Drive-by analysis of accelerated approval is intellectual malpractice

Medical journals, news publications incorrectly conclude half of cancer accelerated approvals are useless

Academic opponents of accelerated approval have cranked out another misleading paper, this time creating the false impression that about half of the cancer drugs granted accelerated approval turn out to be useless.

Presented at this month’s American Association for Cancer Research (AACR) meeting and given additional stature by publication in JAMA, the paper purported to assess the clinical benefit of cancer drugs that were granted accelerated approval, and concluded that over a five-year period 41% “did not improve overall survival or quality of life in confirmatory trials after more than 5 years of follow-up.” For the subset of drugs converted to full approval, 60% of the conversions relied on a surrogate endpoint.

The paper was written by researchers at the Program on Regulation, Therapeutics, and Law (PORTAL) at Brigham and Women’s Hospital. One of the authors, Aaron Kesselheim, was an author on a similar paper published in JAMA Internal Medicine in 2019.

The authors repeatedly note that many cancer drug accelerated approvals are converted to full approval in the absence of data demonstrating an improvement in overall survival. They fail to point out that this isn’t surprising because FDA had not made conversion to full approval contingent on demonstrating an overall survival benefit.

The clear implication is that the lack of evidence of survival improvements necessarily reflects poorly on either the drugs or FDA. That kind of blanket assertion in the absence of an analysis of the reasons why the agency may not have required evidence of a survival benefit is irresponsible.

For reasons described below, and that BioCentury has previously chronicled, overall survival is often not an appropriate endpoint for confirming benefit. The paper’s authors presented the data in a way that obscured this important point.

A summary of “key points” deployed the passive voice to make it seem as if the drugs were to blame: “Although accelerated approval can be useful, some cancer drugs do not end up demonstrating benefit in extending patients’ lives or improving their quality of life.”

Medical journals and news publications took the clickbait and ran with it.

BMJ’s headline announced: “Fast tracked drugs often do not improve clinical outcomes, say researchers.”

Less august medical journals ran headlines such as “Most Cancer Drugs Granted Accelerated Approval Fail Confirmatory Trials — Only 43% showed clinical benefit within 5 years” and “Most Cancer Drugs Granted Accelerated Approval Don’t Confer a Clinical Benefit.”

Business publications were less restrained. Bloomberg folded the PORTAL paper into an account of unrelated controversial approvals and informed its readers that “Americans Are Paying Billions to Take Drugs That Don’t Work.” LinkedIn News transformed this into a question: “Billions spent on useless drugs?”

Biopharma trade publications opted for headlines such as “Most Cancer Drugs Granted Accelerated Approval Do Not Improve Survival” and “Nearly half of cancer drugs granted accelerated approval didn’t end up showing clinical benefit — study.”

You wouldn’t know it from reading the JAMA paper or the stories that were written about it, but there is a world of difference between the finding that an overall survival benefit wasn’t demonstrated in confirmatory trials and concluding that a drug doesn’t confer clinical benefit.

It shouldn’t be difficult to understand that lack of data proving that a drug has extended lives is not the same as proving that it doesn’t extend lives. 

Absence of evidence isn’t evidence of absence.

The PORTAL paper is part of a larger campaign against pathways FDA uses to bring drugs to patients more quickly.

These programs could be improved, but the notion implicit in much of the criticism that FDA should always require large, replicated double-blinded placebo-controlled studies with clinical endpoints is nonsense. 

FDA is not and should not be immune from criticism. The agency needs to do a far better job of defining the criteria for approvals and of applying these criteria consistently. Some recent accelerated approvals, for example of drugs to treat Alzheimer’s disease and Duchenne muscular dystrophy, have been controversial and merit deep scrutiny.

In the arena of cancer, however, where accelerated approval is most frequently used and where FDA has had the most time to work out its procedures and think deeply through the details, a fair analysis of the pathway’s impact would conclude that it has brought important medicines to patients quickly and made possible development of certain therapies that otherwise would have been impractical or impossible to develop. 

Effective drugs

The examples the PORTAL researchers provide of accelerated approvals that FDA converted to full approval based on endpoints other than overall survival are instructive.

These drugs aren’t perfect — unfortunately, there are no perfect drugs — but they aren’t useless. Unlike recent accelerated approvals for neurologic diseases, the approval endpoints are not controversial, they are objective measures that physicians and patients believe are important.

Libtayo cemiplimab from Regeneron Pharmaceuticals Inc. (NASDAQ:REGN) received accelerated approval in 2021 and conversion to full approval in 2023 for second-line treatment of metastatic basal cell carcinoma. Libtayo is the only drug that has demonstrated efficacy in the population. It does not work for everyone, but follow-up data showed tumor shrinkage in 22-26% of patients, with responses durable for at least 12 months in 58-79% of patients.

