Forget about the emerging world of personalized medicine. It’s now a merging world, one in which pharmaceutical companies and diagnostics firms are shaking hands on any number of deals to develop drugs and companion tests for oncology.
Consider these recent partnership announcements:
- Abbott and Pfizer
- Pfizer and Ventana
- Ventana and Bayer
- Life Technologies and GSK
- Roche and, well, Roche
Such unions may not count as kumbaya moments for these medical behemoths, but they do point to budding awareness of how to pull drugs and tests alongside each other and before the FDA.
Along the way, both pharma and diagnostics firms have encountered a few jolts. So has the FDA. This may have slowed things down over the years, though many in the field merely shrug when asked about the pace at which drug-test pairs have come to market. Scientific knowledge may grow quickly, but putting it into a drug and onto a test platform doesn’t happen overnight.
Mara Aspinall, president/CEO, Ventana Medical Systems, has watch-ed personalized medicine unfold in a three-step process. The first step is not the usual premise of promise. It’s fear. “People in the medical and scientific world want to know that personalized health care is not causing any harm,” she says.
Fear has been giving way to step 2 (proving the value of personalized medicine) and step 3 (accepting it in practice). Aspinall’s colleague at Ventana, pathologist Eric Walk, MD, sees medicine now flitting between these two steps. “Industry spent more time than I would have expected at the fear stage,” says Dr. Walk, Ventana’s chief medical officer.
The last five to seven years have been a time of answering questions created by the Human Genome Project, says Ronnie Andrews, president of medical sciences, Life Technologies. Many thought those questions would have been answered long ago. “I’d be lying if I said we weren’t frustrated,” Andrews says. But his optimistic side springs quickly to life when he adds, “The next 10 years will dwarf the last 50 in terms of our ability to treat cancers.” He calls the learning curve of recent years “an adventure. But that rough and rocky road has brought us to a good place.”
That place is filling up with partnerships.
The Ur-partnership was Herceptin and HER2 testing, which wasn’t a sunny, June Allyson-Jimmy Stewart movie marriage. Even with the simultaneous approval of Herceptin and HercepTest in 1998, it took five-plus years for the best testing approach—IHC? FISH? both?—to emerge. “I think we learned a lot from that,” Andrews says. Not the least was that pharma and diagnostics need each other.
That may sound self-evident, but Andrews says pharma companies entered the post-Human Genome Project era thinking diagnostics firms were superfluous to their own efforts. Their inexperience with test development eventually backfired, when their marker sets left highly controlled pharma R&D labs and encountered live patient samples. “Many times pharma will find a series of diagnostic targets as they’re uncovering the drug target, and they’ll put together an assay for their discovery,” says Andrews. “But that assay is not usable and reproducible in the clinical lab.”
So now pharma is chanting a new mantra, which sounds like, “Maybe we don’t want to be in diagnostics. Maybe it’s better to partner.”
At the same time pharma was discovering it was not an island, diagnostics makers smartened up, too. Pharma doesn’t care about technology or testing platforms and is neither impressed nor convinced when a test maker insists on using a specific technology, Andrews says. Pharma cares about the information the technology creates, and whether clinicians can use it to make treatment decisions.
In many ways, this relationship should sound familiar—it could be a microcosm of the pathologist-clinician relationship. “Absolutely,” says Kathryn Becker, PhD, global marketing di-rector for oncology, Abbott Molecular. “A key for a diagnostic manufacturer is to maintain flexibility in meeting the partner’s needs,” she says.
Many of those needs, Dr. Becker says, go back to what the FDA wants. “Overall, I would say that the development is driven by the phase of the therapeutic and plans for the regulatory approval and clinical trials. That’s the underlying driver. We as a diagnostic manufacturer need to ensure that we meet those timelines and are supportive throughout the process.”
“Companion diagnostics is being driven by the FDA,” agrees Debra Leonard, MD, PhD, professor and vice chair for laboratory medicine, Weill Cornell Medical College, and chair, CAP Personalized Health Care Committee.
It’s a quaint word, “companion,” a bit Edith Wharton-y, like a job Lily Bart might have considered during her descent into poverty. But there’s nothing genteel about developing a drug or test. It’s big business, and set to get bigger. No one’s talking poverty here. And with the stakes so high, it’s important to get these partnerships right.
