The study objective was to investigate the economic value of tumor-agnostic therapies when only single-arm effectiveness data are available at launch by applying multiple methodologies to establish comparative effectiveness.
In the absence of direct comparative data, 3 methods were used to estimate the counterfactual: (1) a historical control based on a systematic literature review for each tumor site from the larotrectinib trials, (2) an intracohort comparison using the previous line of therapy time to progression from larotrectinib trials, and (3) a nonresponder control that applied outcomes for larotrectinib nonresponders. Cost-effectiveness was modeled using the partitioned survival approach. Stochastic parameter uncertainty was assessed in a probabilistic sensitivity analysis (PSA). A triangulated estimate of the mean cost-effectiveness result was generated combining all 3 counterfactual estimates.
Incremental cost-effectiveness ratios were similar across the 3 methodologies in the deterministic analysis ranging from £83 868 (95% uncertainty interval [UI] £65 698-£107 668) to £104 922 per quality-adjusted life-year (95% UI £80 132-£139 658). PSA results for each method substantially overlapped when plotted on the cost-effectiveness plane. Weighting PSA results for each method equally in the triangulation method produced an incremental cost-effectiveness ratios of £95 587 per quality-adjusted life-year gained (95% UI £70 449-£137 431).
In the absence of direct comparative data, different methods of estimating a counterfactual are possible, each with strengths and limitations. Triangulating results across the methods provides a composite view of the total uncertainty and a more consistent estimation of the cost-effectiveness of the tumor-agnostic intervention compared with choosing a single method.