Part 1 of 2 in a feature series.Read the second article here.
The head of pharmacogenomics at Takeda Pharmaceuticals, part of the largest drug maker in Asia, recently stood before a crowd of life sciences leaders in Northern California and told them that pharmaceutical companies have been doing it all wrong when it comes to advancing personalizing therapies. If the aim is to deliver "the right drug for the right patient," as the mantra goes, then development shouldn't start with the drug, Eric Lai said at the Personalized Medicine World Conference in January. It should begin with the patient.
Instead of mechanically following the well-worn three-phase drug development paradigm, where biomarkers, pharmacogenomics, and companion diagnostics enter the picture too late in the game, usually when the treatment doesn't appear to work so well in a broader patient population, Lai advised industry colleagues to first identify the molecularly defined patient groups in need of effective treatments and work backwards.
He recommended his colleagues make more use of the big databases in which various groups are collecting an array of information on large cohorts. "You want to identify subgroups of patients from these large databases and from there, you're going to take those subtypes and see if [they] correlate with phenotype in your clinical trial," Lai said. "You have to start from the beginning and identify patients, not just the target."
To illustrate his point, he cited the fact that 35 percent of rheumatoid arthritis patients don't respond to anti-TNF therapies or methotrexate. "What are the characteristics [of these patients], what are the targets for those patients? From there you think about what are your commercial needs. And now you identify the target," he said.
At a life sciences meeting minutes from Google headquarters, Lai was preaching to the choir. After all, it's not a personalized medicine conference until someone laments that the traditional "one-size fits all" drug-making process is broken and then announces that low-cost genome sequencing will save the day. In a twist, Lai said, "Even if we have the technology to do personalized medicine … we don't have enough drugs to give you personalized medicine. I don't care how good your technology is."
Although pharmacogenomic strategies are making a tremendous impact in oncology, in Lai's view, the field of precision medicine won't really take off until it can address the maladies that affect more, if not most, of us. "In order to do this you have to improve the drug development process … and you have to think about it very differently."
In his stark assessment of the challenges for pharmaceutical companies developing personalized medicines, Lai is not alone. There are others in the drug industry who are also speaking out about much needed changes that must happen before individualized care is the care everyone receives, not just those with cancer or rare monogenic diseases. Kenneth Emancipator, director of companion diagnostics at Merck, has been making the rounds at various medical conferences, and making quite an impression, with his talk, entitled "It's not just about the drug!"
It's a talk in which Emancipator is relentless with the message that drug makers need to think from the get go not just about their compound — which is what they're used to — but also about the test that will determine which patients will get their drug. Pharma companies now need to think about the technology of the companion test, how to validate the test, how to get it through regulatory approval, how to market the test, how to educate doctors about the drug and the test, and how to gain reimbursement ... not just for the drug, but also for the test. Because if no one pays for the test, then there is no test, and no one gets the drug.
Then, there are complementary diagnostics to consider. These tests aren't directly tied to making treatment decisions with a particular drug, but they can improve patient diagnosis, gauge which have worsening disease, and which can forgo aggressive, costly interventions. In this way, complementary tests can certainly impact the dynamics of a disease market a pharmaceutical company is selling its drug in.
When a therapy is in preclinical development, there may seem to be no reason to start thinking about a companion diagnostic. In Emancipator's view, however, drug developers can't start early enough when it comes to thinking about a companion testing strategy for an agent. The fact that top drug makers have all hired diagnostic executives like Emancipator to guide their personalized medicine programs certainly signals some change in industry thinking. Still, pharma companies aren't used to developing and marketing their drugs in the context of a diagnostic and they often aren't prepared to face the complexity that a test adds into the process, said Peter Keeling, CEO of consulting firm Diaceutics.
"There's still risk management going on within pharma. They're still trying to develop drugs in all comers first but studying biomarkers within that," Keeling told PGx Reporter. Diaceutics specializes in helping drug makers plan out their personalized medicine programs and has overseen more than 50 such projects over the last eight years, involving drugs such as GlaxoSmithKline's HIV treatment Ziagen (abacavir), Pfizer's HIV drug Selzentry (maraviroc), and AstraZeneca's non-small cell lung cancer agent Iressa (gefitinib).
"Despite the fact that you have some very successful targeted therapies on the market, which are billion-dollar brands, there is still the perception that you are subsetting your market by not going with an all-comers strategy" where the drug is studied in a molecularly-undifferentiated patient group, Keeling said. "That's increasing the complexity and it's increasing the risks."
