PostScript

Archimedes: does caffeine treatment for apnoea of prematurity improve neurodevelopmental outcome in later life? We were surprised by some of the com-ments of Winckworth and Powell1 about the Caffeine for Apnea of Prematurity (CAP) trial.2 The authors concluded that the CAP trial had a ‘large preanalysis dropout rate (up to 15%)’. We correct this serious misunderstanding in table 1 which shows the number of children who had missing data for the outcome of death or disability at 18 months, and for each of the components of this composite outcome. The true ‘dropout rates ’ ranged from 3.1 % for death to 8.3 % for cognitive delay in survivors at 18 months. Of note, the ascertainment rate for cognitive delay will never be 100 % because children with severe disabilities cannot be tested successfully. The authors performed ‘worst-case sce-nario ’ analyses and reported that under these conditions, higher rates of cerebral palsy and cognitive delay would be found in the caffeine group.1 Thirty-fi ve children in the caffeine group and 36 children in the placebo group had missing data for the outcome of cerebral palsy. The probability that by chance all 35 children with missing data for this outcome in the caffeine group would have developed cerebral palsy, and none of the missing placebo children, is 4.5 × 10−21. Without due diligence in the esti-mation of the likelihood of the worst-case scenario, such analyses may be misleading, as in the example provided by Winckworth and Powell.1 Finally, we do not understand the rel-evance to the CAP trial of the authors’ comment about the use of corticosteroids in the fi rst 4 days. However, we are most concerned about the authors ’ conclusion that ‘a substantial number of patients were lost to follow-up’.1 By the standards of most experts in the fi eld, our follow-up rates of 92–97 % at the end of the second year of life were very respectable