It's not just Theralase that has gone through a long and winding road. The development and commercialization of Keytruda has also gone through numerous twists and turns.
So much so that
Keytruda patent protection is set to expire in 2028. And it has taken all this time despite the resources available to deep-pocketed Merck to accelerate development. (500 clinical trials as of 2017). There is a very interesting 2017 Forbes article detailing the history of Keytruda from the beginning. It says this about their 2013 BTD approval and about the decision to select patients with high levels of the biomarker PDL1 expressed in their tumours in order to up the efficacy numbers in their trial results.
...............................................................................
Regulatory Breakthrough (Designation)
Merck had another ace up its sleeve as well. At a safety conference in Japan in 2012, the head of non-clinical safety, Joseph DeGeorge–who earlier in his career had spent 13 years at the FDA–learned from FDA colleagues how they intended to apply and implement a pathway called breakthrough designation (BTD), designed to accelerate and rationalize the approval of new drugs for unmet needs. While BTD does not mandate any specific schedule acceleration for molecules accepted into this category, it fundamentally changes the relationship between the FDA and the company developing the drug. Meetings can occur with greater frequency, and–most significantly–the FDA works in much closer partnership with the company to identify the most efficient path to approval (while not changing the specific approval standards).
With its promising early results in advanced melanoma, Merck applied for and, in January 2013, received BTD status for Keytruda in this indication–a fact it did not immediately publicize, as it didn’t want to alert potentially drowsy competitors to the possibility of using this novel regulatory mechanism.
In speaking with a number of Merck scientists, it appears one of the key benefits of BTD status was to backstop the pembro-inspired shakeup of Merck’s historically conservative and linear drug development mindset and assure Merck that its new approach–involving doing more work at risk, and with far greater agility–was regarded favorably by regulators and represented a move in the right direction.
As the melanoma work was proceeding, the Merck team–motivated by promising results nivo was reporting–considered the potential benefit of the drug in non-small cell lung cancer (NSCLC) and added cohorts to the phase 1 study to explore this possibility, specifically the use of pembro as second-line therapy.
The expansion of the pembro program at Merck created a logistical challenge for the company. At the time, Merck was hardly an oncology powerhouse, and it just didn’t have enough people in its clinical development organization to throw at the growing need. On the other hand, the intrinsic excitement of the program motivated some to step up. The lung program, for example, would be led by a seasoned veteran, Martin Huber, who had led oncology programs in lung cancer more than decade earlier, and had worked his way to a senior role, director of pharmacovigilance at Merck. In this role, he was also on a review committee for compounds in clinical development and quickly recognized both the promise of the drug and the need for his involvement. He stepped out of his management role and into the arena, leading the pembro NSCLC program.
Saved By The Biomarker
When Merck started the lung cancer program, they viewed themselves as years behind BMS and thought their only chance for rapid approval would be to demonstrate unusually high response rates in second-line therapy. The challenge was how to achieve this–and they turned to a biomarker.
Merck’s instinct to utilize a biomarker in clinical development is hardly surprising; a hallmark of Peter Kim’s tenure at Merck was a strong emphasis on biomarkers–molecular signposts that can be used to segment patient populations, either to identify those most (or least) likely to respond or to get an early indication that a new drug is behaving as expected.
(A quick aside: I had a front row seat to Merck’s embrace of biomarkers during the Kim years; I worked in Merck’s newly-formed Department of Experimental Medicine from 2006-2008 and developed a firsthand appreciation for both the possibility and the limitations of biomarkers in clinical development.)
In this case, Merck turned to a molecule called PDL1, which is expressed on the surface of a number of cells, including some tumor cells, and is the natural ligand of PD1. The hope was that by selecting patients with tumors that expressed high levels of PDL1, they could achieve increased efficacy. (This was based on empiric observations, including those reported by nivo researchers, as well as the thought that if a tumor doesn’t express PDL1, it’s less likely that the PD1 pathway has been upregulated.) It was never exactly clear why PDL1 expression should predict efficacy–a relationship that to this day remains unclear and the subject of intensive discussion.
The decision to utilize a biomarker for clinical development is often fraught, and was particularly so here, for a combination of commercial, scientific and humanistic reasons.
