A set of gene-expression profiles may be able to help oncologists better pick chemotherapies in about 80 percent of cases in three types of cancer, according to scientists at Duke University’s Institute for Genome Sciences and Policy.
The IGSP team’s approach differs from existing pharmacogenomic-based methods, such as Genomic Health’s Oncotype Dx, in that it is designed to enable doctors to pick from among seven drugs as opposed to identifying whether a single agent will be efficacious.
The tests would affect how the following chemotherapeutic agents are prescribed: adriamycin, docetaxel, paclitaxel, topotecan, etoposide, 5-FU, and cyclophosphamide.
Prospective clinical trials validating the team’s predictive profiles are slated to begin earlier this month and results “should be available somewhere between 18 and 24 months,” Anil Potti, the group’s lead investigator and an assistant professor of medicine at IGSP, told Pharmacogenomics Reporter this week.
To develop their predictive profiles, IGSP researchers generated expression predictors based on expression data from 60 cell lines that were treated with the above chemotherapeutic drugs, Potti said.
The team then validated the profiles in a second study of 30 cell lines, and examined their predictive power retrospectively in about 350 samples from patients who had already received chemotherapy, he said.
The researchers hope that oncologists will be able to compare gene-expression profiles taken from a patient’s tumor against expression signatures compiled by Potti’s group to find the most effective chemotherapeutic to prescribe.
The tests are accurate on the first try about 80 percent of the time, according to the group’s study, which appears in the November issue of Nature Medicine. Currently, physicians use trial and error to arrive at the most effective agent.
Prospective clinical trial results validating the team’s predictive profiles “should be available somewhere between 18 and 24 months,” told Pharmacogenomics Reporter this week.
In all likelihood, Duke will try to license the predictive profiles to a diagnostics company for commercial development, Potti speculated. The group has not discussed commercializing the profiles with any diagnostics companies yet, but the university has filed patents for the profiles “in the last week,” he added.
“We have actually had multiple discussions with the FDA so that the FDA approval comes in concurrent with the completion of the trials — if the trials are positive,” he said.
The method may be applied to other indications beside breast, lung, and ovarian cancers as profiles are validated, said Potti. The profiles can also be used to choose between a wider array of cancer drugs, including targeted therapies, he said.
“We’ve not shown that in the paper, but we’ve gone on to develop an extended list [of other agents] that we hope to publish in the near future.”
In the United States, between 400,000 and 500,000 patients receive chemotherapy each year to treat breast, lung, ovarian, and colon cancer, said Potti. “If you just looked at breast, lung, and ovarian, you’re probably talking about somewhere between 300,000 and 400,000” individuals, he said.
Previous pharmacogenomic guides for chemotherapy, such as Oncotype Dx, employ gene-expression profiles to predict whether a particular treatment is appropriate for an individual patient. In the case of Oncotype, that treatment is a combination of cyclophosphamide, methotrexate, and 5-fluorouracil — the so-called CMF regimen — for node-negative, estrogen receptor-positive breast cancer.
Another advantage of the Duke team’s test is its high level of validation, assuming it passes clinical trials. “There are a lot of [validation studies] out there that are small sample sets,” said Potti. The IGSP tests will be validated in around 800 patients.
For chemotherapy, Oncotype was validated in about 200 patients, while a predictor for paclitaxil response in breast cancer patients — developed by Lajos Puztai’s group at the University of Texas MD Anderson Cancer Center in Houston — has been validated in about 80 patients.
William Evans and colleagues at St. Jude’s Children’s Research Hospital in Memphis, Tenn., developed a predictor for the response of acute lymphoblastic lymphoma to prednisolone, vincristine, asparaginase, or daunorubicin. That profile has been validated in 98 patients.
“We have actually had multiple discussions with the FDA, so that the FDA approval comes in concurrent with the completion of the trials — if the trials are positive.”
A large validation of Oncotype in 10,046 patients is underway as the Trial Assigning Individualized Options for Treatment, or TAILORx, which will examine the test’s ability to predict response to taxane-containing and nontaxane-based chemotherapeutics.
In January, Potti and colleagues also plan to validate their Metagene lung-cancer gene-expression test in a 1,200-patient trial. Metagene is designed to identify early-stage lung cancer patients who might benefit from unusually early chemotherapy (see PGx Reporter 8/09/06
“In three of those validations, it was a prospective collection for the sake of the validation,” Potti said. However, treatment was not based on tumor genetic profiles, he added.
The IGSP group is “currently in the process of enrolling patients” for prospective, randomized clinical trials, in which patient treatment is determined by tumor expression profiles, to prove the clinical utility of their profiles, said Potti.
Potti and colleagues are performing three studies in lung cancer patients — one in advanced lung cancer, one in the intermediate stage, and one in the early stage — as well as a study in breast cancer patients, and a study in ovarian cancer patients. All of these studies “have either taken off or are taking off in the next month,” he said.
The group’s research is supported by grants from the American Association for Cancer Research, the National Cancer Institute, and the V Foundation. Potti said his team has not received any support from pharmaceutical companies, he said.