May 12, 2006 | GlaxoSmithKline’s head of genetics research, Allen Roses, says that pharmacogenetics is having a profound impact on the stratifying of patients, the minimization of adverse events, and the expedited passage of drug candidates through clinical trials.
“I want to attack the notion that personal medicine is years away,” Roses said.
The centerpiece of GSK’s efforts in finding genetic disease associations is “HitDIP” — the high-throughput human-disease-specific target program. This approach centers on “tractable genes” — a group of 1,800 well-characterized drug target genes, using 7,000 SNPs. Over the past eight years, GSK has collected highly phenotyped patient collections with informed consent and IRB approval, for common diseases including diabetes, asthma, rheumatoid arthritis, coronary artery disease, obesity, and a dozen more.
Last year, GSK initiated full genome analysis with Affymetrix 500K SNP chips. Those data will be compared to the HitDIP screening data and duly published, beginning with data on Alzheimer’s disease and obesity. Roses said they will be publicly available on either GSK’s or an NIH Web site.
While pharmacogenomics plays an increasing role in discovery research, Roses said, “the clinical pipeline is where the rubber hits the road.” A classic example was a restenosis trial called PRESTO, involving more than 11,000 patients, 4 percent of whom developed hyperbilirubinemia. During the Phase III trial, Roses’ team performed an association study focusing on genes associated with liver function. When the study was unblinded, only individuals homozygous for a specific gene variant who had received the drug developed the side effect.
The problem was not observed during the smaller Phase II study, raising the key question: How few patients does it take to recognize a SNP profile related to an adverse event during drug development? Roses said safety SNP profiles could be identified with as few as 10-20 adverse event cases. If you look at a pair of susceptibility genes, the number of cases is even smaller.
A more recent example involves patients taking the highly touted Tykerb, an oral medication for breast cancer that could compete with Herceptin. Fifteen percent of patients reported unpleasant side effects, including diarrhea and rash. An analysis of densely mapped SNPs in cytochrome P450 metabolizing genes revealed a strong association between the side effects and variants in CYP2C19. Only patients homozygous for the *2 allele experienced side effects.
The upshot is that patients identified as having high risk of side effects can be prescribed lower doses. “We don’t have to poison them with 10 times [the level of] the drug they don’t need,” said Roses. “The moderate side effects are predictable at onset of therapy, and can allow adjustment of dose.” Enrollment in the Phase III trial for Tykerb was halted last week following positive results. “That’s what pharmacogenomics can do,” said Roses.
Pharmacogenomics can also guide Phase II efficacy. In an obesity study, GSK typed obesity candidate genes in 80 patients in a weight-loss program. Polymorphisms in three genes were found to correlate with weight-loss effectiveness.
Roses also presented data showing the impact of pharmacogenomics in early drug development. He described data from a European Phase II trial for Alzheimer’s disease using Rosiglitazone (a drug approved for diabetes). The study, involving 511 Alzheimer’s patients, addressed the question: “Do patients without APOE4 allele respond differently than those who carry APOE4?” By separating patients with and without the E4 variant, the data showed that Alzheimer’s patients without the E4 allele respond better to Rosiglitazone.
“For the first time in history, nonresponders were identified for follow-up with follow-on candidate molecules,” said Roses. He added: “We’ve taken a failed study and, with FDA involvement, initiated a Phase III program for a drug that has a clinical effect on Alzheimer’s disease. Why haven’t you read about this [in the press] yet? Rosiglitazone is available on the market, and we want to scrupulously avoid presenting this in the wrong context — it should not be used for Alzheimer’s disease yet.”
Roses also touted GSK’s “druggable genome” RNAi screening laboratory — a fully automated laboratory to screen 18,000 genes. “There is no capability commercially available in 2006. We’ve got lots of cell lines and want to find what genes may tell us...what targets we should be going after. It’s not what if, it’s done.”
One application of RNAi screening has shown that knocking down estrogen receptor levels in cancer cell lines renders the lines susceptible to a specific drug. This led to the design of a new breast cancer trial in which patients receive a different drug regimen depending on their level of estrogen receptor activity. “The next time you read an analyst’s report that says genetics has no impact on drug discovery, it’s 20 years away, read another analyst!” Roses quipped.