Looking Beyond Overall Survival: FDA-AACR Workshop Explored Novel Oncology Endpoints
Clinical trials are designed to evaluate the safety and efficacy of new therapeutics, with a longstanding focus on whether the new therapeutic helps patients live longer. But as advances in cancer research have—fortunately—extended overall survival, measuring it in a clinical trial has become less feasible, simply because it takes so long to reach the median.
“While overall survival remains the gold standard endpoint, it becomes challenging in clinical trials where the curve may take many years to read out,” said Lillian L. Siu, MD, FAACR, who is the President of the American Association for Cancer Research (AACR) and a medical oncologist and researcher at Princess Margaret Cancer Centre in Toronto.
This can be even more protracted in trials where researchers are looking for a very small effect size, she added. These delays mean effective treatments may not reach patients who need them.
“To expedite drug development and move promising treatments to patients faster, we are hoping that we can identify and validate novel endpoints that allow us to measure a treatment’s efficacy much earlier than we can with overall survival,” said Siu.
To this end, the AACR and the U.S. Food and Drug Administration (FDA) cohosted a workshop to examine potential alternative endpoints, as well as the considerations for validating and applying these endpoints in oncology clinical trials. The FDA-AACR Workshop: Approach to Novel Oncology Endpoint Development, held September 11, was chaired by Siu; Nicole Gormley, MD, of the FDA; and Alexandra Snyder, MD, of the Society for Immunotherapy of Cancer (SITC).
“The workshop was an important first step towards understanding and addressing the issues surrounding clinical trial endpoints,” said Siu. “By convening researchers, regulators, vendors, pharmaceutical companies, patients, advocates, and other key stakeholders, we were able to conduct a robust, multidisciplinary discussion on how we move forward with establishing novel oncology endpoints that may ultimately be sufficient for regulatory approval.”
Siu noted that the workshop benefitted from the unique perspectives that the AACR and FDA each contributed.
“AACR is a trusted source of scientific information. Drawing from its more than 58,000 members and its numerous publications, meetings, and educational opportunities, AACR was able to bring the most relevant topics to the table and generate very important discussions in this area. These dialogues may impact regulatory considerations on novel oncology endpoints.”
Which Alternative Endpoints Are Under Exploration?
Researchers are exploring several endpoints that may serve as early indicators of clinical benefit. These include minimal residual disease, pathologic complete response, and progression-free or relapse-free survival.
“I think one of the most attractive endpoints in the setting of early cancers is minimal residual disease,” said Siu, adding that the absence of minimal residual disease is typically a sign that a treatment has been effective and may correspond with positive long-term outcomes. Last year, the FDA deemed minimal residual disease an acceptable early endpoint to support accelerated approvals for multiple myeloma treatments.
Although minimal residual disease can be used to predict responses in patients with hematologic malignancies, Siu noted that, until recently, its application for solid tumors had been limited due to technological barriers. “Minimal residual disease was not feasible to use for solid tumors until the technology advanced to detect microscopic levels of cancer through circulating tumor DNA.”
In certain solid cancer types, pathological complete response (no visible signs of cancer in resected tissue after presurgical therapy) could also help predict overall survival, said Siu. In breast cancer, for example, pathological complete response has been associated with a greater chance of five-year survival.
Siu added that, with further research and validation, artificial intelligence (AI)-based metrics could someday serve as early endpoints as well. “I think AI will ultimately become a part of the picture—as we expect it to be a part of many pictures—but I don’t think, at this point, AI is yet ready for validation. We need to collect and share large amounts of relevant clinical, pathological, and molecular data in order to train AI and machine learning models to accurately predict patient outcomes.”
Distinguishing Between Early Endpoints and Surrogate Endpoints
Gormley emphasized that while endpoints such as minimal residual disease, pathologic complete response, and progression-free survival can be important early indicators of efficacy, they may not necessarily be surrogates for overall survival in all contexts. Even if a drug shows promise based on one of these early endpoints, it may ultimately not be any better—or may even be worse—at extending overall survival than existing treatments.
A true surrogate endpoint, Gormley explained, should serve as a stand-in for overall survival by capturing the full effect of a treatment on overall survival. This means the treatment should not impact overall survival without also impacting the surrogate endpoint, and the surrogate endpoint should not change without a corresponding change in overall survival.
To date, very few oncology endpoints have met this standard, Gormley noted.
As a cautionary example, she pointed to the BELLINI phase III clinical trial, which evaluated the impact of adding venetoclax (Venclexta) to standard treatment for patients with advanced multiple myeloma.
Patients in the trial who received venetoclax were more likely to experience a treatment response, were less likely to have minimal residual disease, and had significantly longer progression-free survival than those receiving placebo. The overall survival results, however, were the opposite: at an interim analysis, there were significantly more deaths in the venetoclax arm than in the placebo. This finding led the FDA to suspend further enrollment in the clinical trial.
The results underscored the need to collect data on overall survival in clinical trials, even when it is not the primary endpoint of the trial and even when early endpoints are available, Gormley noted.
“The use of early endpoints for regulatory decision is associated with risk, including the potential to approve an ineffective or harmful therapy,” she said. “The decision to use early endpoints really must be data-driven and balance potential advantages and risks.”
How Are Candidate Surrogate Endpoints Validated?
Researchers continue to explore alternative endpoints that may serve as true surrogates for overall survival. Gormley explained that meta-analyses can be used to validate candidate surrogate endpoints and verify that they correlate with overall survival at both the individual level and at the trial population level.
Proper evaluation requires that the meta-analysis incorporates patient-level data across multiple clinical trials, includes trials with positive results and trials with negative results, and has consistency across trials in terms of when the endpoints were measured.
But even if a surrogate endpoint is validated, there are still caveats to their use, Gormley said. They may not be appropriate for future clinical trials if the trial population or therapeutic’s mechanism of action are substantially different than the ones included in the meta-analysis validation.
What Were the Workshop’s Key Takeaways?
Workshop participants agreed on the need to address concerns about surrogate endpoints with more data, and they emphasized the importance of collaboration across all parties involved in drug development and approval in the collection of such data. This includes patients, regulators, investigators, hospitals, pharmaceutical companies, biotech companies, and diagnostic and biomarker developers.
Presenters discussed the importance of continuing to engage with the FDA early in the process to confirm whether their trial endpoints are acceptable to support approval, of seeking input from patient advocates, and of initiating confirmatory trials early in the case of accelerated approvals based on early endpoints.
Snyder, who is senior vice president of translational medicine and discovery oncology at Merck, noted the emphasis on approaching the development of oncology endpoints as an iterative process, whereby research and technology advances may lead the field to adopt increasingly accurate surrogate endpoints. She identified breast cancer as an example, where the use of pathological complete response to predict outcomes is evolving toward the use of residual cancer burden.
As far as where to focus efforts, participants raised a “call to action” to prioritize the development and validation of surrogate endpoints for scenarios where new therapies are greatly needed.
“Our collective goal is to conduct smaller, faster studies in order to more efficiently make novel therapeutics or combinations available to high-risk patients with unmet need,” said Snyder. “We must weigh the risks of using novel surrogate endpoints against the risks of not innovating and the associated delays to access to effective treatments.”
As Siu closed out the workshop, she remarked, “The dialogue that we are having today doesn’t end here. One theme that has come through very clearly is that the science of endpoint development is rapidly evolving.
“We are all learning, and we need to learn together.”
