Bayesian multi-parameter evidence synthesis to inform decision-making: a case study in hormone-refractory metastatic prostate cancer

TL;DR

This paper demonstrates how Bayesian multi-parameter evidence synthesis, specifically bivariate meta-analysis, can improve decision-making in health technology assessments by estimating unreported treatment effects to enable more accurate cost-effectiveness models.

Contribution

It introduces the application of bivariate meta-analysis to predict unreported treatment effects, facilitating more detailed multi-state models in health economic evaluations.

Findings

Three-state model yielded lower cost-effectiveness ratio ({ extsterling}21,966/QALY) than two-state model ({ extsterling}30,026/QALY).

Using BVMA allowed inclusion of unreported treatment effects, improving model accuracy.

Advanced meta-analysis techniques can prevent data wastage and enhance decision-making in health technology assessment.

Abstract

In health technology assessment, decisions are based on complex cost-effectiveness models which, to be implemented, require numerous input parameters. When some of relevant estimates are not available the model may have to be simplified. Multi-parameter evidence synthesis allows to combine data from diverse sources of evidence resulting in obtaining estimates required in clinical decision-making that otherwise may not be available. We demonstrate how bivariate meta-analysis (BVMA) can be used to predict unreported estimate of a treatment effect enabling implementation of multi-state Markov model, which otherwise needs to be simplified. To illustrate this, we used an example of cost-effectiveness analysis for docetaxel in combination with prednisolone in metastatic hormone-refractory prostate cancer (mHRPC). BVMA was used to model jointly available data on treatment effects on overall survival (OS) and progression-free survival (PFS) to predict the unreported effect on PFS in a study evaluating docetaxel. Predicted treatment effect on PFS allowed implementation of a three-state Markov model comprising of stable disease, progressive disease and death states, whilst lack of the estimate restricted the model to two-state model (stable disease and death states). The two-state and three-state models were compared by calculating incremental cost-effectiveness ratios, which was much lower in the three-state model: {\pounds}21966 per QALY gained compared to {\pounds}30026 obtained from the two-state model. In contrast to the two-state model, the three-state model has the advantage of distinguishing patients who progressed from those who did not progress. The use of advanced meta-analytic technique helped to obtain relevant parameter estimate to populate a model which describes natural history more accurately, and at the same helped to prevent valuable clinical data from being discarded.