Sir.—In Table 1 of the recent article by Nelson and Ellenberg,1 I am puzzled by some of the data. The authors report a rate of cerebral palsy of 5.7 per 1000 among children with no birth complications and one neonatal sign. This contrasts with the rate of 2.8 per 1000 among children with birth complications and one neonatal sign. This is counterintuitive: why should children with birth complications have a rate only 50% that of children without birth complications? I note a similar phenomenon in terms of the rate of cerebral palsy per 1000 children having five-minute Apgar scores of 0 to 5 and two neonatal signs (Table 3), where a rate of 0.0 children with cerebral palsy per 1000 is reported. I wonder if selection bias was operating with regard to children who died before follow-up or were unavailable for follow-up? If children with birth complications or low Apgar scores were more likely to be omitted from the follow-up data because of death, this would falsely reduce the apparent risk of cerebral palsy. I would appreciate knowing whether this was investigated by the authors.