VALIDATION OF MATERNAL SELF-REPORTED PROXIMITY TO AGRICULTURAL CROPS IN A CASE-CONTROL STUDY OF NEURAL TUBE DEFECTS

Abstract

ISEE-470 Abstract: There is a suspicion among epidemiologists that self-reported perinatal exposures to chemicals or pollutant sources in case-control studies of birth defects may be inaccurate due to misreporting among mothers of malformed infants, non-malformed controls, or both. We tested this suspicion in a study of neural-tube defects delivered in California in 1987-1988. In that study, mothers were asked about proximity to agricultural crops and pesticide applications. We employed historical land-use survey maps to validate self-reported residential proximity within 0.25 mile of agricultural crops. Using map-based proximity to geocoded residences as our gold-standard, we compared case-control odds ratios (ORs) calculated for 1) self-reported and 2) map-based proximity to any crops and to non-permanent field crops, orchards, and vineyards and compared the sensitivity and specificity of maternal reports for cases and controls by geographic region and maternal demographic characteristics. Logistic regression was used to derive adjusted ORs for predictors of underreporting and overreporting of proximity to crops. The OR for self-reported proximity to any crops appeared to be positively biased (1.63; 95% confidence interval (CI): 1.09, 2.45), compared to the OR based on map-based proximity (1.17; 95% CI: 0.79, 1.72). This difference appeared to be primarily a function of control mothers underreporting proximity to crops. That is, sensitivity of self-reported proximity was greater for cases (66.7%) than controls (50.4%), while specificity was slightly lower in cases (87.7% vs. 89.5%). Similar trends were observed for proximity to specific crop types, except for vineyards. Underreporting appeared to be negatively associated with NTD cases (OR: 0.56; 95% CI: 0.29, 1.10) versus controls and rural residents (OR: 0.51; 95% CI: 0.25, 1.06) versus urban residents. Mothers employed during pregnancy (OR: 2.25; 95% CI: 1.14, 4.46) were more likely to underreport than unemployed women. Overreporting of proximity was positively associated with rural residence (OR: 1.88; 95% CI: 0.95, 3.73) and residence in Central California (OR: 2.21; 95% CI: 1.08, 4.53). In addition, we observed a positive dose-response relationship for overreporting with maternal educational level and maternal gestational cigarette smoking. These results suggest differential reporting of proximity to crops between case and control mothers as well as an influence from geographic region and some maternal demographic characteristics on reporting accuracy.