It's been a big week for immunotherapy for lung cancer. Right on the heels of a press release that the PD-1 inhibitor Opdivo (nivolumab) significantly improved survival for patients with advanced non-squamous NSCLC, my friend Dr. Garon from UCLA presented results at the American Association for Cancer Research conference with another PD-1 inhibitor, Keytruda (pembrolizumab). These results are from phase I trial of this agent given at different doses and schedules to patients with advanced NSCLC, either squamous or non-squamous, most but not all patients having received prior chemotherapy, coincident with the online publication of the results in the New England Journal of Medicine.  Taken together, the entire population demonstrated an objective response rate of 19.4% and median overall survival (OS) of 12 months, median progression-free survival (PFS) of 3.7 months. These numbers are not especially impressive, though the median duration of response (DOR) in responders was a little over a year (12.4 months). 

What was most impressive here were the results in a subset of a little more than a third of the patients with tissue testing that showed a high proportion of the protein PD-L1, which is a key mediator of immune system activation vs. suppression. In that subpopulation with at least 50% of their tumor cells staining for PD-L1, the response rate was 45.2%, compared with 16.5% for patients with lower but detectable PD-L1 expression (on 1-49% of tumor cells) and 10.7% for patients <1% of tumor cells positive for PD-L1 expression. OS and PFS also appear to be higher for the high PD-L1 expression subgroup. 

These are very encouraging results, to be sure. But as we are now clearly entering an era where we have multiple immunotherapies for overlapping clinical settings, it's time to ask whether these agents really provide meaningful differences if given to the same patients, or are the differences truly marketing differences of segmenting markets for remarkably similar treatments. Are we really just seeing results for Coke and Pepsi, perhaps with a newly created "Pepsi Generation"? 

I think there's good reason to suspect that Coke and Pepsi are more distinctive than these two agents with the same mechanism of action.  The most significant contribution of the Keytruda data is that it shows in a substantial population for the first time that PD-L1 expression on tumor tissue is a valuable predictive marker of benefit with one of these agents, as results have been mixed. But this doesn't mean that you wouldn't see the same results with Opdivo given to a population with high PD-L1 expression. Instead, this is like Pepsi running a test of consumer-reported refreshment among a large number of people, then breaking them down by which ones were hot and thirsty before the test, and reporting that their level of refreshment was especially high -- this doesn't mean that this is Pepsi-specific -- if you give a Coke to hot and thirsty people, they'll also be especially refreshed. And I strongly suspect that if you give ANY immune checkpoint inhibitor to patients with high PD-L1 expression, you'll see especially favorable results.  Are PD-L1 inhibitors like MPDL-3280A and MEDI4736 significantly different? Perhaps, but I suspect that they are about as different from Coke & Pepsi as 7-UP and Sprite: a little different, but not a remarkably different or better experience. We'll need to learn whether there's a value to transitioning to a different immunotherapy after a patient progresses on their first one. Until there's evidence to show it, I wouldn't presume there to be value of sequential immunotherapies...but let's see what the data show.

In the meantime, if I'm not blown away by the favorable results with Keytruda, they do show for the first time that PD-L1 expression can be of some value. Importantly, though, the positive predictive value (if you have the marker, you're especially likely to benefit) and negative predictive value (if you don't have the marker, you won't benefit) are less stark than with the biomarkers we've readily adopted, like EGFR mutation or ALK rearrangements (though we haven't really tested ALK inhibitors in ALK-negative patients to any meaningful extent).  

So how will we actually choose among the different immunotherapy agents, when we also have Opdivo, which is going to be approved for a broad population, regardless of PD-L1 expression, and likely other entries into the market in the next 1-2 years? If you believe that these agents are likely to produce extremely comparable results when given to similar patients (i.e., Opdivo also works especially well in high PD-L1 expressing patients, if it were to be tested), you would be inclined to not bother testing PD-L1 status, except perhaps if planning to give it as a first line agent.

If Keytruda is approved as a first line therapy for the minority of patients with high PD-L1 expression, it makes sense to consider testing for it up front -- if positive, I suspect most patients would prefer immunotherapy, which is typically well tolerated (and was in the Keytruda results reported) and can lead to a longer response than we'd expect to see with conventional chemotherapy. In other words, I see the distinct potential that PD-L1 testing will be most valuable at initial diagnosis, so that we can identify patients more likely to benefit from immunotherapy than chemotherapy -- though we'd really love to get a randomized trial of Keytruda vs. chemotherapy in high PD-L1 expressing patients, in case high PD-L1 expression is actually a marker of greater chance of benefit from chemotherapy as well.  But if a patient starts with chemotherapy or doesn't test positive for high PD-L1 expression at initial diagnosis, I don't see a great appeal to repeating a biopsy after progression on first line chemotherapy.  Though there is evidence that PD-L1 expression is dynamic and can change over time and treatments, a repeat biopsy incurs a time delay and some risk of complications, as well as cost, and only a minority of patients will test positive. Meanwhile, we can always give Opdivo to patients regardless of PD-L1 expression levels, and if we have no evidence that patients with high PD-L1 expression do better with Keytruda than with Opdivo, giving Opdivo is never a wrong choice and can save time and money and risk.

In practice, we may find that insurers dictate a preferred immunotherapy in the same setting, based on financial decisions, in which case, oncologists and patients likely won't fuss over whether they are approved for Opdivo or Keytruda or another agent that appears on the market in the future any more than they are likely to leave the restaurant when the server asks "is Pepsi OK?" when the customer reflexively requests a Coke.  To see meaningful differences, the definitely test would be a head to head comparison of one to another, but that's not the kind of question that excites anyone. Instead, the next round of questions will center on sequential therapies ("I'll have an Opdivo, with an MPDL-3280A chaser"), or mixed cocktails of different immunotherapy agents. That has the promise of quite a happy hour, as long as you don't get stuck paying the bar tab.