Methodological issues in occupational cohort mortality studies and their implications for a study of mortality in a cohort of workers exposed to beta-naphthylamine

Catherine Dee Axtell, Purdue University

Abstract

A mortality study of workers exposed to aromatic amines, primarily beta-naphthylamine (BNA), while employed at a facility that manufactured synthetic dyes in Augusta, Georgia was updated by adding nearly 15 years of follow-up. Significant elevations in the risk of death due to lung cancer, nonmalignant respiratory diseases, violent deaths, and other causes including unknown causes were found among 413 white males. Among 872 nonwhite males, deaths due to psychoneurotic disorders including alcoholism, cerebrovascular disease, cirrhosis of the liver, and all other causes including unknown causes were significantly elevated compared with rates for U.S. nonwhite males. There was no evidence, however, for an association between any cause of death and length of employment in either race group. The majority of the cohort members were short-term workers, and only 15 percent may have actually been employed in jobs involving exposure to BNA. Mortality may not have been an appropriate measure of the risk of bladder cancer associated with exposure to BNA, since 5-year survival rates for bladder cancer are currently 81% for males, and because a bladder cancer screening and notification program set up for these workers may have increased medical surveillance and reduced the case-fatality rate for bladder cancer in this cohort. Methods for estimating power in the design phase of occupational cohort studies were developed, and their results were compared using the Augusta cohort. Power was accurately estimated by (1) assuming cohort members would die at the age- and calendar period-specific mean death rates for the reference population, and (2) reducing person-time as deaths occurred over the follow-up period to estimate person-years at risk. Methods which incorporated stochastic elements and used sampling distributions of the observed number of deaths to avoid using Poisson density functions, produced comparable estimates and were found to be unnecessary to estimate power accurately in this cohort. Finally, a review of power in 44 occupational cohort mortality studies with SMR analyses which were published in five journals in 1993-1994 found that 63 (78%) of 81 cause-specific hypothesis tests did not have adequate power $(\geq$0.80) to detect a 50% elevation in mortality among the exposed, and 36 (44%) did not have power $\geq$0.80 to detect a 100% elevation.

Degree

Ph.D.

Advisors

Clickman, Purdue University.

Subject Area

Occupational safety|Toxicology

COinS