Diagnostics & Imaging Week

ATLANTA – As a group of well-known politicos prepared for a lunch keynote session at the 2009 Biotechnology Industry Organization convention to discuss President Barack Obama's plan for providing affordable healthcare options, a panel of oncology experts and executives were in an adjacent building at the Georgia World Congress Center reviewing some of the latest developments in predictive biomarkers aimed at getting the most effective treatment to individual cancer patients.

Ideally, that biomarker work would lead to the clear identification and use of specific predictive markers that would allow oncologists to tailor cancer therapies, essentially offering a form of personalized medicine, an approach that has long been hailed as a way to cut overall treatment costs. But, realistically, that development is taking longer than expected and still faces challenge.

There's no question that being able to screen patients prior to treatment would boost the success rate of current drugs – right now even the most successful cancer therapeutic only benefits a third of patients at most – and several recent advances, such as the completion of the Human Genome Project, suggest that developing predictive markers is right around the corner.

But the existing tools have not been effectively integrated into clinical trials, which would help promote the routine and accepted use of biomarkers in cancer, said John Heymach, assistant professor at the University of Texas M.D. Anderson Cancer Center (Houston).

For starters, only a few biomarkers are known. "We know about HER2 and the estrogen receptor signaling pathway" in breast cancer, Heymach said. Genentech's (South San Francisco, California) Herceptin (trastuzumab) is indicated specifically for patients with HER2-positive breast cancer.

Investigators also have identified a mutation in the KRAS gene that correlates to a lack of response with epidermal growth factor receptor (EGFR)-targeting drugs such as Tarceva (erlotinib from OSI Pharmaceuticals and Genentech), Erbitux (cetuximab) and Vectibix (panitumumab).

But it's usually not until after the trial is completed and data are being analyzed that predictive markers are identified.

The determination and testing for biomarkers often "lags behind the trial," said Tim Clackson, chief scientific officer at Ariad Pharmaceuticals (Cambridge, Massachusetts) which, with partner Merck & Co. (Whitehouse Station, New Jersey) is testing its oral mTOR inhibitor ridoforolimus in a Phase III trial in metastatic soft-tissue and bone sarcomas.

Earlier in development, Ariad had identified the PTEN gene as a possible predictive marker. Evidence indicated that the presence of PTEN could influence the sensitivity and activity of mTOR inhibition, but further analysis showed "no correlation with PTEN status and response," Clackson said. "It was not a valid biomarker."

But even when identified, the FDA has yet to accept retrospective analyses to support changes to a product's label. After post-Phase III analyses showed that certain mutations in the KRAS gene correlated to less effective treatment with EGFR-targeting drugs, ImClone Systems (New York) – now a subsidiary of Eli Lilly and Co. (Indianapolis) – and Amgen (Thousand Oaks, California) appeared before the agency's Oncologic Drugs Advisory Committee in December seeking to include retrospective biomarker data in the labeling of their respective drugs, Erbitux and Vectibix. Though the advisory panel discussed the issue at length, they did not reach any conclusions on guidance at that time.

And given the duration and expense of the average cancer study, companies probably aren't going to "repeat a whole trial to validate a biomarker," Heymach said.

But there are a couple of alternatives for companies gearing up to start their large cancer drug studies. One, he said, involves using an adaptive signature design for the study, in which an assay or signature that identifies patients sensitive to treatment can be incorporated into a study in progress.

In an ongoing Phase III study in non-small-cell lung cancer at M.D. Anderson, investigators are expected to conduct an unblinded interim analysis and, if a predictive marker is identified, then "we can do market validation for the rest of the study," Heymach said.

At the end of the trial, all the data is unblinded, and "you look for overall survival in the marker-positive group and overall survival for the whole group," he said. "You could win either way," and it "doesn't require knowing the markers beforehand."

Another design, which Heymach called the "Bayesian" adaptive design, calls for each patient to get a biopsy at the start, with the results to determine which treatment protocol they receive. He referred to the ongoing BATTLE (Biomarker-integrated Approaches of Targeted Therapy of Lung Cancer Elimination) program, which begins a tumor biomarker analysis study, after which patients will be assigned to one of four Phase II studies testing different NSCLC treatments.

New technologies also might help to identify predictive markers earlier in the process. Murray Robinson, senior VP of oncology at AVEO Pharmaceuticals (Cambridge, Massachusetts), said his firm has a platform featuring inducible in vivo cancer models engineered to contain signature genetic mutations that are present in different tumors.

Collecting that information at the preclinical level "really obviates man of the challenges" of identifying predictive markers, he said.

AVEO is using that technology to develop a multigene biomarker for its own lead product, AV-951, an anti-vascular endothelial growth factor inhibitor set to start Phase III trials in renal-cell cancer in the second half of this year.