If you are a payer or a policymaker, should you except the results of a global clinical trial or require additional evidence that a drug works in your country. There are a few options to consider.
- Delay approval until local confirmatory trial. Approach enacted in several Asian countries (see Habr et al. 2024). While this policy reduces uncertainty, it delays access to potentially life-saving therapies.
- Conditionally adopt. In this case, patients gain access, which payers wait for post-market confirmatory trial is conducted in parallel. While this approach is good in theory, manufacturers may be incentivized to delay trial completion and enforcement is often weak (see Drummond et al. 2022)
- Approve. Gets drugs to patients sooner, but does not resolve uncertainty.
- Reject. Strongest demand for evidence and least expensive, but most problematic for patients with high unmet needs/few treatment options.
Clearly, if you have a very large clinical trial, with very strong results for a new medical technology that is not that costly, you should lean towards approving or conditional adoption. If the clinical trial has small sample size, weak results and the treatment is expensive, you would lean more towards delayed approval or rejection. How can we formalize this decision?
A paper by Liang and Jiao (2026) aims to create a decision analytic framework using a value of information (VOI) methodology to measure the pros and cons of each approach depending on the quality of evidence among other factors. Specifically, the authors use the power prior approach (see Ibrahim et al 2015), which is a flexible Bayesian method that down-weights external evidence based on its perceived relevance to the local context. The methodology allows users to put an exponent (alpha) on the Bayesian prior (the results from the trial) where if exponent equals 1, then the external and local data are considered interchangeable; if the exponent equals 0, then the external data is discarded and only local data matters. In between, exponents closer to 1 give more weight to the external trial data; those closer to 0 give less weight to external data. Note that the model does not model jurisdiction specific treatment effects, but focuses on uncertainty in the applicability of the external data to local contexts.
One key contribution of this paper is that they explicitly model that conditional approval requiring a local trial weakens manufacturers incentives to expedite the local trial (since the drug is already covered until the local trial results read out).
What do the authors find?
When uncertainty is high (alpha<0.3, delayed adoption pending local trial yields the highest benefits. However, once alpha exceeds 0.3, conditional adoption with local trial requirement (policy 2) becomes consistently optimal…notably, when alpha >0.6, delayed adoption is even less favorable than immediate adoption without a trial requirement…
shorter trial durations (1-2 years) mitigate the costs of waiting, making delayed adoption the preferred strategy at moderate to high uncertainty levels. However, as trial timelines lengthen to 5 to 6 years, conditional adoption becomes increasingly favored. Longer trial durations amplify the opportunity costs of delay…
When additional delays [in confirmatory trial completion] are modest (1-2 years), conditional adoption remains the preferred strategy across a wide range of uncertainty levels. However, as additional delays extend to 4 to 6 years, delayed adoption becomes increasingly favorable, particularly under moderate to high uncertainty, as the prolonged period of unresolved uncertainty diminishes the NB of early access…
Reducing the launch price by 20% to 30% below the value-based price significantly enhances the relative benefit of conditional adoption, even when uncertainty is substantial…
Very interesting throughout. You can read the full paper here.