STUDY OBJECTIVE: To compare recall of smoking habits during pregnancy 0.5-3 years after delivery across groups defined by recall time (5 six month periods) and pregnancy outcome (pre-eclampsia, pregnancy induced hypertension, intrauterine growth retardation, preterm or post-term delivery compared with controls).

DESIGN: Case-control nested in cohort study.

SETTING AND PARTICIPANTS: A subsample of 503 women from a cohort of 6347 women established between 1989 and 1991 in Aarhus University Hospital.

MAIN RESULTS: Measures of agreement between concurrent and retrospective data on smoking status varied between 0.93 and 1.0 (sensitivity), 0.90 and 0.98 (specificity), and 0.79 and 0.98 (kappa). Spearman's correlation coefficients for number of cigarettes smoked/day varied between 0.87 and 0.97; mean differences were all close to zero. Accuracy of recall tended to diminish with increasing alcohol intake, particularly among women smoking > or = 10 cigarettes/day.

CONCLUSIONS: Information on smoking habits could be accurately obtained retrospectively independent of recall time and the pregnancy outcomes studied here. Accuracy diminished with increasing alcohol intake, particularly among heavy smokers.