The effects of nonresponse and misclassification bias on the odds ratio were evaluated in a case-control study of smoking during pregnancy and chromosomally normal spontaneous abortions. Information from cases (N=650) and controls (chromosomally abnormal spontaneous abortions, N=582) was obtained in two ways--a personal interview soon after the spontaneous abortion and a mail questionnaire 1-3 years thereafter. To evaluate nonresponse, information obtained from women who were interviewed but who did not respond to the mail questionnaire was used to estimate the exposure status of the mail questionnaire was used to estimate the exposure status of the mail questionnaire nonrespondents. Response to the questionnaire varied by exposure status and case-control status. Exposed cases were the least likely to respond (36% response rate), followed by unexposed cases (48%) and all controls (51%). The crude odds ratio for smoking during pregnancy and chromosomally normal spontaneous abortions was 1.2 (p=0.41). After adjusting for the effects of nonresponse, the odds ratio increased to 1.6. Misclassification bias was evaluated by comparing responses to the interview with the responses to the mail questionnaire. This also showed that errors in reporting caused the odds ratio to be underestimated (crude odds ratio, 1.2; adjusted odds ratio, 1.4).
