Robust Tobacco Smoking Self-Report in two Cohorts of Vulnerable Pregnant Women and Adults

Background: Stigma associated with tobacco smoking, especially during pregnancy, may lead to underreporting and possible bias in studies relying on self-reported smoking data. Cotinine, a nicotine metabolite with a ~20h half-life in blood, is often used as a biomarker of smoking. The objective of this study was to examine the concordance between self-reported smoking and plasma cotinine concentration among participants enrolled in two related cohorts of vulnerable individuals: human immunodeficiency virus (HIV)-positive and HIV-negative pregnant women enrolled in the CARMA-PREG cohort and HIV-positive and HIV-negative non-pregnant women and men enrolled in the CARMA-CORE cohort. Methods: For HIV-positive (n=76) and negative (n=24) pregnant women, plasma cotinine was measured by ELISA in specimens collected during the third trimester, between 28 and 38 weeks of gestation. Plasma cotinine was also measured in HIV-positive (n=43) and negative (n=57) women and men enrolled in the CARMA-CORE cohort. Results: Self-reported smokers were more likely to have low income (p<0.001) in both cohorts, and to deliver preterm (p=0.007) in CARMA-PREG. In the CARMA-PREG cohort, concordance between plasma cotinine was 95% for self-reported smoking, and 89% for self-reported non-smoking. In the CARMA-CORE cohort we observed similarly high concordances of 96% and 92% for self-reported smoking and non-smoking, respectively. In this sample, the odds of discordance between self-reported smoking status and cotinine levels were not significantly different between self-reported smokers and non-smokers, nor between pregnant women and others. Taken together, the overall concordance between plasma cotinine and self-reported data was 94% with a Cohen9s kappa coefficient of 0.860 among all participants. Conclusions: Given the high proportion of vulnerable people in the CARMA-PREG and CARMA-CORE cohorts, our results may not be fully generalizable to the general population. However, they demonstrate that participant surveying in a non-judgemental context can lead to accurate and robust self-report data. Implications: Reliable self-reported smoking data is necessary to account for smoking status in subsequent studies. Our results suggest that future studies should ensure that study participants feel sale to speak candidly to non-judgemental research staff to obtain reliable self-report data.