An Economic Model of Surrogate Endpoints

Much has been written about how to make healthcare decisions under imperfect information. Less attention has been paid to the related question of how much to invest in refining the set of information about new medical therapies.  This question arises frequently, because firms developing medical therapies often have to rely on noisy “surrogate endpoints” to proxy for clinical benefits in contexts where the ultimate benefits of interest – e.g., life expectancy – take a long time to observe.  We propose a model of decision making by regulators and payers facing surrogate endpoint evidence.  From a positive standpoint, we find that payers and regulators should in fact rely more on preconceived prior beliefs about clinical benefit, when evidence is noisy.  While intuitive from a statistical standpoint, this often runs counter to the expectation that only the directly relevant clinical trial evidence matters.  Moreover, rather than refusing to approve or reimburse drugs on the basis of noisy evidence, payers and regulators should simply demand a higher level of surrogate benefit.  Prices and social surplus also depend on the noise in measures of benefit:  Noisier measures lead to lower prices paid to manufacturers, as well as lower social surplus earned by payers and patients.  From a normative standpoint, we show that regulators approve an inefficiently high number of new therapies as they fail to consider the economic costs of using a therapy.  Manufacturers base their price on negotiations with the payer based on surrogate endpoint evidence.  Finally, payers grant access for too few new drugs as they focus on the price of therapy instead of its lower marginal cost of production.  This study suggests that allocating public resources towards the development of more precise surrogate endpoints would result in more socially optimal decisions.