Recently there have been calls to reevaluate standards governing asbestos in light of suggestions that chrysotile asbestos is relatively safe. Arguments for the relative safety of chrysotile note that while workers in a South Carolina asbestos textile plant using chrysotile experienced substantial excesses of lung cancer, Canadian chrysotile miners have not. One possible explanation for these divergent findings is that the carcinogenic effect of chrysotile exposure is determined, in part, by fiber size and morphology. The North Carolina Asbestos Textile Study addresses this hypothesis with data on a cohort of workers at four previously unstudied asbestos textile plants. In a project supported by NIOSH (R01OH007803), we have reconstructed historical exposures in these plants, applied novel techniques to characterize the morphology and size distribution of fibers and carried out epidemiological analyses revealing a significant excess of lung cancer relative to the US population (SMR 1.95 95% CI 1.73-2.20). This application is a competing continuation of that project. In contrast to the original study's emphasis on assessing exposures among North Carolina asbestos textile workers, this continuation focuses primarily on quantitative risk assessments using data from the North Carolina cohort and from a cohort of South Carolina asbestos textile workers. We propose a series of analyses that go well beyond those planned for the original project. These analyses are motivated by prior findings that strongly suggest a peak, then decline, in lung cancer risk following asbestos exposure. The analyses we propose here employ advanced modeling approaches to produce estimates of the change in the rate of lung cancer per unit of exposure that vary with time since exposure. If an exposure effect varies over time, then such methods can inform risk assessments and obtain results that are consistent between study populations. Specifically, we propose to: 1) apply flexible statistical models to describe latency effects including exposure time-windows, bilinear, sigmoid, and cubic B-spline latency models;2) apply biomathematical cancer models as a complement to the empirical models of exposure-time-response associations;and, 3) conduct pooled analyses of exposure and epidemiologic data from North Carolina and South Carolina asbestos textile worker cohorts. The work outlined here will substantially improve the quantitative risk estimates derived from historical cohort studies of US chrysotile asbestos textile workers. The proposed work will substantially improve quantitative risk estimates derived from historical cohort studies of US chrysotile asbestos textile workers and will provide much needed information concerning fiber characteristics most strongly predictive of excess lung cancer risk. The results may help to explain some of the heterogeneity that has been observed among cohorts exposed to chrysotile. Pooled analyses of exposure and epidemiologic data from North Carolina and South Carolina asbestos textile cohorts will strengthen the precision of exposure assessment and risk estimates and improve the ability to characterize exposure-time- response associations.
