NEW YORK (GenomeWeb) – Ideally, molecularly informed drug development is supposed to provide pharmaceutical companies with more confidence about how well their therapies work in a specific patient population. But the recent failure of Bristol-Myers Squibb's immunotherapy Opdivo (nivolumab) to meet its primary endpoint in a late-stage non-small cell lung cancer trial highlights the pitfalls drugmakers face amid evolving science and imperfect biomarkers.
The anti-PD-1/PD-L1 drug market is estimated to be worth around $30 billion by 2020. The negative Phase III results for Opdivo in first-line NSCLC shocked market analysts, and pharmaceutical companies developing drugs in this space "are taking stock of the situation" and surely assessing whether they have the right test methods in place in their drug studies, said Peter Keeling, CEO of consulting firm Diaceutics.
Most drugmakers developing anti-PD-1/PD-L1 drugs are evaluating whether their therapies work especially well in patients whose tumor cells express PD-L1, a protein that cancer cells use to hide from an immune system attack. But studies have demonstrated that PD-L1 expression status doesn't neatly bucket responders and non-responders in the same way that testing for EGFR mutations or ALK rearrangements can, making it difficult for pharmaceutical companies to integrate PD-L1 testing for patient stratification in clinical trials.
"PD-L1 is not the ideal marker," said Edgar Braendle, who previously headed up Novartis' companion diagnostics unit and this August became CEO of ARUP Labs. "There may be patients with very low expression and they may respond. That's an issue."
Despite the limitations of PD-L1, some industry observers are of the view that it's better to use it, especially alongside other cancer markers, to learn as much as possible about patients' tumors ahead of treating them. In the case of HER2 testing, which first came to market in 1998 alongside Genentech's breast cancer drug Herceptin (trastuzumab), "you still read about HER2 borderline cases, equivocal cases, and other data you should look at," said Panna Sharma, CEO of Cancer Genetics, which sells molecular diagnostics for personalizing oncology treatment, including PD-L1 tests. "It's imperfect, but it works, and it's better than the alternative environment where you have no data."
In the case of Opdivo, BMS utilized a PD-L1 test to stratify patients in studies, but industry observers have noted that in the CheckMate-026 study the firm didn't segment the population as stringently as it should have, choosing to go after a broader population: first-line advanced NSCLC patients who had PD-L1 expression in 5 percent or more of their tumor cells. "We designed our development program in lung cancer to address the unmet need of every lung cancer patient and our scientific approach is a bold one," BMS said in statement in early August when the disappointing study results were released. The firm didn't respond to a request for comment for this article.
In second-line NSCLC, a study had already shown that patients responded to Opdivo regardless of their PD-L1 expression status. However, it was also clear from this 600-patient study, which the FDA used for market approval, that the more PD-L1 expression patients had in their tumors, the better they did on Opdivo.
In the highest PD-L1 expressing group — those with expression in 10 percent or more tumor cells — median overall survival was 19.4 months compared to 8 months in the docetaxel arm. When researchers looked at response regardless of PD-L1 expression, patients on Opdivo lived an average of 12.2 months compared to 9.4 months on docetaxel.
Although BMS' PD-L1 cutoffs in the first- and second-line NSCLC studies were different, some market analysts have cited the earlier results as a hint that BMS' cutoff of patients with PD-L1 expression in 5 percent or more tumor cells may have been too generous in the CheckMate-026 study. Of course, many of these same analysts had also pegged Opdivo as the winner of the immunotherapy market, since sales for BMS' drug had been far outpacing another NSCLC checkpoint inhibitor, Merck's Keytruda (pembrolizumab).
In the second quarter, Opdivo brought in $840 million in sales, while Keytruda netted $314 million. Many industry observers had attributed Opdivo's success to the fact that it wasn't held back by the requirement for a companion diagnostic, like Keytruda was.
"This is a little like the hare and the tortoise story with Opdivo leveraging test-free status in second-line NSCLC to gain early oncology use and experience, [which is] always valuable in a drug war scenario," Keeling said. "Meanwhile, Merck's dependency upon a PD-L1 test was perceived as a disadvantage. That all just changed."
