To stoke future biomedical innovation, President Obama is proposing to turn the National Institutes of Health into a drug company.

The administration is set to establish a billion-dollar division of the NIH called the National Center for Advancing Translational Sciences that will develop pharmaceutical treatments for diseases the government perceives as underaddressed by private industry.

Unfortunately, recent evidence suggests the government is ill-suited for drug invention. In the case of the advanced cancer treatment Avastin, the NIH -- and the Food and Drug Administraion -- has hindered medical progress. If this new agency is serious about driving drug development, it needs to avoid making the mistakes the government made with this drug.

In the late 1980s, Nobel Prize-winning scientist Judah Folkman identified an antibody called "bevacizumab" that could shut off blood flow to tumors. For years, the NIH rejected Folkman's discovery as unproven and even went so far as to test its own version of the compound to show that it didn't work.

Folkman was open about his frustrations with the government's refusal to give his discovery a fair shake. He'd explained that "some breast cancers, for instance, make only one [growth] factor while others make six. That means that you have to balance the [antibody] dose against a specific tumor, as much as you would adjust the dose of insulin according to your blood sugar levels for diabetes."

But, he dourly noted, such balance is "not the way" the FDA wants scientists to do clinical trials.

It was left to the drug firm Genentech in 2000 to bet billions on Folkman and develop bevacizumab into a workable drug, eventually called Avastin. The FDA quickly approved Avastin for lung- and colon-cancer treatment.

Then, in 2008, the FDA authorized the drug for metastatic breast cancer. That approval came after Avastin showed impressive results with regard to what is known as "progression-free survival" -- a measure of its ability to stop tumor growth. For many trial patients, Avastin proved to restrict or shrink their tumors, improving quality of life and extending life span.

However, last summer, the FDA announced it would reevaluate Avastin under a new standard called "overall median survival," which essentially measures how much longer the average patient lives because of the drug.

This announcement surprised many because the overall survivability standard isn't appropriate for a drug like Avastin when it comes to breast-cancer treatment. The bulk of the drug's benefits fall on a key set of "super responders" that gain immense amounts of additional life. But for the "average" patient, Avastin improves the quality of life but doesn't significantly extend it.

Overall survivability doesn't take into account individual variances in treatment responses. The progression-free standard does.

Nonetheless, the FDA stuck with the "overall survivability" metric, determined Avastin doesn't "provide a sufficient benefit in slowing disease progression" and ruled to revoke its approval. Genentech is appealing the decision.

Avastin had been serving as the last, best hope for thousands of dying breast-cancer patients. But this ruling, if left unaltered, could wipe Avastin off the table as a treatment option, leaving many women to suffer and die prematurely.

Cases like Avastin are all too common. The NIH, FDA and other government bodies haven't done a good job applying reasonable, consistent approval standards to drugs. The result has been a steady decline in the number of new treatments hitting the market.

Indeed, just 21 drugs received FDA approval last year. That's down from 25 in 2009 and 24 in 2008. Applications for approval in 2010 were also lower than in previous years even though the number of products in the pipeline is increasing.

If the government is going to step into the drug-innovation business, it can't treat new drugs like its treated Avastin.

Dr. Robert Goldberg is vice president of the Center for Medicine in the Public Inter est and author of "Tabloid Medicine: How the Internet is Being Used To Hijack Med ical Science for Fear and Profit."