On the surface, these deals resemble the Oklahoma land rush, as companies eagerly sign deals with each other (and, in the case of Roche’s drug and diagnostics sides, within itself). But they’re carefully working out the choreography underneath, as both sides—not to imply an adversarial relationship—learn how the other operates, and, more critically, what the FDA expects of each partner.
Both sides are also learning to appreciate the risks the other faces.
For diagnostics companies, it’s inherently more risky to develop a companion diagnostic than to stick with a core menu of traditional, bread-and-butter assays, says Dr. Walk, who has sat on both sides of the table. Before coming to Ventana, he worked on the pharma side, where, he says, he used Ventana for companion diagnostic development. The diagnostics maker is assuming the risk associated with the drug. “And we all know that there’s a significant failure rate on the drug side for these compounds as they go through development,” he says. “If the drug fails, there’s no place for the companion diagnostic.”
But those risks pale in comparison to the high stakes for drug developers, Dr. Walk says. The technical risk of finding a sensitive, specific antibody, and developing a scoring system, is far smaller than the task facing pharma: validating the clinical utility of a drug and its intended use.
There may also be risk involved in developing a companion diagnostic earlier in the process, Dr. Walk says, in the sense that locking in a diagnostic early may mean that a better test may come along later.
Nonetheless, those partnerships are starting sooner. “The piece that has changed most dramatically in the last five years is that pharma is bringing in their diagnostic partner earlier and earlier in the drug’s development cycle,” says Aspinall.
There may have been no choice in the matter, according to Andrews. “Pharma would get almost to the end of the road, and the FDA would say, ‘Great, show us your biomarker.’ And pharma would say, ‘Oops!’”
Abbott no longer waits until the phase 2 or 3 clinical trials, Dr. Becker says. Early involvement may help make those trials and studies move faster, she says. It also helps diagnostics makers keep a higher level of consistency in analyzing patient specimens and results.
Aspinall sees only an upside. “It’s a lower risk than having a program that does not go forward,” she says. “The largest risk in terms of patients and cost and time is when a therapeutic program fails in stage 2 or stage 3. So understanding that there may be a subset that is most relevant, and what that diagnostic is, to me is the best way to avoid very costly failure.”
With relationships starting earlier, they can’t help but invite chicken-or-egg comparisons. What comes first? The drug, or the diagnostic?
It’s rarely the latter, though Andrews says Life Technologies has approached pharma about targets in a systems biology that might be important for mitigating a recurring cancer. How does pharma take to such suggestions? “There’s interest. And skepticism,” Andrews says.
More typically, pharma makes the first move. Dr. Walk says a pharma company will approach Ventana having already made a decision—based on either preclinical or early clinical research—that it needs a tissue-based companion diagnostic biomarker. Usually that means an IHC assay to detect expression of a particular protein, though he notes that pharma folks increasingly are looking for gene targets, either copy number or amplification status.
Often pharma will have a particular antibody in mind, Dr. Walk continues. Sometimes it will work well in Ventana’s IHC platform; if not, Ventana works with pharma to develop a better antibody, one that will lead to a robust assay that can work reliably in routine pathology practice.
Scientists from both groups will iterate the assay “to the point where we’re both happy with the performance on a technical level,” Dr. Walk says. Later, Ventana will work with pharma’s pathology team specifically to develop scoring systems and cutoffs for assay validation. This results in a so-called design locked assay, which will be used in a clinical trial.
Ventana, for example, is developing an IHC assay for human equilibrative nucleoside transporter 1, or hENT1, for the pharma company Clovis Oncology, which is developing a compound (code name CO-101) that is an analog of the pancreatic cancer drug gemcitabine.
Unlike the parent compound, Dr. Walk explains, the analog is not dependent on a drug transporter (in gemcitabine’s case, the hENT1 protein) to bring the drug into a cell’s cytoplasm. Some 50 percent of pancreatic tumors do not express hENT1 either at all or at very high levels. Clovis Oncology sees an opportunity for a drug targeting patients who have low expression of hENT1.
“We worked with Clovis very early on to develop the antibody and the screening system,” says Dr. Walk. “And now we’re enabling the registration trial for that compound.”
A slightly different partnership occurs when pharma and diagnostics are under one roof.
