A1735
Title: Spatial causal analysis: Case study of testicular cancer and PFAS exposure in Veneto, Italy
Authors: Allegra Sartore - University of Padova (Italy) [presenting]
Dolores Catelan - University of Padua (Italy)
Tony Fletcher - London School of Hygiene and Tropical Medicine (United Kingdom)
Cristina Canova - University of Padova (Italy)
Mirko Berti - University of Padova (Italy)
Giorgia Stoppa - Padova (Italy)
Annibale Biggeri - University of Padova (Italy)
Abstract: Causal inference is essential in epidemiologic research and public health. Ecological studies using spatial or spatio-temporal data face challenges with spatial autocorrelation, violating the ``no interference'' assumption. An ecological regression study analyzed orchiectomy rates for testicular cancer across 21 municipalities in Veneto, Italy, where water was contaminated by PFAS. Exposure data included municipality-specific geometric means of PFOA serum concentrations (GMPFOA), adjusted for individual covariates. A Bayesian hierarchical spatial model was specified with a propensity score model on GMPFOA and socio-demographic variables, plus a shared spatial random component. The disease model included GMPFOA, socio-demographic variables, the shared spatial component, and a specific random component for response spatial patterns. The regression coefficient for the exposure of interest (logRR per 30 ng/L PFOA) was 0.315 (95\% Credibility Interval: -0.018, 0.668). Using the propensity score approach, an adjusted causal regression coefficient of 0.829 (95\% CrI: 0.078, 1.616) was found. This approach incorporates two spatial random components to meet assumptions for causal interpretation of the regression coefficients. Sensitivity analysis discusses structural equation models and instrumental variables. A dose-response association between testicular cancer risk and PFOA exposure is confirmed.
