While the true figures are not well established, outcomes of twin pregnancies are directly dependent on a small number of preterm births between 22 and 27 weeks. Observation of perinatal outcomes in twin pregnancies yields two contradictory results. Firstly, it shows an improvement in perinatal mortality figures. Secondly, it reveals an increase in the rates of preterm deliveries. These findings result from the observation of 783 twin pregnancies followed and delivered in a level 3 perinatal centre in Paris between 1993 and 1998. Women followed since the beginning of pregnancy through the outpatient clinic of the institution are included in this number, as are women who were referred or transferred to the centre at a later date due to complications. This analysis reflects the influence of two contrasting policies. The first, and less recent policy is devoted to the prevention of preterm births, and is reflected by the low number of extremely preterm deliveries at 22–32 weeks. The second is the effect of our new approach to the prevention of foetal deaths in relation to foetal growth retardation in twins which has resulted in increased medical intervention such as the induction of labour or scheduled Caesarean birth. This has resulted in an increase in twin preterm births from 33 to 36 weeks, with the expected result of fewer foetal deaths.