Effects of maternal arsenic exposure on birth outcomes using harmonized data across three birth cohorts

Abstract Objectives To assess the association between urinary arsenic concentrations and birth outcomes by harmonizing data from three independent birth cohorts. Methods We harmonized and analyzed data from the Navajo Birth Cohort Study (NBCS), the New Hampshire Birth Cohort Study (NHBCS), and the PROTECT Center study based on Puerto Rico. Birth outcomes of interest included birth weight, head circumference, birth length, gestational age at delivery, incidence of preterm birth, and size for gestational age. Urinary arsenic concentrations were used as the primary exposure metric. Harmonization involved aligning variable formats, adjusting for differences in laboratory methods, and excluding incompatible covariates, such as income and race. Results Harmonization increased the total sample size ( N = 3222) across cohorts. However, pooled analyses did not consistently demonstrate increased statistical power. Effect estimates for arsenic exposure were attenuated in some cases, and confidence intervals remained wide or even expanded relative to individual cohort analyses. Differences in biospecimen collection and laboratory assay methods required cohort-specific adjustments. Due to missing arsenic speciation data, the PROTECT cohort was excluded from two exposure models. Variability across cohorts limited the interpretability and precision of pooled estimates despite harmonization efforts. Conclusions While harmonizing data across multiple cohorts increased the sample size, it did not necessarily enhance statistical power or strengthen observed associations. Differences in data collection, laboratory methods, and available covariates posed significant challenges. These findings underscore the need for caution when interpreting pooled results from heterogeneous sources and highlight the importance of prospective planning for data harmonization in multi-cohort studies.