MVPBT: R package for publication bias tests in meta-analysis of diagnostic accuracy studies

TL;DR

The paper introduces the MVPBT R package that implements generalized Egger tests for detecting publication bias in diagnostic test accuracy meta-analyses, addressing a gap in existing methods.

Contribution

It provides a practical implementation of advanced publication bias tests tailored for multivariate DTA meta-analysis, enhancing detection power.

Findings

MVPBT facilitates publication bias testing in DTA meta-analysis.

Generalized Egger tests improve detection power by incorporating outcome correlations.

The package offers user-friendly tools with real data examples.

Abstract

Meta-analysis for diagnostic test accuracy (DTA) has been a standard research method for synthesizing evidence from diagnostic studies. In DTA meta-analysis, although publication bias is an important source of bias, no certain methods similar to the Egger test in univariate meta-analysis have been developed to detect such bias. However, several recent studies have discussed these methods in the framework of multivariate meta-analysis, and some generalized Egger tests have been developed. The R package MVPBT (https://cran.r-project.org/web/packages/MVPBT/) was developed to implement the generalized Egger tests developed by Noma (2020; Biometrics 76, 1255-1259) for DTA meta-analysis. Noma's publication bias tests effectively incorporate the correlation information between multiple outcomes and are expected to improve the statistical powers. The present paper provides a nontechnical introduction and practical examples of data analyses of the publication bias tests of DTA meta-analysis using the MVPBT package.