Assessing the effect of using exposure imputation approaches on the association between nitrate concentrations in public drinking water and birth outcomes in Ohio, 2006-2013

摘要

Background: Epidemiologic evidence suggests that in utero nitrate exposures are associated with adverse reproductive outcomes. Study limitations have included limited sampling frequency and exposure misclassification, which may have diminished the ability to quantify exposure-response relationships. Objective: This retrospective cohort study examines the effects of using various exposure imputation methods on the association between nitrate exposures in public drinking water and low and very low birth weight (LBW, VLBW) in infants born in Ohio from 2006-2013. Methods: Sampling frequency of nitrates from 1,045 Ohio public water systems (PWSs) ranged from 1 -92 times per year. Most PWSs sampled annually, wherein an imputation method based on the exposure distribution for each year by quarter was used to estimate quarterly exposures. Birth data included all term singleton births (n=1,085,948). Nitrate data were matched to maternal zip code at time of birth. Nitrate exposure was calculated as an average concentration of exposure during pregnancy. Logistic regression was used to estimate the association between nitrate exposure and both LBW and VLBW. Potential confounders included maternal age, number of prenatal care visits, change in maternal weight (during pregnancy), pre-pregnancy body mass index, marital status, maternal education, smoking, race/ethnicity, gestational age, and parity. Results: Non-imputed and imputed nitrate exposure values were highly correlated (rs=0.87;p<0.05) with an overall range of 0.003 to 57.9 mg/L. When comparing non-imputed versus imputed exposure quartile classifications, the 72% of non-imputed quartile 1 scores that were reclassified were all now in the intermediate quartiles. Quartile 4 results were largely unchanged, with only 7% of non-imputed scores reclassified as quartile 3. When comparing non-imputed versus imputed exposures with VLBW (aOR=0.89; 95%CI: 0.63, 1.26 vs. aOR=1.15; 95%CI: 0.90, 1.48), preliminary results appear to be attenuated. We saw null results for LBW irrespective of the exposure metric used. Discussion: Although our preliminary results showed little evidence of associations between in utero nitrate exposures and fetal growth retardation; we saw changes in VLBW when comparing the use of non-imputed versus imputed exposures. We also saw some evidence of exposure misclassification based on preliminary imputations. Given the potential for seasonal fluctuations in nitrate concentrations, the available monitoring data does not allow for consideration of more narrow exposure assessment windows (e.g., trimesters). Future analyses will consider integration of private water system data, more advanced imputation techniques to address un-captured temporal variability, and consideration of the cumulative impacts of joint exposure to co-occurring pesticides. Disclaimer: The views expressed in this abstract are those of the authors and do not necessarily reflect the views or policies of the U.S. EPA or the ORISE.