There have been several good questions recently about the importance of progression-free survival (PFS) as a key endpoint in our clinical trials. In fact, there have been more trials recently that have assigned PFS as the “primary endpoint”, the main variable that the trial is designed to test to see if there is a significant difference (as described here). There are several advantages to PFS, including the fact that it takes less time to get an answer about whether your treatment changed the outcome. Feedback is provided without a need to follow patients after the end of the treatment. This also highlights another potential advantage of PFS as an endpoint: that it’s “cleaner” and not subject to the diluting effects of subsequent treatments given off protocol, after the end of the study treatment.
But this is also a potential weakness. Most oncologists, and I believe also many patients, report that the most important effect of a treatment is to improve survival, weighed against by the side effects of treatment. So if a treatment improves progression-free survival for a first line treatment, but subsequent treatments make up the difference and overall survival (OS) is the same for both groups, it’s fair to question if it’s worth subjecting patients to a treatment with side effects and extra cost if it doesn’t improve OS.
Another key point is that PFS, like response rate, is defined with some rigid definitions based on measuring the dimensions of particular lesions, and the rules don’t always match perfectly with the more complex situation of the real world. There is usually some discrepancy between the median PFS in a trial as measured by the investigators (who do tend to see their results through the prism of their expectations and hopes for a patient) and those of independent investigators. Overall survival is far less controversial.
But the other key issue is that, unlike overall survival, progression-free survival tends to be clustered in the times when you look for it on a scan. As described in a recent paper (here), if scans are done every eight weeks and the actual progression takes place between scans, the time of the scans will overestimate the actual time of progression:




