Presentation (CTI)

OBJECTIVES: Network meta-analyses (NMAs) are commonly accepted in health technology assessments (HTAs) when treatments are not compared directly. For time-to-event (TTE) outcomes, standard, time-constant NMA may be unsuitable if the proportional hazards (PH) assumption is violated. Several non-PH NMA methods are available, but there is a lack of guidance on which to use. We examined the prevalence, characteristics, and justifications for non-PH NMAs used in National Institute for Health and Care Excellence (NICE) HTA submissions.
METHODS: A systematic review was conducted on NICE technology appraisals (TAs) published between April 2015 and March 2025. Main extraction items included: use of (non-PH) NMA and its rationale, NMA outcomes, type of non-PH NMA, incorporation of results into economic models, and methodological critiques
RESULTS: Of the 347 NICE TAs identified, 164 (47.3%) reported NMAs and 27 (16.5%) were non-PH NMAs. All non-PH NMAs supported oncology TAs. Non-PH NMAs were conducted mainly for progression-free survival (25 [92.6%]) and overall survival (24 [88.9%]). Only four TAs (14.8%) considered more than one non-PH NMA method. Of the 27 non-PH NMA TAs, 21 (77.8%), seven (25.9%), two (7.4%) and one (3.7%) employed fractional polynomials, parametric, spline, and piecewise NMAs, respectively. Although PH violation was the predominant reason for conducting non-PH NMA, only 22 (81.4%) TAs reported some form of assessment, of which seven (31.8%) relied only on visual inspection. While the assessors largely agree with non-PH NMA use in these TAs, there were concerns with model fit and selection. Interestingly, of 38 TAs that conducted PH NMAs on TTE data, 19 (50.0%) did not report or conduct PH assessment.
CONCLUSIONS: When non-PH NMAs were used, there was little justification for the type of non-PH NMA selected, with few TAs comparing multiple methods. No non-oncology TAs considered non-PH NMA. Guidance is needed for the selection of appropriate non-PH NMA methods for TTE outcomes.