Addressing subgroup-specific treatment effects to avoid inefficient care: the conflict between statistical and health economic analyses.

Abstract

Objective: When patients are classified into subgroups based on previously identified heterogeneity, this heterogeneity may affect the cost-effectiveness of health interventions. Whether this heterogeneity is reflected or ignored in cost-effectiveness analysis (CEA) may influence reimbursement decisions. This is illustrated using a simulation study of a hypothetical treatment to prevent disease progression.

Methods: With the DARTH Sick-Sicker Markov model, we analysed the cost-effectiveness of Treatment versus Standard of Care in a population comprising Group 1 (G1) and Group 2 (G2). We compared three strategies for informing reimbursement decisions; 1) ignore evidence on subgroup differences ('ignore subgroup evidence'); 2) test for subgroup differences in trial data at hand ('statistically-guided'), and 3) use all evidence on subgroup differences ('all evidence'). This simulation study varied total sample size, G2 proportion, treatment effectiveness, and baseline mortality risk. For each scenario the net health benefit (NHB) and reimbursement decision (i.e. reimburse in both subgroups, G1 only, G2 only, or no reimbursement) was determined per strategy.

Results: The statistically-guided strategy led to subgroups being ignored except for the largest total sample sizes. At a willingness-to-pay threshold of €50,000/QALY gained, the statistically-guided strategy resulted in an incremental NHB of -1.00 and 0.49 when compared to the strategies of ignoring subgroup evidence and for incorporating, respectively.

Conclusions: When subgroup heterogeneity is known, ignoring subgroups, or taking a statistically-guided approach will result in suboptimal reimbursement decisions and thus fail to optimise societal benefits. Therefore, subgroup-specific CEAs should be informed by all available evidence of subgroup differences.