Indirect treatment comparisons - choosing the right tool for the job
It was great to participate in this webinar on indirect treatment comparisons from the HTA European Special Interest Group. This is a topic that is increasingly relevant within the context of new European Union Joint Clinical Assessment and the development of the EUnetHTA 21 methodological guidelines. I spoke primarily about marginal versus conditional estimands, non-collapsibility and standardization in the specific context of indirect treatment comparisons.
There has been active debate among researchers about whether marginal or conditional treatment effect estimands are more appropriate inferential targets for health technology assessment. Marginal estimands are widely regarded as more appropriate for population-level decision making. In indirect treatment comparisons, challenges arise from the conflation and potential incompatibility of marginal and conditional estimands. This is particularly the case when effect measures are non-collapsible. Model-based marginalization or standardization methods allow one to produce marginal, but covariate-adjusted, treatment effect estimates that are compatible in indirect treatment comparisons.
Established covariate-adjusted indirect comparison approaches, e.g., matching-adjustment, can be viewed as methods that marginalize over the covariate distribution of an external study. I outlined currently available covariate adjustment methods for indirect treatment comparisons, and highlighted some of their advantages and disadvantages. I also discussed the implications of non-collapsibility to currently recommended practices in evidence synthesis, the assessment of effect measure modification and heterogeneity, and the definition of estimands with limited access to subject-level data. Standard practices may require updating in light of conflicting views!
There were also very interesting talks from Anja Schiel (Norwegian Medical Products Agency) about the applicability and acceptability of indirect treatment comparisons in the European HTA landscape, and from Suzy Van Sanden (Johnson & Johnson) about the application of matching-adjusted indirect comparison to a prostate cancer case study. Many thanks to Min-Hua Jen (Lilly) for the invitation to present.
