PT - JOURNAL ARTICLE AU - Walpole, I AU - Zubrick, S AU - Pontré, J TI - Confounding variables in studying the effects of maternal alcohol consumption before and during pregnancy. AID - 10.1136/jech.43.2.153 DP - 1989 Jun 01 TA - Journal of Epidemiology and Community Health PG - 153--161 VI - 43 IP - 2 4099 - http://jech.bmj.com/content/43/2/153.short 4100 - http://jech.bmj.com/content/43/2/153.full SO - J Epidemiol Community Health1989 Jun 01; 43 AB - STUDY OBJECTIVE: to investigate the relationship between alcohol consumption and pregnancy outcome. DESIGN: prospective randomised cohort survey with follow up sample stratified on level of alcohol intake. SETTING: antenatal clinic of large maternity hospital in Western Australia. PARTICIPANTS: 2002 randomly selected pregnant women recruited over 3 year period for questionnaire survey (58% in 1st trimester, 33% in 2nd trimester, 8% in third trimester at recruitment). Only 19 refused participation. Stratified subsample of 665 women followed up, of whom 60 had miscarriage, stillbirth or neonatal death. Subsample was selected on basis of prepregnancy alcohol consumption. INVESTIGATIONS AND MAIN RESULTS: All 2002 women completed a comprehensive questionnaire on demographic, lifestyle, health (including diet) and obstetric factors. The stratified subsample was followed through pregnancy and data were collected on obstetric course and infant outcome. Results showed that beer, wine and spirits drinkers differed significantly in maternal characteristics, nutrition and other important variables such as smoking. Women with stillbirths or miscarriages drank more beer than those with live births, though total levels of alcohol intake did not differ. Beer drinkers were less likely to reduce their consumption in pregnancy than other drinkers if they also smoked more than 20 cigarettes per day. CONCLUSIONS: Studies of effects of maternal drinking must include extensive information on the variables examined in this study or conclusions relating to maternal drinking in pregnancy are likely to be invalid.