In a study of Roche's blockbuster drug, which targets and blocks a protein called VEGF-A, researchers found that different forms of the protein lead to varying responses and Avastin had no benefit in at least half of those taking it.

Avastin, or bevacizumab, has been shown to increase survival from bowel cancer in around 10 to 15 percent of patients and was licensed on that basis, but until now it has been impossible to predict who is likely to benefit.

"This is the only evidence so far that shows which patients will respond to Avastin - or more importantly those who will not respond to Avastin - in the conditions for which it was originally licensed," said David Bates, a professor at Britain's University of Bristol, who led the research.

"We've shown which half of patients don't benefit, so that takes half the patients out of the treatment group," he told Reuters. "But it's also quite possible that it's only a fraction of the other half that do benefit."

Sales of Avastin, which brought in $6 billion for Roche in 2011, have been hit by a decision last year by U.S. drugs regulators to revoke its approval in breast cancer after it decided the drug was not effective enough to justify its risks.

Although Avastin is licensed in Europe for advanced bowel or colorectal cancer, it is not used within Britain's taxpayer-funded National Health Service (NHS) because the health costs watchdog NICE says its benefits to an unknown minority of patients do not justify its costs.

DIFFERENT FORMS

For this study, Bates' team looked at two different forms of the VEGF protein - one called VEGF165 which helps cancers to grow new blood vessels to get food and oxygen from the blood, and another called VEGF165b which has the opposite effect and acts as a brake on this growth.

Analyzing data from 97 samples from patients in a final stage clinical trial of Avastin, they found that those with low levels of VEGF165b survived three months longer without their cancer progressing compared with patients not treated with Avastin.

But patients with higher levels of VEGF165B saw no benefit from Avastin and survived no longer than those not taking it.

A spokesman for Roche, which is the world's largest maker of cancer drugs, said the findings had limitations because they were based on a relatively small number of patients and were retrospective.

"Validation (of these findings) requires a prospective study with an adequate (or) much larger sample size," the spokesman said in an emailed statement.

Bates, whose work was published in the journal Clinical Cancer Research, said the key question now was how regulatory authorities respond.

"They could say we need further clinical trials, or they could say we now need to start screening for VEGF165b to determine whether patients should get (Avastin) treatment," he said in a telephone interview.

Bowel cancer is the third most common cancer in Britain with around 41,000 people diagnosed each year. In the United States, an estimated 143,460 people will be diagnosed with this form of cancer in 2012, according to the National Institutes of Health.

(Editing by David Holmes)

• A Permanent Economic Emergency