Presentation (CTI)

OBJECTIVES: Matching-adjusted indirect comparisons (MAICs) are commonly used to evaluate the efficacy of two treatments when direct, head-to-head trials are unavailable. The objective of this study is to compare statistical uncertainty between unanchored and anchored MAICs using data from patients with ulcerative colitis.
METHODS: Using data from the Phase III UNIFI trial of ustekinumab in moderate-to-severe ulcerative colitis, obtained through the YODA Project, we compared its efficacy with published trial results for vedolizumab, upadacitinib, mirikizumab, guselkumab, etrasimod, and adalimumab. Each comparator trial included a placebo group. We conducted both anchored (using placebo data) and unanchored (excluding placebo data) MAICs for each comparator. Clinical response and clinical remission during induction were assessed and compared across methods.
RESULTS: Risk differences and confidence intervals for clinical response and clinical remission were estimated for ustekinumab versus each comparator using both anchored and unanchored MAIC. On average, the confidence interval width for clinical response was 22.78 percentage points for anchored comparisons and 15.99 for unanchored comparisons. For clinical remission, the average width was 15.40 for anchored and 13.22 for unanchored comparisons. Statistical significance was observed in 2 of 6 unanchored comparisons for clinical response, whereas only 1 of 6 anchored comparisons was significant. For clinical remission, 5 of 6 unanchored comparisons were significant compared to 1 of 6 for anchored comparisons.
CONCLUSIONS: Unanchored and anchored MAICs can yield substantially different inferences, which may lead to contradictory decisions in comparative effectiveness assessments. Confidence intervals from unanchored MAICs are narrower compared to those from anchored MAICs despite leveraging less information. While anchored MAICs are always preferable when feasible, this research highlights differences between the two approaches and stresses the need for caution when interpreting results from unanchored MAICs in situations where anchoring is not possible.