Document resource
Objective The association between exposure to welding fumes and lung cancer has been extensively studied. The most common exposure metric is the cumulative index of exposure (the product of fume concentration and duration of exposure). Individuals with the same cumulative exposure but different temporal exposure patterns may show different risks. Therefore, there is still a need for research to adequately capture the time-varying intensity of exposure and to identify critical time-windows during which exposure has the strongest impact on lung-cancer risk.Material and Methods Latent Class Mixed Models (LCMM) simplify heterogeneous lifetime exposure into more homogeneous classes and identify distinct subgroups of individuals, following a similar exposure pattern. We determined latent classes for welding-fume exposure in two German population-based case-control studies (3,498 lung-cancer cases and 3,539 control subjects) and we used these classes to estimate smoking-adjusted OR with 95% CI via logistic regression. Before applying the LCMM function, exposure levels for each welding activity were determined using a measurement-based job-task-exposure-matrix with estimates from 15,473 personal measurements of inhalable fume taken at welding workplaces.Results LCMM identified four latent classes of welding-fume exposure as the best solution according to fit and diagnostic criteria. The highest lung-cancer risks were observed for the class in which welding-fume exposure in the past 10 years before the interview/diagnosis was highest (median 450 µg/m3) with an average duration of welding of 30 years (OR=1.71, 95%CI 0.92-3.15). Participants in one other class with long-term high intensity (median up to 1,000 µg/m3 experienced more than 20 years before the interview/diagnosis) also showed higher lung-cancer risks compared to non-exposed men (OR=1.26, 95%CI 0.46-3.49).Conclusions The highest relative lung-cancer risks were observed after a recent high exposure to welding fumes. LCMM opens new perspectives of dose-effect relationships and could be employed to complement established methods in occupational epidemiology.