Commentary on “Enteral Lactoferrin Supplementation for Prevention of Sepsis and Necrotizing Enterocolitis in Preterm Infants”

Abstract

Meta-analysis of outcome data from randomized controlled trials can increase the precision of estimates of the effect of interventions. These, however, need to be interpreted carefully as various forms of bias can inflate pooled effect size estimates. In particular, small study bias is recognised as an important limitation to the external validity and applicability of meta-analyses in systematic reviews [1]. Small trials have, on average, lower methodological quality than large trials. Typically, small trials are more likely to contain design limitations in the key domains: generation of random sequences, concealment of allocation, masking of parents, caregivers, and investigators, completeness of outcome assessment, and selectiveness of reporting [2]. Because of the prevalence of these sources of bias, small trials tend to have greater (though less precise) effect size estimates than large trials. Small study bias may, furthermore, be associated with (and be compounded by) “publication bias” – the tendency for reports of statistically significant or clinically important effects to be more likely than others to be submitted and accepted for publication [3]. Consequently, meta-analyses that contain data from many small trials can generate spuriously inflated effect size estimates.