Another important issue that has not been considered previously for this cohort is the ''missing dose" that may have occurred as the result of recording a zero for below-detectable doses, especially for frequent (weekly) film badge readings at X-10. The results presented here are based on a preliminary dose adjustment procedure that was derived from a small sample of hard copy X-10 personnel dose records [2,3,5]. The sample was needed because daily and weekly radiation monitoring data prior to 1957 are not available in electronic form. Future studies that include X-10 workers should use all available dosimetry data to deal with the uncertainty (systematic bias and random measurement error) in these data [32,33,34]. Our preliminary dose adjustment method does not consider other sources of systematic (e.g. neutron exposures) and random measurement error, nor did we consider the effect of measurement error on other covariates (internal exposures, SES) used in the analyses. Results based on adjusted doses are consistent with earlier studies, but suggest that the effect of missing dose was to introduce an upward bias in the dose-response coefficient and the score test statistics. Differences in measurement of external radiation at X-10 and Y-12 are relevant to the results reported in Table X , which show consistently increasing dose coefficients for longer lags only at X-10. The lag-20 coefficients are based entirely on exposure measurements made before 1965. For most of those years, less than one quarter of Y-12 workers routinely wore external dosimeters, while the proportion for white male X-10 workers after 1946 was much higher . If there was an association between radiation exposures and cancer mortality during this period, it would be difficult to detect if few workers were monitored. In addition, recorded external exposures were higher at X-10 than at Y-12 during this period, resulting in a less skewed dose distribution upon which to base the dose- response estimates.