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

Missing outcome data can invalidate the results of randomized trials
and their meta-analysis. However, addressing missing data is often a challenging
issue because it requires untestable assumptions. The impact of missing outcome
data on the meta-analysis summary effect can be explored by assuming a relationship
between the outcome in the observed and the missing participants via an
informative missingness parameter. The informative missingness parameters cannot
be estimated from the observed data, but they can be specified, with associated
uncertainty, using evidence external to the meta-analysis, such as expert opinion.
The use of informative missingness parameters in pairwise meta-analysis of aggregate
data with binary outcomes has been previously implemented in Stata by
the metamiss command. In this article, we present the new command metamiss2,
which is an extension of metamiss for binary or continuous data in pairwise or
network meta-analysis. The command can be used to explore the robustness of
results to different assumptions about the missing data via sensitivity analysis.