That’s the observation of Garret Hampton, PhD, senior director, oncology biomarker development and companion diagnostics, Genentech, which is owned by Roche. Dr. Hampton credits Roche leaders with paving the way by setting corporate goals for personalized health care within the company. “The expectation by the end of 2014 is that [of] the molecules that we bring forward into pivotal development, at least 50 percent have to be associated with a companion diagnostic,” he says.
Walter Koch, PhD, vice president and head of global research, Roche Molecular Diagnostics, notes the Zelboraf story was relatively straightforward, in part because the drug and the companion diagnostic (both approved in August 2011) were a result of a collaboration of Roche scientists. Zelboraf is used to treat patients with late-stage or unresectable melanoma; the companion test is the Cobas 4800 BRAF V600 mutation test.
The question pharma posed to diagnostics was direct, Dr. Koch says: Can you develop an assay for this mutant protein?
The Roche pipeline at the time did not have any assays targeting single mutations in oncogenes, Dr. Koch says, though it did have a sequencing test for the p53 protein. But the biggest challenge came from the pharma side, when it became clear the drug itself would have to be reformulated. This put the diagnostics team into hurry-up-and-wait mode. Once the exposure levels were sufficient to inhibit the protein and cause tumor regression, however, “Everything jumped into high gear,” Dr. Koch says. This was at the end of the phase 1 extension trial, in 2009.
Roche Molecular was also in the midst of a change, creating another speed bump. The test, which had been developed for a platform that was being phased out, had to be shifted to the newer system and revalidated.
Within a few months of the start of the phase 2 trial, the phase 3 trial began. “So now things were moving very, very quickly,” Dr. Koch says. The lesson, to him, is clear: Pharma and diagnostics need to start working together early, so they’re prepared when pivotal trials get underway. “It’s not just an assay. It’s also the system, the software, the hardware,” Dr. Koch says—all of which need to be clinically validated in clinical trials supporting drug efficacy as well as the diagnostic result.
It all boils down to the trials. It’s not easy, even within one company, to link a potential drug with the most likely diagnostic companion. Roche tries, by phase 2 trials, to integrate the “lead hypothesis”—that is, the diagnostic hypothesis its scientists believe is most likely to correlate with a drug’s efficacy or lack of it, says Dr. Hampton. At the same time, it tries to integrate its exploratory diagnostics into clinical trials as well, in case the need arises for an alternative diagnostic.
But the hunt has to start even earlier. Whenever there’s scientific evidence of subpopulations of patients who might benefit from a potential molecule, “We’re chasing that at a very, very early stage,” Dr. Hampton says—even when the risks are relatively high, “sometimes when we’ve no evidence of the drug’s activity clinically.”
They’ve really no choice, he says. The time it takes to create a technically validated assay is significant, and having to match the timelines of diagnostic and drug developments means they’re often starting before phase 1.
Molecules need to have a companion diagnostic at phase 2 to enable a phase 3 trial, he says. “It means that we have to have a prototype assay of sufficient technical validity in order to unblind safety, efficacy, and diagnostic together at the end of the phase 2 trial.” That permits them to create an in vitro diagnostic assay before phase 3, allowing them to prospectively enroll patients for a phase 3 trial. “There’s a significant focus on getting those timelines right.”
Genentech has unblinded several phase 2 trials in which the diagnostic hypothesis has borne fruit, Dr. Hampton says. One is MetMAb, an antibody against the Met receptor, being tested in advanced non-small cell lung cancer. In the intent-to-treat population—roughly 120 patients—there was no evidence of efficacy by adding MetMAb to Tarceva in patients enrolled in the study; in contrast, when those patients were split into two groups according to a prespecified diagnostic cutoff, it “led to a terrific result,” Dr. Hampton reports. “We saw significantly extended progression-free and overall survival in patients with high met expression.”
With the diagnostic (developed by Ventana), a drug that appeared to be negative in an intent-to-treat population has been shown to be active and worth further investment. Do this enough times, Dr. Hampton says, and the industry may see a significant drop in the attrition rate for candidate medicines.
Even with these success stories, building a diagnostic test versus building an oncology therapeutic can still look mysterious to the other partner.
