Risk of publication bias in therapeutic interventions for COVID-19

Abstract

Systematic reviews and meta-analyses are secondary research designs that constitute a reference to guide decision-making. Faced with the pandemic, many investigations are published at an accelerated rate, which is why various organizations have proposed to keep them updated through “live reviews”. Even if developed with great methodological rigor, publication bias can represent a threat to their validity. The characteristics of publication bias, the regulatory ways to avoid it, and statistical tools to suspect it are described. Publication bias is defined as hiding or delaying publication and / or withholding data from research studies. Up to half of the controlled trials that are conducted remain unpublished, due to various interests. A recent and shocking example was that of Oseltamivir during the H1N1 pandemic. The delay in publishing results of studies funded by the industry led to the purchase of a drug that, later, it was learned, had no relevant beneficial effects. Various initiatives propose regulating the publication and registration of clinical trials  to reduce this bias, and some statistical techniques allow us to suspect it. It is exemplified by a statistical analysis to assess publication bias in three therapeutic interventions related to COVID-19: Corticosteroids, Ivermectin and Tocilizumab.