With private funding and Duke’s help, Roses focuses on discovery and development.
June 10, 2008 | Last year, Allen Roses left his position as senior vice president, pharmacogenetics, at GlaxoSmithKline to return to Duke University Medical Center, where he is director of the Duke Drug Discovery Institute (and Jefferson-Pilot Professor of Neurobiology and Genetics and a member of the Duke Institute for Genome Sciences and Policy). Ricki Lewis caught up with Roses, who keynoted Bio-IT World Expo in 2006 (See “Personalized Medicine’s Rosy Picture,” Bio-IT World, May 2006), to review pharma’s approach to genome-wide screening, his new freedoms back in academia, and the latest on pharmacogenomics and Alzheimer’s research.
Bio-IT World: How is the pharmaceutical industry using genome-wide association studies (GWAS)?
Roses: Genome wide screening for pharma will be most important in confirming candidate gene variants that differentiate patients with efficacy, using the particular end-points of the clinical trial. This is most important at the critical proof-of-concept (POC) step. Let’s say a molecule has made it through preclinical safety and first time in humans. The first efficacy indication would come from a small Phase IIA trial, then a larger Phase IIB proof of efficacy trial. During these smaller trials, an extensive list of polymorphisms from candidate genes, immunological genes and HLA antigens would be tested for possible associations. At this early stage, genome wide screening—and correction for the number of tests performed—would not be productive. The candidate list is small and more focused. The efficacy PGX experiment is designed to compare patients who met the proposed clinical endpoints against patients who did not. In this way early hypotheses could be incorporated into Phase III registration studies and, if re-confirmed, provide more information for targeting therapy.
How do genome-wide scans compare with a candidate gene approach?
Genome wide screening may be particularly useful early for identifying safety signals. We’ve initiated side-effect or adverse event (AE) experiments with candidate gene panels. We were able to show, in several clinical trial programs, that genome wide methods can later be helpful to confirm candidate genes and to recognize regions of the genome where additional candidate genes associated with AEs might lie. This takes at least 15 to 30 individuals who had experienced the AE. Genome-wide screens are better used to confirm candidate gene associations, or in a hypothesis-generating mode, or if nothing comes up with a particular candidate gene screen. This is also useful for drug surveillance, but the process of surveillance should actually start during development clinical trials.
Should GWAS be built into clinical trials prospectively, or analyzed retrospectively?
It is important that the data supporting a hypothesis are built into the clinical protocol for a predictive test. By funding the association with an AE, or with efficacy during a trial, regulators classify studies as “exploratory of hypothesis generating” versus hypothesis testing. The latter is preferred, so that the “generating” experiment can also be used for registration instead of mandating another repeat study. As development continues, each successive clinical trial can be used to confirm positive signals and sequentially eliminate false positive signals.
GWAS is most important in generating associated variants for use as a companion diagnostic. The experiments must be performed during development from a company’s point of view. It can make the difference between a targeted therapy versus a borderline or negative efficacy result when only a few patients get excellent responses. Companion diagnostics should be timed with registration by the FDA, so that targeted therapy differentiation can be considered for pricing and, more importantly, reimbursement.
How did the situation with Iressa, which FDA cleared for marketing in 2003, highlight the value of genotyping?
Timing is key. Iressa [AstraZeneca’s small cell lung cancer drug] was labeled as a second- or third-line therapy because the efficacy response rate was low across all comers. After marketing, several academic research groups observed that a few specific patients showed remarkably good improvement. Had that been studied before registration, efficacy would have been easier to demonstrate, albeit in a small subgroup. The company could have sought registration as a targeted therapy for cancer patients carrying specific mutations. The finding changed the risk/benefit picture for a definable group of patients. Post-marketing is too late to change the commercial proposition. The end result was to take a small market and make it smaller, without increasing the price for a truly targeted therapy. If the company had that data during development, they could have registered the drug as a targeted therapy—with similar considerations of cost and pricing as with orphan drugs.
Won’t genotyping limit market size by stratifying patients?
Large studies can hide subgroups who respond in a particular, genetically determined way. A question in industry is why genotype, restricting the market, when you can get the whole market? Iressa works well in some people. You barely get an average signature if you look at many patients, but get a very strong signal in a small identifiable subset of them. That’s enough for a company to do well if it has developed a companion diagnostic and had it in the label. There was FDA Guidance for Industry in April 2008, providing definitions to be used in upcoming additional guidance documents for pharmacogenetics and pharmacogenomics. Interestingly, the April 2008 document also came with a “black box” on page one suggesting that variance from the suggested guidelines comes with some additional risk for approvals.
