CHEST Guidelines-Research-Methods-for-Zero-Event-Proportion-Studies

Research-Methods-for-Zero-Event-Proportion-Studies

Pdf Summary

The document discusses a meta-analysis published in CHEST, focused on the safety of thoracentesis and tube thoracostomy in patients with uncorrected coagulopathy. The analysis by Fong et al. found a pooled incidence of zero percent for major bleeding and mortality, suggesting the procedures are safe. However, Xu and Doi questioned the methods used in the meta-analysis, particularly the Freeman-Tukey double arcsine transformation and Miller’s method, noting that these introduced a bias towards zero, leading to an underestimated incidence.<br /><br />Xu and Doi argue that using the correct back-transformation method and a different denominator reveals a more accurate incidence rate of 0.63%, implying 6.3 events per 1,000 observations. They also suggest the Doi plot to detect publication bias, identifying asymmetry not captured by the funnel plot and the Egger test, which Xu and Doi find unsuitable for zero-event studies. They propose the inverse variance heterogeneity model over the DerSimonian-Laird model, to account for publication bias and offer more accurate synthesis results.<br /><br />In their response, Fong et al. acknowledge the criticism, recognizing the inherent limitations of their methods. They clarify that, even when using Xu and Doi's methods, the incidence remains low, supporting the safety of the procedures. Both the original authors and the critics agree on the general safety of thoracentesis and tube thoracostomy despite methodological differences in estimating complication rates. The exchange underscores the importance of choosing appropriate statistical methods to accurately report and interpret clinical data.<br /><br />No conflicts of interest were declared by the authors, and the research was conducted independently of the sponsors, according to the acknowledgments.

Keywords

thoracentesis

tube thoracostomy

coagulopathy

meta-analysis

bleeding incidence

statistical methods

publication bias

Doi plot

inverse variance heterogeneity

clinical safety