What BMS’s Cancer Drug Failure Means for the Future of Precision Medicine
On the morning of August 5, 2016, pharmaceutical giant Bristol Myers-Squibb (BMS) announced that its drug, Opdivo (nivolumab), failed a major lung cancer clinical trial. Almost instantaneously, BMS lost $20 billion in market value — its largest drop in over 14 years.
Unlike many other clinical trials where failure is indicative of a drug’s poor efficacy or dangerous safety profile, Opdivo’s failure has a lot more to do with its study design and strategy towards “precision medicine.”
Precision medicine is an emerging approach to treatment whereby drugs are custom-tailored to a patient (a topic I touch on here.) This concept has gained significant traction among oncology drugs, where directing the most effective drugs to the right patients is especially important as treatments are both costly and urgent.
Academic, regulatory, and investor uncertainty have swirled around whether Opdivo, and drugs like it in this class of “PD-1 inhibitor” drugs, need to be targeted and precise. And although PD-1s are by no means the first class of targeted oncology drugs, the highly publicized failure of Opdivo’s trial and the success of its rival’s (Keytruda by Merck & Co.) highlight today’s gray area of precision medicine.
PD-1s: A Gray Area for Precision Medicine
The ideal goal of “precision medicine” is to have a binary determination of whether a drug will work in a certain patient. If Drug X only works on patients with a specific gene mutation, then only those patients with that mutation should receive the treatment.
Many examples of this exist in oncology already. Herceptin and Tarceva are both existing oncology drugs that require a genetic or proteomic test for a HER2 gene or EGFR mutation, respectively, prior to its use. Such targets are known as “predictive biomarkers” — biological measures used to predict if the drug treatment will work for the patient.
Such binary measures are not always evident. The ill-defined PD-1 biomarker is only probabilistically indicative of whether the drug will work. High protein expression of the PD-1 biomarker (known as PD-L1) means high likelihood of the drug being effective. Low PD-L1 protein expression means low likelihood. Little clarity exists beyond these directional indicators on the PD-L1 expression cutoff at which a drug will or will not be effective.
In light of this uncertainty, Merck opted to take a pro-biomarker position for its Keytruda trials in which patients would need to be tested prior to using Keytruda. BMS, on the other hand, took a more generalized approach for Opdivo.
Why BMS Slumbered when Merck Soared
The outcome of Merck’s and BMS’s contrasting biomarker positions took the front page when Opdivo’s failure to meet clinical endpoints was announced. But why did this failure end up shaking the investor community so greatly?
Merck’s and BMS’s bet on the role of PD-1 biomarkers was of course not exclusively academic in nature. Rather, their respective positions carried immense commercial implications — implications that helped grant BMS a lead of $942M for Opdivo over $566M for Keytruda in 2015 sales alone.
To fully appreciate the commercial implications of Merck’s and BMS’s biomarker strategy, I’ll start first with a brief overview on the immuno-oncology market and how BMS’s ill-fated Opdivo trial transpired.
Immuno-Oncology Drugs are the Next Big Thing
Oncology drugs are seen as the next frontier of life sciences with immuno-oncology (IO) drugs at the forefront. The excitement over IO drugs revolve around their superior efficacy and safety over existing chemotherapies, many developed as early as the 1960s. 1-year survival rates of IOs have been seen to nearly double over chemo and side effects are often less severe than that of chemotherapy.
PD-1’s and PD-L1’s such as Opdivo and Keytruda are a subset of IO drugs grouped by their shared mechanism of actions. PD-(L)1’s alone are predicted to reach a global market of $31 billion by 2020 according to Goldman Sachs. Of that universe, non-small cell lung cancer (NSCLC), a highly prevalent disease with high unmet need, represents a significant $15 billion market opportunity. Today’s trials by BMS and Merck are a race to see which PD-(L)1 company can capture the largest piece of this overall pie.
Opdivo’s Initial Commercial Success Over Merck Reflects BMS’s Biomarker Strategy
Prior to the failure of Opdivo’s “first-line” trial on August 2016, BMS had seen immense commercial success in “second-line” NSCLC therapy all through 2015. Line of therapy represents the order in which drugs are administered to patients. “Second-line” therapy indicates the second intervention that can be used if the first drug is ineffective. It is often the first “indication” (an FDA-approved use) pursued in a cancer-type given its lower clinical benchmark and faster market access.
Opdivo’s commercial success in second-line had very much to do with Merck’s and BMS’s contrasting study approaches. In an ordinary trial, one might just compare patients who receive the PD-1 drug to patients who receive the chemotherapy. However, in light of uncertain evidence that the drug works best among patients with high PD-L1 biomarker expression, one may be more inclined to plan a study that evaluates only this subset of patients. After all, if in fact the PD-L1 biomarker is meaningful and the study included patients with low expression, there’s a immense risk that the study would fail among these patients not responding to the drug.
Thus, two strategies emerged:
- Merck’s Risk-Averse Trial Strategy: Examine a subset of NSCLC patients with high PD-L1 biomarker expression and evaluate if they respond better to Keytruda over standard chemotherapy.
