Proceedings of the 6th International Symposium on Epidemiology and Occupational Risks, Braz, Austria, April 22-24, 1998. Vienna, Austria: International Section for Research, 1998 Apr; :128-131
Many of the approaches available for quantitative risk assessment of occupational hazards have been developed for analyses of animal bioassay data. Methods that have been developed for animal bioassay data n-e often not applicable to epidemiologic data, primarily because of differences in study design and data structure. In the past, epidemiologic data has been often underutilized and misused in quantitative risk assessment (QRA). The explanation for this appears to be that risk assessment has been a field dominated by toxicologists and statisticians, and perhaps more importantly that few epidemiologists have taken an interest in this area. This situation appears to be changing as there has been a growing recognition of the importance of epidemiologic studies for cancer and non-cancer risk assessment. This trend has in part been fueled by the growing skepticism towards the use of toxicologic data for predicting human risks [e.g. AMES and GOLD, 1990]. It also appears that more epidemiologists are becoming interested in participating in the QRA process. SMITH  has suggested that the primary limitation of epidemiologic data for QRA is the quality of the exposure data, and the primary limitation of toxicologic data for QRA is that it is based on the wrong species. He has also argued that the uncertainty surrounding the exposure estimates for epidemiologic data is generally much smaller than the uncertainties surrounding the extrapolation of data from animal studies to predicting human risks. There are, however, many other potential sources of uncertainty in using epidemiologic data for QRA [STAYNER, 1992]. Potential for confounding and other sources of bias (e.g. selection bias) may often severely limit, if not completely invalidate the use of epidemiologic data for QRA. Inadequate length of follow-up may also be a. serious limitation, particularly since we are generally interested in QRA in estimating lifetime risks and few epidemiologic studies follow a cohort for an entire lifetime. The fact that humans general1y have multiple exposures complicates QRA for single agents; however, the fact that effects of hazards are studied in. a real world .context may also be viewed as an advantage that epidemiologic studies have over toxicologic studies. Small sample size and resulting low statistical power may be the most severe limitation of epidemiologic studies particularly for detecting low level risks that are of potential public and regulatory concern. For example, the sample size that would be needed to have an 8 percent chance (power) of detecting a lung cancer excess in a cohort study for varying levels of excess risk are presented in this slide. Most cohort studies would have an adequate sample size to detect an excess risk of 1 in a 100, which is estimated to require approximately 4,000 workers. However, very few cohort studies would include the more than 400,000 workers that would be required for detecting an excess lifetime risk of 1 in a 1000 which is a threshold for setting occupational standards in the U .5. [INFANTE, 1995]. A study size of approximately 400 billion workers would be needed to detect a 1 in a 100,000 risk, which is greater than the world population (approximately 6 billion). Obviously, cohort studies are simply incapable of detecting excess lung cancer risks less than 1 in a 100,000 that are generally of concern in setting environmental standards. Several attempts have been made in the past to use epidemiologic studies as a basis for validation of risk assessment models that have. been developed using animal bioassay data [ALLEN et aI., 1988; ZEISS, 1994]. Methylene chloride is a classic example in which several authors have used different methods for making comparisons between animal based RA models and epidemiologic findings, and have reached very different conclusions about their consistency [STAYNER and BAILER, 1993]. In this figure, the standardized mortality ratios and confidence intervals for lung and liver cancer reported in a study of Kodak workers exposed to methylene chloride is contrasted with those predicted from a QRA model based on an animal bioassay for liver and lung cancer. It may be seen from this graph that the confidence intervals from the epidemiologic study clearly encompass the intervals from the animal based QRA model at all of the exposure levels reported in this study. This result is actually what maybe expected in using negative epidemiologic studies for testing QRA models based on animal data, since the variability surrounding the epidemiologic findings is often large.
Quantitative-analysis; Laboratory-animals; Bioassays; Epidemiology; Statistical-analysis; Risk-analysis; Risk-factors; Cancer-rates; Occupational-hazards; Occupational-health; Work-analysis; Work-environment; Worker-health; Workplace-studies; Liver-disorders; Lung-cancer; Lung-disorders; Reproductive-system-disorders; Pulmonary-system-disorders
Proceedings of the 6th International Symposium on Epidemiology and Occupational Risks, Braz, Austria, April 22-24,1998