Last year, an article came out in the Journal of Clinical Oncology that noted a gradual shift away from the "hard endpoint" of overall survival (OS) for lung cancer trials toward progression-free survival (PFS), a "softer endpoint" that is much more subject to interpretation.  The analysis also noted that many of the trials that were technically negative for the primary endpoint, the prospectively defined emphasis of the trial, were written up as if they were positive, highlighting a secondary endpoint that looked favorable, or a subset of patients who demonstrated a more favorable result than was seen in the overall trial population.

The accompanying editorial implied that this shift represented a regrettable relaxing of appropriate standards, but I actually didn't and don't see it that way.  First, we've seen that some trials haven't reported on survival because patients were responding and continuing without progression for so long that it was self-obvious that survival was improving.  For instance, the US FDA approved XALKORI (crizotinib) back in August, 2011 for ALK rearrangement-positive patients, based on the terrific results in a single-arm, phase I/II trial that showed that nearly two thirds of patients had significant tumor shrinkage and that the median PFS was in the range of 8-10 months. Both of these results were about twice what we would expect from standard chemotherapy regimens.  It wasn't until several years later that we saw the definitive evidence that XALKORI was superior to chemotherapy in terms of OS. How many ALK-positive patients would have been denied this clearly beneficial therapy over the additional 3 years just so that we could run a randomized trial that is the equivalent of running a trial to confirm the value of parachutes for people falling out of an airplane (technical term -- "gravitationally challenged")?  Most of the lung cancer community rightly considered XALKORI the clear standard of care in the first line setting without waiting for the subsequent trial to prove a survival benefit, because the conclusion was already self-evident based on a profound PFS benefit.  The FDA has actually just put out a new document noting that they will continue to consider approving drugs for lung cancer based on secondary endpoints like PFS.

Back in 2000-2005, there was much more reason to question the value of PFS as an independent and predictive endpoint.  Before trials with targeted therapies, the magnitude of PFS differences that you'd see between two arms of a trial was usually only up to 1.5-2 months. If that isn't accommodated by at least as much of an improvement in OS, it's hard to get excited about that. Yes, patients can argue that a few extra weeks before being told that their scan looks worse is valuable, and while that's justifiable for drugs that don't cost $5000-10,000/month or more, it's fair for society to expect much more compelling value for cancer drugs when everyone ends up sharing the cost burden by paying taxes or rising insurance premiums.  On the other hand, I have firmly believe that seeing a PFS of more than 6 months in a randomized trial of Tarceva (erlotinib) with or without Avastin (bevacizumab) in favor of the combination is very clinically meaningful and predicts for a high probability of a survival benefit with additional follow up.   Imagine having two homes for sale that are next door to each other, very comparable, but one is in a color you clearly prefer and one is a less preferred color. Does color matter? Yes, and if they cost the same, you should jump at this minor factor. But if the home in your preferred color costs 25% more, it's absolutely not worth it. It's a very minor factor. On the other hand, if the color of the home also correlated with something that really matters, such as if the home in the color you prefer was also bigger and much nicer, it's clearly worth paying the 25% premium.  

So why should PFS matter in one trial and clinical setting but not another? Because there's a major difference between a 1-2 month difference in progression-free survival (which should, in this case, be designated "pfs" instead of "PFS", I think) and a 6-12 month PFS effect seen with therapies that profoundly change the trajectory of the disease and are much more likely to track with a clinically meaningful OS difference. 

PFS can be an important surrogate endpoint, but it remains a controversial, debatable endpoint because there's a slippery slope between a paltry pfs difference and a clinically meaningful PFS effect. And we don't have the luxury of indulging the pretenders when their drugs are priced as if they make a more substantial difference than they've really demonstrated.