Study Quality

This is a pooled re-analysis of data collected from children who participated in seven international population-based longitudinal cohort studies (Lanphear et al., 2005, 2019). Children were followed from birth until 10 years of age. The pooled analysis comprises a diverse study population, including children from the United States, Mexico, Australia, and the former Yugoslavia. Venous or fingerstick capillary blood samples were collected and the BLL most proximal to the neurocognitive testing was used in this re-analysis. Full-scale IQ was measured using a version of the Wechsler Intelligence Scales for Children administered between 5 and 10 years of age. The authors considered a range of potentially important confounders, relying on a priori criteria for final model specification. The authors also evaluated interaction terms between BLLs and each included covariate.

ISA Relevance

This paper extends the analysis of the international pooled data first examined by Lanphear et al. (2005, 2019) and later by Crump et al. (2013). Van Landingham et al. identified "highly likely" confounders (HOME score, maternal education and maternal IQ) using a combination of correlation analysis and backward selection with a criteria of p=0.15 to retain covariates in the model. Van Landingham et al. (2020) also included interaction terms between BLL and each included covariate, regardless of the significance level of the interaction term. The authors fit predictive models and calculated the IQ decrement for various increments of BLL, and across levels of each of the covariates. The interpretation of the covariates and the predicted IQ decrements at various combinations of BLL decrement and covariate levels that are presented in the paper, however, potentially conflate the interpretation of direct versus total effect estimates and may not inform causality (i.e., the table 2 fallacy; Westreich and Greenland, 2013).

Van Landingham, 2020, 7295317

Study evaluation