Incorporating External Opinion in Survival Extrapolations to Inform Long-Term Survival Projections for Health Technology Assessment
Author(s)
Liu Jingyun, BA1, Anne Meiwald, MSc2, Chamath Perera, MSc2, Alex Hirst, BSc, MSc2, Louise Heron, BA, BSc, MSc2.
1University of Sheffield, Sheffield, United Kingdom, 2Adelphi Values, Bollington, United Kingdom.
1University of Sheffield, Sheffield, United Kingdom, 2Adelphi Values, Bollington, United Kingdom.
Presentation Documents
OBJECTIVES: Accurate long-term survival estimation is critical for health technology assessments (HTA), yet many trials lack sufficient follow-up, necessitating extrapolation to inform decision making. This study assessed the incorporation of expert opinion on survival extrapolations, using data from CheckMate-214 investigating nivolumab plus ipilimumab (NIVO+IPI) for advanced renal cell carcinoma as a case study.
METHODS: A three-state partitioned survival model was developed using digitized trial data to evaluate overall survival (OS) and progression-free survival (PFS). Standard parametric survival models were initially fit to these data and assumed hypothetical expert opinions were then incorporated by penalizing the likelihood function. Cost-effectiveness was assessed from the UK National Health Service perspective over a lifetime time horizon. Both deterministic and probabilistic sensitivity analyses were used to assess the robustness of the results.
RESULTS: Incorporating hypothetical expert opinion at one time point which indicated a clear survival benefit for the intervention arm reduced uncertainty in OS predictions and lowered the incremental cost-effectiveness ratio (ICER) for NIVO+IPI from £41,893 / quality-adjusted life year (QALY) to £27,685 / QALY. Patients receiving NIVO+IPI experienced increased incremental QALYs due to prolonged survival. The ICER fell below the £30,000 willingness-to-pay threshold, suggesting cost-effectiveness under these hypothetical survival assumptions. Deterministic and probabilistic sensitivity analyses were in accordance with the base-case results.
CONCLUSIONS: In the absence of long-term trial data, expert opinion can enhance long-term accuracy and reduce uncertainty in survival extrapolations. In this case study, expert-elicited insights sufficiently improved the ICER to fall below the commonly accepted threshold. However, only one method for incorporating expert opinion was examined, warranting further exploration of alternative methods to validate and expand these findings. Linking these efforts to structured expert elicitation, could enhance the reliability of findings compared to informal methods. Further research should refine these methods and assess their broader impacts on HTA decision making.
METHODS: A three-state partitioned survival model was developed using digitized trial data to evaluate overall survival (OS) and progression-free survival (PFS). Standard parametric survival models were initially fit to these data and assumed hypothetical expert opinions were then incorporated by penalizing the likelihood function. Cost-effectiveness was assessed from the UK National Health Service perspective over a lifetime time horizon. Both deterministic and probabilistic sensitivity analyses were used to assess the robustness of the results.
RESULTS: Incorporating hypothetical expert opinion at one time point which indicated a clear survival benefit for the intervention arm reduced uncertainty in OS predictions and lowered the incremental cost-effectiveness ratio (ICER) for NIVO+IPI from £41,893 / quality-adjusted life year (QALY) to £27,685 / QALY. Patients receiving NIVO+IPI experienced increased incremental QALYs due to prolonged survival. The ICER fell below the £30,000 willingness-to-pay threshold, suggesting cost-effectiveness under these hypothetical survival assumptions. Deterministic and probabilistic sensitivity analyses were in accordance with the base-case results.
CONCLUSIONS: In the absence of long-term trial data, expert opinion can enhance long-term accuracy and reduce uncertainty in survival extrapolations. In this case study, expert-elicited insights sufficiently improved the ICER to fall below the commonly accepted threshold. However, only one method for incorporating expert opinion was examined, warranting further exploration of alternative methods to validate and expand these findings. Linking these efforts to structured expert elicitation, could enhance the reliability of findings compared to informal methods. Further research should refine these methods and assess their broader impacts on HTA decision making.
Conference/Value in Health Info
2025-05, ISPOR 2025, Montréal, Quebec, CA
Value in Health, Volume 28, Issue S1
Code
P41
Topic
Methodological & Statistical Research
Disease
SDC: Oncology