Mapping the Occupational Exposome for Respiratory Health

Abstract

Occupational respiratory diseases arise from complex and heterogeneous workplace exposures that vary across occupations, industries, and time. Establishing causal relationships between occupational exposures and respiratory health remains challenging because individual studies often have limited statistical power, occupational exposures frequently occur in combination, and accurate exposure assessment is difficult. This thesis addresses these challenges by developing and applying an integrated framework for mapping the occupational exposome in respiratory health research. It combines methodological evaluation, large-scale pooled epidemiological analyses, systematic evidence synthesis, and discovery-driven multi-exposure modelling. First, the thesis evaluates scalable approaches for occupational exposure assessment by comparing three automated job coding tools, AUTONOC, CASCOT, and LabourR, with manually coded occupational histories from the AsiaLymph study. Agreement was modest at detailed occupational coding levels but improved for broader occupational groupings and when job codes were translated into occupational exposures using job-exposure matrices. These findings demonstrate the potential of automated coding to improve the efficiency of large-scale occupational studies while highlighting the need for study-specific validation and further methodological development. Second, hypothesis-driven analyses were conducted within the SYNERGY pooled case-control dataset to investigate occupational exposure to chlorinated solvents and benzene in relation to lung cancer. For chlorinated solvents, there was limited evidence of an overall association, although suggestive exposure-response patterns were observed for some exposure metrics. In contrast, occupational benzene exposure was consistently associated with increased lung cancer risk, with exposure-response relationships observed for cumulative exposure and duration. These associations persisted after adjustment for smoking and other occupational carcinogens. A subsequent systematic review and meta-analysis of 13 studies, comprising 366,975 participants and 17,030 lung cancer cases, further supported this association. The pooled relative risk for occupational benzene exposure was 1.14 (95% CI: 1.03 to 1.27), with evidence of increasing risk with greater cumulative exposure. Together, these analyses strengthen the epidemiological evidence linking occupational benzene exposure to lung cancer. Finally, the thesis extends beyond conventional single-exposure analyses by applying an occupational exposome framework to longitudinal lung function decline in the European Community Respiratory Health Survey. Forty-nine occupational exposures were examined simultaneously using ensemble machine learning and complementary causal inference methods. Strong correlations between exposures reflected the complex co-exposure patterns characteristic of real-world workplaces. Exposure groups related to ergonomic stressors, physical stressors, gaseous substances and fumes, and particulates and fibrous dusts were among the most important predictors of accelerated decline in forced expiratory volume in one second (FEV1). Although individual causal effect estimates remained uncertain, this analysis demonstrates the feasibility of integrating high-dimensional exposure data, ensemble learning, and group-based interpretation methods in occupational epidemiology. Overall, this thesis advances both substantive and methodological understanding of occupational determinants of respiratory health. It shows how complementary approaches, ranging from scalable exposure assessment and pooled hypothesis-driven analyses to systematic evidence synthesis and discovery-driven modelling, can be combined to address different dimensions of the occupational exposome. The findings strengthen the evidence for benzene as an occupational lung carcinogen and provide a broader analytical framework for investigating complex, multifactorial exposure-disease relationships in occupational and environmental epidemiology.