By John Russell
June 11, 2009 | Very soon, FDA will restart its End of Phase 2A (EOP2A) meeting program during which sponsors and FDA collaborate often using modeling and simulation (M&S) – pharmacometrics (PM) in FDA parlance – to analyze current data, refine clinical trial design and to inform industry’s critical go/no-go decisions on projects. Significantly, the reason the pilot EOP2A was halted in 2007 wasn’t so much to absorb its early lessons, though that was important; it was also because rising demand for PM reviewers on NDA (new drug application) approval decisions meant there simply weren’t enough PM resources to go around.
In a bit more than a decade, pharmacometrics at FDA has grown from a small effort that was regarded with early skepticism and whose activities were mainly restricted to population PK questions into one of the agency’s most promising priorities. A small cadre of champions, some immediately recognizable (e.g. Dr. Lawrence Lesko, Director, Office of Clinical Pharmacology) and others toiling quietly in the trenches, have worked steadily inside FDA to prove PM’s value and to coax the agency to embrace it.
Happily, their efforts are paying off as was decisively shown in two surveys* of early PM efforts inside FDA. Of roughly 244 NDAs submitted between 2000 and 2004 in cardio-renal, oncology, and neuropharmacology divisions, 42 included pharmacometrics review. Of those reviewed for drug approval (26), PM was deemed pivotal to the decision in 54 percent (14) and supportive in 12 (46%). The results were similar for labeling with PM considered pivotal in 57 percent (21) and supportive in 30 percent (11). In only five labeling cases did PM play no role. A second survey of 31 NDAs covering February 2005 to June 2006 showed similar results.
The message that this PM stuff works and brings value is starting to sink in.
Leading the modeling and simulation charge at FDA now is Joga Gobburu, a ten-year agency veteran, who heads the 17-person PM group embedded in CDER review divisions. Roughly a year ago, Gobburu presented **Pharmacometrics 2020, an ambitious manifesto in which he suggested PM will help industry and FDA reduce late phase attrition from 50 percent today to 15 percent by 2015 and to 10% by 2020. What’s more, he says the practice of pharmacometrics and the restart of the EOP2A program will help transform FDA’s working relationship with sponsors.
“Look at other meetings we have such as the End of Phase 2 meeting,” says Gobburu. “The sponsor says ‘this is where we’re going; this is our registration trials design; these are the end points; and we have these ten questions for you to answer.’ We look at the ten regulatory questions and we provide qualitative responses usually yes or no and whatnot. Another similar kind of meeting is what’s called the triple zero protocol meeting, the first time in human study. They say, ‘we want to administer the compound to humans, do you have any objections?’ We just say yes or no, often because of relatively limited available data to delve more into details.” These are important meetings, but they are primarily regulatory in nature.
“The EOP2A meeting is not like that. The theme is to let us think through trial design together,” he says. “What is the best trial design, can we think about alternative designs? The focus is more scientific and the meeting outcomes are not binding; so we want to take the binding thing off the table because we wanted the scientists to talk to each other.”
In Gobburu’s 2020 roadmap, he asks who will be PM’s customers and what will its products be. “My personal opinion is the customers of this type of technology and science are going to change. Right now we are talking about drug developers, but the same information about exposure/response, dose selection, etc. could be useful to many others. The recipients of this technology won’t just be FDA. It is possible we’ll interact, for example, with CMS (Centers for Medicare and Medicare) or CMS forms its own PM group or the VA (Veterans Administration) forms one or Kaiser Permanente could form one.”
It’s good to have stretch goals and the agency seems to agree. A planned move to elevate the PM group to division status within CDER is currently on hold while the Obama administration reviews all such proposed organizational changes - most observers expect the change will go through. Gobburu is actively hiring at a time when most are paring staff back. PM’s computational resources have been expanded - reviewers used to do all their work on personal laptops but recently a 48-cpu computer cluster has been added to their arsenal.
This a far cry from PM’s modest start at FDA in the mid 1990s. Then, the Office of Clinical Pharmacology (OCP) had three evaluation divisions aligned with various therapeutics areas. Each had one or two ‘pharmacometricians.’ By 2000, preliminary success and the desire to make PM practices more consistent prompted FDA to centralize PM into single group within OCP.
Gobburu divides PM’s journey into three decade-long stretches. From 1990 to 2000, he says, PM growth was marked by technology development, early success, and the evolution of regulatory policy. 2000 to 2010, is being dominated by growing influence within and outside the agency and the development of more formal organizational structures. 2010 to 2020 will be about standardizing processes, increasing staff size, and expanding PM’s scope.
“Very early this field was an uncharted territory, nobody could even think of using modeling information to make a key agency decision,” he says. The first population-PK guidance was issued in 1997 and it dominated thinking until about 2002 when the exposure/response guidance was issued. “With the exposure-response guidance our focus was more on new drug applications (NDAs) and that really started getting people’s attention. We started to get into dose selection, trial design and approval-related decisions.”
Indeed, from 1995 to 2000, labeling and research was virtually the only work PM did. By 2005 the situation was quite different. The volume of labeling work had continued its impressive growth, but it now accounted for only 50 percent of the total workload as PM reviews also dealt with drug approval decisions and entered the trial design and approval policy arena. By 2007, work on approvals reached parity with labeling at roughly 30 percent each, and of course the total volume of PM work was still growing. “We got into opportunities where modeling played the total role in making a decision of whether to approve the drug or not, substantially judging its effectiveness or not.”
It should be noted that the 2005 International Conference on Harmonization issued guidance (E14 Variance) for evaluating the QT risk of new compounds. FDA responded by establishing a group inside CDER to review all QT-related submissions. That group is IRT (Interdisciplinary Review Team) for QT. Gobburu’s PM group drew primary responsibility from the clinical pharmacology perspective to review the QT studies, causing a spike in PM’s workload and adding to the scope of PM reviews.
