group of major pharmaceutical companies will share pooled data from failed clinical trials in an attempt to figure out what is going wrong in the studies and what can be done to improve drug development.

In the first wave, data from 4,000 patients across 11 failed Alzheimer's-drug clinical trials from Johnson & Johnson, GlaxoSmithKline PLC, AstraZeneca PLC, Sanofi-Aventis and Abbott Laboratories will be publicly available as of Friday.

Data from additional drug makers and the National Institutes of Health will be added in the future. The coalition aims to create similar pooled databases for Parkinson's disease and tuberculosis, said Marc Cantillon, executive director of the Coalition against Major Diseases, which spearheaded the project, funded by the Food and Drug Administration and Science Foundation Arizona.

The data will be available to all the participating drug makers, as well as outside researchers with a valid scientific question, Dr. Cantillon said.

"Companies said they're running into a stone wall with Alzheimer's and Parkinson's," said Ray Woosley, chief executive of the Critical Path Institute, which oversees the coalition. "We really believe drugs are failing because we honestly don't understand the disease."

The hope is that this large database will help answer some questions that individual trials with just a couple of hundred patients can't answer, such as how the disease progresses and whether there are differences in subgroups in the population.

"Innovation no longer happens solely in one company's lab," said Frank Casty, AstraZeneca's vice president of technical evaluations. "It is happening through constant interaction between scientists in the biopharma industry, patient advocates, academia and government."

For the FDA, the project is something of a counter to criticism that it has been slowing down drug development because it is too focused on patient safety as opposed to on how well a drug works. Rather than debate whether the benefit-risk calculation for new drugs should be changed, the FDA wanted to help the industry develop better methods for determining if drugs are safe and effective, according to Mark McClellan, a former FDA commissioner who helped develop and support the coalition.

"The whole point of this type of effort is to get out of that debate," Dr. McClellan said.

FDA Deputy Commissioner Joshua Sharfstein said, "I think the FDA recognizes that one thing that can accelerate drug development is sharing information that is relevant to a disease."

If the database works as hoped, it may enable drug makers to build more sophisticated computer models to help design more efficient clinical trials that require fewer patients and less money, Dr. Cantillon said.

With current studies, if the results show no difference between patients getting an experimental treatment and those getting a placebo, it often isn't clear if the drug failed or if there was something wrong in the study design that prevented the effect from being picked up in the statistical analyses.

The most recent example is Pfizer Inc.'s Dimebon, originally a Russian cold medicine that had shown promise in treating Alzheimer's patients in a small, Russian-based trial several years ago. But, a larger, multinational late-stage study recently showed that the drug had no effect on this new sample of patients.

"People are left wondering why it didn't work," said Dr. Cantillon.

If a better model could be developed to predict which patients would respond, "it's possible" that Pfizer could go back and look at Dimebon's response in those subgroups, he said.

A Pfizer spokesman said the company "is participating in a number of public-private partnerships with governments, independent medical and scientific groups, companies and advocacy groups. Collaborations such as these can lead to more treatment options and better care for patients and the people who care for them."