OBJECTIVES : Response to treatment is a clinically meaningful outcome for patients with cancer. Economic models that capture the dynamics between treatment response and disease evolution can be well suited for immuno-oncologic (I-O) therapy. The probability of being in response function (PBRF) is a composite endpoint that combines time to response, response rate and duration of response. This study investigated the impact of incorporating the PBRF into a partitioned survival model (PSM) of nivolumab versus everolimus in previously treated advanced renal cell carcinoma.

METHODS : A 4-state PSM was developed by further partitioning the progression-free (PF) state of a standard 3-state PSM into two separate states: “PF with response” and “PF without response” using the PBRF concept. Patient-level data from CheckMate 025 was used for the utility and survival analyses. Aggregated health-state utility values (HSUV) were estimated from EQ-5D-3L trial data using a Canadian-based tariff. Best fitting parametric survival models were used to extrapolate outcomes beyond the 5-year trial follow-up. Quality-adjusted life-years (QALYs) over a 10-year time horizon were estimated using the 4-state PSM with PBRF and the 3-state PSM.

RESULTS : Mean HSUV for PF “with response” and PF “without response” were 0.91 and 0.84, respectively. Over a 10-year horizon, the 4-state PBRF model generated 0.84 QALYs in the PF state for nivolumab (0.54 [64%] with response; 0.30 [36%] without response) versus 0.54 for everolimus (0.08 [14%] with response; 0.46 [86%] without response). The incremental QALYs comparing nivolumab with everolimus in the PF state were 10% higher in the 4-state PBRF model versus the 3-state PSM.

CONCLUSIONS : The PSM augmented with PBRF provided a more detailed assessment of QALYs gained in the PF state for nivolumab and everolimus than the 3-state PSM. Model developers should consider this approach when there are marked differences in the response profiles of competing interventions.