Toward Progression-Free Survival As a Primary End Point in Advanced Colorectal Cancer

Abstract

In this issue of the Journal of Clinical Oncology, Buyse et al evaluate progression-free survival (PFS) as a surrogate end point for overall survival (OS) in advanced colorectal cancer (CRC). They employ a general methodology for validating surrogate end points across multiple trials, as proposed by Buyse et al in 2000. They considered trials where leucovorin-modulated fluorouracil (FU) was compared with either FU without leucovorin or with raltitrexed. In these trials, they were able to show modest correlation between the proportion alive and progression free at 6 months and the proportion alive at 12 months, as well as good correlation when data across all time were considered. More importantly, they showed that the effect of treatment on each of these end points was highly correlated both when data across all time were considered and when the PFS and OS data were censored at 6 and 12 months, respectively. This latter step is an important advance over previous attempts to establish surrogacy. The model relating treatment effects between the PFS and OS end points was validated on a set of three contemporary trials in which leucovorin-modulated FU was compared with regimens that added irinotecan or oxaliplatin. The model was used to predict the treatment effect on OS given the observed treatment effect on PFS. The observed treatment effect on OS for the validation trials fell within the prediction limits for all three trials. The authors conclude that PFS is an acceptable surrogate for OS in advanced CRC and provide a surrogate threshold effect (STE), which is the treatment effect size for PFS which, if exceeded, would rule out equivalence with 95% confidence for the OS end point based on the model’s prediction limits. PFS has now been validated as a surrogate end point for OS in advanced CRC, but OS is not a perfect end point for this disease. In advanced CRC, the goal of therapy in order of desirability of outcome is to eliminate cancer, reduce cancer burden, stabilize cancer, or slow cancer progression. Prolongation of life is secondary to the goal of therapy. As more and better options for secondand third-line therapies in advanced CRC become available, the relevance of OS as the primary end point is increasingly coming into question. Let’s take a moment to compare and contrast the virtues of PFS and OS as potential primary end points for first-line treatment of metastatic CRC. OS is usually defined as time to death from any cause and may be considered a composite end point based on the cause of death where the event is the first of death related to the same cancer, death related to another cancer, death related to treatment, or death resulting from other causes. PFS is a composite end point where the event is the first of progression resulting from increasing size of tumor, progression resulting from formation of a new tumor, death related to the same cancer, death related to another cancer, death related to treatment, or death resulting from other causes. The component events are listed in Table 1 for the PFS and OS end points along with their relationship to the disease process under study or the therapies being studied. Events related to the disease process under study or to treatment toxicity provide important information for comparing potential therapies. Events that are unrelated to the disease process under study or to treatment toxicity are essentially background noise and provide little useful information for comparing potential therapies. We don’t want to ignore these likely irrelevant events, because sometimes treatments have unintended consequences. The underlying rates of these so-called background events should be equivalent on the treatment arms, because they are by definition unrelated to the disease process or treatments. Some desirable attributes for a primary end point in advanced CRC are listed in Table 2. Event status determination is extremely reliable in OS. The difficulties associated with determining