and costly to do so if it feared the FDA might ultimately not agree with its interpretation that the 10-gene signal is real.
To overcome this problem, the FDA established a program around 2004 whereby companies have been submitting genomic data to the agency on a voluntary basis. The goal of the program was for companies to feel comfortable discussing their complex, still-premature data with the agency without fear that doing so might have regulatory consequences, while at the same time allowing the industry and FDA to figure out what kind of biomarker information might have to be included in formal submissions to the agency.
The conversations that resulted over the past few years—including 45 face-to-face meetings between companies and the FDA—have been immensely productive, says Federico Goodsaid, who is the leader of the Genomics Group at the agency’s Center for Drug Evaluation and Research. “We are the ones that learned the most,” he told me.
The article gives a refreshingly candid description of several meetings between the FDA and various companies, as both parties tried to define what their data meant and how it might be used to help guide approval of drugs and diagnostic tests.
For example, the article described efforts at Novartis to identify a biomarker that predicts poor response in patients with kidney transplants. The company analyzed the expression levels of 12,000 genes and identified 10 whose expression predicted poor outcomes after transplant. Together with the FDA, it worked to define parameters for a possible diagnostic test that might allow doctors to identify these patients on time, the hope being to begin early therapy in cases of likely organ rejection.
The article also mentions AstraZeneca’s efforts to find gene markers predicting an adverse liver reaction it had observed in clinical trials of an anti-clotting drug. There was also a program at Roche aimed at finding gene variants that might predict which patients will suffer toxic side effects from a chemotherapy drug. Pfizer had a project to find a more efficient test to define whether a drug candidate is a possible carcinogen.
This isn’t the kind of sexy stuff likely to make the front page of the New York Times. But the fact that industry and FDA have begun to define how these types of genomic data will be used to guide clinical development is, I think, a much bigger step than any one genomics-era blockbuster in the making.
