Abstract 1312 Poster Session II, Sunday, 5/2 (poster 86)

Background: The choice of the primary outcome and the specification of its minimum detectable treatment difference under the null hypothesis are crucial steps in the design of a randomized clinical trial (RCT): They strongly affect sample size as well as the clinical utility of the study results. In general, a difference in the rate of a binary outcome event (e.g. death) has greater clinical relevance than a mean change in some continuous measure (e.g. oxygenation index). Moreover, since interventions in newborns may alter subsequent growth and development, the timing of the primary outcome assessment demands special attention in neonatal RCTs.

Objective: To perform a systematic review of primary outcomes in recently published neonatal RCTs.

Methods: Medline was searched in May 98 for all neonatal RCTs which had been published since 1993 in Arch Dis Child, BMJ, J Pediatr, Lancet, N Engl J Med and Pediatrics. Excluded were trials in which newborns were allocated to a diagnostic test, and all ancillary reports of previously published RCTs. The primary outcome was identified wherever possible. For primary binary outcomes, the minimum detectable treatment difference was calculated as the hypothesized relative risk reduction (RRR), if the pertinent information had been provided in the methods section of the report.

Results: 129 reports were eligible for review. A primary outcome could not be identified in 32 (25%) reports. A continuous measure was the primary outcome in 48 (37%) trials. The 49 (38%) trials with a primary binary outcome are described in the table:

Table 1 No caption available
Full size table

Conclusions: Most authors of recently published neonatal RCTs in six high-profile general and pediatric journals either failed to specify a primary endpoint, or chose a continuous measure rather than a clinically important outcome event. Trials with a primary binary outcome tend to have a short duration of follow-up and unrealistically large minimum detectable treatment differences. Therefore, important but modest effects on death or serious non-fatal events as well as long-term morbidities were likely missed by a large majority of recently published neonatal RCTs.