UNCLASSIICD sure data on emploveesatdifferentsites, or within sites over time, have varied widely. In the earlier years only a fraction of the workers were monitored, and there was s¢nousrisk of under-reporting exposures. At some sites, for many workers, it is impossible to distinguish between unmonitored years and years with a zero dose. At othersites, a zero dose was recorded for any exposure below the threshold of film badges. In one case spurious “correction factors” were invented to lower exposure figures and give some workers a “negative radiation dose,” something that does not exist in nature. One research team has concluded that there is “no constant relationship berween recorded doses. . . and actual workers, and the informationis typically out of date by five to seven years or more. This is reflected by the limited length of follow-up reported in manystudies. For most nuclear weapons workers covered by the published studies, follow-upis far short of the period required before many forms of cancer, especially solid tumors, appear. Such studies are therefore radically incomplete, and the reported absence ofsignificant findings may constitute a false reassurance. These deficiencies are more scrious in view of a few recent studies finding more cancer deaths during extended follow-up periods. It is noteworthy that those more recent DOE epidemiologic studies cover only a relative handful of the 76 nuclear weapons . research, production ond testing sites. While some of the DOE’s published srudies may acknowledge the healthy workereffect, taey rarely regard excess, but not statistically significant, worker death rates as warning signals. Instead, they tend to DEAD RECKONING i 4 Re Of the cumulative total of approximately 600,000 nuclear weapons workers, large numbers are not represented in published DOEstudies. From 1947 to 1978 at somesites, no exposure data were kept on the employees of subcontractors. Data on thousands ofworkers are incomplete. By 1990, only 250,000 workers . were represented in computerized databases. At one site involved in a study ofall workers exposed to 5 rem of external radiation in any one year, records are so confused that the true number of workers exposed at that level may be three times greater than the number included in the study, and the numberexposedat 4 to 5 rem (many of whom mayin fact have had higher exposures) is ten times greater. The published DOE epidemiologic studies cover only a relative handful of the 76 nuclear weaponsresearch, production andtesting sites. Because DOE sites vary in the industrial processes they employ, and average radiation exposures vary widely at different sites, the published research Many of the benign or dismissive interpretationsof excess cancerrisk in nuclear weapons workers as compared to the general population— interpretations that are consistently found in DOEsponsored studies—give insufficient weight to the “healthy worker effect,” which predicts /ower risks of disease for workers. The workforce almost always has low mortality in comparison to the populationatlarge, since thelatter includes many more people at high risk of poor health, who are too sick to work, who lack good medical care, who have lower average socioeconomic status and higherrates of smoking, etc. Years of research has taught that overall death rates, and death rates from specific diseases such as cancer, will be lower among workers than in the general population. For example, any comparative increase in death rates for cancer among workers runs counter to expectations and calls for further investigation and follow-up. Despite widespread knowledge of the healthy worker effect, studies that are subject to this form ofbias continue to be conducted; the majority of published DOE studies are plagued by this problem. rnp Coverage of the Workforce and of DOE Sites The “Healthy Worker” Effect and the Lack of Moritdity Data wd 7 : pare workers with higher radiation exposures to those with Jower or no exposures, the only proper method of analysis. Thereis also a pervasive lack of data on workers’ medicalirradiation histones, smoking and other factors which could distort or confuse findings. ods tend to report higher cancer mortality rates and morefindings that are statistically significant. piememe worse the data, the harderit is to com- studies which cover longer time peri- ie classification bias (mixing exposed and unexposed workers together, which would dilute the estimated effect). The There have been long and inexplicable delays in gathering death data on many nuclear weapons plant UNGL effects and raising the possibility of mis- Length of Follow-up of Worker Cohorts The published DOE doses.” At five important DOEsites, no radiation exposure data are available for epidemiologic studies; at others, computenzation of exposure data and linkage to individual workers are years out of date. The great majority of published DOEstudies do not present any individual-specific exposure data, thereby limiting the analyses of health findings may have overlooked someserious hazards to health.