The use of structural equation modeling to estimate the effect of low-level arsenic exposure on birth weight in Chilean newborns

Abstract

Introduction: Prenatal arsenic exposure and pediatric outcomes have been previously explored. Although some studies have shown an inverse association between birth weight and arsenic exposure (iAs), the relationship remains unclear. The main objective is to use structural equation modeling to estimate the effect of low-level iAs exposure on birth weight while controlling for potential mediation effects and exogenous effects. Methods: This study is a cohort of pregnant women, recruited from 5 primary health care centers in Arica, Chile during 2013. Sociodemographic information and urine samples were collected during the 2nd trimester. iAs exposure during pregnancy was estimated by determining the sum of arsenical species. Birth weight was obtained from birth records in grams (g). Structural equation models were then used to estimate total iAs effects, indirect and direct effects, on birth weight. Results: The median of total urine iAs was 14 μg/L (IQR=12.3 μg/L). The median of birth weight was 3350 g (IQR=480 g). After adjusting for risk factors, iAs was no longer associated with birth weight (β=4.373, 95%CI: -2.02, 10.77). The Chi-square test model had a p-value of 0.063 and the root mean square error of approximation (RMSEA) had a p-value of 0.062. Moreover, total effects of iAs on birth weight had a borderline association (p-value: 0.051). The indirect effects were not mediated by gestational age or gestational diabetes. In this sample, gestational age (β=213.5, 95%CI: 163.6, 263.4), gestational diabetes (β=-112.8, 95%CI:-192.2, -53.4), maternal age (β=11.9, 95%CI: 0.91, 22.9) and newborn sex (β=264.3, 95%CI: 137.1, 391.5) were associated with newborn birth weight. Conclusion: In our study, iAs exposure was not directly associated with birth weight. However, iAs total effects did show a tendency towards having an effect. Gestational age, maternal age, and newborn sex were positively associated with birth weight, which is consistent with previous literature.