OBJECTIVES: Economic evaluations based on clinical trials may lack generalizability if heterogeneity is not accounted for. We explore different methods to account for covariates to obtain cost-effectiveness results in different populations and assess the expected value of perfect information (EVPI).

METHODS: Data from the TASMINH-2 trial were analyzed to estimate the cost-effectiveness of self-management of hypertension. Covariate selection was based on fitting univariate and multivariate regression models for costs and quality-adjusted life years (QALYs) including all baseline characteristics as covariates. Important covariates from a statistical and clinical perspective were considered in flexible Bayesian generalized linear models. The generalizability of the results to the target population that consisted of patients with hypertension recruited in real-world evidence (RWE) studies in the UK was explored by averaging over the joint covariate distribution in the target population via numerical integration methods.

RESULTS: Three covariates (baseline EQ-5D, IMD score, chronic kidney disease [CKD]) were identified as important predictors and were included in the analysis. Patients recruited in RWE studies had lower baseline EQ-5D, were slightly more deprived and were more likely to have CKD compared to patients recruited in the TASMINH-2 trial. At a cost-effectiveness threshold of £20,000/QALY, self-management of hypertension was associated with an incremental net monetary benefit (INMB) of -£351 in patients recruited in TASMINH-2. When generalizing the results to patients recruited in RWE studies, the INMB was estimated at -£551. Population EVPI for a 10-year time horizon was estimated at £77million for patients recruited in TASMINH-2 and at £39million for patients recruited in RWE studies.

CONCLUSIONS: We propose a methodological framework to assess the generalizability of cost-effectiveness and value of information estimates to the real-world setting in the presence of heterogeneity. Such methods can guide optimal resource allocation and future research decisions.