OBJECTIVES

Modelling the cost-effectiveness of histology-independent technologies (HITs) creates a number of challenges as a consequence of the breadth of the population covered by the marketing authorisation, the evidence of effectiveness typically gathered across many tumour types/histologies powered for response assessment, and the potential for heterogeneity. In this study, we explore the implications of different approaches to decision modelling for the economic evaluation of HITs.

METHODS

A simplified landmark response-based modelling approach was used to incorporate survival distributions, conditional on response for a hypothetical HIT. Response rates used in the regulatory approval of the HIT, larotrectinib, were used as inputs. The modelling approaches were: (i) all histologies have the same, average response rate observed in the trial; (ii) all histologies have independent response rates based on those observed in the trial; (iii) all histologies have exchangeable responses, generated from the posterior response probability estimates from a Bayesian hierarchical model.

RESULTS

The results of (i) suggest the hypothetical HIT has a 78% probability of being cost-effective for all histologies at a threshold of £50,000 per QALY. The results of (ii) suggest a range of cost-effectiveness probabilities for the histologies represented in the trial: 95% for thyroid cancer, 0% for cholangiocarcinoma. Finally, (iii) shows less extreme histology-specific results compared to (ii): 93% and 11% probability of being cost-effective for thyroid and cholangiocarcinoma, respectively. Unrepresented tumour types have a 52% probability of being cost-effective.

CONCLUSIONS

The results have implications for HIT modelling. The assumptions in (i) do not account for heterogeneity in the effectiveness data and the resulting high probability of cost-effectiveness could be misleading. The assumptions in (ii) are based on extremely limited and uncertain data, and do not provide cost-effectiveness estimates for unrepresented histologies. Assumptions in (iii) allow for heterogeneity in effectiveness results and generate evidence for the unrepresented tumour types.