Methods of multi-indication meta-analysis for health technology assessment: A simulation study

Area of Science:

• Health Economics and Outcomes Research
• Biostatistics
• Oncology

Background:

• Bevacizumab and similar oncology drugs are approved for multiple indications.
• Health technology assessment (HTA) traditionally appraises drugs within single indications, limiting evidence scope.
• Existing HTA methods may not fully leverage data from all drug indications.

Purpose of the Study:

• To explore and evaluate multi-indication meta-analysis methods for synthesizing evidence across indications.
• To assess the performance of different synthesis models in predicting overall survival (OS) in a target indication.

Main Methods:

• Conducted a simulation study using a multistate disease progression model.
• Evaluated univariate (mixture and non-mixture) and bivariate surrogacy models.
• Simulated datasets with varying heterogeneity, outlier indications, and OS data availability.

Main Results:

• Univariate multi-indication methods reduced uncertainty without increasing bias when OS data was available in the target indication.
• Mixture models offered no significant performance improvement over univariate methods for HTA.
• Bivariate surrogacy models showed potential for bias correction in scenarios lacking target OS data and with outlier indications.

Conclusions:

• Multi-indication meta-analysis methods, though complex, can enhance HTA by reducing uncertainty.
• Univariate methods are effective when target OS data is present.
• Bivariate models warrant further investigation for complex HTA scenarios.