States indicated lower risk estimates than the com-Objective-To study cause specific mortality mission recommends, and when the American data of radiation workers with particular reference to are combined with our analysis the overall risks are associations between fatal neoplasms and level of close to those estimated by the commission. This exposure to radiation. first analysis of the National Registry for Radiation Design-Cohort study. Workers does not provide sufficient evidence to Setting-United Kingdom. justify a revision in risk estimates for radiological Subjects-95 217 radiation workers at major sites protection purposes. of the nuclear industry. Main outcome measure-Cause of death. Results-Most standardised mortality ratios were Introduction below 100: 83 unlagged, 85 with a 10 year lag for ali Estimates of the risks of ionising radiation rest causes; 84 unlagged, 86 lagged for all cancers; and 80 mainly on evidence from Japanese atomic bomb for all known other causes, indicating a "healthy survivors and from people exposed for medical reasons. worker effect." The deficit of lung cancer (75 These groups provide information on risks from unlagged, 76 lagged) was significant at the 0-1% exposure to high doses at high dose rates. There is little level. Standardised mortality ratios were signifi-direct evidence of the effects of lower doses and dose cantly raised (214 unlagged, 303 lagged) for thyroid rates typical of occupational exposures. To provide cancer, but there was no evidence for any trend with such direct evidence the National Radiological Protecexternal recorded radiation dose. Dose of external tion Board, after extensive consultation with the radiation and mortality from ali cancers were weakly nuclear industry and other interested groups, set up correlated (p= 010), and multiple myeloma was the National Registry for Radiation Workers in 1976 as more strongly correlated (p=0.06); for leukaemia, the national study of radiation workers, following National Radiological excluding chronic lymphatic, the trend was signifi-individuals through different employments.' Protection Board, Chilton, cant (p=0-03; all tests one tailed). The central The first analysis of the registry covers over 95000 Didcot OXII ORQ estimates of lifetime risk derived from these data radiation workers whose collective dose from external were 10-0% per Sv (90% confidence interval <0 to radiation is about 3200 man Sv. The essentials of the GM Kendall,PHD, principal 24%) for all cancers and 0-76% per Sv (0-07 to 24%) for study are described in this paper; more details can be scientific officer leukaemia (excluding chronic lymphatic leukaemia). found in a separate report.2 C R Muirhead, PHD, These are, respectively, 2-5 times and 1-9 times the BMJlf 1992;304:220-5 combined cohorts of radiation workers in the United (a) and (d) from

The mortality from leukaemia and other cancers in the National Registry for Radiation Workers is analysed to determine the subcategories of leukaemia in which the trend with dose is most marked. Differences between external (SMR) and internal analyses (of trends with external radiation dose) are assessed for the various leukaemia subtypes. The significance of the trend for leukaemia mortality observed in the previous analysis is largely accounted for by the group of chronic myeloid leukaemias. Sensitivity analyses are performed relating to various uncertain parameters to examine the robustness of the previously published findings. The magnitude (and significance) of the trend with dose for leukaemia is robust to a reasonable range of assumptions concerning the lag period; if a lag of 5 years is used the trend retains borderline levels of statistical significance (p=0.07); if lags of 0 or 2 years are employed rather more significant trends are apparent (p=0.03).

No abstract

No abstract