Survival Extrapolation in Cancer Immunotherapy: A Validation-Based Case Study.

Area of Science:

• Health Technology Assessment
• Survival Analysis Modeling
• Oncology Drug Efficacy

Background:

• Immune-checkpoint inhibitors offer long-term survival benefits, potentially through a cured patient proportion.
• Health technology assessment bodies often lack sufficient data to confirm these long-term benefits.

Purpose of the Study:

• Revisit the National Institute for Health and Care Excellence assessment (TA319) of ipilimumab in melanoma.
• Evaluate the accuracy of extrapolation methods used in TA319 with updated trial data.
• Compare previously unused survival extrapolation techniques to identify more accurate projection methods.

Main Methods:

• Compared survival projections from the TA319 piecewise model with alternative models (parametric, spline, mixture, mixture-cure) using a minimum 3-year follow-up.
• Validated model projections against a 5-year follow-up data cut and external data.
• Assessed various extrapolation techniques including mixture-cure models.

Main Results:

• The TA319 model and a mixture-cure model accurately predicted 5-year survival.
• Standard parametric, spline, and non-curative mixture models underestimated 5-year survival.
• Survival projections from the TA319 and mixture-cure models diverged significantly beyond 5 years, remaining unvalidated.

Conclusions:

• Models incorporating external information (cure fraction, background mortality, registry data) accurately estimated 5-year survival.
• Flexible models lacking external information extrapolated poorly, failing to capture long-term survival trends.
• Accurate long-term survival estimation for immune-checkpoint inhibitors necessitates incorporating external data sources.