Selection bias introduced by informative censoring in studies examining effects of vaccination in infancy.