Pharma and diagnostics don’t always speak the same language, though Aspinall suggests their lexicons have begun to overlap in recent years. “‘Personalized health care’ has become almost ubiquitous,” she says. As the partnerships progress, pharma has come to appreciate that diagnostic sensitivity and specificity are critical to setting expectations for drug usage, she says; likewise, diagnostics makers are gaining an understanding of how a drug works.
That’s the science side. The business side has “also been an education,” Aspinall says. “It’s critical that each group understands how its product or service is marketed and how it’s reimbursed.”
Pharma has had a blind spot when it comes to the FDA’s approval process for assays. Drug approval, they get; diagnostics, less so. “There’s still a little bit of a gap in appreciating what we need to go through at the FDA,” says Dr. Walk. Some pharma companies, for example, didn’t realize that for a registration trial for a companion diagnostic, the FDA looks for biomarker status in negative patients as well as positive patients. “There’s a challenge for the diagnostic company if you run a biomarker positive trial because the FDA’s stance is, ‘How do you know that the companion diagnostic works at all?’” he says.
The biggest misunderstanding pharma has had of diagnostics, Andrews says, “is that you build a diagnostic test the same way you build a drug.”
Declining to name names, he mentions a pharma client he worked with in previous life at Clarient, where he was CEO and segment leader, GE Molecular Diagnostics. “They mandated that we lock down the genes that we were interested in using to direct their trials during early feasibility studies.” Clarient did so, using what it thought were the right targets based on the discovery of pharma’s compound.
“We took that set of genes all the way through to the FDA,” Andrews recalls. “But along the way we were looking at and uncovering new information as we put more and more patient samples through that test.” Ultimately, the test that earned FDA approval was not the most relevant set of genes for patient selection, he says.
The pharma partner is now trying to take a second diagnostic test through the approval process so potential candidate patients aren’t overlooked. “So the understanding we all need to come to is, how do we marry up a very long drug development process, which is fairly rigid at phase 1, versus a diagnostic assay development?” Andrews asks. “As you pressure test that assay with real patient samples, there are codons and different genetic nuances that we need to tweak along the way, so that the test that is put out there at the time the drug is released is the most appropriate test to select the most patients.”
If pharma can be guilty of tunnel vision, it’s not as if the diagnostics side has been a model of clairvoyance. Test makers don’t always appreciate the precision the FDA requires of pharma, Andrews says, perhaps because their own world shades toward gray, especially in oncology testing. “But that doesn’t afford pharma the precision they need to get their trials done in the time frame they have.”
Finally, Dr. Koch points to one unanticipated challenge during the Zelboraf study. At many of the sites bringing patients into the clinical trials, physicians were screening patients with laboratory-developed BRAF tests prior to enrollment. That created tremendous bias. Normally, only about 50 percent of patients would have the mutation, he says; some sites had much higher percentages of patients showing up. “Obviously, when the FDA saw that, they told us it was inappropriate,” he says. “You can’t prove the performance of the test in conjunction with the drug, for that intent-to-treat population, if you’ve biased it in any way. We had to tell the clinical sites to stop doing that prescreening. And that’s something I think none of us had contemplated.”
Two already complicated maneuvers—developing a clinically useful drug and developing a clinically useful in vitro diagnostic—becomes a 3D chess game when drug and diagnostic come before the FDA seeking approval.
If the FDA took the Myers-Briggs test, it would be a straight ‘S’ (sensing), with an overwhelming preference for dealing with concrete information. It wants to see facts, not potential. “But there aren’t too many examples that have come before them,” Dr. Hampton says.
There are some, of course. The recent pairings of Zelboraf/Cobas BRAF, as well as Xalkori (Pfizer)/Vysis ALK FISH test (Abbott), have given the FDA two strong examples of how to move targeted therapies and companion diagnostics through the approval flume ride. Says Dr. Koch, “When you know what the biomarker is, and the hypothesis is extremely strong—and even perhaps phase 1 data exist—that association is going to play out.”
But there are still plenty of scenarios where the biomarker hypothesis is a bit more jiggly at phase 1, and may firm up—with sufficient statistical backing—only in phase 2. “That’s where things get more challenging,” Dr. Koch says. “Because now you have to wrap up quickly to do the pivotal phase 3 selection. And if you didn’t put enough de-risking activity”—this is a popular phrase among industry types, by the way—“into a diagnostic early, then you may not be there in time.”
