Pharmaceutical officials behind the development of two of last year's personalized medicine success stories recently shared some of the lessons learned along the path to regulatory approval of their drugs and companion diagnostic products.
At Cambridge Healthtech Institute's Molecular Medicine Tri-Conference last week, officials from Roche/Genentech and Pfizer provided a behind-the-scenes view of the development and regulatory steps leading up to the US Food and Drug Administration's approval of their respective Rx/Dx combination products last August. Within a two-week span, the FDA approved Roche's melanoma drug Zelboraf (vemurafenib) alongside the company's internally developed companion test, the Cobas 4800 BRAF V600 Mutation Test; as well as Pfizer's non-small cell lung cancer treatment Xalkori (crizotinib) along with a companion test developed by Abbott Molecular called the Vysis ALK Break Apart Fluorescence In Situ Hybridization Probe Kit (PGx Reporter 9/7/2011).
FDA officials have pointed to these Rx/Dx combination products, which were approved well ahead of projected review timelines, as an example of how drug and diagnostic firms can work together with the agency to bring personalized medicine products to market quickly. Last month, Elizabeth Mansfield, who leads the Office of In Vitro Diagnostic Device Evaluation and Safety in FDA's Center for Devices and Radiological Health, cited the Zelboraf and Xalkori approvals as evidence that "codevelopment works," and noted that the agency "didn't really have a lot of major issues" with either approval (PGx Reporter 2/1/2012).
Pharma executives involved in the process appear to hold the same view. During the Tri-Conference session, three officials from each firm provided detailed case studies of the development of the drugs and diagnostics as well as the regulatory paths they followed to ensure rapid approval. There was a consensus that working with the FDA early in the process led to the success of these particular products, and all the officials who spoke noted that both CDRH and the Center for Drug Evaluation and Research were accommodating in terms of ensuring that all stakeholders were present for meetings.
The two firms pursued a number of similar strategies that helped speed the approval process. Both Pfizer and Roche gained orphan designation for their drugs, which provided them with a number of financial incentives. Both also used the fast-track rolling review designation for the drug and the modular pre-market approval process for the diagnostic, which allows the FDA to start reviewing paperwork as it was ready, further speeding the process. Both firms also adjusted their clinical trials as they went along in order to adapt to new data regarding their companion diagnostics.
There were a number of differences, however. While Pfizer needed to find an external partner to develop the ALK companion diagnostic for Xalkori, Roche relied on its own diagnostic arm to develop the BRAF test for Zelboraf. In addition, Pfizer was informed by the FDA during its Phase 3 trial that a PMA-approved companion test would be required in order to get the drug approved, leaving it little time to work with Abbott to get the test developed and approved. Roche, meantime, planned from the beginning to have a PMA-approved companion diagnostic for Zelboraf. In fact, the original developer of the drug, Plexxikon, sought out a partnership with Roche's diagnostic arm to develop the test before signing a co-development and marketing deal with the company's drug-development arm.
In addition, while the BRAF V600E mutation was first described in 2002 — well before the first clinical trials began for Zelboraf in 2007 — the EM4-ALK fusion gene was discovered a year after clinical trials began for crizotinib in 2006. Prior to that, the company had been investigating the use of the drug as an inhibitor of cMet and ALK.
Despite these differences, however, both firms were able to align their diagnostic and drug development timelines and negotiate a regulatory roadmap with the FDA in order to reach the finish line at the same time last year.
Zelboraf: Modified Statistical Analysis
Officials from Roche and Genentech explained that they needed to modify the statistical design of a clinical trial in order to keep up with evolving data on the response of BRAF mutation-positive patients to Zelboraf.
Chris Bowden, vice president of oncology clinical development at Genentech, said that based on early data, the company put together a "package" of overlapping trials that included an ongoing Phase I, a single-arm Phase II study with 122 patients called BRIM2 that began in the fall of 2009, and a randomized Phase III trial versus dacarbazine with 675 patients called BRIM3 that began in early 2010. All enrollees in both BRIM2 and BRIM3 harbored the BRAF V600E mutation, as measured by the Cobas test.
Bowden explained that the data from the Phase I trial, which showed an overall response rate of 56 percent — unheard of for metastatic melanoma — was published in the New England Journal of Medicine in August 2010, just as the company was enrolling patients for BRIM3. For that trial, patients randomized to the dacarbazine arm were not given the opportunity to cross over to the Zelboraf arm, but the fact that the positive Phase I data was now public — coupled with an article that ran in the New York Times questioning the ethics of randomized trials for drugs where there are no other treatment options — led to a "chorus" of patients and physicians asking for a change in the trial design.
Bowden said that the company was able to submit a revised statistical plan to the FDA by the fall of 2010 before the data from the Phase II trial was reported and before the last patient was enrolled in BRIM3 in December.
While the original primary endpoint for BRIM3 was overall survival, the revised plan, "based on evolving data from BRIM2," used two co-primary endpoints — overall survival and progression-free survival, which allowed Genentech to perform an interim analysis based on a cutoff of 98 patients in January — much earlier than would have been possible under the original plan, he said.
