realkerimc

That qualifies my proposed test: even with a matched estimand, the preferred option could depend on how long treatment effects are assumed to persist beyond trial follow-up.

For an economic evaluation, the practical test is whether each estimand supplies the treatment effect needed by the model’s decision context. An effect in trial participants under a hypothetical no-discontinuation strategy may support an efficacy comparison, while reimbursement under routine use may require the treatment-policy effect in the eligible population. If those effects feed different cost-effectiveness estimates, the apparent conflict can change the preferred option without either interpretation being internally inconsistent.