Absence of small study effects in neurosurgical meta-analyses: A meta-epidemiological study

Abstract

BackgroundSmall studies are prone to lower methodological quality and publication bias, and are more likely to report greater beneficial effects. A meta-epidemiological study was undertaken to investigate and quantify the impact of small study effects on meta-analyses in the neurosurgical literature. MethodsA PubMed search was used to procure meta-analyses from Journal of Neurosurgery, Neurosurgery, Spine, Acta Neurochirurgica and Journal of Neurotrauma. Outcome data were extracted from meta-analyses the effect of study size was estimated by calculating the ratio of odds ratios (RORs) between small and large studies. Results16 meta-analyses of 229 primary studies and 90,629 patients were included. All but two included pooled outcomes were significantly different from 1. On average small studies did not demonstrate greater beneficial effects, with an estimated pooled ROR of 1.32 (95% CI, 0.89 to 1.75). Stratification by meta-analysis effect size and heterogeneity yielded similar findings. ConclusionsThe absence of small study effects in meta-analyses of neurosurgical studies may reflect widespread poor quality of the neurosurgical literature affecting both large and small studies, rather than an absence of publication bias.