimate, the 90% CIs just overlaps, but does not support that based on the NRRW-full data for the post-1959 hire period. However, it is consistent with the NRRW-external workers hired in the same period.

For both the NRRW-full data and the NRRW-external workers sub-cohort, the risk estimates for all cancers excluding lung cancer among the pre-1965 hired workers are higher than those for workers hired later but the differences are not statistically significant (P = 0.11 and P = 0.15 respectively). For the NRRW-full data and workers hired before 1965, a LE model is a better fit to the dose-response than a linear model.

The risk estimate using the NRRW-full data for the pre-1970 hire workers is also higher than the respective post-1969 value but not significantly so, while among the NRRW external workers pre-1970 and post-1969 values are the same. For all solid cancers excluding lung, the INWORKS risk estimate for the post-1964 workers is larger but consistent with the INWORKS-full data. However it is not consistent with the post-1964 hired workers of the NRRW-full cohort (table 1) and is higher by a factor of 11 compared to the estimate for the post-1964 hired workers in the NRRW-external workers.

In considering these results it must be remembered that the US and INWORKS study results are based on mortality data. The analyses detailed here are primarily of cancer incidence, but because cancer incidence data were not available before 1971 earlier deaths were also included in the analyses. However, the deaths included only represent 3.6% of the total number of solid cancer cases that contributed to the solid cancer incidence analysis from 1955 till 2011. When the same analysis was repeated using the mortality data in the updated NRRW-3 cohort for all solid cancers and also for lung cancer, the findings supported the results from the current study of incidence data (not shown here, but see Haylock et al [2] for mortality risk estimates for all years of hire) and the estimates from the mortality analysis for those first hired after 1960 or 1965 were not consistent with the corresponding estimates in the INWORKS. However, cancer incidence data are of much higher quality and provide more detailed diagnostic information, while mortality data are relying only upon death certification and what is recorded as the underlying cause of death.

There are differences between the NRRW, the US study and INWORKS such as the use of colon doses by INWORKS, and differences in adjustment factors such as neutron monitoring flag, age cut-offs and follow-up dates. One notable difference between the US study and NRRW is that follow-up began up to 11 years earlier in the US study. The US cohort included 101 363 workers and a total of 12 069 solid cancer deaths that occurred through to the end of 2016. Just over half of the workers (56%) were first hired in 1960 or later. The number of deaths was lower among this later hired group (3,712 deaths) compared to those first hired pre-1960 (8357 deaths). In contrast, the updated NRRW-full cohort includes 172 452 workers of whom most (86%) were first hired post1959, and of the 18 310 cases of solid cancers, again most cases occurred among those who started work in 1960 or later (72%). In INWORKS, the cohort includes 309 932 workers of whom 77% were first hired post-1957 and the corresponding value for workers first hired post-1964 was 61%. Of the 28 089 deaths from solid cancer, 53% occurred among those first hired post-1957, but a lot lower percentage occurred among workers hired post-1964 (29%). The NRRW-external workers sub-cohort comprised 130 178 workers of whom 90% were first hired in 1960 or later, and of the 13 199 cases of solid cancers, 82% occurred among those who started work post-1959.

It should also be recognised that not all workers in the updated NRRW-3 analysis were included in INWORKS, which only includes the main nuclear industry groups [9]. The analysis in the updated NRRW cohort includes workers from other smaller sites, while the US study involves five main facilities: Portsmouth Shipyard, Hanford, Idaho National Laboratory, Savannah River and Oak Ridge National Laboratory. There are differences in these UK and USA facilities, both within countries and between countries, of the types and patterns of exposure; some are nuclear power plants to generate electricity and others are reprocessing plants such as Sellafield, and shipyard workers are also included in both the US and the NRRW studies and have the potential for significant asbestos exposure, especially in the early years. Such exposures might well impact on lung cancer, and certainly on pleural cancer.

Doses from neutrons and intakes of radionuclides are not generally computed in the NRRW or in the US study. Adjusting for neutron monitoring or considering those workers who had potential internal exposures makes interpretation of external radiation (x-ray plus gamma) effects uncertain in the NRRW and the US studies, as well as in the INWORKS [1, 3, 9, 10]. Annual recorded whole-body dose in the NRRW includes x-ray and gamma exposure plus neutrons and tritium, but the information about neutrons and tritium is limited. The assessment of the neutron component is problematic because the standard types of dosemeter (film badges) used in personal monitoring to measure photons also measure thermal neutrons (i.e. very low energy neutrons), which can be inaccurate and miss fast neutrons that can carry most of the dose and risk. Some fast neutron exposures were measured in the UK, but information about fast neutron doses and their relative biological effectiveness is limited. Corrections for missing neutron contributions for workers in the NRRW cannot be made.