Can you give an example of how a pharmacogenetic approach improved the safety profile of a particular drug?
The Abacavir story [4-5% HIV patients taking the reverse transcriptase inhibitor develop hypersensitivity] has had a major impact on the ability to regulate safety, putting a focus on pharmacovigilance, and not requiring whole genome data. If you can select appropriate candidate genes for safety experiments (including some knowledge of the drug’s mechanism), the candidate strategy works…
The general impact is that pharmacogenetics enables us to make drugs safer. Abacavir isn’t the only example that has happened, just the first with a prospective clinical trial for measuring the predictive value of the test. Similar data have been developed for two other GlaxoSmithKline drugs [including Tranilast]…
Why was the pharma industry slow to embrace pharmacogenetics, given that discussions about sequencing the human genome began in the late 1980s?
Pharma has reacted to the period of time when the sequencing of the human genome was hyped and hyped (by NIH, Celera, and grant or investment seekers) about what miracles were just around the corner. They all acted as if they had broken a bubble, but the bubble was artificial. Even if the whole genome became known, we wouldn’t know what to do with the information. Pharma and the venture marketplace viewed the lack of immediate effects from the genome project as part of that hyperbole. In the meantime, GlaxoSmithKline, for example, recognized the importance that pharmacogenetics could have and invested in the experiments important to a pharmaceutical pipeline. Early followers are now engaged to achieve a competitive advantage while creating drugs that can be predicted to be safe for most people.
You’re best known for the Apolipoprotein E (APOE) story in Alzheimer’s disease (AD). To what extent has APOE’s role in the disease been validated?
We originally found the association with APOE4 [the most serious version of a gene linked to Alzheimer’s] in familial as well as sporadic cases in 1992—the age of onset of Alzheimer disease changes as a result of this genotype. We looked at tissues from homozygotes for APOE4 and found that they had a lot more amyloid than homozygotes for APOE3. That was a phenotypic association.
Now we have a mechanism that explains what goes wrong in neuronal cells that allows the AD phenotype to emerge over time. There is a 10 to 15 year difference between APOE4/4 individuals and APOE3/3 individuals in what gradually progresses over time. But companies lose interest, because anything more that we learn about AD genes would not result in a new drug for another ten years. That is why I’ve returned to Duke, with private funding, to discover and develop a drug or drugs in a virtual pharmaceutical company...
Can you describe the new Alzheimer’s gene your group recently discovered?
Nice try! It is proprietary for now. But our discovery of a second gene still needs to be confirmed. The results will be published. It may take up to two years to confirm the data but, in the meantime, we are developing a screening assay. If the data are not confirmed, we’ve wasted time, but if they are, and we find what we think we will, we’ll be two years ahead.
Can you explain the surprising findings about APOE4 in the rosiglitazone treatment trials for Alzheimer’s disease?
The operative word in your question is “surprising.” We have predicted that result [the drug was not efficacious for the whole population, subsequent analyses revealed that people who did not have an APOE4 allele improved with all three experimental doses in the study protocol]. [There’s] an article in Forbes from April 4, “Attacking Alzheimer’s”. It begins: “The drug industry has bet heavily on one theory about the disease. What if that theory is wrong?” Well, I have thought it was wrong for almost 20 years. People in the Alzheimer’s field are wrapped tight around amyloid as the cause, and are apparently “surprised” by the clinical trial. Amyloid and APOE are in fact interconnected with amyloid being a downstream consequence of APOE-induced mitochondrial toxicity, secondary to different rates for APOE3 and APOE4.
Should the APOE4 association be used to develop a predictive test for Alzheimer’s?
My concentration is to develop a preventive therapy. If rosiglitazone works in patients with Alzheimer’s and makes them a bit better, then the next opportunity is a prevention study attacking the same mechanism. It is a very interesting economic problem, from a company’s point of view. The patent life for rosiglitazone will be over soon. It is now up to others in academia and the Alzheimer’s community to find a way to focus on a prevention trial, perhaps at a low dose. One of the institute’s goals is to organize a prevention trial. Another alternative might be to incentivize companies for continuing expensive Phase IV preventive trials.
This article appeared in Bio-IT World Magazine.
Subscriptions are free for qualifying individuals. Apply Today.