Because of the important changes in medical care and the social context of pregnancy, the relevance of well-established risk factors of preterm delivery (< 37 weeks) needs to be re-assessed. The objective of this study was to determine whether the associations between maternal characteristics and preterm delivery in France in 1995 were similar to those in 1981. We used two national representative samples of births, 5577 newborns in 1981 and 13 318 in 1995. Univariable and multivariable analyses were applied to identify variables associated with increased risk of preterm delivery. Some risk factors were similar in 1981 and 1995, including age > 34 years and previous adverse obstetric history. Other factors that were observed in 1981 but which were not linked to preterm delivery in 1995 included the mothers being very young, single or foreign. However, parity and previous induced abortion were associated with preterm delivery in 1995, but not in 1981. These results show that the definition of high-risk groups used in prevention programmes should be brought up to date regularly.