Presentation (CTI)

OBJECTIVES: Immune checkpoint inhibitors (ICIs) may maintain treatment benefit after treatment discontinuation. However, the magnitude and duration of the effect remains uncertain and frequently debated during National Institute for Health and Care Excellence (NICE) technology appraisals. This uncertainty is exacerbated by the immature evidence often available at appraisal, necessitating assumptions regarding treatment effect duration and waning when extrapolating long-term survival. As longer-term trial data have become available for earlier ICI appraisals, this study evaluated whether treatment effect waning (TEW) assumptions used in decision-making were consistent with subsequently observed evidence.
METHODS: A targeted literature review identified 13 NICE single technology appraisals (STAs) of ICIs with longer-term trial follow-up. TEW assumptions were extracted, and the pivotal trials informing each STA were identified. Survival analyses were conducted using data cuts available during appraisals and corresponding long-term follow-up data to generate time-varying implied hazard ratios (HR) and investigate potential evidence of TEW. Observed long-term outcomes were compared with TEW assumptions applied during the STAs.
RESULTS: Findings suggest some form of TEW is likely to occur, although results were associated with substantial uncertainty. In 3 of 13 STAs, TEW assumptions applied in STAs aligned with observed long-term outcomes. In the remaining 10 STAs, assumptions either overestimated (n=4) or underestimated (n=1) treatment effects relative to long-term data, or findings remained inconclusive (n=5). Several factors, including treatment switching, subsequent therapies, stopping rules, and built-in selection bias and non-collapsibility of HR, may influence estimates of TEW. Immature data cuts presented during STAs increase uncertainty in estimated long-term treatment effects. This may partly explain why a minority of assumptions made during STAs were consistent with observed long-term outcomes.
CONCLUSIONS: TEW of ICIs is likely, but previous applications of TEW assumptions in STAs may not suitably reflect long-term treatment benefits. This may impact cost-effectiveness results and resource allocation decisions.