OBJECTIVE: Evaluation of policy and treatment of deliveries at the limits of viability in the Netherlands and resulting survival figures. DESIGN: Cohort study. METHOD: Within the framework of the European 'Models of organising access to intensive care for very preterm births in Europe' (MOSAIC) study, data was collected on all 512 births in 2003 (terminations excluded) following 22-31 weeks gestation in the catchment areas of the perinatal centres in Nijmegen and Utrecht, the Netherlands. RESULTS: Gynaecologists and neonatologists practised a reserved policy for the active treatment of pregnancies under 25 weeks (5/77; 6%); all infants died. At 25 weeks, an active obstetric policy was used in one quarter of pregnancies, but none of the infants survived. Even at 26 weeks pregnancy, the obstetric policy was reserved and the mortality relatively high (9/31; 29%). From the neonatal deaths, 86 out of 92 (93%) were preceded by a decision either not to start or to discontinue treatment. CONCLUSION: Dutch obstetricians and neonatologists practised a reserved policy at the limits of neonatal viability. There is more need for active antenatal transfer to perinatal centres for those at the lower limit of neonatal viability to enable well-balanced decisions to take place. The parents' wishes should always be taken into account.