OBJECTIVE: To determine whether customised birthweight standard improves the definition of small for gestational age and its association with adverse pregnancy outcomes such as stillbirth, neonatal death, or low Apgar score. DESIGN: Population based cohort study. POPULATION: Births in Sweden between 1992-95 (n = 326,377). METHODS: Risks of stillbirth, neonatal death, and Apgar score under four at five minutes were calculated for the lowest 10% birthweights according to population-based and customised standards, and were compared with the data from the group with birthweights over this limit. Population attributable risks for stillbirth using various birthweight centile cutoffs were calculated for the two standards. OUTCOME MEASURES: Odds ratios and 95% confidence intervals for stillbirth, neonatal death and Apgar score under four at five minutes, and population attributable risks for stillbirth at different birthweight centiles. RESULTS: Risks of stillbirth, neonatal death, and Apgar score under four at five minutes and population attributable risks of stillbirth were consistently higher if 'small for gestational age' was classified by a customised rather than by the population-based birthweight standard. Compared with infants who were not small for gestational age by both standards, the odds ratio for stillbirth was 6.1 (95% CI 5.0-7.5) for small for gestational age by customised standard only, whereas it was 1.2 (95 % CI 0.8-1.9) for small for gestational age by population standard only. CONCLUSIONS: Compared with the population-based birthweight standard, a customised birthweight standard increases identification of fetuses at risk of stillbirth, neonatal death and Apgar score under 4 at 5 minutes, probably due to improved identification of fetal growth restriction.