In general, commercial teams tend not to favor biomarkers and seek to avoid them wherever possible. They would rather target an entire population than a subpopulation, and also recognize the inconvenience of requiring a patient to get tested prior to using a drug. All things being equal, a doctor would prefer to prescribe a drug immediately, without waiting for a test to be ordered and the results received and interpreted.
But all things are rarely equal; the ability to define a population that responds well to a drug can result in clinical trials that yield far more impressive results–and require far fewer patients to demonstrate that the drug is effective. In this case, the Merck team felt they were years behind BMS in lung cancer and thought their only way to stay relevant was to use an effective biomarker to identify a subgroup of patients that would respond in a such a compelling fashion that regulators would be impressed and would want to ensure the drug was available as soon as possible to serve this defined group of patients.
Scientifically, the challenge is that biomarkers are rarely perfect, and the PDL1 biomarker assay was perhaps a bit messier than most–yet Merck believed that it had developed a reliable assay that could support clinical development and incorporated this in yet another arm of its ever-expanding phase 1 study.
Humanistically, the challenge is that by requiring the use of a companion diagnostic, or biomarker, patients who test negative generally cannot receive the drug. This makes sense if the biomarker works perfectly, but if not, it means you are effectively denying the drug to patients who might benefit. As Alise Reicin, a Merck executive who oversaw much of the early pembo program (and is now SVP and head of global clinical development at EMD Sorono) recalls, “We heard from physicians that late-line lung cancer patients had no options, so even if they had only a 7-8% chance of responding" because the tumors did not express high levels of PDL1, the doctors still "wanted the option to use the drug.”
At the time the pembro team was engaging in these internal debates, Roger Perlmutter replaced Peter Kim as head of Merck Research Labs; many of Perlmutter’s former Amgen colleagues soon joined him, and one of their first tasks was reviewing the pembro development program. According to a number of scientists, the new leadership team was very concerned about the biomarker approach that had been adopted. Some speculated that perhaps the new leadership team felt it had been burned by previous challenging experiences with biomarkers at Amgen, possibly during the development and commercialization of Amgen’s Vectibix.
But many were convinced the main worry about the use of the PDL1 assay was scientific–which was how Perlmutter himself explained his concerns. “Decisions in early research are based on scientific, not commercial, considerations,” Perlmutter told me, adding that from the outset, he embraced the idea of a biomarker for the lung program, but wanted to be sure it worked.
As Perlmutter explained, while PDL1 is the natural ligand of PD1, the biology is not so straightforward. Drugs that block PD1 may be especially effective in tumors with high levels of PDL1 because PDL1 itself may be a surrogate marker for local inflammation–it’s upregulated in the presence of inflammatory cytokine interferon gamma, Perlmutter points out.
After considerable review and often-anguished debate, Perlmutter and colleagues backed the continued use of the biomarker in the NSCLC second-line program (the first studies had already kicked off by the time Perlmutter joined in April, 2013)–as well as in another, extremely important (and hugely valuable) set of studies evaluating the use of Keytruda as first-line therapy for NSCLC.
Even after this decision was made, it wasn’t clear it was the right choice; BMS, for example, had elected to develop its PD1 agonist for lung without a biomarker–a decision that as recently as March 2016 was seen as yielding better results; consider this lede from the Wall Street Journal: “Bristol-Myers Squibb has sprinted to an early lead in the race to sell a new class of cancer treatment by bucking the trend toward precision medicine.” The authors go on to observe that “Bristol has outflanked Merck partly by sticking to the old, mass-marketing approach.”
However, these results were for second-line therapy; when the results for pembro and nivo for first-line NSCLC were reported in 2016, there was a shock: Merck’s study was strongly positive, while the BMS study–which used a biomarker, but with a far lower threshold–failed, a result that stunned many observers but seemingly vindicated Merck’s approach.
The Merck team also used PDL1 levels to prioritize which tumors to target next; a preliminary study of the top four–head and neck, bladder, triple negative breast cancer and gastric cancer–led to promising results and dedicated studies in each of these indications (although the phase 3 program in advanced head and neck cancers reported negative results this week). An even more expansive program–evaluating pembro in a number of other advanced tumors, selected based on measured PDL1 level, soon followed. ...............................................................................
The Startling History Behind Merck's New Cancer Blockbuster