WASHINGTON (AP) - A new study suggests it may be possible to modify an experimental vaccine for Alzheimer's disease, making it safe for humans. The vaccine was withdrawn earlier this year because of side effects in test subjects.

A refined form of the vaccine for the degenerative brain disease was effective in mice, raising the possibility it might not produce the inflammation in humans that brought an end to the earlier clinical trials.

But that remains only a possibility, requiring more research, said JoAnne McLaurin of the Centre for Research in Neurodegenerative Diseases at the University of Toronto, first author of the study being published Tuesday in the online edition of the journal Nature Medicine.

In January, the Irish drug company Elan Corp. suspended a 360-patient experiment with the vaccine after 15 patients suffered serious brain inflammation. The company said in March it was abandoning the vaccine, although officials said it plans to continue seeking ways to slow the worsening of Alzheimer's.

Dave Morgan, an Alzheimer's researcher at the University of South Florida, said other groups have also been working on the ways to refine the vaccine.

The problem, Morgan said, is that it is not clear why the patients in the now-halted trial developed brain inflammation - whether immune cells entered the brain to attack the vaccine or if the vaccine provoked a reaction in the brain itself causing the inflammation.

"Before we put any more vaccines in people we need to know what the problem was," said Morgan, who is not part of the Toronto research team.

The goal of the vaccine is to clear away tangles of amyloid-beta protein, called plaques, that accumulate in the brains of Alzheimer's victims.

In tests on mice, the vaccine produced antibodies directed against the AB proteins that cleared the plaques and reversed symptoms of brain degeneration.

McLaurin, who is not connected to the Elan research, explained that her experiments in mice were aimed at finding out what the active antibody is and how it works.

Her team of researchers found a way to produce antibodies against AB protein in mice using a vaccine that included only a small portion of AB rather than the entire molecule.

This raises the possibility that a vaccine using only a portion of AB would be less likely to cause the inflammation that brought the earlier trial to a halt. It might also lead to development of a small molecule that mimics the vaccine, sidestepping the potential problems of the tested vaccine, McLaurin said.

The response to the vaccine seems to target a specific group of AB in the body known as protofibrillar precursor, the study found.

McLaurin's team included researchers at the University of Toronto and the University of Konstanz, Germany.

In a separate paper in Nature Medicine, researchers headed by Roger M. Nitsch of the University of Zurich, Switzerland, reported that the human patients in the Elan trial did accumulate antibodies against AB.

They said the antibodies recognized AB in tangles, diffuse AB deposits and AB in blood vessels of the brain and were able to cross the blood-brain barrier, which suggests they might be able to directly destroy plaques in the brain.

However, the antibodies did not attack the longer form of AB that occurs in the nerve cells of healthy people as well as Alzheimer's patients.

The function of that type of AB is unknown, but healthy nerve cells can contain a lot of it. This selectivity is good news, since an attack on the longer form of AB could result in complications.