The FDA approved Opdivo last year in second-line NSCLC with a so-called complementary diagnostic, which unlike a companion diagnostic is not required for the safe and effective use of a drug, but can "help physicians determine which patients may benefit most from treatment with Opdivo." In press statements, BMS noted that "biomarker testing is not required for Opdivo."
Keeling believes that based on a view that PD-L1 is an incomplete or under-developed marker, and helped by the flexibility afforded by the complementary diagnostic category, BMS pursued a commercial strategy in second-line NSCLC where it could promote Opdivo as not requiring prior PD-L1 testing to identify which patients would respond to the drug. This view on the biomarker may have also ended up hurting the company's Opdivo studies in first-line NSCLC. "BMS saw the opportunity to say that the PD-L1 science was imperfect," he said. "As a result of that, they decided clinically and from a business perspective to seek a broad patient population [for Opdivo]."
But the results from CheckMate-026 should also be a lesson to drugmarkers that there is a complexity to incorporating biomarkers alongside a drug, according to Keeling, and not managing that efficiently could have financial consequences. Most drugmakers still want to "avoid a test when they can, and embrace a test if necessary," he said. "Until now the financial exposure to these mistakes was opaque."
We're beginning to understand the nuances of certain biomarkers in immunotherapy and are applying these key learnings to future studies.
Dealing with imperfection
Drugmakers have historically viewed response markers and diagnostics as limiting market share and sales. But a poorly managed diagnostic strategy, even with an imperfect biomarker, could also lead to financial losses.
For example, retrospective market analysis suggests that challenges related to immunohistochemistry-based HER2 testing for Herceptin may have resulted in up to $3 billion in lost revenues. Similarly, Diaceutics modeled the impact of PD-L1 testing on immunotherapy revenues and concluded that suboptimal access to testing could result in $744 million in lost drug revenues over five years just in second-line NSCLC.
Although Keytruda sales were lagging behind Opdivo in the second-line setting, after BMS' setback in CheckMate-026, it now looks like Merck will enter the first-line NSCLC market before BMS. The company this week announced that the FDA has granted priority review and breakthrough therapy designation to Keytruda, and the agency could decide whether to approve the drug before the end of the year. Merck is still pursuing a companion diagnostic strategy in first-line NSCLC based on Phase III data showing that patients with high PD-L1 expression (defined as having a tumor proportion score of 50 percent or more) lived longer on Keytruda than on chemo.
Most other pharmas developing drugs in this space are all evaluating responses using PD-L1 but the challenge for regulators, labs, and ultimately doctors, is that each company has taken markedly different approaches toward integrating testing in studies. For example, the FDA has approved four PD-L1 tests — a companion diagnostic for Keytruda in lung cancer and three complementary diagnostics for Opdivo in lung cancer and melanoma, and for Tecentriq in urothelial cancer — but these tests have different cutoffs for PD-L1-positive and -negative patients and employ different antibodies.
Although there is an initiative underway to understand how comparable these tests are, "there is a lot of unclarity about what are we measuring," as well as uncertainty about the best cutoff value for defining PD-L1 positive/negative patients, ARUP's Braendle said. Adding to this complexity is the fact that some tests employ algorithms factoring in PD-L1 expression just on tumor cells, while other tests are also considering expression on tumor-infiltrating immune cells.
Genentech is one company using IHC to gauge PD-L1 expression on tumor and immune cells in trials for its anti-PD-L1 drug Tecentriq (atezolizumab). "When it comes to the science of immunotherapies, we still have much to learn," Dan Chen, global head of cancer immunotherapy development at Genentech, told GenomeWeb. "We recognize that not only is the biology complex, but IHC assays can differ in both their performance and how reliably they can be read by pathologists."
According to Chen, Genentech is dealing with these challenges by trying to learn as much as possible about PD-L1 by studying responses in patients with different expression levels across Phase III Tecentriq studies. Genentech is also working with pathologists to ensure they can read test results. "We're beginning to understand the nuances of certain biomarkers in immunotherapy and are applying these key learnings to future studies," Chen said.
