A Novel Framework to Assess Noninferiority in National Institute for Health and Care Excellence (NICE) Cost-Comparison Evaluations (CCEs) Where Network Meta-Analyses (NMAs) Report Nonstatistically Significant Differences
Author(s)
Ben Burgess, PhD, Sophie Ip, PhD, Nicole Downes, BSc (Hons), Archie Walters, MSc, Clare Dadswell, PhD, Steve Edwards, DPhil.
BMJ Technology Assessment Group, London, United Kingdom.
BMJ Technology Assessment Group, London, United Kingdom.
OBJECTIVES: NICE CCEs require that a new treatment is clinically non-inferior compared to relevant comparators. However, confirming this assumption is difficult when the non-inferiority of a new treatment is assessed using statistically non-significant NMA results. Accordingly, this research sought to develop a new framework to help NICE committees determine whether a new treatment is non-inferior when faced with statistically non-significant NMA results.
METHODS: A new framework was developed as an ‘add-on’ to a typical Bayesian NMA. This framework utilises empirical cumulative density functions to determine the probability that a treatment is non-inferior to a comparator. Non-inferiority is assessed relative to a non-inferiority margin (NIM), or a value of no difference (i.e., 0 or 1) where a NIM is not available. This framework produces a ‘point-and-density plot’ that showcases the point estimate (and 95% credible intervals [CrIs]) of a treatment comparison alongside the distribution of iterations, and the probability of non-inferiority. This framework was applied to a recent NICE CCE (TA1019) for crovalimab to treat people with Paroxysmal Nocturnal Haemoglobinuria by conducting an NMA for the co-primary endpoint of transfusion avoidance.
RESULTS: Based on point estimates, non-inferiority of crovalimab relative to both eculizumab and ravulizumab was not demonstrated due to 95% CrIs overlapping the reported NIM. However, the novel framework reported an 86.3% and 95.9% probability that crovalimab was non-inferior to ravulizumab and eculizumab, respectively. Based on a threshold of 95%, it is likely that, for transfusion avoidance, crovalimab may be considered non-inferior to eculizumab but not to ravulizumab.
CONCLUSIONS: The novel framework can aid NICE committees in determining whether an intervention is non-inferior to comparators where statistically non-significant NMA results are reported. Through point-and-density plots, this framework provides easily interpretable outputs, which can be considered alongside other factors (e.g., similar mechanisms of action) to support a conclusion of non-inferiority.
METHODS: A new framework was developed as an ‘add-on’ to a typical Bayesian NMA. This framework utilises empirical cumulative density functions to determine the probability that a treatment is non-inferior to a comparator. Non-inferiority is assessed relative to a non-inferiority margin (NIM), or a value of no difference (i.e., 0 or 1) where a NIM is not available. This framework produces a ‘point-and-density plot’ that showcases the point estimate (and 95% credible intervals [CrIs]) of a treatment comparison alongside the distribution of iterations, and the probability of non-inferiority. This framework was applied to a recent NICE CCE (TA1019) for crovalimab to treat people with Paroxysmal Nocturnal Haemoglobinuria by conducting an NMA for the co-primary endpoint of transfusion avoidance.
RESULTS: Based on point estimates, non-inferiority of crovalimab relative to both eculizumab and ravulizumab was not demonstrated due to 95% CrIs overlapping the reported NIM. However, the novel framework reported an 86.3% and 95.9% probability that crovalimab was non-inferior to ravulizumab and eculizumab, respectively. Based on a threshold of 95%, it is likely that, for transfusion avoidance, crovalimab may be considered non-inferior to eculizumab but not to ravulizumab.
CONCLUSIONS: The novel framework can aid NICE committees in determining whether an intervention is non-inferior to comparators where statistically non-significant NMA results are reported. Through point-and-density plots, this framework provides easily interpretable outputs, which can be considered alongside other factors (e.g., similar mechanisms of action) to support a conclusion of non-inferiority.
Conference/Value in Health Info
2025-11, ISPOR Europe 2025, Glasgow, Scotland
Value in Health, Volume 28, Issue S2
Code
MSR8
Topic
Health Technology Assessment, Methodological & Statistical Research, Study Approaches
Disease
No Additional Disease & Conditions/Specialized Treatment Areas