Of the projects Diaceutics has had input in, 60 percent have been drugs that are being co-developed with a companion test, and 40 percent are enabled by complementary tests for disease screening, monitoring, and prognosis. The company estimates that 20 percent of the Rx/Dx projects it has worked on have already come to market, approximately 50 percent are still under development, and 30 percent of the projects have been discontinued because the biological hypothesis didn't pan out.
The global pharmaceutical market is estimated to be worth $300 billion. Personalized medicine revenues currently total $20 billion, Takeda's Lai estimated. "Personalized medicine is definitely growing in healthcare, but it's still a small number," he said at the conference.
'It's not a lack of wanting to'
Drug developers' unwillingness to cut into profits with a molecularly guided treatment strategy is an oft-cited reason for the disappointing pace of personalized treatment development. There has been much handwringing about the fact that more than a decade after the Human Genome Project, the majority of medicines are not molecularly targeted. There are many reasons for why this is the case. The main one being that improvements in our ability to read a genomic sequence haven't necessarily enabled us to decipher what the code says about disease.
Lai, who a decade earlier led the team of GSK researchers that discovered the association between the HLA-B*5701 allele and hypersensitivity reactions with HIV therapy abacavir, said that as of last year, there are more than 10 million known SNPs. Since the first sequencing of the human genome in 2003, thousands of human genomes have been fully sequenced.
But there are only some 120 drugs (many of them generics) with pharmacogenetic information in their US Food and Drug Administration-approved labels. Lai recalled a paper some years ago that claimed that in order to do a genome-wide association study researchers needed knowledge of 3 million SNPs. "And everyone looked at that guy [who wrote the paper] and said, 'You've got to be out of your mind. There are not that many SNPs in the human genome,'" Lai recalled. "Well, we were wrong."
Given the complexity of the human genome, the first personalized medicine success stories — Roche's Herceptin (trastuzumab) and Novartis' Gleevec (imatinib) — occurred in cancer, because researchers could home in on the tumor (or cancerous blood or bone marrow cells in the case of leukemia) and identify the genomic abnormalities driving the disease. In other conditions, like dementia or diabetes, there is no tumor in which researchers can look for biomarkers, and the disease-related markers that they have identified to date aren't so clear cut. So, for drug makers, identifying the right target has been more challenging.
While DNA information is just a piece of the puzzle for developing personalized therapies for common, complex ailments, most major pharmaceutical companies are now using genomics tools in their early research and discovery efforts. For example, J&J last year launched a project to sequence the genomes of 450 rheumatoid arthritis patients who were involved in a clinical trial of its drug Simponi (golimumab). Through this effort, J&J is hoping to discover genes that correlate with disease predisposition, as well as new drug targets.
Within the immunology division at J&J's Janssen Pharmaceuticals, drug development teams begin thinking about biomarkers and diagnostic tools very early in the process with the goal of one day advancing precision therapies, according to Mark Curran. "We do that because we fundamentally recognize that no one drug works for all patients, and we only want to treat the patients who are going to respond and have a good efficacy profile," said Curran, VP of systems pharmacology and biomarkers in the immunology therapeutics area at Janssen. "Unfortunately, that's an aspirational goal right now, because we don't understand the science well enough to make it reality in all cases ... It's just difficult science. It's not a lack of wanting to."
In cancer, now there are a handful of fresh precision medicine success stories beyond Gleevec and Herceptin, with the launch of Roche's BRAF-mutated melanoma drug Zelboraf (vemurafenib), Pfizer's Xalkori (crizotinib) for ALK-positive non-small cell lung cancer, and next-generation HER2-positive breast cancer drugs, Roche's Perjeta (pertuzumab) and Kadcyla (ado-trastuzumab emtansine). But even in oncology, where there are the most examples of individualized agents, the science doesn't always pan out.
Even though retrospective studies had suggested that pancreatic cancer patients with low hENT1 expression didn't respond well to gemcitabine, Clovis Oncology found no difference in patients' survival between high and low hENT1 expressing patients in a Phase II trial of the agent CO-101. Clovis researchers had hoped that its drug, a gemcitabine-lipid conjugate, would enter tumor cells through passive diffusion and as such the drug's efficacy wouldn't be hindered by hENT1 expression. But hENT1-low patients did not respond any better to CO-101 than they did to gemcitabine in Clovis' study.
