Progression-Free Survival: Helpful Biomarker or Clinically Meaningless End Point?

It seems obvious that patients with cancer would be better off when the burden of malignancy is less rather than more. Yet serial studies have failed to show that tumor shrinkage—objective response rate (ORR)—is predictive of patient outcomes. If tumor shrinkage is not prognostic, perhaps control of tumor growth would be more informative, given that cancer progression represents an ominous march toward death from malignancy. This is the basis for the use of progression-free survival (PFS) as an end point for determining whether a therapeutic intervention is likely, or unlikely, to make a meaningful impact in a patient’s survival. By extension, it also would seem true that the longer it takes for cancer to progress, the longer a patient will live. With this belief, cancer intervention trials of past decades used tumor progression as a surrogate end point of overall survival (OS). And in fact this may well have been valid in the early decades of cancer research, because the absence of cross-sectional or positron emission technology imaging usually meant that progressive disease was determined by declining performance status and worsening symptoms, signaling clinically relevant cancer progression. Given that the most effective therapies were generally employed upfront, survival beyond progression was likely to be shorter than the first-line PFS. The search for surrogate end points, of course, is critical if we are to determine the ultimate effectiveness of therapy without following patients until death. As an example, the end point of disease-free survival (DFS) was shown a decade ago to predict for OS in patients treated with adjuvant therapy for stage III colon cancer and enabled the oncology field to embrace the current standard chemotherapy regimen of infusional fluorouracil, leucovorin, and oxaliplatin without having to wait several years for the data of the MOSAIC trial (Multicenter International Study of Oxaliplatin/5-Fluorouracil/ Leucovorin in the Adjuvant Treatment of Colon Cancer) to mature (and a decade later, the findings were confirmed). If PFS was valid in the past, can it be trusted today? It is not valid with the pseudoprogression that is described in immune-based therapies, where tumors may enlarge before shrinking. Even for conventional treatments, the meaning of PFS has changed. The rigorous assessment of end points in contemporary clinical trials include frequent quantitative imaging, so PFS is almost always determined by strict radiographic tumor measurement, or in the case of positron emission technology, the assessment of FDG avidity by the radiologist, rather than by the clinical status of the patient. Because patients are typically enrolled on studies with much lower volume metastatic disease compared with years ago—tumor burden migration (Will Roger’s phenomena)—not only are most patients asymptomatic when given the news that the cancer is progressing, but most will live much longer postprogression than before. This is a tribute both to advances in treatment beyond first line and to the lead-time bias resulting from more sensitive imaging techniques. If not ORR or PFS, then what? New measurements, such as time to maximal response and depth of response and image reconstruction to compute volume rather than RECIST measurements, remind us how out of favor response rate has become. Similarly, the energy and money spent to collect data on tumor behavior beyond progression reflect a growing concern that as tumors evolve under the pressure of treatment, early gains may be negated by later losses. For example, the downstream effect of the withdrawal of an agent such as bevacizumab has been recently reported in a randomized study. In other words, improved PFS may not always correspond with longer survival. As more agents become available to treat the common cancers, the meaning of the results of the initial treatment is more difficult to assess. In considering the optimization of resources for the successful development of better treatment options in metastatic colorectal cancer (mCRC), we are faced with different challenges. On the one hand, we continuously question as above if PFS correlates with OS and quality of life, the ultimate goals of treating our patients. On the other hand, we strive to develop and validate better efficacy end points that, besides PFS, could provide guidance as to whether it is fair to continue developing a new drug or treatment based on the data generated in small proof-of-efficacy randomized phase II studies (Fig 1). In this issue of Journal of Clinical Oncology, two different approaches that further inform the debate regarding surrogate end points for mCRC trials are reported. Shi et al analyzed the raw data of conventional outcomes from the Analysis and Research in Cancers of the Digestive System (ARCAD) database of mCRC patients included in 22 clinical trials conducted between 1997 and 2006. In contrast, Sharma et al worked with the original data set from the historical N9741 clinical trial and explored dynamic end points that were not considered a decade ago, such as time to tumor growth (TTG) and log JOURNAL OF CLINICAL ONCOLOGY E D I T O R I A L VOLUME 33 NUMBER 1 JANUARY 1 2015