Associations Between Early-Life Indoor and Outdoor Air Pollution Exposure and Childhood Asthma

BackgroundEarly-life air pollution exposure has been associated with childhood asthma, but its relative contribution is poorly characterised in cohorts with repeated indoor measurements. Short sampling windows and seasonal bias limit how well any single measurement can represent cumulative exposure. ObjectiveTo evaluate the association between early-life indoor and outdoor pollutant exposure from birth to 18 months and asthma development at ages 7, 12, and 18 years, while testing potential exposure-exposure and gene-environment interactions. MethodsIn the COPSAC2000 birth cohort (n = 411), indoor PM2.5, black carbon, NO2, NOx and VOCs, and modelled outdoor PM2.5, NO2 and NOx, were assessed at 6, 12 and 18 months of age. Indoor metrics were corrected for pump-flow drift and seasonal bias via a Gaussian process with a periodic kernel. Hierarchical Bayesian logistic regression(HBLR) and calibrated random forests (CRF) were evaluated using stratified nested cross-validation for discrimination using AUC-ROC, Calibration using integrated calibration index (ICI), and Brier skill scores. Exposure-exposure and gene-environment interactions were quantified via Shapley interaction indices. ResultsDiscrimination was weak at all endpoints with test AUC-ROC: 0.58-0.67, BSS near zero (0.012 to 0.05) and ICI: 0.043-0.061 for random forest and AUC-ROC 0.53-0.61, BSS equal to zero and ICI: 0.054-0.067 for hierarchical Bayesian logistic regression (HBLR). No pollutant reached a highest density interval excluding OR = 1. The polygenic risk score(PRS) was the only term non-null at all three ages in the adjusted models. While gene-environment interactions were detected, effect sizes were negligible and statistically indistinguishable from zero. SignificanceWhere ambient pollution is low, cumulative early-life exposure adds little to long-term asthma prediction beyond genetic and contextual factors, and an improved exposure correction does not uncover an effect. ImpactRepeated bedroom measurements across infancy are rare, and short sampling windows make them hard to read as cumulative exposure. We corrected measured concentrations for pump-flow drift and for seasonal sampling bias on the pollutants and used two complementary machine learning models to predict interaction and risk. In this low-exposure Danish cohort, genetic risk and residential context dominated, and the measured pollutants added almost nothing to the asthma risk at 7,12 and 18 years. These null results indicate that early cumulative exposures of indoor and outdoor air pollution contribute little to nothing to asthma up to 18 years.