SAN DIEGO – With less than one-fourth of all clinical-stage oncology drugs making it through to market – at the same time that trial costs are escalating dramatically – it's more important than ever for companies to select the right compounds to advance into large studies.

But, according to a panelist at the annual meeting of the American Association of Cancer Research (Philadelphia) here, relying heavily on the use of biomarkers is not the way to go. Despite hype surrounding the potential of biomarker data to predict clinical outcome, the reality often has come up short, said Mark Ratain, professor of medicine at the University of Chicago, during an educational session on how to make "go/no-go" decisions in clinical development.

Overall, the oncology space has a disappointing track record vs. other disease areas, he said, when it comes to successful drug development. The biggest issues for companies developing cancer compounds are the low Phase III success rate, the failure to report sufficient efficacy to justify the steep – and sometimes controversial – pricing on cancer drugs and the difficulties in determining optimal dosing in early trials.

Yet biomarkers – defined as measurements or indications of biologic processes – are "neither necessary nor sufficient to meet these needs," Ratain said. He said that the presence of a biomarker effect alone is "insufficient justification to initiate Phase III testing."

The only exception he cited is if the biomarker is "reasonably likely to correlate with a clinical endpoint," and "that's the important hurdle" here.

For example, appropriate biomarker assessments in Phase I testing might look at dose-dependent changes in tumor size, or dose-dependent toxicity measures such as the neutrophil count following treatment with mitotic kinase inhibitors or blood pressure readings following VEGF inhibitor therapy.

An analysis conducted by the FDA looking at a non-small-cell lung cancer trial indicated that a change in tumor size at eight weeks "is a good biomarker," Ratain said, because it's "reasonably likely to predict clinical benefit."

A trickier use of biomarkers involves testing patients' tumors for specific cell expressions or mutations. It's a method some companies have turned to in recent years, believing that biomarkers will allow them to screen patients, ultimately enrolling a treatment population most apt to respond to treatment, and reduce clinical trial costs.

The downside is that the firm might miss the drug's broader opportunity. Ratain offered Nexavar (sorafenib), the multikinase inhibitor sold by Onyx Pharmaceuticals (Emeryville, California) and partner Bayer Pharmaceuticals (Montville, New Jersey), as an example. Researchers initially were looking at Nexavar as a treatment in metastatic melanoma, particularly in patients with RAF-dependent tumors, but they still included patients with other types of cancer in early studies.

Because of that, "they were able to discover a signal in renal cancer," Ratain said.

Nexavar won approval in renal-cell carcinoma in 2005 as its first indication. The drug later was approved in liver cancer.

If that early Nexavar trial had been limited to colorectal cancer patients with RAF tumors, "it would have been a false negative trial," Ratain said, and could have hindered the drug's market opportunity.

At most, he said, biomarkers can "augment but not replace carefully designed Phase II trials."

Hopefully, at by at least the conclusion of those Phase II studies, there is clear indication, one way or the other, of the drug's activity and efficacy for the next "go/no-go" decision prior to launching large and costly Phase III trials. Still, to date, only about half of the oncology drugs entering Phase III development succeed.

It's "still a very tough area," said Peter Ho, senior VP of oncology for GlaxoSmithKline (GSK; Research Triangle Park, North Carolina). An average of four to five new chemical compounds get the FDA's blessing every year, and so far, no one company "has been able to go to the well very frequently in short period of time."

GSK, though, has done fairly well for itself, securing the approval of Tykerb (lapatinib) in breast cancer last year and progressing in pivotal development with pazopanib, a pan-VEGF inhibitor, in renal-cell cancer.

Between 2002 and 2005, the company made 14 program decisions. Six of those programs, including Tykerb and pazopanib, were chosen to continue.

Ho attributes the company's success to a careful "go/no-go" strategy. He said that proof-of-concept data, naturally, drives most of those decisions, but he added that designing randomized trials early on, even in Phase I development, can give companies much-needed information while development is less costly.

Firms should "strongly consider randomized trials in Phase I and Phase II," he said, to help them "arrive at the go/no-go decision with greater confidence."