The recent US study [3] that found evidence of an increasing trend in the risk of lung cancer mortality, with an ERR/Sv = 0.65, 95% CI: 0.09; 1.30, (number of deaths = 3871) is substantially higher by a factor of about 9 compared with the estimate in the previous US study, ERR/Sv = 0.069 (95% CI: −0.43; 0.66) (number of deaths = 3514), which reported no evidence of an increasing linear trend for lung cancer with follow-up to 2005 [10]. This large difference is reported to be due, among other things, to longer duration of employment and excluding tritium exposure in the later study [3]. However, it is also important to note that in the previous US study [10], the recorded whole-body dose included neutron (with a weighting factor of 10 for fast neutrons) and tritium components, unlike the recent study [3], which used a category-based neutron monitoring factor that may well explain, in part, this large difference. In addition, the recent US study also showed that the ERR/Sv for lung cancer was increased slightly when the analysis excluded neutron-monitored workers (ERR/Sv = 0.76, 95% CI: 0.02; 1.61). The analysis in INWORKS [4, 9] also uses an adjustment for neutron monitoring status that is a category-based neutron monitoring surrogate variable [8], which in a previous analysis led to a sizeable increase in the estimated ERR/Gy for the grouping of all cancers excluding leukaemia by a factor of about 2.5 compared with the model with no neutron adjustment [9], albeit only 13% of workers with potential exposure to neutrons were identified [8]. The ERR/Gy estimate for all solid cancers with no neutron monitoring status adjustment is not reported in the recent INWORKS study so the difference with the estimate from the model with this adjustment is not known [4].

Currently the NRRW does not have any internal exposure data which means it is difficult to assess if the size of the internal dose received is important. The NRRW study has shown that the use of a non-linear relationship based on the LE model to estimate excess risk from external exposure over the whole dose range appears to be problematic and may not provide a reliable description of risk for the average worker. Excluding workers monitored for internal exposure leads to good evidence for increasing linear trend and the evidence of non-linearity (LE model) disappears. However, there may be an impact on the external dose response of simply being monitored for internal exposure, regardless of internal dose. Internal monitored workers tend to have much longer employment periods than external monitored only workers and these internally monitored workers make up a substantial proportion of the workers with higher external doses in the study population, e.g. about 84% of workers with doses over 400 mSv. The publication in 1998 by Carpenter et al [11], concluded that further investigations of the relationship between radionuclide exposure and cancer in nuclear industry workers are needed, which still rings true today.

There is also a concern in the NRRW cohort about over-estimation of external dose in early years, especially for those internally monitored workers with high cumulative external doses. If true this would certainly be an important effect, as recorded annual external doses in the older plutonium process plants were often close to 50 mSv. This could lead to attenuation of the dose response for internally monitored workers, resulting in an underestimation of the risk per unit dose. In the medium term we expect to be able to acquire internal dose estimates for a significant proportion of the NRRW cohort. These data will then allow the investigation of this issue to be taken further.

An epidemiological study of US Nuclear Power Plant workers who were part of the Million Person Study includes 135 193 workers, the great majority of whom were first monitored after 1969 and exposed to penetrating external gamma radiation and with smaller neutron or internal exposures [12]. There were 9329 solid cancer deaths and 3334 lung cancer deaths. The mean absorbed dose to the colon was 43.7 mGy and 43.2 mGy to the lung. The mean duration of follow-up was 30.2 years. The ERR/Gy for solid cancer was 0.10 (95% CI: −0.3; 0.5) and for lung cancer the ERR/Gy was −0.4 (95% CI:−1.1; 0.2), which are consistent with the estimates in the NRRW workers hired after 1969 using both the NRRW-full data and NRRW-external worker sub-cohorts, which also found no evidence of excess risk for either solid cancer or lung cancer.

The risk estimates in the INWORKS and the 'US component of INWORKS' study from recent workers are inconsistent and much higher than the life span study (LSS) of Japanese atomic bomb survivors for solid cancers mortality (ERR/Gy = 0.32; 95% CI: 0.01; 0.50) based on male survivors who were exposed at ages 20–60 years [4], but all the risk estimates for recently hired workers in the updated NRRW shown in table 1 are consistent with risk estimates derived from the LSS for solid cancer, although the LSS risk is expressed in terms of colon dose whereas NRRW is expressed in terms of recorded dose (i.e. Hp(10) dose).

These analyses reported here demonstrate that only for lung cancer is there strong evidence for a difference in the risks for first employment pre-1965 and post-1964 using the NRRW full cohort, but in the other calendar period breakdowns and for the other cancer groups the differences in the risks were not statistically significant. The NRRW estimation of risks between later and earlier hired workers is not generally consistent with the US study or the INWORKS evaluations that recently hired workers are at much higher risk than early workers. It is important to stress that these early workers accumulated larger doses and have a longer follow-up for late effects. There are enormous differences in risk estimates between the NRRW and the US study, between the NRRW and INWORKS, and also within the US study and INWORKS, when the sub-cohorts of recent and early workers according to different calendar periods are used. However, the findings from those workers first hired in 1960 or later in the US data appear to be influencing the results for workers hired post-1957 in INWORKS.

If the aim of the analyses of calendar period of first hire in the US and INWORKS studies is to investigate the impact of dosemeters used in the early years that involve uncertainties associated with individual film badge assessment, selecting a calendar period for those first hired before and after a certain date is not necessarily the right method. In the NRRW, many of the earlier hired workers stayed at work for long periods and accumulated significant dose in later years (post-1959) when the technology to measure external doses to workers improved markedly. The fact that they received doses in later periods makes the distinction between early vs. late workers somewhat blurry regarding their radiation exposure.

The conclusion that the INWORKS and US studies demonstrate a real difference in excess solid cancer risk from external radiation exposure between earlier and later hired workers is premature. The results presented here should also be treated with caution because of the limited corroborating evidence from other published studies. Information on internal doses, neutron doses as well as non-radiation factors such as smoking and asbestos exposure would be needed to make definitive inferences.

The data cannot be made publicly available upon publication because they are owned by a third party and the terms of use prevent public distribution.