PFS: The Endpoint We Love and Love to Hate

Abstract

One of the major challenges in oncology drug development is the identification of appropriate endpoints that indicate direct benefit to the patient. The U.S. Food and Drug Administration (FDA) has clearly defined meaningful clinical benefit as improvement in patient-reported symptomatology or improvement in survival and has considered these patient-directed endpoints in clinical trials as the basis for registration of new agents (1). However, the FDA has also recognized that such endpoints may be difficult to obtain and so has allowed the establishment of surrogates that may substitute for the above endpoints, often then requiring later confirmation through further clinical study. The most widely used and perhaps the most controversial surrogate is progression-free survival (PFS), which measures the amount of time from study enrollment to a prespecified point at which a tumor is no longer considered to be affected by the administered therapy, generally a 20% increase in tumor size above the starting value or an achieved nadir. Interpretation of PFS remains a challenge despite its use as an endpoint in oncology for over 40 years, illustrated in the 1969 Moertel and colleagues PFS plot (Fig. 1; ref. 2). Parenthetically, it should be noted that this 20% cutoff has never been proven to actually indicate treatment failure, as the boundary may be crossed despite ongoing tumor growth delay—efficiently termed a “cytolentic,” or cell-slowing effect. More widely discussed is the point that the determination of progression is susceptible to multiple sources of error. Those errors can introduce missing or incomplete data in randomized trials. Missing data, in turn, can cause informative censoring. This CCR Focus section delves deeply into PFS measurement with John J. Crowley, former Director of the Statistical Center of SWOG (previously known as the Southwest Oncology Group), as Guest Editor. Experts in the field identify sources of error in PFS determination, propose strategies to limit the error, and introduce two new approaches in clinical trial methodologies. These approaches, embracing an audit strategy to reduce missing data events in randomized trials, and using a new phase II/III trial design, should expedite drugs through the development pipeline and deserve a close look. As with every edition of CCR Focus, it is hoped that these articles will stimulate discussion among colleagues, inspire those already working in drug development, and inform those desiring a more in-depth understanding of this important challenge in oncology.Susan E. BatesDeputy Editor, CCR FocusNational Cancer Institute