How Similar is Similar Enough? Assessment of Indirect Treatment Comparisons to Support Similarity for National Institute for Health and Care Excellence's Cost Comparison Route.

Objectives: Health technology assessment bodies are increasingly using cost comparison (cost-minimization) analysis to manage rising demand. Such an approach requires demonstration of clinical similarity. Although head-to-head comparisons from equivalence or noninferiority studies are usually accepted as evidence of similarity, guidance is lacking on when equivalence may be assumed from indirect treatment comparison (ITC), either quantitative or qualitative.

Methods: We reviewed methods articles and previous National Institute for Health and Care Excellence appraisals to assess methods for determining equivalence when no head-to-head trial exists, identify how National Institute for Health and Care Excellence and their committees decide if a cost comparison route is suitable, and explore how best to present evidence of similarity to aid decision making.

Results: We identified 5 methods articles, 41 case studies, and 33 appraisals that used cost comparison based on ITC. All 33 appraisals were recommended, although none applied any of the formal methods identified from the literature. Instead, companies used narrative summaries to assert similarity, often based on a lack of significant differences, leading to committee uncertainty that was typically resolved through clinical expert input. The most promising methods we reviewed were the estimation of noninferiority ITCs in a Bayesian framework followed by a straightforward, probabilistic comparison of the indirectly estimated treatment effect against a prespecified noninferiority margin.

Conclusions: Formal methods to assess equivalence in ITC-based cost comparison are emerging but have not yet been applied in practice. We provide recommendations for how these methods and qualitative methods, such as evaluation of the plausibility of class effects and clinical expert input, could help address uncertainties in assuming equivalence.