OBJECTIVES:

Nivolumab was approved by NICE in 2016 for treatment of second-line renal cell carcinoma (2L-RCC) based on the 2-year CheckMate-025 trial data. In 2019, updated 3-year trial data were published. This analysis explores the impact of using publicly available 2- or 3-year CheckMate-025 data on cost-effectiveness of nivolumab versus everolimus in 2L-RCC.

METHODS:

A partitioned survival cost-effectiveness model with four health states (progression-free, progressed, terminal, death) comparing nivolumab and everolimus in the 2L-RCC UK setting was built following a targeted literature review. Published 2- and 3-year CheckMate-025 data were digitized, and pseudo-patient level data generated using the Guyot algorithm. Parametric survival curves were fitted to the data to extrapolate outcomes over a lifetime time horizon. Utility values were extracted from the NICE submission. Drug costs were based on list prices, with no treatment stopping rule applied. Time on treatment (ToT) was assumed equal to progression-free survival due to lack of published data.

RESULTS:

Based on fit to data, the same parametric curves as the Evidence Review Group preferred ones were selected. The model estimated a 16% increase in both life years (LYs) and quality-adjusted life years (QALYs) gained between the earlier and later data (0.51 and 0.61 LYs; 0.38 and 0.44 QALYs, respectively) comparing nivolumab and everolimus. The incremental cost-effectiveness ratios (ICERs) were £117,187 and £129,221 per QALY in the 2- and 3-year data scenarios.

CONCLUSIONS:

Reliance on published data made the results not directly comparable to the NICE appraisal results. The digitization procedure likely biased downwards the estimates for differential LYs (-39%) and QALYs (-38%) compared to extrapolations based on patient-level data. ICERs were stable between scenarios, with differences mostly driven by the increased survival estimates. This analysis demonstrates stability in the long-term clinical and cost-effectiveness estimates between results based on the 2- and 3-year CheckMate-025 data.