The original study was carefully designed to avoid selection bias, because all cases of perinatal deaths during the study period were included and controls were randomly selected among survivors of the neonatal period identified from live-born birth certificates. We compared the data on sex, birth weight, and gestational age available in the birth certificates for those whom we did not include in the study (~16% of cases and controls) with the included infants and found no statistical significant differences among them, confirming that there was no selection bias.