The FDA, like industry, is learning. “We’re very pleased with how things went with Zelboraf and the Cobas test,” Dr. Koch says. “They worked with us throughout the process to streamline the procedures, allowed us to have modular filing of the PMA sections, and a host of things that probably hadn’t been done before—at least that we didn’t have experience with before.”
Abbott likewise reports a satisfying encounter with the FDA during the approval process for Xalkori and the FISH test. “We had multiple interactions, as you might imagine, with the FDA, where Abbott and Pfizer approached both branches together,” says Dr. Becker. Abbott has done the same with other pharmaceutical partners as well, meeting with the FDA at each step. “They’ve been nothing except responsive,” she says.
Part of the problem is obvious but bears repeating: Oncology doesn’t stand still. Says Dr. Leonard, “We can make discoveries faster than we can move them into clinical use because of the complexity of the pathway.” Understanding a tumor’s molecular makeup is one matter; understanding its proteomic output is another. And any new drug requires physicians to look at multiple genetic mutations that are being transcribed upstream through multiple pathways.
The FDA, however, requires pharma companies to run site-specific trials—a breast cancer trial, for example, or a lung cancer trial. “But what we know today that we didn’t know five to seven years ago is that cancers have similar pathways and similar genetics, regardless of what site they come from,” Andrews says. “It’s very plausible that you could have a lung cancer that looks like a colon cancer at the molecular level.”
Dr. Koch raises an important question: Is it sustainable to have one oncology drug partnered with one diagnostic? In recent years larger cancer centers have been using multiplex testing to make what is essentially a differential diagnosis, of, say, non-small cell lung cancer into subtypes that are divided by their driver oncogene mutations or other aberrations. It’s a sensible way to work with a limited amount of sample, and delivers useful prognostic as well as therapeutic information. Trying to get the FDA to approve such a test, with its multiple drugs and drug partners, will be no traipse through a meadow. “But it’s the future many of us see,” Dr. Koch says.
The FDA is starting to expand its views. “We don’t want to criticize the FDA,” Dr. Walk makes clear. “The FDA’s been very good about working with industry, both diagnostics and pharma.” FDA realizes that speed is important, he says. “They deserve some credit for working with industry to make that happen.”
Out of necessity, Dr. Walk says, the FDA has begun joint meetings with its drug side—Center for Drug Evaluation and Research and Center for Biologics Evaluation and Research—and its diagnostics side—Center for Devices and Radiological Health. Likewise, pharma and diagnostics have begun coordinating their meetings with the FDA.
In early March, Genentech’s Dr. Hampton and colleagues from Roche met with the FDA—CDER and CDRH—to talk about the company’s portfolio, its commitment to personalized health care, and its impact on diagnostics. They also told the FDA that (like General MacArthur) they shall return—“again and again and again, with our new medicines and proposals for companion diagnostics.” Dr. Hampton and his colleagues asked questions about biomarker cut-point thresholds, and multiple biomarkers, though he makes clear these were informal talks only, and not intended to extract guidelines. Nevertheless, he says, “I walked out of the room, as did many of my colleagues, believing that the FDA in general wants to be engaged in this. And they recognize the challenges that exist on both the pharma and the diagnostics sides, and having to deal with two different [FDA] divisions.”
Dr. Hampton is encouraged by the FDA’s apparent willingness to align the pharma and diagnostics regulatory arms. Further, he says, “They’re genuinely interested in enabling the use of diagnostics to identify patients who will or won’t gain benefit from drugs. There’s no question about that.” The agency also seems to realize that as multiplex diagnostics become more the norm than the exception, it will be challenging to interpret data and integrate companion diagnostics into phase 2 and phase 3 trials.
It should be obvious by now there is no beau ideal for the perfect pharma-diagnostics partnership. Dr. Hampton sees the value of being part of a larger company that has the technology platforms and expertise to develop companion diagnostics. But at the pharma-diagnostics table, there’s room for smaller players as well, he says.
Smaller players may benefit from being more agile. Companies such as Foundation Medicine, a molecular information technology (which last month announced its own collaboration with Array BioPharma), focus on generating relevant genetic data for cancers and some noncancers—no small matter in personalized health care. “It’s not just about developing the companion diagnostic,” Dr. Hampton says. Someone has to look at the diagnostics calls and decide what therapy is likely to be beneficial.