- BMS’s Risk-Taking Trial Strategy: Examine all NSCLC patients regardless of biomarker expression and evaluate if they respond better to Opdivo over standard chemotherapy.
Ultimately, both drugs passed their respective trials and gained approval that time around. Keytruda achieved a second-line indication for patients with high PD-L1 expression and Opdivo achieved the same indication for the general population.
This has a lot of implications. For one thing, Opdivo now has a wider market opportunity in that it can be used for any second-line NSCLC patient, regardless of PD-L1 expression. Meanwhile, Keytruda use would be restricted to only the 60% of NSCLC patients that have high PD-L1 expression.
Secondly, physicians are much more likely to prescribe a drug that does not require an extra diagnostic test for the patient to test for PD-L1 expression. When time is of the essence, extra tests are often avoided by physicians.
Taking BMS’s Second-Line Strategy to First-Line (and to Opdivo’s Demise)
The next natural step after acquiring a second-line NSCLC FDA approval is to move into first-line where the stakes are even higher. The stakes are higher for two reasons: a) because there are more patients in the first-line group meaning a larger market and b) a first-line indication prevents rival products from eroding their market share if patients can take their drug first before any other drug. Patients tend to stay on the same drug once treatment is initiated, so patients receiving Opdivo first are unlikely to try another PD-1 next, effectively protecting their market.
The success of this indication was pinned on one trial each. Opdivo’s trial, named CheckMate 026, was initiated March 2014. Keytruda’s trial, named Keynote 024, was initiated August 2014.
Merck’s risk-averse strategy at targeting patients with high PD-L1 expression levels (expression of >50%) vs. BMS’s riskier strategy at targeting a generalized population (this time choosing a biomarker, but of only expression of >5%) continued into how they approached their first-line NSCLC clinical trial design.
Investors believed that BMS’s dominance over a more generalized population would continue just as it had with second-line.
BMS’s Biomarker Strategy Fails in First-Line
In spite of investor confidence that Opdivo would pass its first-line trial and expand its sales lead, the trial failed. As it turned out, high PD-L1 expression in patients is more important in the clinical success of first-line therapy.
Why did BMS’s strategy work in second-line but not in first-line NSCLC?
At the end of the day, the clinical thresholds in first-line therapy were higher and thus, more difficult to surpass. This is partly because the standard of care comparator in a clinical trial for second-line is weaker (e.g. docetaxel, which was used in Opdivo’s second-line trial, generally has low response rates, making it an easier benchmark to overcome). Furthermore, the unmet need is higher for second-line as treatment options are limited, thereby lowering the threshold of approval for a new indication.
Low PD-L1 expressing patients not responding to Opdivo likely dragged down the trial results. Conversely, Merck’s targeting of high-expressing PD-L1 patients, given its narrow focus, responded positively to Keytruda.
The Uncertain Future of Precision Medicine and Biomarkers
The Opdivo and Keytruda case study highlights a few key questions:
- Until precision medicine evolves to a point of binary determination, how does one deal with gray areas of imprecise determination?
- Particularly in life-threatening diseases like oncology, who determines whether low PD-L1 expressing patients shouldn’t access the drug, even if just for the remote chance of success?
- What happens when drug development outpaces the precision of biomarker diagnostics?
The outcomes of the study and its impact among the investor community point to a large divide in how different types of stakeholders consider precision medicine.
In the case of Keytruda, the FDA states only that it’s approved for PD-L1 expressing patients, with no further clarity on what level is required. Rather, the FDA shifts the burden of this determination towards the diagnostics used to test for PD-L1 levels. The lack of a definitive biomarker benchmark is less so a problem of the drug, but the accompanying diagnostic. After all, there are a number of ways PD-L1 expression can be tested, screened, and calculated, all of which can lead to different measures.
Among the clinical community, the National Comprehensive Cancer Network (NCCN), an authority on U.S. cancer treatment recommendations, took a stance on this issue in its guidelines for second-line NSCLC. Even though Keytruda had only been proved to be efficacious in patients with PD-L1 expression of >50%, NCCN recommended its use even in patients only expressing 1% PD-L1.
And finally, there are the payers for which precision medicine represents a huge opportunity to control healthcare costs. Payers, who make reimbursement decisions based on clinical data, are suddenly beholden to such “what-if” scenarios of patients with poor biomarker expression benefiting from the drug, however low the probability.
How this debate will evolve over time remains uncertain. Perhaps this is merely an ugly transition phase of precision medicine and a waiting game for when diagnostics evolve to be more predictive. Precision medicine represents a true opportunity to cut healthcare costs and improve patient safety by which expensive medication is administered only to patients that’d respond. However, a lot of work remains before that vision is achieved.
Sources for graph:
- IO market at $31 billion.
- NSCLC market estimated at $15 billion.
- First-line NSCLC market estimated at $12 billion.
- PD-L1 positive patients roughly make up 60% of NSCLC patients.
