BACKGROUND: The longer follow-up programs last, the greater the loss to follow-up. These losses to follow-up may undermine the completion of health care goals and the validity of study results. OBJECTIVES: 1) To compare neonatal characteristics in children easily followed-up and in those lost to follow-up. 2) To trace and assess children lost to follow-up. 3) To estimate the occurrence of cerebral palsy in children easily followed-up and in those lost to follow-up. METHODS: From 1991 to 1997, 601 neonates with a birth weight under 1,500 g were admitted to the Neonatology Department. At discharge, 447 infants were included in the follow-up program. Moderate-to-severe cerebral palsy was assessed when the children were aged 2 years. A specific search strategy was implemented to find those children lost to follow-up. Data on the development of those traced were updated through a standardized telephone questionnaire. RESULTS: Twenty percent of the children were lost to follow-up before the age of 2 years. Fifty-seven percent of those not available at this age were assessed by telephone interview. No differences were found in the neonatal characteristics of infants easily followed-up and those lost to follow-up except in situations of critical social disadvantage: 10 % in followed-up infants, 41 % in infants lost to follow-up. Disabling cerebral palsy was observed in 7 % of children easily followed-up and in 23 % of those lost and traced (relative risk: 3.1, 1.5-5.5). CONCLUSIONS: The risk of having disabling cerebral palsy is three times higher in children lost to follow-up than in those easily followed-up. Dismissing this source of bias may underestimate disability rates when assessing health care programs or when interpreting study results.