Mice in the early stage of Alzheimer’s disease had some of their brain abnormalities reversed and their declining mental function restored when they were given low doses of a rarely used cancer drug.

The drug, bexarotene, stimulated the removal of ­beta-amyloid, a substance whose accumulation in the brain appears to be the main cause of Alzheimer’s dementia. After treatment, the animals fared better in tests of memory and social behavior, according to a study published online Thursday by the journal Science.

The findings were dramatic, but their relevance to people with Alzheimer’s disease is unknown. But because the drug is already approved for human use, finding out may be easier than if the drug were an entirely new chemical compound.

“It has to work in humans like it works in mice or we can pick up and go home,” said Gary E. Landreth, a neuroscientist at Case Western Reserve University School of Medicine, who headed the experiment.

A study of bexarotene’s effects in normal brains will start in a few months. Clinical trials in people with early Alzheimer’s — or at high risk for the disease for genetic reasons — will take years. If the drug’s effect in human brains turns out to be different, it may never move on to be tested in Alzheimer’s patients.

Nevertheless, the researchers think speed is important, as they fear that people might start using the drug before it is fully evaluated. A few days ago, Landreth got a call from a physician in another city. A person who had heard of the drug through the grapevine had asked for a prescription for bexarotene.

“We’ve got to work fast, and we have got to be right. We can’t screw this up,” Landreth said.

About 5.4 million Americans have Alzheimer’s disease, which is the leading cause of dementia. About 5 percent of people in their late 60s suffer from it, and possibly half of people in their late 80s. By 2050, 1 in 85 people worldwide will have the disease, according to one estimate.

Bexarotene is in the retinoid family of compounds, which are all chemically related to Vitamin A. Retinoids affect cell division and growth, immunity and other essential biological functions.

Sold under the trade name Targretin, bexarotene is approved by the Food and Drug Administration as a treatment for cutaneous T-cell lymphoma, a disease diagnosed in about 3,000 Americans a year. It is an “orphan drug,” a designation that gives drug companies an incentive to develop medicines for rare diseases. Orphan drugs stay under patent protection longer than regular drugs. It has been tried against other forms of cancer, without impressive results.

Numerous “candidate” Alz­heimer’s drugs, which looked promising in lab studies, have washed out in recent years.

Last month, Pfizer and a smaller partner announced that they were abandoning a drug called Dimebon as a possible Alzheimer’s drug after it failed to show benefit in a clinical trial. In 2010, Lilly halted development of a different compound, sem­agacestat. At the Alzheimer’s Association, an advocacy organization, the new study’s results were greeted with that history noted.

“This is an early study and it was in mice,” said Maria C. Carrillo, the organization’s director of scientific relations. “We need to be cautiously optimistic and pursue this lead as we would any other.” She added that what makes “this an exciting study is that it involves a repurposed drug.”

The Case Western experiments were conducted by Paige E. Cramer, a graduate student in Landreth’s lab. The mice used had genetic defects that allowed beta-amyloid to accumulate in their brains, which in turn changed their behavior. The animals, however, are an imperfect model for the human disease. They don’t lose brain cells after beta-amyloid accumulates as people with Alzheimer’s do. (In people, dementia can begin even before cells die, as beta-amyloid disrupts the normal firing of nerve cells.)

Bexarotene spurred the production of “apoE,” a protein that breaks down beta-amyloid molecules floating in the watery fluid between brain cells. Beta-amyloid levels fell by 25 percent within a few days of a single dose. The drug also stimulated housekeeping cells called microglia to consume plaques of solidified beta-amyloid.

While the density of plaques fell by 75 percent over a few weeks, eventually they reaccumulated. Curiously, however, the animals’ behavior didn’t regress to its former state.

“It appears that the measurable activity of the brain is not affected by the presence of the plaques. I think that is a really interesting finding that will stimulate a lot of new science,” Landreth said.

The behavioral tests involved an animal’s ability to remember a cage in which it had gotten a shock, to find a submerged platform in a pool of water, to identify a smell and to build a nest. After treatment with the drug, all those activities returned to normal.

Several physicians with patients taking bexarotene say they’ve never noticed mental improvement from the drug, although they have few, if any, demented patients under their care.

“I have treated over 500 patients with cutaneous T-cell lymphoma in studies and in practice with bexarotene and have not heard any beneficial effect on the brain or cognitive function,” Madeleine Duvic of M.D. Anderson Cancer Center in Houston wrote in an e-mail.

“We really didn’t see anything like this,” said Heather Wakelee, a cancer researcher at Stanford University. She noted, however, that demented people were excluded from studies of bexarotene in advanced cancer.