Summary The relation between smoking and preterm delivery is not totally known. Our aim was to determine whether smoking during pregnancy was associated with preterm birth among women at different risk according to their obstetric history. The study was based on data from the 1998 French national perinatal survey. Of the 13073 singleton live births, 4.7% were preterm; 15% of the pregnant women were moderate (one to nine cigarettes per day) and 10% heavy smokers (at least 10 cigarettes per day). Smoking heavily was related to preterm birth (crude odds ratio [OR] = 1.35, 95% confidence interval [95% CI]: [1.04, 1.74]). Multivariable logistic regression showed a relation between smoking and preterm birth among multiparae without previous adverse pregnancy outcomes; the associated adjusted ORs (AORs) were 1.25 [95% CI 0.83, 1.87] among moderate smokers and 1.46 [95% CI 0.98, 2.20] among heavy smokers. The corresponding AORs were 0.69 [95% CI 0.46, 1.05]) and 0.96 [95% CI 0.59, 1.56] for primiparae and 1.11 [95% CI 0.63, 1.93] and 0.50 [95% CI 0.25, 0.98] for multiparae with previous adverse pregnancy outcomes. Our study showed a relation between heavy smoking during pregnancy and preterm birth mostly for women with low obstetric risk.