• William O Tarnow-Mordi.
• Department of Neonatal Medicine, University of Sydney, Westmead Hospital, and The Children's Hospital at Westmead, New South Wales, Australia.
• Pediatr Crit Care Me. 2005 May 1; 6 (3 Suppl): S146-9.

ObjectiveTo identify strategies to enable randomized, controlled trials in neonatal sepsis to recognize therapies that increase disability-free survival.MethodsRapid literature review.ResultsRandomized, controlled trials are the gold standard for testing a therapy because they minimize bias. However, randomized, controlled trials must also minimize random error if they are to detect moderate (and realistic) improvements in rates of disability-free survival reliably. This requires surprisingly large samples (i.e., thousands rather than hundreds). Against this perspective, most neonatal trials have been too small to be conclusive, so most neonatal therapies remain incompletely evaluated. As in specialties like cardiology and obstetrics, achieving reliable trials in neonatal sepsis will require international collaboration, simpler data sets, more cost-effective recruitment strategies, less exclusive criteria for selecting collaborators and patients, and an appreciation by clinicians and data-monitoring committees that substantially greater sample sizes are needed to limit the play of chance. Prospective meta-analysis using individual patient data is a promising strategy. It requires researchers to obtain funding from their national agencies to conduct similar trials according to an agreed protocol with prespecified hypotheses, interventions, power calculations, data sets, and measures of outcome. Prospective meta-analysis combines the methodologic advantages of a single "megatrial" with the practical advantages that financial burdens are spread internationally and different national funding cycles can be more flexibly accommodated.ConclusionsProspective meta-analysis using individual patient data is a promising strategy for achieving large-scale, randomized evidence in neonatal sepsis.

9 keywords...

hide…