"We got enough information so that we didn't need to randomize any more patients into dacarbazine," he said. "That's the key point — we ended the trial early and avoided randomizing patients to a therapy that offered no hope for them."
Linda Burdette, director of drug regulatory affairs at Roche, added that the company was able to clear the new protocol for BRIM3 across 100 global sites in eight weeks. In its discussions with FDA regarding the modified statistical design, the agency was most concerned about "consistency between the Phase I and Phase II trials," she said.
Overall, Burdette noted, Roche received valuable feedback from FDA across the entire development process, though she noted that there was sometimes disagreement between CDRH and CDER.
For example, she said that based on early Phase I results in which 11 mutation-positive patients out of a total of 16 showed an objective response to the drug, the company asked the agency if it would need to collect data in the marker-negative population in order to show that there was a difference in response between marker-positive and marker-negative subjects.
"CDRH had a good point that obviously you need the marker-negative patients to validate the test, but they also were very quick to point out that if we didn't have marker-negative data, we would not be able to make any predictive claims on the label," she said. CDER, on the other hand, "said that we had insufficient safety data to support the benefit risk assessment in the marker-negative patients at this point, so from this point forward our clinical development program was solely focused on marker-positive patients."
She added that Roche also received "invaluable feedback from CDRH on how to submit a modular PMA, which was what was required in order for CDRH and CDER to stay in sync in terms of having joint approval of the drug and diagnostic."
This advice "was very helpful to us [in order] to go back to our [development] teams who had said, 'Oh no problem. They’ll go ahead and approve the drug and let the diagnostic come along later.' So it was a clear message to say, 'No, we have to stay in sync.'"
CDRH "also told us that it was important to have the commercial version of the test available for our clinical studies, and if, by chance, we had to modify that test, then we would have to do concordance studies to demonstrate the performance characteristics."
On the diagnostic development front, Laura Hashimoto of Roche Molecular Systems' genomics and oncology group said that the Cobas BRAF test was validated in more than 25 verification studies in more than 600 specimens.
The company has touted the reproducibility of its test but faces a commercialization challenge in a marketplace that already bears a number of lab-developed BRAF tests (PGx Reporter 9/28/2011). Even though Zelboraf's label states that an "FDA-approved" test is required for determining which patients will respond, it is likely that many labs will continue to run their home-brews until the FDA issues firm guidelines regarding its plans to regulate such tests.
In response to a question from a conference attendee regarding the prevalence of BRAF LDTs, Hashimoto acknowledged that some labs are likely still using them but said she believes that "in the next few years there will be a shift toward emphasizing the use of FDA-approved tests."
Xalkori: Homing in on a Targeted Therapy
Hakan Sakul, executive director and head of diagnostics at Pfizer, said he views FDA approval for companion tests as an incentive for diagnostic developers.
"Under the typical model, you develop the diagnostic product, put it on the market, and then after intense competition for a few years you begin to see a drop in income," he said. "In this case, when you have a diagnostic label referring to a drug, and [required] for the drug to be prescribed, you're actually riding the wave of the drug throughout its patent life. Of course there will be competition, but I believe it does represent a bit of a different model for diagnostic companies than the one that has been traditionally available."
Sakul noted that in late 2008, "we found out through our interactions with the FDA that we needed a PMA-approved diagnostic assay to go with our drug. That was a very clear message. … Once we knew that, and since we don't have a diagnostics division to make one for us, we went out and searched to find the company that was in the best situation to develop the diagnostic test in time for our product."
One factor in picking Abbott as a diagnostic partner was the company's marketing reach, he said. "Global availability of diagnostics is very important because we don't just sell in the US and EU, but anywhere patients need it," Sakul said. "We always look at what capabilities companies have. They may have the best technology, and it would be perfectly fine to work with them, but we also look at how then they will go out and market and make the test globally available."
He added that Pfizer had to work "very closely and on a very short time frame with Abbott Molecular, and they were very good about that."
Erling Thor Donnelly, director of worldwide regulatory strategy for Pfizer, estimated that the company's accelerated approval pathway "cut 18 to 24 months off the standard approval timeline, which forced Abbott to shift resources rapidly to keep up with the progress we were making on the drug side." He added that Abbott did a "tremendous job" in condensing its development timeline to meet Pfizer's requirements.
[ pagebreak ]
Unlike Zelboraf, which was developed specifically to inhibit the mutant BRAF protein, the path that ultimately matched crizotinib with its ALK fusion companion test was not so clear cut. Keith Wilner, senior director of oncology clinical development at Pfizer, noted that at the time the first clinical studies of the drug began in 2006, the company was considering it as a treatment for gastrocarcinoma patients with cMet-positive tumors, since it was a known inhibitor of the c-Met/hepatocyte growth factor receptor tyrosine kinase. The compound was also known to inhibit ALK fusions so the company was also considering it as a treatment for anaplastic large-cell lymphoma, which had been associated with ALK mutations.
