An interesting paper from some of my colleagues at Precision Health Economics:
Approval of new drugs is increasingly reliant on “surrogate endpoints,” which correlate with but imperfectly predict clinical benefits. Proponents argue surrogate endpoints allow for faster approval, but critics charge they provide inadequate evidence. We develop an economic framework that addresses the value of improvement in the predictive power, or “quality,” of surrogate endpoints, and clarifies how quality can influence decisions by regulators, payers, and manufacturers. For example, the framework shows how lower-quality surrogates lead to greater misalignment of incentives between payers and regulators, resulting in more drugs that are approved for use but not covered by payers. Efficient price-negotiation in the marketplace can help align payer incentives for granting access based on surrogates. Higher-quality surrogates increase manufacturer profits and social surplus from early access to new drugs. Since the return on better quality is shared between manufacturers and payers, private incentives to invest in higher-quality surrogates are inefficiently low.
Source:
- Katalin Bognar, John A. Romley, Jay P. Bae, James Murray, Jacquelyn W. Chou, Darius N. Lakdawalla. The role of imperfect surrogate endpoint information in drug approval and reimbursement decisions. Journal of Health Economics. Volume 51, January 2017, Pages 1–12
I often find that when economist discuss commons problems like this (i.e. shared benefits), they implicitly take the continuous benefits model (e.g. the more I contribute the larger the gain to society) as the correct model, when in most cases the discreet benefits model (e.g. as long as the amount of contributions/effort given reaches x amount, then the good is provided). In the latter case, there is usually an equilibrium in which one party is willing to supply the good. If the share of benefits that would accrue to manufacturers is large enough, then they have a private incentive to use high-quality surrogate end-points.