Using latent profile analysis to examine associations between gestational chemical mixtures and child neurodevelopment

In this study, we introduced Latent Profile Analysis (LPA) as a novel technique for studying gestational chemical mixtures. Using data from the Maternal-Infant Research on Environmental Chemicals Study, a longitudinal birth cohort study of pregnant Canadian women and their children, we examined the relationship between 30 gestational biomarkers and Verbal IQ, Performance IQ, and Full-Scale IQ. We generated five latent profiles: A Reference profile, a High Level profile, a Low Level profile, a High Organophosphate Pesticides profile, and a Smoking Chemicals profile. Multiple regression analysis showed strong negative associations between the Smoking Chemicals profile and IQ scores. We also found positive associations between the Low Level profile and IQ, and a negative association between the High Level profile and Verbal IQ. However, all 95% confidence intervals spanned the null. After conducting sensitivity analysis comparing LPA with k-means clustering, we concluded that LPA is a promising alternative to other clustering methods.