Is It Time to Formally Recognize That PFS Is More Than a "Surrogate" End Point?

Maurie Markman, MD

It may appear remarkable to some that the debate regarding acceptable primary end points in randomized cancer trials is unending. The FDA, generally considered a conservative regulatory agency, has appropriately accepted a statistically significant improvement in progression-free survival (PFS) in the presence of acceptable toxicity as such an end point. However, peer-reviewed articles and opinion pieces published in medical journals continue to either vigorously challenge this perspective or insist on labeling this objectively measured outcome as nothing more than a potential surrogate for what the manuscript’s authors claim are effects representing actual clinically meaningful results.

For example, in a recent “viewpoint” article published in JAMA Oncology the authors note: “Both the FDA and EMA [European Medicines Agency] accept a significant improvement in PFS for the registration of drugs for most types of cancer, although PFS is rarely a surrogate for OS [overall survival].” They add: “For many new drugs that have been shown to improve PFS, subsequent analysis has demonstrated no improvement in OS or QOL [quality of life], but these drugs are rarely withdrawn from the market.”¹

In a peer-reviewed manuscript published in JAMA Network Open exploring FDA backing of 49 therapeutic agents approved by the regulatory agency in 2020, the authors noted that more than half of the drugs were “supported by a single pivotal trial,” and that compared with nononcology approvals, oncology approvals were more likely “to use at least 1 surrogate measure as a primary end point [94% vs 14%].”² Finally, based on their analysis the authors, rather remarkably, concluded: “The FDA and consumers may benefit from a revised approach that better balances time to market with ensuring that approved drugs show evidence of efficacy.”² I must note that the authors are discussing drugs used by patients with cancer. How much time do these individuals have to wait beyond the current assessment of efficacy and safety employed by the FDA?

It is perhaps a little disturbing that these (and other) manuscripts continue to be published with little or no recognition of the strong counterargument: As critically important as the length of survival is to patients with cancer and their families, the increasing reality of care for advanced malignant disease in the United States makes achieving a statistically significant improvement in OS in a population of participants in a randomized clinical trial extremely difficult if not impossible to achieve.

Why?

The simple answer is the objective evidence that many advanced cancers are rapidly becoming serious but more chronic disease processes where survival from time of diagnosis is commonly measured in years rather than months. The reason for this clinically relevant development is the increasing number of new therapeutic options—including molecularly targeted drugs, more focused radiation, and innovative surgical interventions—that may favorably effect outcomes in appropriately and carefully selected patients.

Although not written with this intent, a recent article discussing new antineoplastic drug approvals from 2016 to 2021 has highlighted the point made above.³ The manuscript noted that of the “207 FDA cancer drug approvals in oncology and malignant hematology during this time period [42% were for] use in the second-[line], third-[line,] or laterline settings.”³ This emphasizes evidence of both safety and efficacy—as determined by the FDA—for multiple lines of treatment in these malignancies. It should be acknowledged that these novel regimens are rarely, if ever, delivered with curative intent. They do, however, favorably effect clinical outcomes in a proportion of patients with cancer in a specific setting to a sufficient degree to achieve regulatory approval.

What does this have to do with the inability of cancer clinical trials to demonstrate a statistically significant improvement in OS? One point is that other treatments received by patients following completion of trial-based therapy may have a potential effect on survival and may influence the calculation of an OS end point.

In a landmark paper rarely mentioned by those who deny PFS as an essential primary cancer study end point, 2 well recognized cancer biostatisticians attempted to model the effect of “cancer as a more chronic disease process” on the conduct of randomized clinical trials.⁴ The question being addressed in this exercise was the total number of trial participants required to potentially be able to document an improvement in OS assuming one of 2 possible median posttrial survival outcomes. The hypothetical trial had 80% power to detect a 3-month statistically significant improvement in median PFS and included a population of 280 patients.⁴

The first group in this model was projected to experience a 2-month median posttrial survival—an outcome that might be expected following progression after first-line chemotherapy for highly chemoresistant malignancies. The second group was modeled to experience a median posttrial survival of 24 months—an outcome not much different from women with disease progression after first-line treatment for advanced ovarian cancer who achieved an objective response to platinum-based chemotherapy.

The model revealed that only 350 patients would be required in the first setting to detect a 3-month difference with the anticipated short survival, because of the objectively limited effect of subsequent treatment approaches. Conversely, the model suggested 2400 individuals would be necessary to attempt to find an effect on OS in the setting of the extended 24-month median posttrial survival. Recognizing that all such hypothetical models can be criticized for assumptions required to develop and conduct the exercise, the results speak for themselves regarding the point being made in this commentary.

Do we want to design trials of novel therapeutics that demand this number of participants, considering the time, effort, and cost involved, as well as the multiple competing studies that may be available? Finally, even with this large number of patients, there is no guarantee that multiple possible and uncontrolled treatment strategies used after individuals leave the trial with varying outcomes (ie, length of use, efficacy, toxicity) will permit a legitimate examination of the effect on the ultimate OS of the trial-based therapy.

It is appropriate to conclude there is nothing in the preceding discussion that declares a novel cancer therapeutic cannot be shown to effect OS in a randomized trial. Rather, the sole point being made is that the failure of an investigative program to successfully achieve this end point does not indicate the absence of a favorable effect on survival, but only the inability to demonstrate this result.

Maurie Markman, MD, is editor in chief of Oncology Live and president of Medicine & Science at Cancer Treatment Centers of America, a part of City of Hope.

References

1. Tannock IF, Pond GR, Booth CM. Biased evaluation in cancer drug trials-how use of progression-free survival as the primary end point can mislead. JAMA Oncol. 2022;8(5):679-680. doi:10.1001/jamaoncol.2021.8206
2. Mitra-Majumdar M, Gunter SJ, Kesselheim AS, et al. Analysis of supportive evidence for US Food and Drug Administration approvals of novel drugs in 2020. JAMA Netw Open. 2022;5(5):e2212454. doi:10.1001/ jamanetworkopen.2022.12454
3. Benjamin DJ, Xu A, Lythgoe MP, Prasad V. Cancer drug approvals that displaced existing standard-of-care therapies, 2016-2021. JAMA Netw Open. 2022;5(3):222265. doi:10.1001/jamanetworkopen.2022.2265
4. Broglio KR, Berry DA. Detecting an overall survival benefit that is derived from progression-free survival. J Natl Cancer Inst. 2009;101(23):1642-1649. doi:10.1093/jnci/djp369