Modeling Is Not a Drug Approval Pathway: Why Merck Flunked Keytruda’s TNBC Advisory Panel
Executive Summary
Advisory panel agrees with US FDA that Merck’s pathological complete response data is not strong enough for accelerated approval in high-risk, early-stage triple-negative breast cancer patients and says modeling of expected outcomes data cannot make up for this weakness.
The US Food and Drug Administration’s Oncologic Drugs Advisory Committee voted unanimously in favor of delaying an approval decision on Merck & Co., Inc.’s Keytruda (pembrolizumab) for treatment of high-risk, early-stage triple-negative breast cancer, siding with the agency’s position that modeling cannot substitute for completed clinical trial data.
The 10-0 panel vote on 9 February was also a reminder of the limits FDA has laid out around the use pathological complete response as a surrogate endpoint.
Merck is seeking accelerated approval for the immunotherapy to be used in combination with chemotherapy as neoadjuvant treatment, then as a single agent as adjuvant treatment after surgery based on interim results of the KEYNOTE-522 trial, a randomized double-blind placebo-controlled trial comparing Keytruda to placebo in combination with chemotherapy in the neoadjuvant setting and as a monotherapy in the adjuvant setting.
FDA has warned the company that the application is unlikely to succeed as the study showed only a small absolute improvement in the pCR in the neoadjuvant setting, which the agency does not believe is clinically meaningful.
The agency also dismissed the sponsor’s efforts to model the likelihood the drug will eventually meet its other primary endpoint of event-free survival or its secondary endpoint of overall survival in the ongoing study. (Also see "Trial Design Issues, Failure To Follow Guidance Likely to Hurt Merck At Keytruda Advisory Panel" - Pink Sheet, 7 Feb, 2021.)
“Substantial evidence is required for either accelerated approval or conventional approval. And substantial evidence is a statistical persuasive effect on an endpoint. This is not about a guessing game, whether the drug works or whether a model shows something. This is basically an effect on an endpoint that is reasonably likely to predict a clinical benefit,” said Richard Pazdur, FDA’s Director of the Oncology Center of Excellence and Acting Director of the Office of Oncologic Diseases.
“We can’t be put in a predicament of potentially approving a placebo here or [there] being very, very marginal benefits,” he added, particularly given the toxicity associated with Keytruda and the long duration of use of the drug being proposed.
Pazdur emphasized that his remarks on modeling would apply to other cancer settings as well.
“With metastatic disease, we wouldn’t take someone coming in with interim analysis at any time point, just saying we’re modeling this to determine whether an effect is there. We need to see that effect.”
‘That’s Why You Play The Game’
Almost every advisory committee member made similar comments when discussing their vote against the supplemental biologics licensing application being ready for review.
“I think [Pazdur’s] comments are supporting my concern that we can’t really predict what's going to happen without actually seeing the real data. We can make a logical guess, but it’s like predicting a football game. One team looks better than the other but that’s why you play the game,” said Daniel Hayes, a professor breast cancer research at University of Michigan.
“Trends and signals toward improvement,” in the trial can change with time, said Committee Chairperson Philip Hoffman, University of Chicago.
Hayes also was concerned about “maintaining the integrity” of the trial including patients in the placebo arm if the drug was granted approval before the study completed. The issue of keeping patients in a study after a fast approval has recently been a hot topic of COVID-19 vaccine regulation. (Also see "Placebo-Unblinding Should Be Uniform Across COVID Vaccine Trials, Advisory Committee Says" - Pink Sheet, 17 Dec, 2020.)
Advisory Committee Vote
10-0 that a regulatory decision on pembrolizumab in combination with multi-agent chemotherapy for neoadjuvant treatment followed by pembrolizumab monotherapy for adjuvant treatment of high-risk early-stage TNBC be deferred until further data are available from future analyses of KEYNOTE-522.
While many of the advisory committee members said they were sympathetic to the pleas from patients for more treatment options during the open public hearing, they also cautioned against simply providing false hope.
“I'm a little baffled by why the rush here despite FDA saying let’s wait for the complete data. We really have a responsibility to patients to not prematurely offer treatment and therefore hope when we don’t have solid evidence and benefit,” said patient representative Natalie Compagni Portis.
Keytruda’s safety data, including four deaths FDA said are likely attributable to the immunotherapy, also weighed heavily on many committee members.
Hoffman said that while he has seen “spectacular” clinical results with immune checkpoint inhibitors like Keytruda, he has also seen patients that can’t enjoy those results because of the associated toxicities they experience.
“Even under the intense scrutiny of a clinical trial, almost 1% of patients who were treated with this died as a result of this treatment and in the general community that rate of mortality would likely be higher. … So I think that it’s the most prudent thing and the thing that's safest for our patients who clearly need new therapeutic options is to make sure that we're not giving them false hope or treating them with things that can hurt them more than it can help them,” said Deborah Armstrong, of Johns Hopkins Sidney Kimmel Comprehensive Cancer Center.
High Bar For pCR
The advisory committee may serve as a reminder to sponsors of the strength of data needed to seek accelerated approval using pCR as an endpoint in high-risk, early-stage breast cancer.
Pazdur has previously highlighted the uncertainty with the endpoint, and the need to have strong efficacy and safety in another line of therapy. Other FDA officials have emphasized the endpoint was only meant for a minority of breast cancer drugs. (Also see "Revised pCR Guidance Likely To Include Failed ALTTO Trial, FDA Says" - Pink Sheet, 18 Jun, 2014.)
An additional challenge with the Keytruda application is that the immunotherapy is also given in the adjuvant stage in the trial before the event-free survival and overall survival data is collected, which will make it harder to determine the predictivity of the pCR data on the hard outcomes.
“I think we heard repeatedly that the relationship between pCR and outcome is tenuous at best. We don’t fully understand that. And worse yet, in this case, there's adjuvant therapy between the pCR and the EFS events so that we really don't have a good tie between the pCR result and what that likely means for the EFS,” said Stan Lipkowitz, the chief of women’s malignancies branch at the National Cancer Institute’s Center for Cancer Research.
Panelist Matthew Ellis, the Breast Center director at the Baylor College of Medicine contrasted this application with the panel’s 2013 vote in favor of Genentech, Inc. ’s Perjeta (pertuzumab) for accelerated approval in the neoadjuvant breast cancer setting using pCR. (Also see "Neoadjuvant Surrogate Endpoint Is Tough Sell, But FDA Panel Says Perjeta Worth A Chance" - Pink Sheet, 23 Sep, 2013.)
A key factor in that panel was that Perjeta had already shown a survival benefit in metastatic disease, he said.
“I use the pertuzumab standard and put that data in that context. Obviously, I think we all agree it was not as compelling story as that,” said Ellis.
FDA said the data provided for Keytruda from the metastatic triple-negative breast cancer setting are not supportive of clinical benefit in the neoadjuvant setting.
The next interim analysis of the Merck trial, which is driven by the calendar, not study events, is expected in the third quarter, the company said. The application’s user fee date is 29 March.