Misclassification of maternal smoking status and its effects on an epidemiologic study of pregnancy outcomes.

Reliance on self-reported smoking status among pregnant women can result in exposure misclassification. We used data from the Calcium for Preeclampsia Prevention trial, a randomized study of nulliparous women conducted from 1992 to 1995, to characterize tobacco exposure misclassification among women who reported at study enrollment that they had quit smoking. Urinary cotinine concentration was used to validate quit status, and factors associated with exposure misclassification and the effects of misclassification on associations between smoking and pregnancy outcomes were evaluated using logistic regression. Of 4,289 women enrolled, 508 were self-reported smokers and 771 were self-reported quitters. Of 737 self-reported quitters with a valid cotinine measurement, 21.6% had evidence of active smoking and were reclassified as smokers. Women who reported having quit smoking during pregnancy were more likely to be reclassified than women who reported quitting before pregnancy (p<.001). Among smokers, factors independently associated with misclassification of smoking status included fewer cigarettes smoked per day and fewer years smoked. After reclassification the odds ratio for a small-for-gestational-age birth among smokers decreased by 14%, and the smoking-related reduction in birth weight decreased by 15%. Effects of misclassification on the association with hypertensive disorders of pregnancy were present but less dramatic. In conclusion, use of self-reported smoking status collected at the time of study enrollment resulted in the introduction of bias into our study of smoking and pregnancy outcomes. The potential for this type of bias should be considered when conducting and interpreting epidemiologic studies of smoking and pregnancy outcomes.