Logistic regression model of neonatal mortality with the impact of twin pregnancy formulated as a function of GA. Adjusted for sex, maternal smoking, maternal age, mother's health, parity, prior stillbirth, prior neonatal death, congenital abnormality, year of birth, and second order polynomials of weight and length SDS. OR=odds ratio. Black line=log odds ratio; red lines=95% confidence intervals. A horizontal reference line of zero indicates no difference between twins and singletons.

In the GA range of 29-36 weeks twins had a lower neonatal mortality than singletons. From 39 weeks onward they had a higher neonatal mortality.