OBJECTIVES

TA366¹ submission to NICE (pembrolizumab for ipilimumab naïve advanced melanoma) utilised trial data from KEYNOTE-006² with limited follow-up. Consequently, there was uncertainty regarding the survival models that represented the most plausible long-term overall survival (OS) and progression-free survival (PFS). A more mature data cut of KEYNOTE-006, with 5-year follow-up, is now available. This study aims to conduct a comparison of the survival projections adopted in TA366, and the 5-year trial data.

METHODS

OS and PFS projections were taken from the TA366 manufacturer submission and the Evidence Review Group report and compared to the recently published 5-year KEYNOTE-006 data. All survival data related to the 3-week cycle (Q3W) intervention arm of KEYNOTE-006. These data were then compared at two time points³: the maximum follow-up point of the original data cut (TP1), and the mid-point between the maximum follow-ups of the original data cut and the mature data cut (TP2).

RESULTS

TP1 was 17.2 and 14.5 months, and TP2 was 38.5 and 35.6 months after trial start, for OS and PFS respectively. At TP1 the manufacturer’s extrapolation underestimated OS by 8% and PFS by 19.8% while at TP2, OS and PFS were underestimated by 13.9% and 28.8% respectively. At TP1 the ERG’s extrapolation underestimated OS by 2.7% and PFS by 26.2%, while at TP2, OS and PFS were underestimated by 3.6% and 26.1% respectively.

CONCLUSIONS

Long term extrapolation of survival outcomes for TAs has been found to be reasonably aligned to later available, mature data³. Nevertheless, this study’s findings demonstrate the underestimation of the manufacturer and ERG extrapolations for both OS and PFS, during TA366. This indicates that the current manufacturer and ERG survival modelling approaches may be too conservative for the long-term extrapolations of OS and PFS, for novel immunotherapy drugs.