OBJECTIVES: Match-adjusted indirect comparison (MAIC) is a method commonly used to perform indirect treatment comparisons. MAIC involves reweighting the Individual Level Data (ILD) from one study to align to the aggregate characteristics of a study with only Aggregate Level Data (ALD) available. Several packages have emerged to implement MAICs using the R programming language; “maic”, “MAIC”, “Maicplus” and “maicChecks”. This work compares the results and usability of the different packages.

METHODS: Unanchored MAICs were performed matching ILD (N = 1,000 patients) to ALD (based on N = 1,000), using data from a published simulation study. Each R package was used to perform several MAICs with different combinations of matching variables. The variables included means, medians, and proportions; all of which are typically seen in MAICs. Comparisons of effective sample sizes (ESSs), and weighted survival outcomes were made.

RESULTS: When matching on ALD means or proportions, all four packages generated identical ESS values and weighted outcomes. The only package that could natively handle ALD medians was the “maic” package. Investigating the source code, we found all packages use the same underlying code, sourced from the National Institute of Health and Care Excellence (NICE) Technical Support document 18. Other differences in usability were also noted. A secondary finding was that ESS was higher when matching to medians than means, a methodological finding not previously observed in the literature.

CONCLUSIONS: The R packages currently available produce identical outcomes, though do have differences in usability – most notably around the use of medians, which are frequently reported in clinical trial publications. Although demonstrated in unanchored MAICs, these results would translate to anchored MAICs. Further research is needed to determine whether different analytical code for implementing MAICs produce variations in estimates, such as ESS or outcomes.