You may have seen the recent commercials for Merck's immuno-oncology drug Keytruda (pembrolizumab), with the cancer patient talking about the biomarker test she took that determined she had high levels of PDL1, allowing her to receive Keytruda as a first line therapy for her lung cancer over chemotherapy.

It's this test and this distinction that have helped make Keytruda a success story in non-small cell lung cancer (NSCLC), while competitor and former front-runner, Bristol-Myers Squibb's Opdivo (nivolumab) has fallen behind.

Immunotherapies like Opdivo and Keytruda have changed the cancer treatment landscape, but it's becoming increasingly clear from trial results that these potent therapies need to be targeted to the right patient populations to have the greatest impact.

Until this summer, Opdivo had been considered the market leader; the drug had been rapidly gaining approval for indications — melanoma, second-line non-small cell lung cancer, renal cell carcinoma, classical Hodgkin lymphoma, squamous cell carcinoma of the head and neck. Initially approved in 2014, the drug brought in just shy of $1 billion in 2015, jumping to sales of $3.8 billion in 2016. Keytruda, on the other hand, brought in only $566 million in 2015 and hit $1.4 billion in 2016. Opdivo seemed unstoppable, and Merck's Keytruda seemed forever on its heels. 

Yet, Bristol-Myers hit a setback late last summer when Opdivo failed in a clinical study as a first-line treatment for patients with non-small cell lung cancer (NSCLC). This clinical trial changed everything and has set Merck up to take over the market. Shortly after the Opdivo failure, Keytruda went on to get approval from the Food and Drug Administration for the first-line indication.

The differentiator

While PD-1 inhibitors are far from the first drugs that potentially need to be targeted to a certain patient population, their emergence on the market over the last two years has had investors and analysts questioning whether these hot new immuno-therapies need to be targeted to only patients with the PD-L1 biomarker. So far, it's unclear whether PD-L1 expression is an indicator of whether the drugs will definitively work, but there is evidence that patients who express high levels of the protein are more likely to respond to the PD-1 inhibitors, while those with low, or no, expression are not. One of the challenges though — companies don't know where that threshold of PD-L1 expression begins and ends.

For savvy business reasons, Bristol-Myers opted to target a broad patient population, hoping for the widest approval for Opdivo possible. It was suggested in early 2016 — that despite no data or approval in the first-line setting — physicians were already prescribing Opdivo off label to about 20% of first-line NSCLC patients. Analysts predicted the first-line setting could be a $12 billion market. Evercore ISI analyst Mark Schoenebaum previously estimated that Opdivo would bring in approximately $9 billion in revenues by 2019, with more than half of that coming from the NSCLC indication.

Merck, on the other hand, decided to take the more conservative approach. The New Jersey drugmaker opted on the side of precision medicine and used a pro-biomarker strategy, testing patients before trials and only allowing those patients into clinical trials that expressed certain levels of the PD-L1 biomarker. Investors weren't initially keen on this strategy — it limited the potential first-line indication to just 30% of that market, or $4 billion. Now, it seems to be paying off.

So what went wrong?

Clinical trials fail all the time — even for highly anticipated drugs like Opdivo, but usually due to efficacy reasons. In this case, (at least some of) the blame fell on the Bristol-Myers' clinical trial design and the broad patient population included in the trial.

Merck reported positive results for its Keynote-024 trial first, setting the drug up for approval and raising expectations for Bristol-Myer's CheckMate-026 trial. The key difference, though, was Keynote-024 focused on patients with PD-L1 expression higher than 50%. The trial showed Keytruda reduced the risk of death in this population by 50% compared with chemo and lowered the likelihood of disease progression by 40%.  

Meanwhile, the CheckMate-026 trial included patients whose tumors expressed levels of PD-L1 higher than 5%. CheckMate-026 failed to its primary endpoint of progression-free survival. In fact, in the overall patient population of the trial, it didn't even beat chemotherapy. 

"The missed result likely reflects the fact that Bristol-Myers pushed the envelope too far in designing its trial," wrote Bernstein analyst Tim Anderson in August. "By making it 5%, Bristol-Myers was in essence trying to broaden the patient population where it could claim a benefit (had results been positive, of course), but failed results suggest they likely made it too broad, meaning they enrolled patients with too little PDL1 expression, and this soured the overall analysis."

"This is a MAJOR SURPRISE — possibly the biggest clinical surprise of my career," wrote longtime biopharma analyst Schoenebaum in an Aug. 5 note to investors.

Maybe Bristol-Myers should've seen it coming (although no one did).

The company initially got approval for squamous NSCLC in March 2015 based on the CheckMate-017 trial that covered call patients, regardless of PD-L1 status. The late-stage trial found that Opdivo showed better overall survival across PD-L1 expression levels over docetaxel.

Next, the CheckMate-057 trial showed a similar outcome over docetaxel in the non-squamous setting, a larger market than squamous NSCLC. Yet, once the data was analyzed further, it showed that those patients with higher levels of PD-L1 expression had better survival.

This is where things got tricky for Bristol-Myers. Analysts and investors started to ask questions about why there were different results in -017 vs -057, but Opdivo was still performing so well (with data to support it) that their concerns didn’t create much headwind. In fact, in 2015, the biomarker stratification wasn’t considered all that important and investigators were still debating the importance of the PD-L1 biomarker at that year's American Society of Clinical Oncology conference.

Fast forward to August 2016. Opdivo fails, and now biomarkers are very important.

Ultimately, this won't only impact the lung cancer market, but the way cancer drugs will be developed going forward, including the multitude of combinations that are currently in the clinic.

Bristol-Myers has now shifted its strategy; more recently, competitor AstraZeneca changed its trial design for its PD-L1 inhibitor. Expect further changes to come in the I/O market as the science of biomarkers catches up.