RANKING OF TREATMENTS IN MULTI-LEVEL NETWORK META-REGRESSION (ML-NMR): INCORPORATING MINIMALLY IMPORTANT DIFFERENCES
Author(s)
Tristan Curteis, MSc1, Augustine Wigle, PhD2, Christopher Michaels, PhD3, Adriani Nikolakopoulou, PhD4.
1Head of Statistics, Costello Medical, Manchester, United Kingdom, 2Department of Epidemiology, Biostatistics and Occupational Health, McGill University, Waterloo, ON, Canada, 3Costello Medical, London, United Kingdom, 4Department of Hygiene, Social and Preventive Medicine and Medical Statistics, School of Medicine, Aristotle University of Thessaloniki, Thessaloniki, Greece.
1Head of Statistics, Costello Medical, Manchester, United Kingdom, 2Department of Epidemiology, Biostatistics and Occupational Health, McGill University, Waterloo, ON, Canada, 3Costello Medical, London, United Kingdom, 4Department of Hygiene, Social and Preventive Medicine and Medical Statistics, School of Medicine, Aristotle University of Thessaloniki, Thessaloniki, Greece.
OBJECTIVES: In NMA, ranking metrics (e.g., probability best and Surface Under the Cumulative RAnking curve [SUCRAs]) often ignore relative effect magnitude, which may mislead medical and HTA decision-making by prioritising statistically different, but clinically comparable, treatments. Methods to adjust for minimally important differences (MID) in ranking metrics have been proposed for NMA. Despite supporting interpretation when a shifted null hypothesis may be useful or required (as per, in certain cases, EU Joint Clinical Assessment [JCA] indirect treatment comparison guidance), software for implementation of these methods is not available for ML-NMR, a method for population-adjustment in NMA.
METHODS: We extended the R package mid.rank.nma (an extension of multinma) to apply the MID-based treatment ranking methodology developed for Bayesian NMA by Curteis et al. (2025) to ML-NMR. As a proof-of-concept example, we applied mid.rank.nma to an ML-NMR in plaque psoriasis. Specifically, IXORA‐S demonstrated superiority of ixekizumab Q2W over ustekinumab based on a non-inferiority margin of -12.6% of difference in risk of PASI 90, assuming a 43% response rate for ustekinumab. We applied the same margin (converted to the logit scale) to generate IXORA‐S population-specific MID-adjusted ranking statistics from an ML-NMR of 7 treatments across 9 studies in achieving PASI 90.
RESULTS: Even accounting for an MID of -12.6% of difference in risk of PASI 90, ixekizumab Q2W retained a 100% probability of being best or equal best. With no MID, ixekizumab Q4W, secukinumab 300 mg, secukinumab 150 mg and ustekinumab had distinct median ranks 2 to 5 respectively. However, with MID adjustment, ixekizumab Q4W and secukinumab 300mg (equal median rank 2.5), and secukinumab 150 mg and ustekinumab (equal median rank 4.5), were tied.
CONCLUSIONS: The R package mid.rank.nma now supports MID-adjusted ranking statistics for ML-NMR, facilitating MID-adjusted treatment ranking metrics to inform decision making in a population-adjusted context.
METHODS: We extended the R package mid.rank.nma (an extension of multinma) to apply the MID-based treatment ranking methodology developed for Bayesian NMA by Curteis et al. (2025) to ML-NMR. As a proof-of-concept example, we applied mid.rank.nma to an ML-NMR in plaque psoriasis. Specifically, IXORA‐S demonstrated superiority of ixekizumab Q2W over ustekinumab based on a non-inferiority margin of -12.6% of difference in risk of PASI 90, assuming a 43% response rate for ustekinumab. We applied the same margin (converted to the logit scale) to generate IXORA‐S population-specific MID-adjusted ranking statistics from an ML-NMR of 7 treatments across 9 studies in achieving PASI 90.
RESULTS: Even accounting for an MID of -12.6% of difference in risk of PASI 90, ixekizumab Q2W retained a 100% probability of being best or equal best. With no MID, ixekizumab Q4W, secukinumab 300 mg, secukinumab 150 mg and ustekinumab had distinct median ranks 2 to 5 respectively. However, with MID adjustment, ixekizumab Q4W and secukinumab 300mg (equal median rank 2.5), and secukinumab 150 mg and ustekinumab (equal median rank 4.5), were tied.
CONCLUSIONS: The R package mid.rank.nma now supports MID-adjusted ranking statistics for ML-NMR, facilitating MID-adjusted treatment ranking metrics to inform decision making in a population-adjusted context.
Conference/Value in Health Info
2026-11, ISPOR Europe 2026, Vienna, Austria
Value in Health, Volume 29, Issue 12S
Code
MSR234
Topic
Clinical Outcomes, Methodological & Statistical Research, Study Approaches
Disease
No Additional Disease & Conditions/Specialized Treatment Areas, Sensory System Disorders (Ear, Eye, Dental, Skin)