PFS IS ONLY A SURROGATE END POINT

As we continue developing new therapies for lung cancer, and increasingly give them in sequence, evaluating the benefit of one line of therapy will by necessity become more focused on time-limited assessments. We have rightly moved epidermal growth factor receptor (EGFR) and anaplastic lymphoma kinase (ALK) inhibitors into the first-line setting for patients with the appropriate driver mutation; however, this was not based on a proven benefit in overall survival (OS), which has been very challenging to demonstrate when the majority of patients assigned to initial chemotherapy cross over to targeted therapy.

But in other settings, such as maintenance docetaxel in chemotherapy-treated non-small cell lung cancer (NSCLC)^[1] or the addition of bevacizumab to erlotinib in EGFR mutation-positive NSCLC,^[2] our practice did not change on the basis of a progression-free survival (PFS) benefit in the absence of an OS benefit.

In a recent Medscape commentary on practice-chancing studies published in 2017, Mark Kris, MD, asked readers to focus on PFS as the most important endpoint in managing lung cancer, predicated on the concept that this is the route to curing patients, including those with metastatic disease.^[3] I think it's critical for us to step back and clarify the limits of this approach.

We should remember that PFS is a surrogate for the most important endpoint: how long our patients live.

Although I agree that PFS is a critical endpoint and should be sufficient to define a new standard of care in certain settings, we should remember that this is a surrogate for the most important endpoint: how long our patients live. Our treatments, increasingly $15,000 or more every month, should not be given with the goal of making scans look transiently better if this doesn't translate to an improvement in OS. When the difference in PFS is more than 6 or 9 months, as in many targeted therapy trials, we can infer that the clinical benefit is very significant.

We should also recognize, however, that if OS is the same, presumably because patients are receiving the most effective treatments later rather than earlier, it isn't truly critical that these treatments be given first-line, even if it is sensible to give the most effective treatments earlier simply to ensure access to them. It is also appropriate to prioritize on the basis of PFS when a novel therapy offers a more favorable side-effect profile, better intracranial disease control, or lower cost (stop laughing–-cost savings is at least a theoretical benefit of a new treatment option).

We have been right to be judicious about accepting PFS benefit without an improvement in OS when the benefits are more modest or the new treatment comes with additional toxicity or cost. Having the scans look good is what our patients and we want, but I would contend that when these interventions cost tens of thousands of dollars or pose significant risk, they are not justified by a wish, however sincere, for an imagined, potential OS benefit. They are justified by an actual, demonstrated one. Healthcare systems have enough difficulty broadly providing the care that confers a proven OS benefit, without diverting limited resources to approaches with a merely speculative OS benefit. For those interventions where the OS benefit is very modest or the data are too immature to call, it is far more cost-effective to address patient anxiety about progression by doing fewer scans (and offering We must also be cautious about extrapolating PFS results, even out to 2 or more years, that they represent a cure for our patients with metastatic lung cancer. In fact, the vast majority of patients responding well to a targeted therapy, immunotherapy, or (rarely) chemotherapy and without progression several years into treatment will still eventually develop acquired resistance and die of their cancer.

This is not to say that we cannot hold out the hope that our patients will be outliers, with more and more of them going years without progression. But to imply that curing our patients with metastatic lung cancer is a broadly feasible expectation creates painful false hope for the vast majority of patients. It also invites gratuitous, inappropriate treatments (and associated costs) based on irrational exuberance by oncologists. It is better to underpromise and overdeliver for the lucky few who defy the sobering odds—odds that should continue to foster our humility.

Individualized recommendations for cancer treatment are now more appropriate than ever. Thus, it is critical to be cautious about defining a standard of care. As an illustration of how far the field has come, Kris noted that there is now some evidence to support local consolidation therapy (LCT) for patients with oligometastatic disease and minimal residual disease after effective initial systemic therapy.^[4,5]Although it should be no surprise that PFS is improved when you resect or ablate the sites most prone to progression^[6]—just as we would expect prophylactic appendectomy to lead to lower rates of appendicitis—the most encouraging finding from early work on LCT was that it can significantly prolong the time to development of new metastatic sites.^[4] The indication that LCT can alter the trajectory of metastatic lung cancer progression justifies favoring this approach for the limited number of patients who have up to three sites of residual disease. But these data do not demonstrate that patients are being cured.

Moreover, the fact that no reported trial of LCT has enrolled more than 50 patients makes it debatable (or, to be more clear, incorrect) to contend that LCT for metastatic NSCLC is standard of care for anyone today. My own survey and detailed discussion with 12 nationally recognized thought leaders in lung cancer^[7] revealed that they overwhelmingly consider the data on LCT to be applicable only for rare patients and not relevant for general practice.

Although I actually feel LCT should be considered for the "right" patient, as Kris noted, we need to underscore that the term "oligometastatic" or "oligoprogression" does not apply when there are more than three sites of active disease to target. But any oncologist who has followed the practice of stereotactic radiation for brain metastases transition from one to three lesions to "up to 10," and to essentially no upper limit today, should recognize the potential loss of rigor in patient selection. Then LCT will be practiced on the majority of patients with overtly "polymetastatic" disease—patients who will receive nothing but the ill-conceived satisfaction of receiving inappropriate, futile, and wasteful treatment based on the reflexive premise that more is better.

PFS is a valuable shortcut.

Let me finish by turning to the setting in which PFS may very possibly be correlated with ultimate cure. The PACIFIC trial, in which patients with unresectable stage III NSCLC who completed chemoradiation and were then randomly assigned to receive either the immune checkpoint inhibitor durvalumab or placebo intravenously every 2 weeks for 1 year,^[8] serves as a critical example of how PFS may be sufficient to change practice (or not, depending on other variables). I agree with Kris that on the basis of PFS alone, this study should change practice. Specifically, the PFS benefit with durvalumab after chemoradiation is both dramatic and prolonged, and although we don't yet have results for OS, durvalumab also confers a marked benefit in "time to new metastatic disease or death," an endpoint that should serve as a valuable surrogate for OS.

But we don't yet know whether durvalumab will improve long-term survival. If we find through extended follow-up that durvalumab merely delays relapse in patients not cured by chemoradiation and delivers unnecessary, prolonged treatment, toxicity, and expense for the approximately 15%-20% whom we expect to be cured before moving on to consolidation, we should revise our conclusions and, I believe, no longer broadly recommend routine durvalumab. Even remarkably prolonged PFS is no great benefit if it is achieved by merely treating patients with incurable, metastatic disease before we can detect it, provided it is not accompanied by a significant improvement in OS and particularly if it also requires patients already cured to submit to an additional year of treatment.

As with so many difficult questions, our assessment of the value of PFS must be nuanced. I strongly believe that PFS can be sufficient to change routine practice if the magnitude of that benefit is overwhelming and either OS is immature or the new treatment also delivers a more favorable tolerability or other benefits, such as intracranial disease control or lower cost. We increasingly recognize the value of PFS when prolonged survival enables crossover and treatment with multiple active therapies over time. But we are letting the tail wag the dog if we focus more on PFS than OS, and we should be wary about pointing to PFS as a predictor of cure, particularly on the basis of outcomes less than 4-5 years out and in a setting in which cures have been extremely uncommon. PFS is a valuable shortcut, but we need to confirm we arrive at the right destination. as needed).