Objectives The main aim of this study was to ascertain the performance of BW and ultrasound EFW discordance in the prediction of perinatal loss in twin pregnancies.

Methods A retrospective study of all twin pregnancy births from a large regional cohort of 9 hospitals over a ten year period. Inter-twin BW and ultrasound EFW discordance were related to stillbirth and neonatal death of one or both twins obtained from a mandatory national register. Receiver operating characteristic (ROC), survival and logistic regression analyses were performed to evaluate the contribution of weight discordance in determining perinatal loss.

Results A total of 2,161 twin pregnancies were included in the analysis. A BW discordance of ≥25% was associated with the highest AUC for the prediction of stillbirth and neonatal death (BW AUC = 0.72; 95% CI: 0.65–0.80). The perinatal loss in twins with a BW discordance of ≥25% was significantly greater (60.9 per 1,000 fetuses) than that in twins with an EFW discordance <25% (8.6 per 1,000 fetuses) (p < 0.0001); the predictive accuracy was similar using either BW or ultrasound EFW discordance (EFW AUC = 0.69; 95% CI: 0.62–0.77, p = 0.62). Logistic regression analysis demonstrated that BW discordance and gestational age, but not chorionicity or individual fetal size percentile, were independently associated with perinatal mortality.

Conclusions An EFW discordance of ≥25% represents the optimal cut-off for the prediction of stillbirth and neonatal mortality irrespective of chorionicity or individual fetal size. A policy of increased fetal surveillance commencing from 26 weeks’ gestation and elective delivery by 38 weeks might be reasonable.