monitored for exposure to radionuclides (Rooney and others 1993) in the period 1946–1979. No such increase was observed in the NRRW (Muirhead and others 1999), which included all of the monitored workers in these two studies.
A significant positive association with lung cancer was observed in the AWE and ORNL studies (Beral and others 1988; Wing and others 1991), particularly among those exposed to radionuclides in the AWE and in nonmonthly workers at ORNL. Information on tobacco smoking was available systematically in these studies. A few other significant associations were reported in single studies (Table 8-4). Given the number of associations examined, some of the significant results observed may have been due to chance.
Several points must be kept in mind when making comparisons of these worker-based risk estimates and confidence intervals with those based on high-dose-rate studies. The most important are possible biases and uncertainties in dose estimates, errors in outcome data, and inadequate adjustment for confounders.
Among the very large and potentially most informative cohort studies reviewed in this chapter, two present a number of problems that limit their informativeness. In the Canadian NDR study (Ashmore and others 1998), the very low standardized mortality ratio (SMR)3 for all-cause mortality (61) suggests that record linkage procedures between the Canadian National Dose Registry and the Canadian Mortality Data Base may have been imperfect. There could have been some confounding of the dose-response because of associations between the probability of successful linkage and factors (e.g., socioeconomic status [SES]) associated with occupational radiation dose. This is the only study in which associations have been observed between radiation dose and all-cause mortality, all cancer mortality (without any clear relation to specific cancers), mortality due to cardiovascular diseases (males and females), and fatal accidents (males only). Moreover, no information is available on SES—a factor that has been shown in a number of previous cohorts to be a confounder of the association between radiation dose and cancer risk. Almost three-quarters of the cohort consists of radiation workers employed in different settings (dentistry, medicine, industrial radiography), where radiation control may be very different (possibly less uniform and systematic due to the much smaller numbers of persons monitored in individual workplaces) than in the nuclear industry.
In the Japanese NDR study (ESGNWJ 1997), SES information is also not available. Further, because of difficulties in carrying out vital status follow-up in Japan, the very large cohort had to be restricted to those who were employed in the previous 5 years; hence the follow-up time of this cohort is very short, and older workers as well as workers with higher doses (who were employed in early years and left employment more than 5 years in the past) have been excluded from the follow-up. Consequently the study has little power to estimate possible health risks associated with occupational radiation exposure.
The accuracy and precision of individual dose estimates in the nuclear industry is a function of time, place, radiation energy and quality, the geometry of the radiation exposure, and the location of the dosimeter on the body of the worker.
Efforts were made in some of the studies (AEA, Sellafield, ORNL, U.S. DOE combined analyses, three-country study, Saclay site in France) to assess the importance of dosimetric errors due to administrative practices adopted by facilities to calculate and record personnel dose based on technical, administrative, and statutory compliance considerations (Adams and Langmead 1962; Smith and Inskip 1985; Taylor 1991; Telle 1995; Tankersley and others 1996; Mitchell and others 1997; Watkins and others 1997; Telle-Lamberton and others 1998). Results of reanalyses of data using different approaches to estimate doses from missing dosimeters or below-threshold readings have yielded similar results to the analyses based on original data (Inskip and others 1987; Little and others 1993).
In the three-country combined analyses, a retrospective dosimetry study was carried out to identify the various sources of biases and random errors in dosimetry for workers in each of the facilities included and to estimate the magnitude of these errors. As a result, it was concluded that for the majority of workers with predominant high-energy (300–3000 keV) photon exposures at levels greater than the detection threshold of the dosimeter, there is no strong reason to believe that available dose estimates substantially underestimate or overestimate deep dose. The dose estimates were judged to be compatible across facilities and over time. However, available dose estimates may have overestimated dose to the bone marrow by up to 20%. Estimation of leukemia risk adjusting for this overestimation yielded an ERR of 2.6 Sv−1 instead of 2.2. For deep organs, the factor is likely to be smaller, of the order of several percent. Random errors in dose estimates are likely to bias the risk estimates downwards, compared to estimates from high-dose studies, which have been based on organ doses. At lower exposure levels however, practices for recording subthreshold doses have resulted in a slight underestimation of doses from predominant higher-energy photon exposure (Fix and others 1997).
At the Hanford plant in the United States, based on experiments and expert assessments, efforts were made to