“Once PM was established to constitute a big success by most people in the Office of Clinical Pharmacology it branched out and that’s when the centralized group was formed. We started having what are called scoping meetings. Every time there is a new NDA, we have an interdisciplinary scoping meeting where we go through the submission and come up with the key questions. If the key questions happen to be something that can be answered by a PM analysis, even if the sponsor has not performed such analysis, then we assign a PM reviewer to the submission.”
Gobburu says the experience of participating on NDA reviews clearly identified poor dose selection and poor trial design as major problems. “These were two of the main determinants of the success of a particular clinical trial and there were several trials where a simple PM analysis could have identified a better way to design the study. So we introduced the concept of End of Phase 2A (EOP2A) meetings.”
FDA rolled out the pilot program in 2004. There were about a dozen, he says, and the main focus was on dose selection and trial design for late phase trials. Prior knowledge and heavy use of modeling and simulation was employed to design future trials.
At least one EOP2A participant has had its drug approved. “There was an issue about the dose selection for a compound under development. The sponsor was heading in one direction about the dose selection. The studied doses failed to achieve the clinical significance. Then they came to the EOP2A meeting and we worked with them on the possibility of an alternative dosing which would enhance the success of the trial. The sponsor embraced that. They tested that dose in the next trial and bingo, it passed. And we approved it.”
“The beauty is that since our reviewers are very familiar - through the EOP2A meeting - with the drug development program, it makes it much easier for us to review the NDA.”
During the initial pilot, FDA often provided its own models, developed using historical FDA data, for sponsors to use. When program restarts, the intent is for sponsors to take more responsibility for model development. For FDA, the challenge is workload.
“EOP2A meetings have very tight deadlines. Our NDA timeline on average has 6 to 10 months; on EOP2A meetings we hardly have 3 to 4 weeks for the person who is doing the analysis. So it is very intense. We could not cope with the resources to manage both of them and could not sustain it any longer. We stopped accepting them until we got more resources.”
The results of the pilot are well summarized in an FDA paper, Leveraging Prior Quantitative Knowledge to Guide Drug Development Decisions and Regulatory Science Recommendations: Impact of FDA Pharmacometrics During 2004-2006, J. Clin. Pharmacol. 2008; 48; 146. This paper was accompanied by an excellent perspective, Communicating With the FDA: The "Third Rail" of a New Model for Drug Development, J. Clin. Pharmacol. 2008; 48; 144, Donald R. Stanski and John J. Orloff.
Final EOP2A Guidance is still pending, but Gobburu says FDA will start accepting new requests soon.
Gobburu says, “The key success basis for our group is not really the technology because the technology was taken care of before us. The academic institutions mostly and the commercial vendors made sure products were out there and were widely available to do most of our analysis. We continue to use those tools, but our focus was more on generating success stories and not so much about the technology. Our focus was more on how, the concept of PM added value to decision-making for the whole enterprise,” says Gobburu.
That said there are technology issues now. “There is a burning need for us to streamline the way we manage our knowledge. That’s something we are focusing a lot on here. For example, just imagine you have the next EOP2A meeting for an anti-diabetic compound, and yes we have a plethora of data here, but it is not easy to access that, and I don’t think it is much different for the industry. We need smarter tools to bring previous data in your preferred format to you to make decisions in an efficient manner. Right now, the bottleneck is to get the data in the right format; so there is more think time.”
Both industry and FDA need to come up with better tools to manage knowledge, he says, and his group has two fulltime staff devoted to technology evaluation and development. Armed with a bigger cluster, the PM group is also creating SQL databases of trials in those areas which “we think are of high priority to us. We cannot do it for every trial. Take, for example the QT workload. The immediate question [when work surged] was how can make the learnings from these 100s of QT reviews more efficient.”
One staff member developed an automated tool which streamlined access to the input data, the modeling, the output data and the reports, “so by the click of a button” the analysis, standard graphs, standard tables and report are generated. “We cut down a review which takes about 4-5 days to probably 15 minutes. By doing that, not only are we increasing the efficiency of the review but also we are streamlining the way we store the data. Now I can go back, or any reviewer can go back, and say hmm, what are the QT changes in the placebo arms for the last 50 trials and they can make a table or a chart at the click of a button,” he says.
In any event FDA has bigger plans to create a scientific computational center and the PM group will no doubt plug into that effort and benefit from its efforts. Gobburu expects firmer budget and plans for that project to be announced soon.
Another challenge is the lack of trained pharmacometricians. There are not many schools in the U.S. which train personnel “we can take readily off the shelf. You cannot just take any PhD, but for one with the right skills, I would say a couple of years is reasonable to gain the pharmacometric skills, like a postdoctoral program and that’s what we’re doing actually. One of the 2020 goals is to have 25 trained pharmacometricians through these two-year fellowships by 2020. We already have six of them. Maybe we don’t need 2020 to reach that goal,” he says.
Given the investment, early positive results, and growing demand from pharmacometrics, it seems clear that FDA has high expectations for modeling and simulation. So does Gobburu.
*Impact of Pharmacometric Reviews on New Drug Approvals and Labeling Decisions – a Survey of 31 New Drug Applications Submitted Between 2005 and 2006, Nature, Feb. 2007
*Impact of Pharmacometrics on Drug Approval and Labeling Decisions: A Survey of 42 New drug Applications, AAPS Journal, October7, 2005
** American Conference on Pharmacometrics 2008
This article first appeared in Bio-IT World’s Predictive Biomedicine newsletter. Click here for a free subscription.