When a paper was published in Nature in 2007 showing that the EML4-ALK fusion was a strong oncogenic driver for non small-cell lung cancer, the company shifted gears and amended its trial protocol to enrich for patients who were positive for the ALK translocation. "We tried to make it as broad as we could because there could have been other ALK fusion proteins out there or another tumor type with the ALK fusion," Wilner said.
One new patient — a 31-year-old male with EML4-ALK NSCLC — responded well to the drug, so the company again modified its trial design to concentrate on ALK-positive NSCLC and added two new clinical trial sites. "We were still interested in cMet-positive patients, but for gastrocarcinoma we had not been too successful. We kept screening and screening and couldn't find amplified patients in gastric cancer," Wilner said. He noted that it's likely that the percentage of cMet-positive gastric cancer was probably much lower than the 20 percent the company had been led to believe, and was probably closer to 2 percent.
ALK mutations aren't common either, however. Only about 5 percent of NSCLC cancer patients are believed to be positive for the gene fusion, "so the difficulty was finding these patients," he said. "We had to screen thousands of patients in enrolling our trials."
While the company had the Abbott diagnostic in hand by the time it began enrolling patients in late-stage trials, it used a demographic shortcut to help winnow through the large numbers required to identify eligible subjects. Wilner noted that the demographics for ALK-positive NSCLC are very different than EGFR-positive or wild-type NSCLC, with patients more likely to be younger, non-smokers or never smokers, and with a histology of adenocarcinoma.
Pfizer asked its clinical trial sites to use these demographics as an initial screen before running the Abbott test. "By doing this, we had a positive hit rate of 21 percent," Wilner said. "We expected a rate of 10 percent to 15 percent."
He noted, however, that it's possible that the number was higher because some of the sites started to do their own pre-screening with the Abbott test. "We didn't want it to happen, but it did," he said.
Pfizer is still working on post-marketing studies in order to convert its accelerated approval to full approval status. One requirement is that it has to look at the response to crizotinib in ALK-negative NSCLC — namely to see if there is a response in patients who are ALK-negative, but positive for cMet or a recently discovered mutation in ROS that is associated with only about 1.7 percent of NSCLC cases.
Donnelly said that the FDA "stressed that we need to get that data." While Pfizer submitted "some" ALK-negative data in its new drug application, "continuous questions came up," he said, so "we eventually decided it was best to handle it as a post-marketing commitment."
Lessons Learned, Unanswered Questions
Pfizer and Roche/Genentech execs were unanimous in their recommendation to meet early with the FDA — a "no brainer," according to Pfizer's Sakul. "Everybody says so. I hope everybody does so," he said.
He added that initiating conversations with potential diagnostic partners is also important, but the timing can be tricky. "If you do it too early, you may end up making a lot of extra investments, start too late and you may regret it and never do it again, so finding that sweet spot is I think going to continue to be difficult, especially for companies who do not have a diagnostics division."
With regard to diagnostic partners, "fostering an open and true partnership environment is really important," Sakul noted. "If you have a fee-for-service deal, that's exactly what you get. You get the service for which you pay. But if you engage as a partner, you can move mountains. We did engage with Abbott as a partner and there were some pretty impressive timelines."
Pfizer's Donnelly agreed that "early and frequent" interactions with FDA are critical to a successful drug/diagnostic development program, but noted that "the onus is really on the pharma company to ensure that CDER, CDRH, and your diagnostic partner are present at every meeting."
FDA is "trying to put some processes in place to ensure that CDRH folks are included in all meetings that touch upon a diagnostic, but it really is up to the pharma company to request that to make sure you get sound advice along the way," he said.
Furthermore, he said that the agency is not open to the LDT route when it comes to companion diagnostics. "We heard the message very clearly and loudly at the very beginning of our development program from the FDA," he said. "They said point-blank that the NDA would not be approved unless we had a commercialize-able diagnostic that was PMA-approved at the same time. We had to have a diagnostic that would pick the patient population where we consistently have a positive benefit risk assessment, so I think that's definitely the take-home message and I think industry is getting that."
Roche's Burdette echoed many of these recommendations, and also stressed the importance of informing trial sites that they cannot use their own tests to select patients for trials. "Many investigators are very familiar with the local test and they're very comfortable with the outcomes, so they don't understand why they have to use your test and only your test," she said. "So, you really have to educate them on the regulatory requirements and how, although they are doing this with the best of intentions, they actually could negatively hurt the trial and ultimately the approval of both the drug and the diagnostic."
Burdette also raised a few issues that the agency will likely need to address in future guidance.
"What do you do if the approval of the drug precedes the approval of the companion diagnostic? How is that reflected in the product labeling?" she asked. Another issue will come up when developers seek to use these drugs in combination therapies, she said, noting that it's not clear whether the FDA would require a new PMA for that indication.
Have topics you'd like to see covered in Pharmacogenomics Reporter? Contact the editor at btoner [at] genomeweb [.] com.