O26-1 An analytical approach for the estimation of causal effects of occupational exposures in left censored cohorts. (1st September 2016)

• Record Type: Journal Article
• Authors: Izano, Monika A; Costello, Sadie; Picciotto, Sally; Neophytou, Andreas M; Brown, Dan; Eisen, Ellen A
• Is Part Of: Occupational and environmental medicine. Volume 73(2016)Supplement 1
• Page Start: A49
• Page End: A49
• Publication Date: 2016-09-01
• Subjects: Medicine, Industrial -- Periodicals; Environmental health -- Periodicals; 616.980305
• DOI: 10.1136/oemed-2016-103951.132
• Languages: English
• ISSNs: 1351-0711

Abstract

Abstract : Metalworking fluids (MWFs) – complex mixtures of mineral oils, PAHs, and chemical additives widely used to cool and lubricate metal machining operations – have been linked to a number of cancers. With an estimated 4.4 million U.S. workers exposed to MWFs in 1997, and millions more worldwide, MWF exposure poses a major potential cancer hazard. The United Autoworkers-General Motors (UAW-GM) mortality study of 46, 000 workers in the automotive manufacturing industry, followed from 1941 to 2009, provides a unique opportunity to examine the causal relationship between quantitative exposure metrics for MFWs and incident cancers. Cancer incidence follow-up in the UAW-GM cohort starts when the Michigan Cancer Registry was established in 1985, up to decades after all workers were hired. Since only workers still alive in 1985 are eligible for incidence follow-up, the sub-cohort may be a biassed sample of the full cohort. Because we have data on all subjects in the mortality, the UAW-GM study offers an opportunity to address the potential bias in the left-censored cancer incidence sub-cohort. We establish an analytical framework for the estimation of effects of MFW exposure on incident cancers, under specific assumptions. First, we show in a simulation study that analyses based on left censored subgroups tend to be biassed and that the bias may increase when aspects of the healthy worker survivor effect (HWSE), such as heterogeneous susceptibility and time-varying confounding, are present. Next, we apply an estimation approach designed to reduce left-censoring bias and report its performance in the presence of the HWSE. We conclude with a roadmap for the application of the procedure in the UAW-GM cohort to estimate the causal effects of MWF on incident cancers, under dynamic interventions that satisfy the experimental treatment assignment assumption.