A retrospective, hospital-records-based study of neonates screened for retinopathy of prematurity (ROP) was undertaken to determine whether the inclusion criteria for screening could have been safely altered to reduce the numbers of babies screened whilst not missing any stage III disease. Babies from six neonatal intensive care units in Birmingham were screened by a single examiner. Between November 1989 and November 1995, 1611 babies were examined and 1429 of these fell within the inclusion criteria of current guidelines for ROP screening produced by the Royal College of Ophthalmologists and the British Association of Perinatal Medicine--any baby < or = 1500 g birthweight or < or = 31 weeks gestational age. Thirty-nine (39) babies developed stage III ROP of which 31 (2.2%) were from Birmingham. In addition 8 babies with stage III ROP were referred from elsewhere. All 39 babies with stage III ROP had a birthweight < or = 1250 g or a gestational age of < or = 29 weeks, but 2 fell outside one criterion, indicating the need for both to be used. Had these inclusion criteria been utilised during this period, then 30% fewer babies would have been examined (432 of 1429). The importance of using both birthweight and gestational age as inclusion criteria is discussed, and the dangers of altering the indications for national screening on the basis of one study population is emphasised.