Statistics and Its Interface
Volume 13 (2020)
Small-study effects: current practice and challenges for future research
Pages: 475 – 484
Meta-analyses and systematic reviews are highly valued as evidence for clinical decision- and policy-making. However, inference in these settings may be invalid if the studies do not come from the same underlying distribution. Small study effects is one form of heterogeneity that can lead to biased estimates, particularly if it arises due to the selective publishing of studies, a phenomenon known as publication bias. In this paper we discuss landmark methods for diagnosing the presence of small-study effects and correcting for them, as well as the limitations of each method. We also identify ongoing challenges and key areas in need of methodological innovation.
small-study effects, publication bias, outcome reporting bias, meta-analysis, selection models, funnel plot
A. Marks-Anglin’s effort was supported in part by NIH grant 1R01LM012607. Y. Chen’s effort was supported in part by NIH grants 1R01LM012607 and 1R01AI130460.
Received 12 December 2019
Accepted 4 July 2020
Published 31 July 2020