Name: Andrew Allen
Position: Vice President, Oncology Development, Head of Oncology Therapeutic Unit, Chiron Corporation
Education: MD, Oxford University, United Kingdom; PhD, Imperial College, London, United Kingdom
The most visible part of pharmacogenomics involves the use of diagnostics for drugs and treatments that are already on the market, but some in the industry see an opportunity for pharmacogenomics methods in the early stages of drug development. As techniques for identifying responding populations or those likely to experience adverse events are refined, this technology could help reduce the cost of drug development overall.
Andrew Allen, vice president of oncology development at Chiron, spoke at the Feb. 21-24 Cambridge Healthtech Institute Molecular Medicine Tri-Conference about the use of pharmacogenomics early in the drug pipeline, and Pharmacogenomics Reporter spoke to him this week to hear more about this less-public side of the field.
What has happened in the last year that you felt was important enough to bring up in your talk?
I think … the FDA draft guidance that was published in April 2005 was very helpful to give us a very clear sense of where the FDA was going with this concept. And so it really laid out some useful guiderails for the way that we should be approaching companion diagnostic development. Of course, the final decisions always come down to the review division, which is obviously appropriate, and so what we've been doing is trying to take that guidance, and then apply that to some of our compounds as we develop them in the early stages of drug development. I wrapped this up under the umbrella of trying to de-risk the drug development programs early on, obviously when it [is] cheaper and smaller and faster, to be making those go/no-go decisions.
So we're using companion diagnostics to help or we use our translational medicine capabilities to generate these hypotheses, which we then explore using a companion diagnostic approach.
Can you describe how Chiron is employing pharmacogenomic technologies in the early drug-development process?
We have a small-molecule kinase inhibitor that inhibits a variety of kinases, including the VEGF receptor and the PDGF receptor, which makes it look a little bit like [Pfizer's investigational anti-cancer compound] Sutent, but then the thing that makes it look different is that we also inhibit the FGF receptor, so there is an important family of kinases inhibited by this drug on top of VEGF and PDGF. So, that was the drug, and the question is, 'How can one develop that in a targeted and lean manner, rather than running very large, randomized Phase II studies to determine an impact on time-to-disease-progression? Are there smaller, more focused subsets of patients who have disease [that are] perhaps more driven directly even addicted, to use the jargon to targets hit by this drug?'
So as we did our survey, one of the [groups of patients] has multiple myeloma about 20 percent of them who have a particular translocation, which is the T4-14 translocation, which moves pieces of chromosomes around such that a gene enhancer ends up sitting next to the FGF receptor-3 gene. The consequence of that is that FGF receptor-3 protein is expressed on the cell surface of these myeloma cells, and I've only seen as 20 percent having that specific translocation; it's normally not present on the cell surface. And in collaboration with some external partners, including the Multiple Myeloma Research Consortium, who are actually critical for this effort, we obtained primary isolates from patients with myeloma, grouped them into those that did and did not have this particular translocation. And in vitro, when you put the drug Chiron-258 onto the cells that did have the FGF receptor-3, you saw a significant induction of apoptosis, and when you put the same drug on the cells that did not have FGF receptor-3 on the cell surface, you saw no apoptosis.
And obviously, that's relatively small, but it sets up a very testable hypothesis. So we then took the drug into the clinic, and we are running a phase 1-2 trial, whereby we are dose escalating in traditional fashion to establish the maximum tolerated dose, and then we selectively expand two cohorts of patients one with, one without translocation. Then in relatively small numbers of patients you have 15, 20 patients in each arm you can see whether there's a significant, striking difference in the response rate between those two arms.
And that's the model. And if that works, the consequence is that then you can potentially run a pivotal next-phase study. But to do that ultimately, you actually then need a diagnostic for patient-selection purposes ready around the beginning of that phase 2 study. So one of the corollaries of the FDA [VGDS] guidance is that you need to be working on your validated, approvable diagnostic through phase 1, such that at the onset of a pivotal phase 2B study, you have diagnostic that is robust, reliable, and basically you'll have one that you'll be seeking market approval for.
Is a companion diagnostic always produced by this method of drug development?
Many of us use biomarkers for multiple purposes. If you are using the biomarker just to ask the question, 'Are we inhibiting the target of the drug, either in the target tissue, or surrogate tissue to the extent and duration required for efficacy?' Then that, typically, is a biomarker [that] will not become a diagnostic product that's used for internal decision-making purposes.
But patient selection that leads to a treatment choice, you're clearly talking about a marketed diagnostic product. I guess one question is, 'As an industry, do we tend to do this with partners, or do we try to do it all in-house?' We take the partnership model, and I think that's pretty common now.
Another question is, 'Within the indication of choice, do we actually need to develop, launch, market, and distribute a diagnostic product, or has the community evolved such that they are actually diagnosing these patient subgroups using the tools available for them in a way that's accessible to the agency?' And I guess the example there would be the 5q minus subset of patients with [Myelodysplastic Syndromes] who are treated with [Celgene's] Revlimid, for whom a companion diagnostic was not needed because it was regarded as standard of care to do that molecular classification.
Have pharmacogenomic techniques saved as much time and money as you expected?
I can't say that we at Chiron can answer that, because we have not lived through this yet, so it's a work in progress. I think if you look at Herceptin, the data are very clearly there to show that the size of the trial needed the approval of Herceptin. If they hadn't focused down on the Her-2 population, it would've been limited and it would've taken decades to complete.
So I think we all believe that yes, there are circumstance where it's going to be mandatory to use a targeted approach. Obviously, the magic is in knowing when that is, and when it's not.
So you know that at the end of the experiment. Revlimid with Celgene they ran the experiment and they were successful. I guess one could look at AstraZeneca and the EGF receptor and wonder whether things might've been different had they taken a different approach. That's an open question, obviously.
We're strong believers in patient selection segmentation, but as I was saying, we are at the stage of doing the work, and I can't tell you the outcomes yet.