INWORKS study: risk of leukaemia from protracted radiation exposure - authors' reply

Details

• Personal Author:
  Cardis E ; Daniels, Robert D. ; Gillies M ; Hamra GB ; Haylock R ; Kesminiene A ; Laurier D ; Leuraud K ; Moissonnier M ; O'Hagan JA ; Richardson DB ; Schubauer-Berigan, Mary K. ; Thierry-Chef I
• Description:
  We thank Shigenobu Nagataki and Fumiyoshi Kasagi, and Mohan Doss for their interest in our study on leukaemia risk among radiation monitored nuclear workers in France, the UK, and the USA. The authors ask for additional details on the way the study was done and raise several issues that, they suggest, seriously limit our conclusions. In the following, we address their points to clarify why we disagree with that suggestion. Nagataki and Kasagi pose three questions. First, why was the study conducted on the basis of three countries from the 15-country study? Because of the need for efficiency in project coordination, INWORKS included three major partners from the previous 15-country study (France, the UK, and the USA); however, less than 20% of deaths from leukaemia had been contributed by the other 12 countries. The cohorts included in INWORKS were predominantly externally-exposed, had the most complete and up-to-date information on cohort eligibility and occupational dose, and were each updated for mortality since the 15-country study. As a result, the INWORKS analysis yields much more precise estimates of association than the earlier 15-country study. Second, why were workers with neutron exposure and internal contamination included in our study contrary to the 15-country study? A concern raised regarding exclusion of these workers in the previous 15-country study was that the practical effect was to exclude a large number of workers with high cumulative external photon doses. Therefore, in INWORKS we included these workers and then evaluated the impact of their inclusion on associations of interest; we found little evidence of potential bias due to neutron exposure or internal contamination on the estimated radiation dose-leukaemia mortality associations. Third, Nagataki and Kasagi asked how we dealt with potential confounders such as medical exposures and smoking. Although we could not directly adjust for individual histories of medical exposures or smoking histories, we controlled for temporal trends in leukaemia risk factors, such as diagnostic x-ray exposure, in the analysis. In addition to covariate control, we note that it is unlikely that smoking is a strong confounder in view of its relatively weak association with leukaemia. Doss also raises three concerns. First, he argues that use of a linear model for analysing dose-response associations in this study is not justifiable and he advocates for a model in which low dose radiation would reduce cancer. We did not simply assume that the data fit a linear model. We examined the relative risk of leukaemia (excluding chronic lymphocytic leukaemia) across categories of cumulative dose, observed that the trend in the excess relative risk of leukaemia (excluding chronic lymphocytic leukaemia) with dose was well described by a linear function of cumulative dose, and noted that a higher order polynomial function of dose did not substantially improve the model fit. Second, Doss notes that, in our summaries of our findings, we report that radiation exposure was associated with leukaemia. We agree with Doss that the summary statements in our abstract and conclusion could be modified to clarify that our main finding pertains to leukaemia (excluding chronic lymphocytic leukaemia); however, we expect most will correctly infer this because we clearly refer to results for leukaemia (excluding chronic lymphocytic leukaemia) in the preceding sentence of the Summary and in the Discussion. Third, Doss contends that another concern is the absence of information on doses from medical procedures, which on a population level varied over time. Individual information on radiation doses from medical procedures is unavailable in INWORKS, as is the case for most occupational epidemiology studies. However, as noted above in our response to Nagataki and Kasagi, our analyses estimate the association between occupational radiation dose and mortality within strata defined by the cross-classification of country, attained age, sex, and calendar period. Thus, for confounding to occur, medical radiation exposures would need to be associated with occupational doses, within strata defined by these adjustment factors, which is unlikely to be the case. Doss incorrectly asserts that we stated that we had considered the model stratified by calendar year. We fit no such model. Rather we fit a model that adjusted for calendar period (using a statistical method known as background stratified Poisson regression), yielding a summary estimate of the excess relative rate per gray that represents the average association over the calendar periods, which was the parameter of interest. As pointed out by Doss, INWORKS did not evaluate dose effect modifiation by calendar period, but we note that an analogous evaluation (for birth cohort-dose interaction) in a component cohort found no evidence of effect modification for leukaemia excluding chronic lymphocytic leukaemia. In summary, the INWORKS study (like most observational epidemiology studies) has limitations, which we believe have been adequately described in our Article, and which, in our opinion, do not greatly affect its conclusions. [Description provided by NIOSH]
• Subjects:
  Morbidity Mortality Neoplasms Occupational Health Radiation Risk Assessment Risk Factors Safety Workplace
• Keywords:
  Cancer Exposure-levels Models Morbidity-rates Mortality-rates Neutron-radiation Nuclear-hazards Risk-factors Work-environment Workers
• ISSN:
  2352-3026
• Document Type:
  Text
• Genre:
  Journal Article
• Place as Subject:
  North Carolina ; Ohio ; OSHA Region 3 ; OSHA Region 4 ; OSHA Region 5 ; Pennsylvania
• CIO:
  National Institute for Occupational Safety and Health (NIOSH)
• Division:
  DSHEFS - Division of Surveillance, Hazard Evaluations, and Field Studies
• Topic:
  Workplace Safety & Health
• Location:
  North America
• Volume:
  2
• Issue:
  10
• NIOSHTIC Number:
  nn:20047397
• Citation:
  Lancet Haematol 2015 Oct; 2(10):e405-e406
• Federal Fiscal Year:
  2016
• NORA Priority Area:
  Services
• Peer Reviewed:
  False
• Source Full Name:
  The Lancet Haematology
• Collection(s):
  National Institute for Occupational Safety and Health
• Main Document Checksum:
  urn:sha-512:aac95fa0bfc004c02a07a8c667b1b680d96d58791c589cfe8fb89aeba408261bcfaa7530a9df550ef6f790819d45d2df1290a3bf619b0171000c20443763ff80
• Download URL:
  https://stacks.cdc.gov/view/cdc/202472/cdc_202472_DS1.pdf
• File Type:
  [PDF - 38.22 KB ]