Prices for specialty drugs in the United States are out of control, with spending rising much faster than in many other health care domains. Some state Medicaid programs have been driven to the brink by the cost of new drugs for diseases such as hepatitis C, for which 12 weeks of treatment with Sovaldi can cost nearly $100,000.
Many people are cheering a new potential solution: paying for drugs according to how well they actually work. This commonsense idea comes in a few flavors, which I’ll discuss later, but the underlying principles are the same. A drug that works is worth something; one that doesn’t is not. If a new drug works no better than an older one, the two have equal worth. If a drug costs a lot, that’s OK only if it makes people so healthy that it reduces their spending on other forms of health care.
like a move toward paying smarter or, in health care lingo, “paying for value.” But one big problem plagues a lot of these (and other) ideas in health care: They just don’t scale. In a vacuum, a particular idea might work for a specific drug, but only if it focuses on isolated factors and ignores the long-term context of treatment, different types of treatment, and multiple (often subjective) aspects of value.
Before I unpack the flaws of existing drug-pricing proposals, let me share a prototype that my colleagues and I at Memorial Sloan Kettering Cancer Center developed. It’s an interactive drug-pricing tool — the DrugAbacus — that integrates objective information about cancer drugs while empowering users to define what value means to them.
Yes, everyone agrees that a drug’s value is tied to its clinical benefit for the patient. But how do you measure the impact of, say, drug side effects? If they’re bad, is the drug less valuable? What about if the drug costs a lot to develop or if its novel mechanism of action breaks new scientific ground? What if it treats a rare disease? Do those considerations add value? The DrugAbacus lets you decide how much these and other factors are related to a drug’s value. The tool’s intended users are policymakers who wish to explore the idea of finding fair prices for drugs while respecting the complexity and subjectivity of what “value” means.
Right now, the DrugAbacus incorporates data on many attributes of 54 recently approved cancer drugs. It still lacks some potentially important information, such as how patients describe their experience with the treatment, and my colleagues and I will address those limitations. Nevertheless, by combining objective data about a drug with the tool user’s own assessment of value, the DrugAbacus has a distinct advantage over alternative drug-pricing proposals:
The current system. The present-day method of drug pricing has the benefit of simplicity. Manufacturers charge whatever they want. But the problem is obvious: excessively high prices that defy market forces. For example, in today’s dollars Novartis charged $4,540 for a month of the leukemia drug Gleevec as recently as 2001, but now it costs $8,500 in the United States, even though it is just $4,500 and $3,300 a month in Germany and France, respectively. I didn’t cherry-pick this example. Cancer-drug prices in the U.S. are up more than fourfold in the past two decades, after adjustment for inflation.
Paying for drugs when they “work.” One alternative to the current system proposes paying for a drug only when it benefits the patient. The concept has been tried: In an agreement with the UK’s National Health Service, the maker of the blood-cancer drug Velcade has to rebate money for the drug when it does not change a patient’s blood count as intended.
However, such an approach does not work when treatment involves multiple drugs or, more important, when the marker of benefit to the patient is more opaque. Indeed, the largest gains in cancer care come from giving chemotherapy and other drugs to patients after the cancer appears to (but actually may not) be gone. Many experts say that more than half of our progress in combating breast cancer comes from this type of treatment, called adjuvant therapy. But we can’t definitively pinpoint its success in an individual patient and then price a drug accordingly.
Paying for drugs when they yield “overall savings.” Other pricing proposals consider how much money a drug ultimately saves by improving a patient’s overall health. Consider Novartis’s plan to charge more for its heart-failure drug Entresto only if the drug succeeds in reducing often-costly hospital admissions for heart failure. And Procter & Gamble and Sanofi-Aventis have entered a program with Health Alliance whereby the drug makers pay for bone fractures that occur when a patient is on their drug Actonel, which is designed to prevent such events. This idea has four drawbacks:
- It can’t be applied consistently across drugs. If we link prices for some drugs to the overall savings they produce, many clearly useful drugs will appear to lack value. For instance, drugs that allow leukemia patients to get well enough to progress to a potentially curative but very costly bone-marrow transplant would have a poor price profile because, in the end, they increase total spending. Should we really consider a drug that starts a patient on a path to a possible cure to be of lower value and thus deserving of a lower price?
- It does not take a truly long-term view. After all, most interventions that extend life increase total spending in the end, because a patient who lives longer ends up needing more health care. Even Sovaldi, Gilead’s hepatitis C drug, comes out in every analysis as costing, not saving, the system money.
- It mistakenly assumes that the savings themselves are correctly priced. Remember that we pay too much for most types of care in the U.S., not just drugs. Pricing a heart-failure drug according to how well it prevents an overpriced hospitalization for heart failure just doesn’t make sense.
- It, like other drug-pricing proposals, is at odds with our true goal for health care spending — to innovate our way to a lower-cost system. Overall savings from a drug should be passed through to society at large, not passed back only to the drug maker.
A modest next step. Current drug-pricing proposals simply don’t take the big picture into account. If we try to build a separate pay-for-performance structure for each new drug, we will quickly discover that we can’t come up with a practical one for most drugs. The details for a viable alternative pricing system are not easy to outline, of course. But we can aim for two things:
First, start with a price for a drug given how well it works in initial clinical trials that are intended to win FDA approval. Down the road, recalibrate the price using data on how well the drug helps real patients in practice, potentially captured in medical records, patient registries, and clinical-outcomes research. The new data on the drug’s performance could then be built into a new calculation of value. That kind of approach would reward the most successful innovators. It would also strongly encourage manufacturers to work with providers and insurers to maximize the real-world efficacy of their products so that reimbursement levels are maintained. That means launching more programs to help patients adhere to their prescribed drug regimens — and curbing instances of off-label use that have no health benefit.
Second, work to develop a pricing formula that links price with the ever-elusive “value” of a drug. That’s where the DrugAbacus comes in. The tool makes explicit which features and advantages of a cancer drug get incorporated into its price — and shows how that drug compares with its peers. In short, the tool provides context for value-based decision making about drug pricing.
The DrugAbacus has limitations, as I discussed earlier. But I believe that it provides a template for how we need to start thinking about value in drug pricing — by capturing some of the inherent complexity of value-based decisions without making the fundamentally flawed assumptions that are embedded in other drug-pricing proposals.