Presentation (CTI)

OBJECTIVES: CCEs within the NICE health technology appraisal programme rely on the strong assumption that new treatments are clinically no worse than comparator treatments. When conclusions about the similarity of treatments rely on statistically non-significant differences from ITCs, validating this assumption is challenging. This research aimed to identify and compare methods currently used within NICE CCEs to interpret statistically non-significant results from ITCs and make conclusions regarding clinical similarity, to determine whether inconsistencies exist and identify emerging methods used within CCEs to address this uncertainty.
METHODS: A systematic literature review (SLR) involving searches of the NICE website was performed in April 2025. NICE appraisals using the CCE approach from the outset, relying on statistically non-significant results from ITCs for at least one comparator and with published final guidance were included. Detailed narrative comments made by companies, External Assessment Groups (EAGs) and committees were the key information extracted, with a narrative synthesis performed to summarise findings.
RESULTS: In total, 41 NICE appraisals were included. Companies often relied heavily on ITC results being statistically non-significant but EAGs commonly highlighted the uncertainty associated with these results. To improve confidence in decision-making in such cases, one method emerging from the included CCEs was the use of thresholds such as minimal clinically important differences (MCIDs) or non-inferiority margins (NIMs). In some cases, these thresholds were incorporated into Bayesian analyses to calculate probabilities of the new treatment being non-inferior or clinically equivalent to comparator treatments.
CONCLUSIONS: Clinically relevant thresholds, such as MCIDs or NIMs, may be useful in supporting the interpretation of statistically non-significant ITC results and decision-making in CCEs, with the application of these thresholds within Bayesian analyses to calculate probabilities of being non-inferior or clinically equivalent being a particularly robust use of these thresholds.