Skip to Main Content

Outrage about the rising prices of prescription drugs has put cancer drugs in the spotlight. But there’s an important question that needs to be asked: What is the best pricing metric to use for them?

Economists like us spend a lot of time worrying about this in a more general context. When the government computes its consumer price index, for example, it thinks about the quality of the goods that households purchase. A computer purchased 10 years ago is almost unrecognizable compared to one purchased today, so the consumer price index must adjust for CPU speed, memory, storage capacity, and more.


Similar issues arise in the housing market. The price of a home reflects a number of features — bedrooms, bathrooms, location, and the like. Which features matter the most? Consumer sites like Redfin and Zillow simplify the listing price into one metric: price per square foot. This won’t tell you if the house has a mountaintop view though, and no one would question paying more for such an amenity.

When it comes to cancer therapy, one metric dominates: median gain in survival. It tells you the gain in survival from a new drug therapy at the point when half the participants in a clinical trial have died. Take, for example, Yervoy (ipilimumab), a drug to treat metastatic melanoma. To gain FDA approval, it was tested in a clinical trial against the next best alternative, Freund’s adjuvant. Half of the patients in the control group had died after six months. In contrast, it wasn’t until 10 months that half the patients in the treatment group had died. Thus, Yervoy generated a median survival gain of 4 months.

Critics of the high launch prices of new cancer drugs cite median survival gain as a warning to Medicare and private insurers that they are paying too much. When Yervoy first launched in 2011, the treatment cost more than $120,000. How can drug companies justify charging so much when their drugs extend life by just a few months for half the clinical trial participants?


But as in real estate, one cancer treatment metric doesn’t tell the whole story. Taken alone, median survival gains can make a price look like gouging. But it doesn’t capture the value accruing to the other half of the participants in the clinical trials. Returning to the Yervoy example, 50% of patients in the treatment group survived significantly longer than the median point of 10 months. In fact, 10% of the treatment group was still alive after 2 1/2 years, while almost everyone in the control group had died. These longer-term gains in survival — the cancer drug equivalent to the mountaintop view — are ignored by the metric of median survival gain.

We need a better measure to resolve the raging debate over the value of new and highly effective cancer therapies compared to their high costs.

Over the last two decades, the rise in cancer drug launch prices in the U.S. has quickly surpassed the growth in household incomes. Almost all new cancer drugs that enter the market have price tags higher than $100,000 per year of treatment. Notably in 2017, Kymriah, a chimeric antigen receptor (CAR) T cell therapy, launched at $475,000 per treatment. Even with negotiated discounts, such a cost poses challenges to anyone responsible for the bill.

Some patients with blood cancers who are given CAR-T treatment don’t respond to the therapy. Others who have only weeks to live have seen their tumors sent into durable remission.

Median survival gain can’t accurately capture that varied experience. Although it is a hugely convenient measure, because it means a trial can be discontinued once half the patients have died, it is blind to the survival gains of the remaining patients, and collecting data on their survival can require lengthy follow-up periods or approximations through statistical models.

Research has shown that patients care more about long-term cures than median survival — a phenomenon our economist colleague Darius Lakdawalla has described as the “value of hope.” When cancer patients face poor survival odds, they value treatments that give them a shot at a durable cure. These additional months or years can be better represented through mean survival gain, which considers the survival gains of all patients in the sample.

The difference between mean and median matters when it comes to new treatments. Median represents the midpoint in a range of values. Mean represents the average across all values.

From 1995 to 2017, new drugs increased median survival gains by an average of about six months, while mean survival gains increased by almost a full year. During this time period, the cost of cancer drugs rose faster than median survival gains but didn’t outpace growth in mean survival gains. Put another way, if we adjusted cancer drug prices the same way we adjust the prices of housing or computers, we wouldn’t see any increase at all.

International health regulators get this point. The United Kingdom’s National Institute for Health and Care Excellence relies predominately on mean survival gains to appraise new drug technologies. The need to account for longer-term survival has also been recognized by institutions such as the American Society of Clinical Oncology and the Institute for Clinical and Economic Review.

Promising therapies await more and more people with cancer. Striking a balance between affordability and future innovation will require policies that, among other considerations, accurately relate a drug’s price to its total value. We need to take into account the mountaintop view, not just the price per square foot.

Alice Chen is an assistant professor at the Sol Price School of Public Policy and the Leonard D. Schaeffer Center for Health Policy & Economics at the University of Southern California. Dana Goldman is professor of public policy, pharmacy, and economics at the University of Southern California and director of the Leonard D. Schaeffer Center for Health Policy & Economics. He is a scientific advisor to and holds equity in Precision Health Economics, a consultancy to life sciences companies, some of which produce cancer drugs.

  • Are cancer drug producers playing God? The sheer audacity of putting a monetary value on whatever length of extension of life is utter gouging, and it is despicable. That such arbitrary “pricing indices” put the drugs FAR out of reach for the “average” cancer patient seems not to matter. What does that make those drug developers? lecherous devils. Not the heros they could be.

  • The downside is that clinical trials measuring mean survival gain would take longer and cost more. Isn’t the mean survival gain estimated (for example, by fitting a statistical distribution to the survival data)?

    • Yes, due to the downside that you point out, mean survival gains are mostly estimated from statistical extrapolations of existing clinical trial data. Our embedded link to the UK’s “National Institue for Health and Care Excellence” provides a practical assessment of the validity of different statistical models that have been used to estimate full survival benefits of new cancer therapies.

    • Thanks, that’s what I figured. If you have access to the individual patient data from the trials, finding a statistical distribution that fits is a lot easier that trying to reconstruct it from the Kaplan-Meier curves (even with patients at risk data).

Comments are closed.