DIFFERENTIAL MISCLASSIFICATION OF ALCOHOL AND CIGARETTE CONSUMPTION BY PREGNANCY OUTCOME

Background. The validity of the results of studies using retrospective ly collected information on exposures is often criticized, because cas es may report differently from controls even ii their true exposure st atus is the same. This study was performed to quantify the extent to w hich this effect (differential misclassification) may occur for alcoho l and cigarette consumption by pregnancy outcome. Methods. Prospective as well as retrospective information on alcohol and cigarette consump tion was collected for 2806 mothers resident in all 12 provinces of th e Netherlands, who gave birth between 1978 and 1979. Changes in mean r eported consumption and changes from user to non-user based on retrosp ective and prospective information were compared for cases and control s. This was done by calculating absolute differences (retrospective mi nus prospective) in reported consumption and by calculating 'misclassi fication odds ratios'. Further, conventional odds ratios based on retr ospective information were compared with those based on prospective in formation. Outcome measures were stillbirth, small for gestational age (SGA), congenital malformations, preterm birth and low birthweight. R esults. The only statistically significant result was found for smokin g and SGA. Mothers with an SGA child retrospectively reported a higher number of cigarettes smoked than they had prospectively, more so than mothers oi a control child. However, the odds ratios of the relation between SGA and smoking based on prospective and retrospective informa tion, respectively, were virtually the same. Conclusions. Our results suggest that information bias is unlikely to have a large influence on effect estimates in studies using retrospective information on alcoho l and cigarette consumption.