| This analytic data file set consists of 66 files that were used by Cary Suzanne Lea for an updated cohort mortality analyses of white males employed at ORNL. The CEDR data set ORX10A02 was the primary source of data (demographic, work history, radiation monitoring). The purpose of this dissertation was to address the following research questions. Can excess cancer be explained as due to: (1) confounding by time-related factors, (2) selection bias resulting from choice of empirical induction period, (3) grouping of cancer outcomes, and (4) impact of high-dose outliers in a cohort of workers experiencing low-level cumulative dose? |
| ORNL began monitoring personnel for exposure to external penetrating radiation, primarily gamma rays, in 1943. Pocket chambers were used until June 1944, when film dosimeters (film badges) became the primary dosimeter. Film dosimeters were used for personnel monitoring until 1975, when they were replaced with TLDs. From 1943 until the early 1950s, the usual practice was to provide personal dosimeters to only those workers entering designated areas where there was a potential for exposure. Subsequently, all workers at ORNL were monitored. |
| This dissertation re-analyzed data for a cohort of 8,318 white males employed at ORNL for at least 30 days between 1943 and 1972 with follow-up through 1984. This re-analysis was motivated to understand more fully the results published by Wing et al., 1991 (JAMA, 1991), since mortality results differed from historically similar cohorts of nuclear energy workers. |
| Due to rounding, there are 213,950 person-years in ORX10A04, whereas in ORX10A02 there are 215,680 person-years. All other aspects of the source data for ORX10A04 are the same as ORX10A02, as described below. |
| During the course of the prior study (ORX10A02), vital status was ascertained for 91.8% of the cohort (96.5% of potential person-years of follow-up), and 1,524 deaths were identified by the end of 1984, the study end date. Death certificates were obtained for 1,490 (97.8%) of these deaths. External radiation monitoring data were used to compile annual and cumulative whole-body doses. For the 4.9% of the work-years for which external monitoring data were not available, doses were estimated. |
| The results of this dissertation analysis were the following: (1) confounding by time-related factors -- year of hire was significantly related to mortality risk, but duration of employment and length of employment were not. (2) Selection bias resulting from choice of empirical induction period -- latency refers to removing deaths and person-years during some pre-determined interval of follow-up, say the first 5 years of follow-up. Lagging refers to shifting dose forward by the number of years in the lag interval, say 5 years. There was no difference in mortality risk using either of these approaches for incorporating an empirical induction period. (3) Grouping of cancer outcomes -- grouping by cancer outcome indicated that smoking-related cancer sites, excluding lung cancer, showed the highest mortality risk. |
| The analytic data for this study are partitioned into separate data files according to specific factors of interest (all cancers, solid cancers, lung cancers, smoking-related deaths, and leukemia) in various combinations and over selected lag and latency times. |