In addition to ongoing efforts to improve knowledge of PD-L1, researchers are also investigating other diagnostic approaches using next-generation sequencing, for example, to look at tumor mutational burden and microsatellite instability. But because immunotherapies are so novel and there is much unknown about the optimal response markers for this class of drugs, Keeling recognized that pharmas have a significant challenge in having to lock down a diagnostic within a drug development program, knowing that the science on the biomarker may change.
BMS will surely incorporate what it has learned about the PD-L1 marker into future studies, and has said it will provide more details on the CheckMate-026 trial at an upcoming scientific meeting. The company is investigating the combination of Opdivo plus Yervoy (ipilimumab) for PDL-1 positive patients, and Opdivo plus Yervoy, or Opdivo plus chemotherapy in PD-L1 negative patients.
Industry observers GenomeWeb spoke to believe that BMS is likely exploring how it might reevaluate the patient population for Opdivo in first-line NSCLC. "With monotherapy first line, [BMS] didn't get the data they wanted. But they do know that it does work in a subset of the population," Sharma said. "I'm sure there are a lot of intelligent people working rescuing the data that they already have."
What we've done is told doctors to sort it out. That's not a responsible way to develop precision medicine in the long term.
With the approval of Opdivo in second-line NSCLC last year, the FDA greenlit the first complementary diagnostic. Some industry players feel the new category provides necessary regulatory flexibility since the science behind immunotherapies is still emerging.
"For cancer immunotherapy, the status of key biomarkers is nuanced and not as clear as identifying a mutation in a single gene," Genentech's Chen said. "Complementary diagnostics assess the potential for a response and could be critical to identifying those people that will best respond to immunotherapies. As such, they provide information to aid in the physician and patient decisions related to treatment, but do not provide black and white information regarding when to test."
Based on preliminary data from a Phase III study looking at Tecentriq in second-line NSCLC, Genentech highlighted last week that advanced NSCLC patients, regardless of PD-L1 expression status, lived significantly longer on its immunotherapy compared to those on docetaxel. Genentech would not say if it is pursuing a complementary or companion diagnostic strategy for Tecentriq in the second-line setting.
While the complementary diagnostic category provides flexibility in conducting studies and informing treatment strategies using emerging biomarker science, in Keeling's view the regulatory nuance isn't necessarily furthering personalized medicine, since the commercial operations at some pharma companies have interpreted "complementary" as "test-free prescribing." Furthermore, Diaceutics' surveys of physicians show they are confused by the messages around complementary and companion PD-L1 testing and whether to test patients ahead of prescribing the drug.
"What we've done is told doctors to sort it out," Keeling said. "That's not a responsible way to develop precision medicine in the long term."
Labs are also challenged by having to decide whether to invest in validating and offering all four FDA-approved PD-L1 tests. "If you look from a laboratory perspective, in the ideal world, you need one test, and clear instructions about the algorithm and cutoff values to assign patients to treatment," Braendle said. "Four different tests creates quite a confusing situation for the labs and the physicians."
In the absence of a universal diagnostic, Cancer Genetics is one lab that does offer all the FDA-approved companion and complementary PD-L1 tests, but also spends a lot of time educating doctors and pathologists on the differences between these tests and which to order for a particular immunotherapy. "There's a big need for our pharma cousins to really push more knowledge about the associated diagnostics and tests and how to use them," Sharma said. "That's the only way they're going to get uptake in the community setting."
Despite the muddled messages around PD-L1 testing, healthcare providers and researchers nonetheless seem interested in PD-L1 testing as part of the tumor profiling workup for patients. Diaceutics' surveys show a sharp uptick in the number of labs offering PD-L1 testing over the past year-and-a-half and 52 labs in the US offer at least one PD-L1 test. The company also reviewed biomarkers being studied in 95 Phase II/III NSCLC, and found that approximately half are incorporating patients' PD-L1 status either alone or in combination with other markers, such as EGFR and ALK mutations.
At Cancer Genetics over the past year, there has also been a notable ramp up in orders for PD-L1 testing for lung cancer patients, but also for melanoma and head and neck cancer patients. "One our biggest volume increased tests this year has been PD-L1 testing," Sharma said. "We think there is a lot of opportunity for significant additional growth."