The accelerated approval of Jempereli dostarlimab from GSK plc (LSE:GSK; NYSE:GSK) as a second-line treatment for mismatch repair deficient (dMMR) recurrent or advanced endometrial cancer was based on an objective response rate of 45.4% (95% CI: 37.0, 54.0), with a 15.6% complete response rate and a 29.8% partial response rate. More than half of the patients had a duration of response of at least 24 months.

Critics of the accelerated approval pathway point to FDA’s acceptance of progression-free survival (PFS) to support conversion to full approval as examples of the agency’s lax approach to confirmation of clinical benefit.

According to this thinking, Gleevec imatinib, a drug that transformed chronic myelogenous leukemia (CML) from a death sentence into a manageable disease, should be consigned to the dustbin of medical history.

Gleevec was granted accelerated approval in 2001 based on surrogate endpoints and two years later FDA granted full approval based on PFS. There never has been and never will be a randomized, controlled study of Gleevec or any other effective CML therapy designed to detect an overall survival benefit. Hundreds of thousands of Americans are alive today because of Gleevec and drugs it inspired.

Two of the products the JAMA paper highlights as examples of conversions from accelerated to full approval that lacked survival or quality-of-life data were treatments for CML.

In both cases, the endpoints supporting full approval were major molecular response and complete cytogenetic response. These are not only clinically meaningful measures, given the effectiveness of approved treatments and the relatively slow progression of CML, they are the only endpoints that could be used to assess efficacy. If FDA required overall survival, no new drugs would ever receive full approval for CML.

In some cases, PFS is inherently beneficial, for example, when it keeps patients healthy enough to receive a curative treatment such as a CAR T. In other situations, patients can achieve extended benefit by jumping from one drug that provides a real but limited duration PFS to another similar drug, like frogs crossing a pond on lily pads.

Multiple myeloma (MM) is a good example. Since 2003, FDA has approved 17 drugs to treat MM, almost all of them through the accelerated approval pathway. None of these have been shown to extend survival, so according to the narrative proposed in the JAMA paper, they haven’t benefited patients.

In fact, collectively, drugs that received accelerated approval for MM, including many that subsequently received full approval, have “transformed this disease,” Richard Pazdur, director of FDA’s Oncology Center of Excellence, noted at an advisory committee meeting last week. Overall survival for newly diagnosed MM patients has gone from two to 10 years. Few, if any, companies would find it feasible to run a trial that required a decade of monitoring to learn the answer. Pazdur argued, and FDA’s advisers agreed, that use of a non-clinical endpoint, minimal residual disease, will likely make it possible to move some treatments into frontline use where they could produce even better results.

‘A failed trial does not mean a failed drug’

There are many reasons why it isn’t necessary, and in many cases isn’t possible, to demonstrate an improvement in overall survival for a highly effective cancer drug. These include a lack of equipoise, confounding bias from cross-over study designs that are needed to meet ethical obligations to patients, and medical consensus on the predictive value of non-clinical endpoints.

If a drug that receives accelerated approval provides a real and obvious benefit over existing therapies, it can be impossible to conduct a post-market randomized trial in the same indication because patients are not willing to accept the chance that they will receive an inferior treatment.

One strategy for coping with the lack of equipoise is to allow sponsors to conduct confirmatory trials in a different but related indication. When it is successful, this approach helps patients by developing evidence about new uses. 

Failure to demonstrate clinical benefit in a different indication does not, however, necessarily mean that the drug provides no benefit.

As Pazdur likes to say, “a failed trial does not mean a failed drug.”

There are numerous reasons why a trial of an effective drug can fail. The sponsor may be unable to recruit enough patients, patients may drop-out because another drug became available, there may be a high rate of cross-over from the control to the treatment arm that dilutes the statistical power.

Patients may accept randomization to standard treatment contingent on the ability to cross over to an investigational treatment after accelerated approval. In this scenario it may be impossible to document a survival difference because few if any patients are willing to sacrifice themselves to provide more easily interpretable statistics.

One solution that has been used in the past is to conduct a randomized trial in a country where a drug hasn’t been approved. FDA, however, has determined that this approach is unethical.

There are reasonable explanations for the absence of data demonstrating an overall survival benefit for the other conversions to full approval cited in the JAMA paper.

This doesn’t mean each decision was correct. The point of accelerated approval is not perfection; it’s to allow risk-taking when there’s evidence that a therapy is reasonably likely to benefit patients who are suffering from a serious condition. 

It does mean that assessment of the merits of granting full approval must be grounded in a detailed understanding of the clinical realities.

A drive-by analysis based solely on the lack of a statistically significant demonstration of overall survival is intellectual malpractice.

Signed commentaries do not necessarily reflect the views of BioCentury.