Indirect Treatment Comparisons: Current Practice and the Added Value of Multi-Level Network Meta-Regression

OBJECTIVES: Recently methods to conduct multi-level network meta-regressions (ML-NMR) have become widely available. These methods complement existing methods for indirect treatment comparisons such as matching adjusted indirect comparisons (MAICs) and simulated treatment comparisons (STCs). The potential benefits of ML-NMRs over anchored MAICs and STCs come from the possibility of comparing multiple treatments simultaneously and estimating population average treatment effects for the population of interest. In this study, we assess the potential of ML-NMRs to complement methods currently used for indirect treatment comparisons.

METHODS: A targeted literature review was performed to identify technology appraisals (TAs) submitted to NICE between April 1, 2021- March 31, 2024. TAs that reported a MAIC or STC for time-to-event (TTE) data were included. Information for these TAs was extracted relating to the methodology used (MAIC/STC, anchored/unanchored) along with the recommendations made by the evidence review groups (ERGs).

RESULTS: Of the TAs published in the relevant timeframe, 31 met inclusion criteria. Most analyses were unanchored comparisons (84%). All TAs reported results for MAICs with STCs sporadically used (N=3). MAICs were rarely conducted alongside NMAs, suggesting that either conducting an NMA is not feasible or not required. An important concern raised repeatedly by ERGs was the discrepancy between the population in which efficacy was assessed in the MAIC and the population of interest for decision makers. Due to the vast number of unanchored analyses, ML-NMRs could have replaced MAICs or STCs in less than 20% of instances.

CONCLUSIONS: Although ML-NMRs will not address all challenges faced when conducting ITCs, they can complement MAICs and can ensure the estimation of effects in the population of interest to HTA decision makers. However, a bigger concern is the widespread use of unanchored analyses which cannot be replaced with ML-NMRs.