Current Edition

Follow-up studies often lacking for cancer drugs approved early

  • Cancer drugs are often approved quickly by the Food and Drug Administration. In many cases, however, initial evidence of efficacy is never confirmed by studies testing for a survival benefit, according to new research published Tuesday in JAMA Internal Medicine.
  • Just a fifth of cancer drug approvals between 1992 and early 2017 were eventually supported by clinical trials demonstrating an improvement in patient survival, researchers found. More than 40% of the time, the follow-up trials used to confirm early results tested the same surrogate measure, such as progression-free survival. 
  • Surging industry investment in oncology, coupled with the FDA’s desire to speed patient access to new cancer drugs, has resulted in a steady stream of approvals from the agency. Many of those drugs come with annual price tags stretching to six figures, making what patients are getting for the cost a critical question. 

Cancer drugmakers are frequently the beneficiary of accelerated approvals from the FDA. Over the past four years, 17 of the 22 treatments granted an early, conditional OK by the agency were cancer treatments.

And under recently departed agency chief Scott Gottlieb, the FDA intensified its efforts to speed new therapies to market.

Gottlieb has defended the agency’s evolving approach to evaluating new cancer drugs, citing advances made in targeted therapy and in selecting the right patients for clinical study. 

“We’re trying to balance competing needs,” he said in a speech last June. “The need for high degrees of statistical certainty, and the need for access to care.”

Soon after Gottlieb’s speech, the FDA published a list of surrogate endpoints used by the agency to approve cancer drugs, hoping be more transparent in how it decides to clear drugs without evidence of a survival benefit. 

Two studies published this week in JAMA Internal Medicine, however, call into question that approach and argue for stricter management of how drugs are OK’d on an accelerated basis. 

One, from researchers at Brigham and Women’s Hospital, Harvard Medical School and Queen’s University in Canada, details how few drugs approved on an accelerated basis ever end up being supported with survival data.

The study authors examined 93 cancer drug approvals for 64 separate products which were conditionally granted by the FDA between 1992 and May 31, 2017. Clearance was given based on surrogate measures like progression-free survival and overall response rate, which some have argued are poorly correlated with clinical outcomes. 

Just 19 of those approvals were eventually backed up by confirmatory studies showing an improvement in overall survival. Researchers found many of the studies used by drugmakers to confirm initial findings used surrogate endpoints as well — and, also in 19 cases, the same surrogate.

Notably, in only five cases was a conditional approval withdrawn as a result of negative confirmatory data. 

The other JAMA study, conducted by researchers at Oregon Health and Science University, found that for 85 cancer drug indications the median response rate used to support approval was 41%. Fourteen of those approvals were granted based on data showing tumor response rates less than 20%.

Taken together, the two study findings “raise several doubts about the success of the accelerated approval process,” wrote Sarah DiMagno and colleagues from the University of Pennsylvania in commentary also published Tuesday.

They argue that the FDA should not permit drugmakers to use the same surrogate endpoint in confirmatory study, and call for more rapid withdrawal of approval if follow-up results show significant toxicity or no clinical improvement.

“The reliance on surrogate end points has real costs for patients and society,” they wrote. “Drugs with unproven effectiveness sell false hope to desperate patients, who are likely paying thousands of dollars out of pocket for them.”

Researchers in both studies did acknowledge, however, the wide range of differences in cancer types, differences that can change the context for use of a surrogate measure or for interpreting response rates.