Maurie Markman, MD, discusses the debate between surrogate end points of overall survival and progression-free survival in clinical trials within oncology.
The virtual revolution over the past several decades in our understanding of the basic molecular biology of malignancy has resulted in a substantial increase in novel antineoplastic agents for the treatment of patients in numerous clinical settings. The FDA has responded to this development by permitting nontraditional study designs and what is labeled accelerated approval as relevant components in its regulatory paradigm. One critical component of this process has been the use of so-called surrogate end points, which, in FDA parlance, means study outcomes likely to predict clinical benefit, assuming that the surrogate end point itself is not directly meaningful to patients for some reason. One can appropriately debate the value of achieving a prospectively defined and objectively measured clinical end point.
However, the specific issue to be highlighted here is that use of such surrogates may substantially decrease the time required for a novel antineoplastic agent to be approved for noninvestigative commercial sale. This is compared with the standard process in which it is mandated that the agent achieve success by demonstrating statistically significant improvement in the declared gold standard of overall survival (OS) in a phase 3 randomized trial.
Unfortunately, some experts—mostly from the realm of academic medicine or individuals with little direct knowledge or experience with clinical cancer medicine—have raised strenuous objections to the use of surrogate end points for full regulatory approval, including an improvement in progression-free survival (PFS), an objective primary study end point increasingly employed in the FDA’s decision-making process.1
The commentary focuses on the reason why PFS is often likely to be the most appropriate primary study end point, recognizing that, unquestionably, patients and their families ultimately desire to improve survival, not simply the time until the cancer progresses.
The major issue is the fundamental question of what happens to individuals on the study after they have completed therapy on a clinical trial? Specifically, how might a comparison in OS between experimental and control management arms in a regulatory-defined phase 3 randomized trial be influenced by subsequent therapy in the study patients?
In an era where there are limited or no known beneficial “next-line” treatment options available for patients completing study-based therapy, or at least options that have any realistic opportunity to measurably influence survival (eg, objective response rates of < 10%, with durations of approximately 1 to 3 months in individuals achieving a response), it is reasonable to anticipate that a statistically significant improvement in PFS in a randomized phase 3 trial should be able to be translated into an improvement in OS. In fact, in this specific scenario, it would be rational to suggest PFS would be serving as a surrogate end point for a subsequently observed improvement in OS from the time of initiation of therapy on the clinical trial.
However, in a clinical setting where there are a number of possible therapeutic options available that may favorably affect the subsequent survival of patients (who were previously research subjects) including surgery, radiation therapy, commercially available antineoplastic agents, or promising investigative drugs, it is difficult or objectively impossible to know how such therapy may influence the ultimate measure of the OS of the study population. Unique clinical features may help determine next-lines of treatment available to patients previously treated on study, such as surgical removal of a progressing solidary lung or abdominal cavity mass; localized radiation for a progressing isolated pelvic wall mass; or anti-cancer treatment with multiple possible agents revealed in published results of phase 2 trials to have a meaningful degree of clinical activity in this setting.
Depending on the status of the investigative drug in question, a study may permit crossover to the experimental arm of a trial (if the patient desires such therapy) or, if the agent is commercially available, the patient’s oncologist may simply elect to employ the drug in future care.
It is essential to note that what is described above has nothing to do with the clinical utility of the investigative arm in question but rather the natural history of disease and its treatment after a research subject receiving protocol-defined therapy has re-entered the domain of standard of care or innovative nonresearch clinical management options.
The effect of the availability or lack of availability of possibly beneficial next-line therapy options on the survival outcome of patients participating in randomized trials was highlighted in a report of the relationship between PFS and OS in the initial evaluation of the first generation of truly useful antineoplastic agents for metastatic melanoma.
In an analysis of 9 randomized trials that employed dacarbazine (an almost completely inactive chemotherapeutic agent given for several decades in the treatment of this condition) as a control arm that permitted no crossover to the investigative drug, a very close relationship (correlation coefficient 0.96) between the study defined PFS, and the subsequently documented OS was documented.2 As noted above, this outcome resulted from the lack of known clinically beneficial next-line therapeutic options in this era for the management of metastatic melanoma.
However, in a report published 6 years later, investigators examined trial outcomes in which the experimental arm was checkpoint inhibitor therapy and the control arm was dacarbazine, but there was no difference in OS where subsequent therapy with the immunotherapy was permitted, despite a striking difference in PFS between the study arms.3 Does this mean the checkpoint inhibitor was not effective in the illness?
Absolutely not. In fact, patients randomized to the experimental regimen and the ineffective chemotherapy (dacarbazine) benefitted from their series of treatments. It would be a travesty to conclude otherwise, and yet one must note that the results of this study revealed no improvement in OS between the participants randomized to the investigative and control arms.
Multiple examples can be provided of the inability of a favorable effect of PFS to be converted into a statistically significant improvement in OS, but this should not be necessary to understand the point of this commentary. Advanced and metastatic cancers are increasingly being converted into serious, life threatening, and likely (although certainly no longer universally) fatal chronic diseases, where extended survivals are often measured in many years because of the availability of increasingly effective, often multimodality, approaches. As a result, it will become much more difficult to define the effect on OS of a single trial-based management strategy.
Although OS should certainly be a secondary end point of a randomized trial examining therapeutic efficacy in individuals with advanced and metastatic cancers, it should not be the mandated primary study end point, where reasonable therapeutic options after a patient completes trial-based therapy currently exist.