In glioblastoma, the average survival time since diagnosis has hovered around one year despite efforts to study various combination treatments. Even attempts to stave off this aggressive disease with molecularly targeted strategies have been fruitless. As part of the development program for the integrin inhibitor cilengitide, Merck KGaA had prospectively tested close to 3,500 patients for methylated MGMT under a collaboration with diagnostics firm MDxHealth. But at a major cancer conference last year, study leader Roger Stupp, president-elect of the European Organization for Research and Treatment of Cancer, announced that his team couldn't find a single subset of patients with a response that suggested activity with this agent in a Phase III trial.
Methylated MGMT has shown to be a prognostic indicator of positive outcomes in glioblastoma patients, and also has shown to be predictive in identifying best responders to the chemotherapeutic temozolomide. And while the marker is being explored as part of most major glioblastoma drug trials, it didn't prove to be a predictive marker of response for cilengitide.
Merck KGaA pursued a molecularly defined hypothesis in the Phase III trial based on the available evidence at the time, but also because the drug appeared to have limited activity in an undifferentiated glioblastoma patient cohort. In a Phase II study involving recurrent glioblastoma patients, Merck KGaA had reported in 2008 that cilengitide had “modest” antitumor activity. The lead author of the Phase II trial, Dana-Farber Cancer Institute's David Reardon, wrote in a 2012 review of cilengitide in Genes & Cancer that a subsequent single-arm Phase II study of around 50 patients showed that those with MGMT-methylated tumors achieved a longer median progression-free and overall survival rate.
Roche subsidiary Genentech has also looked for biomarkers of response to Avastin (bevacizumab) in newly diagnosed glioblastoma patients in a Phase III trial, including VEGF-A and VEGFR-2, but these weren't found to predict improved patient survival. After looking at data from these high-profile failures in glioblastoma, experts believe that there are key differences between the nature of the cancer in newly diagnosed and recurrent glioblastoma that yields varying treatment responses in patients, suggesting that developing a precision treatment will be a much more complex proposition than identifying a disease-associated marker. Some experts have even suggested that when it comes to anti-angiogenic drugs like Avastin, there may not be clear cut predictive markers.
Roche also recently announced it would stop a late-stage trial of onartuzumab in combination with Tarceva (erlotinib) as a NSCLC treatment for patients with abnormal MET signaling. In the Phase III METLung trial, an independent data monitoring committee found that the onartuzumab/Tarceva combination didn't have "meaningful efficacy" over Tarceva in previously treated, MET-positive NSCLC patients. When the cell surface MET protein binds to another protein, called HGF, it causes MET to dimerise and signal the growth and spread of cells ─ behavior that could promote resistance to EGFR inhibiting drugs, studies have suggested. Onartuzumab is a monoclonal antibody that is designed to target the MET receptor and halt this aberrant signaling.
Roche pursued the Phase III strategy after a Phase II study showed that the drug in combination with Tarceva tripled median survival in patients with high MET expression compared to those receiving just Tarceva. Earlier results "were consistent with the scientific hypothesis of the interaction of the MET and EGFR pathways that drives the growth of lung cancer tumors," a Genentech spokesperson recently explained. "The decision and the design of the [METLung] study were informed by discussions with leading oncologists."
Given these examples, critics will say that pharmaceutical companies are depending too heavily on data from small trials and retrospective analysis, which can be misleading. Back at the personalized medicine conference, Takeda's Lai illustrated how using his "backwards" development strategy, drug makers could potentially avoid these pitfalls. His team identified molecularly defined subsets of patients from within a large database and ran simulations to see if the groups would remain if the patient cohort got smaller and smaller. By the time researchers reached the size of a typical Phase II trial, the subsets could no longer be detected.
"So, think about … when you formed these clusters in Phase II using small sample sizes and said, 'This is it, I'm going to do a Phase III trial,'" Lai said. "Those are not really stable clusters. That's why we should do it the opposite way. We should take big dataset, form stable clusters … and apply it back to our clinical trial."
Risky business
If the pharma company manages to get the science right, then come the perils of launching a personalized drug – traditionally coupled with a companion test that picks out best responders – on the market. The regulatory path for getting a drug and test combination product approved for commercial use is still under construction. And the reimbursement environment for molecular tests is so dim that in several countries outside the US, drug developers are having to pay for testing just so their drugs can reach the patients in that market. (See part two of this article next week for more on this.)
These challenges make investing in personalized medicine risky. "Often drug companies' commercial arms want a guarantee that if the company takes a personalized medicine strategy that it'll be better than if they pursued an all comers strategy," Diaceutics' Keeling said. Although there are a number of personalized drugs on the market now, Herceptin and Gleevec, available for over a decade, remain the primary examples of molecularly targeted treatments that are also blockbusters.