There should also be room for big-small partnerships. Roche Molecular has worked with startup and smaller pharma companies that have two or three drugs in the pipeline, all in a specific therapeutic area. “They may want more than one technology, but they’re very focused in one area,” Dr. Koch says.
It’s too early to say if one relationship will lead to deeper, ongoing ones, Dr. Koch says. Roche Molecular has been approached by pharma companies about umbrella agreements, which to him suggests that a good track record will help make a diagnostic maker a preferred partner for future relationships. The bigger the pharma company, he points out, the more they want global reach—and a diagnostics partner that can reach equally far.
That sounds reasonable to Abbott’s Dr. Becker. She predicts deeper relationships between pharma and diagnostics. “Rather than doing one-off partnerships for a single marker and a single therapeutic, it makes sense that once companies understand how best to work together, that they leverage those processes across multiple markers and multiple trials,” she says.
Pharma may also prefer a diagnostics partner that’s not bound to one type of test—especially since it may not be clear, at first, what the best test will be. Diagnostics companies are also getting that message. Susan Jewell, PhD, associate director, scientific affairs, Abbott Molecular, concedes that early on, her company looked at oncology primarily through the FISH lens, so to speak; now it’s drawing on multiple technologies (PCR, RT-PCR, sequencing, methylation) as well as its experience with different disease states, including infectious disease and genetics. “We’ve pulled in some of our R&D expertise from other areas and brought it into oncology,” Dr. Jewell says.
Andrews suggests there may be room for an added partner. He sees a triangular relationship down the road between pharma, a high-end reference lab (either a molecular pathology reference lab or an academic lab), and a diagnostics platform company. “You almost need all three to make this work,” Andrews says. The lab can develop the testing application, and the diagnostics company would take the LDT through the FDA as an IVD (proving, if nothing else, that medicine can almost keep pace with the Army when it comes to using abbreviations). “Pharma doesn’t have knowledge of the platforms, and how to service and sell them, and neither do the labs. But labs know how to manage patient accessions, they know how to run chain of custody, and they know how to run these applications. But they don’t understand the precision pharma needs, and they don’t know the FDA process that we go through.”
Here Andrews hits on the touchiest subject of companion diagnostics—laboratory-developed tests.
Andrews calls himself a “big fan” of LDTs, not surprising given his former stint at Clarient. But in his tenure there, he says, it was becoming obvious that as pharma develops its drugs, an LDT would not pass muster with the FDA.
Others echo that refrain. Dr. Koch effortlessly lists a handful of reasons why medicine needs LDTs, but halts at an unavoidable fact: There’s no information about how well they work clinically with a targeted therapy. The only way to prove that is in a clinical trial—and those haven’t happened.
A second strike against LDTs is pharma’s worldview. Access, in other words. Pharma wants comprehensive access to its drugs—and the only way to provide that is to move markers to platforms that can be used across the United States and around the world. An LDT won’t cut it, he says.
None of this helps labs, of course, which have plenty of reasons to want to use LDTs.
Many times the tests being developed as a companion diagnostic already exist as LDTs. “The FDA doesn’t like that,” says Dr. Leonard. The FDA, naturally, wants the FDA-approved test to be used, as do insurers. But that can create all sorts of problems for laboratories that are already using LDTs, and may have been doing so for years.
Musing on the matter, Dr. Leonard says she’s unaware of any LDT that has caused harm in a clinical setting. “And yet there’s this fear factor out there about the harm being done by laboratory-developed tests, or the potential harm,” she says. “But we’ve been doing laboratory-developed tests for more than 20 years now. And so I keep asking, Where’s the harm?”
She’d like to see a clear approval pathway for LDTs, but, with a sigh, says she doesn’t see one emerging anytime soon. It would have to come from test developers, both academic and the IVD industry; the FDA; the Centers for Medicare and Medicaid Services; third-party payers; clinicians; and laboratorians. “The list of stakeholders just goes on and on,” she says. “So I really don’t know what the answer is.”
For now, then, the key stakeholders are pharma and diagnostics firms. And the most personalized part of personalized medicine just might be the relationships the two are creating with each other.
Karen Titus is CAP TODAY contributing editor and co-managing editor.