Lagged Weighted Quantile Sum Regression

BACKGROUND AND AIM: We aim to develop a method for identifying critical exposure windows relating exposure to environmental mixtures and health outcomes that may differ by subject characteristics (e.g., sex). We extend weighted quantile sum (WQS) regression with its variations (e.g., stratified interaction WQS regression) to a lagged version that accommodates heterogeneous exposure times for each subject. METHODS: Briefly, time-specific weights are estimated and the weighted index calculated (i.e., WQS(t)) across the time frame with at least moderate sample sizes. Then a reverse distributed lagged model (rDLM) is similarly parameterized and estimated using a nonlinear mixed model. The rDLM allows for different exposure times per subject while correcting for within-subject correlated observations. Critical windows are identified where the family-wise confidence intervals for the association parameter(s) exclude zero. RESULTS:The approach is illustrated using perinatal exposures measured in shed teeth with later life health effects in a lagged stratified interaction WQS model. CONCLUSIONS:Identifying environmental chemicals of concern in critical exposure windows, that may vary by subject characteristics, may be useful in intervention strategies to protect public health.