Risk-adjusted value based prices
A new research study outlines an approach for how uncertainty in clinical benefits can be used to set value-based prices that incentivizes stronger evidence.
Many drugs that receive market approvals from EMA, FDA, etc., have large uncertainties regarding the data on clinical efficacy/effectiveness. Therefore, any cost-effectiveness modeling will have equally large uncertainties (if not more). The uncertainty is oftentimes larger today compared to 10-20 years ago, given the increasing reliance on surrogate endpoints and single-arm trials in many pharmaceutical trials. I have written several posts about how we often do not really know if a newly approved drug makes patients live longer and/or better (for example, blog posts here and here and academic publications such as this one).
So, should the degree of uncertainty around the clinical benefits impact the market price of a new drug? The argument against this is that since most payers (especially in national single-payer systems) are very large entities that can effectively pool risks, payers should be risk-neutral and only focus on maximizing expected benefits. In cost-effectiveness settings, this would mean that the only focus should be evaluating the most likely QALY and cost impacts and not care about the uncertainty around these estimates.
It makes sense to care about uncertainty
A new research study by Jiao et al. (Value-Based Pricing for Drugs With Uncertain Clinical Benefits) outlines several of the arguments for why it actually makes sense, also for large payers, to care about uncertainty and to be risk-averse - especially in fixed budget settings. Another argument for caring about uncertainty not directly mentioned in the paper is that study design choices are endogenous to reimbursement and pricing decisions. What I mean by this is that if large payers are willing to pay, all else equal, the same price for drugs with more uncertain evidence, we will obviously see more and more studies with weaker designs (because they are less expensive to run). So, we want payers to incentivize good study designs.
The price impact from modeling risk-aversion
The study by Jiao et al. assumed a payer trying to maximize net benefits and, in an expected utility framework, having some reasonable level of risk aversion. In a situation with uncertainty, a payer can choose between:
Reimburse the drug today with price p based on limited evidence
“Wait and see” and potentially reimburse the drug in the future with price p’ when the evidence is stronger
Choosing (1) implies potentially wasting money and health by implementing a drug that is not cost-effective, whereas choosing (2) implies potentially losing health due to a delayed implementation of a drug that is actually cost-effective. A way to reduce the cost of the current uncertainty is to propose a lower price p today for drugs with higher uncertainty (it reduces the cost of making the wrong decision).
Based on a hypothetical drug with marketing authorization from an accelerated approval process, the authors provide some calculations for a risk-adjusted value-based price (defined as the price where the ICER equals the cost-effectiveness threshold). In a scenario with no uncertainty, the price would equal $2000 per month, which would vary between $1500 and $1900 (with some acceptable level of uncertainty) with varying risk-aversion assumptions. See further details in the paper (open access). In this scenario, we would thus consider up to a 25% lower price to account for the uncertainty.
Implementation in practice
The paper touches upon a very current topic. As the evidence base for new drugs is frequently relatively poor, it becomes increasingly important for HTA agencies and payers to decide if and how uncertainty should feed into pricing negotiations. The paper mentions that, for example, the Medicaid system in the US requires larger rebates for drugs approved in the acceleratad approval pathway. In a recent post, I wrote that the Swedish government agency (TLV) will start to consider the uncertainty when deciding on the price premium allowed for ultra-orphan drugs.
I don’t think any payer is willing to be explicit about a utility function and risk-aversion parameters. Still, I think it is reasonable to consider some factors that should require, all else equal, a price discount. For example, I think the following factors could be explicitly used as determinants of price discounts:
Clinical trial data relying on surrogate endpoints (especially if not properly validated for the exact indication)
Single-arm trials
“Small” sample sizes
Trial data from a healthcare context that is substantially different from the payer’s (where the standard of care is not comparable, etc.)
A payer evaluating a new drug for reimbursement that checks all the above boxes should be considering demanding a substantial price discount.
We want innovation in drugs that are truly changing the lives of patients. To do so, we should pay a lot for drugs that can credibly show this, meaning that we also need to pay less for drugs where this is more uncertain.