The FDA in recent years has approved a number of personalized treatments, notably Kalydeco (ivacaftor) for cystic fibrosis patients with various mutations in the CFTR gene; Tarceva (erlotinib) and Gilotrif (afatinib) for first-line treatment of metastatic NSCLC patients with EGFR mutations; two new HER2 treatments, Perjeta and Kadcyla; as well as new options for BRAF-mutated melanoma, GlaxoSmithKline's Tafinlar (dabrafenib) and Mekinist (trametinib). Chronic myeloid leukemia patients, whose disease is characterized by BCR-ABL gene fusions, have options beyond Gleevec, with Tasigna (nilotinib), Bosulif (bosutinib), and Sprycel (dasatinib).
Many of these newer drugs are nowhere near blockbuster status. But as they mature on the market, some of them stand to bring in significantly more revenues than they currently generate, since patients receiving them have good outcomes and live longer than they would on a non-targeted agent. "These are … drugs that people take for a long time," David Carbone, director of the James Thoracic Center at Ohio State University, told PGx Reporter. At OSU, doctors routinely test lung cancer patients for a panel of predictive markers before administering any treatment. "Even though the number of patients [who can receive a personalized drug] is low, if you're successful at treating them, they take it for a long time," Carbone added. "So, there's a big market."
In a recently published paper, Carbone and others identified that a mutation in the ARAF gene, estimated to be present in 1 percent of NSCLC patients, may predict response to Nexavar (sorafenib). This is a drug that Bayer/Onyx was investigating in lung cancer but doesn't appear to be any longer after it failed to extend survival compared to placebo in two trials. Using whole-genome and RNA sequencing, Carbone and colleagues described finding the ARAF S214C mutation in a single "super-responder" in a Nexavar lung cancer trial who remained progression-free and asymptomatic for five years while on the drug. They then worked backwards to identify cancer patients in other cohorts with mutations in ARAF and in a related gene. They also conducted in vitro analysis that suggested that treating cells expressing these mutations with Nexavar or the MEK inhibiting melanoma therapy Mekinist could have an anti-cancer effect.
Based on this analysis, they proposed that drug developers pay closer attention to the rare study participant who has an unusually good response to a drug that wasn't effective in a larger patient cohort; deeply analyze the genomic and molecular characteristics of this one patient to find out why he or she responded so well; and try to find others who are similar. This, they said, could be an effective way of identifying new molecularly defined patient groups and treatment options for them.
In proposing Mekinist as a possible option in lung cancer patients with ARAF mutations, the work of Carbone's group also highlights the value of looking for additional subgroups that can respond well to an already marketed PGx drug. In this way, pharma companies can hope to grow the market, as well as the revenue potential, for these agents. "There are a [handful of] different companies making second-generation ALK-targeting drugs" after Pfizer's Xalkori, Carbone observed. "So, obviously [drug makers] think that market is valuable." Approximately 5 percent of NSCLC patients have ALK mutations in their tumors. However, Xalkori will likely not be limited to only ALK-positive NSCLC patients, since studies have found it is a promising option for those with even rarer mutations in ROS1 and RET.
The market dynamics for every molecularly targeted drug is unique, and some will never be blockbusters, or even nichebusters – the optimistic term for profitable PGx drugs. For example, the pharmacogenetic indication for the anti-coagulant Plavix (clopidogrel) came as the drug was getting close to losing patent protection and newer anti-platelets were entering the market. Bristol-Myers Squibb and Sanofi never supported the development of a companion test that gauged CYP2C19 polymorphisms associated with variable responses to the agent. Moreover, they sponsored studies that has become part of a body of literature shedding doubt on the impact of CYP2C19 markers on Plavix response. This has kept doctors from using PGx testing to administer the drug. After losing patent protection in 2012, Plavix last year saw its worldwide revenues plummet by 90 percent due to generic competition, from $2.5 billion in 2012 to $258 million in 2013.
When his pharma clients are reluctant to take a personalized medicine strategy over developing a drug in all-comers, Keeling tries to explain that it is possible to get a good, if not better, return on investment on a molecularly-targeted drug compared to a treatment for an undifferentiated population. But achieving this will mean dealing with market dynamics pharma is not used to.
"The problem is, there is a complexity to doing it well," Keeling said. "And the minute I use that word, 'complexity,' their eyes roll back and they want to go back to doing what they know."