When we evaluate new drugs in oncology, we, of course, want to determine their impact on survival or its converse, mortality. But there are various ways to define survival. Years ago, a common measure was cancer-specific mortality—the time from initiation of a drug until death from the cancer. This makes the most logical sense as we are primarily interested in how a drug affects death from cancer as opposed to death from other causes. But there can be ambiguity in this measure. When someone dies in a coma from liver failure due to liver metastases, that is clear enough. If someone dies from COVID while undergoing treatment for myeloma, is that a cancer-related death? If someone is driving to Holy Name Hospital to receive their experimental therapy and is hit by an 18-wheeler and dies, is that a cancer-related mortality? If someone has received their experimental therapy and is returning home fatigued and a bit woozy and is hit by an 18-wheeler, is that a cancer-related death?
Each trial had expert committees to make these judgments but obviously there was a large degree of subjectivity. The end result has been that this measure has been subsumed into overall survival (OS) or mortality from any cause. This is considered more objective and unbiased. Of course, OS includes deaths from other causes, such as diabetes or heart disease or lupus that are common among adults, but the logic is that randomization will equalize these deaths between the two arms of a trial. And the cancer deaths will predominate the determination of the OS in a drug trial. The FDA has traditionally approved cancer drugs if they were shown in randomized trials to improve OS or to make a significant impact on quality of life for cancer patients. These seem like perfectly reasonable criteria for this decision. They are objective and unbiased, have stood the test of time, and are desirable outcomes for patients and providers—patients want to live longer and better.
But problems have arisen with OS—as survival times for certain cancers have become increasingly prolonged, it becomes more difficult to ascertain marginal or small gains in OS, and so randomized trials have had to go on for years in order to demonstrate small gains for a new drug. Investigators, patients and manufacturers must wait, and by the time the drug can be brought to market many patients have lost the opportunity to benefit.
The best alternative would be to define intermediate or surrogate endpoints that could provide valid information that would substitute for OS. To date, this function has been served by progression-free survival (PFS). Progression-free survival is defined as the time from treatment initiation until evidence that the disease has begun to regrow or progress. Progression is evidenced in one of several ways—an increase in size of measurable tumor; the appearance of a new metastatic tumor; or death.
Generally speaking, if an intervention does not improve PFS, it will not improve OS, but the converse is not necessarily true. Patients are happy when they are told they have had prolongation in their PFS, but is it truly meaningful or important if it does not also have an impact on OS? Temporary improvement in PFS must be weighed against the added toxicity incurred by the treatment as well, so is this temporary psychological boost indeed worth the use of the new drug in the absence of a true OS improvement?
At the present time, 80% of clinical trials use PFS as their primary endpoint. And both the FDA and the European Medicines Agency have begun to use PFS for oncology drug approvals. Studies show that the use of PFS as the criterion accelerates drug approval by approximately one year. Many studies have been done to evaluate the predictive value of PFS for OS. In most circumstances, the correlation has not been very good. And the longer the survival, the worse the correlation.
As noted earlier, if there is no benefit in PFS, there will almost surely be no benefit in OS. If there is a benefit in PFS, then about 30-35% of the time there will subsequently be a corresponding benefit in OS. For circumstances where the survival time of a tumor is poor, it seems inappropriate to use PFS as a surrogate for OS. Then OS can and should be utilized. Thus, for pancreatic cancer and brain tumors, for example, OS should almost always be the primary endpoint for drug studies. PFS is most valuable for metastatic diseases or adjuvant studies where it would take multiple years for OS data to mature.
An example is the recent randomized trial of capivasertib, an AKT inhibitor, which when added to fulvestrant, an endocrine therapy for women with advanced hormone-sensitive breast cancer, may prolong survival. In a randomized trial, 708 women with hormone receptor positive,
HER2-negative breast cancer who had already progressed on hormonal therapy were randomized to fulvestrant with either capivasertib or a placebo. In the subgroup that had a marker that made them particularly sensitive to the drug, the PFS was 7.3 months for capivasertib versus 3.1 months for placebo with very acceptable toxicities. There was no difference in OS at the time of this analysis. Nonetheless, the FDA and the EMA both approved this drug for this indication based on this study.
Alfred I. Neugut, MD, PhD, is a medical oncologist and cancer epidemiologist at Columbia University Irving Medical Center/New York Presbyterian and Mailman School of Public Health in New York. Email: [email protected].
This article is for educational purposes only and is not intended to be a substitute for professional medical advice, diagnosis, or treatment, and does not constitute medical or other professional advice. Always seek the advice of your qualified health provider with any questions you may have regarding a medical condition or treatment.
