Allowing for informative missingness in aggregate data meta-analysis with continuous or binary outcomes: Extensions...

Area of Science:

• Biostatistics
• Clinical Trials Methodology

Background:

• Missing outcome data poses a significant challenge in randomized trials and meta-analyses, potentially invalidating results.
• Addressing missing data often requires untestable assumptions, complicating analysis.

Purpose of the Study:

• To introduce the Stata command metamiss2, an extension of metamiss.
• To enable exploration of the impact of missing outcome data on meta-analysis summary effects using informative missingness parameters.

Main Methods:

• The metamiss2 command allows for sensitivity analysis by specifying informative missingness parameters.
• These parameters explore the relationship between observed and missing participant outcomes.
• The command supports binary or continuous data in both pairwise and network meta-analyses.

Main Results:

• The command facilitates sensitivity analysis to assess the robustness of meta-analysis results under different missing data assumptions.
• Informative missingness parameters, though not estimable from observed data, can be specified using external evidence like expert opinion.

Conclusions:

• The metamiss2 command provides a valuable tool for researchers to rigorously evaluate the impact of missing data in meta-analyses.
• It enhances the reliability of findings by enabling sensitivity analyses for both pairwise and network meta-analyses.