WASHINGTON — A widely available cancer drug has shown remarkable success in reversing Alzheimer's disease in mice, raising hope of a breakthrough against incurable dementia in humans, US researchers said Thursday.

Mice treated with the drug, known as bexarotene, became rapidly smarter and the plaque in their brains that was causing Alzheimer's started to disappear within hours, said the research in the US journal Science.

"We were shocked and amazed," lead author Gary Landreth, a professor in the Department of Neurosciences at Case Western Reserve University School of Medicine in Ohio, told AFP.

"Things like this had never, ever been seen before," he said.

The drug works by boosting levels of a protein, Apolipoprotein E (ApoE), that helps clear amyloid plaque buildup in the brain, a key hallmark of Alzheimer's disease.

"Think of this as a garbage disposal," Landreth said.

"When we are young and healthy, all of us can basically get rid of this (amyloid) and degrade it and grind it into small bits and it gets cleared.

"Many of us will be unable to do this as efficiently as we age. And this is associated with mental decline or cognitive impairment."

Six hours after mice got the drug, which works through the liver to boost retinoid X receptors (RXR), stimulating production of ApoE in the brain, soluble amyloid levels fell 25 percent, ultimately reaching a 75 percent drop.

The effect lasted up to three days, said the study.

Soon after taking the drug, mice began performing better in tests, showing that they were able to remember things again, were more social and were able to smell again, a sense that is commonly lost in Alzheimer's.

Also, unlike normal mice, Alzheimer's mice will not usually build nests if given tissue paper in their cage, as if they have forgotten to associate paper with the opportunity to nest.

But 72 hours after treatment, the Alzheimer's mice began to build nests again.

"They are not great nests but they are nests nonetheless," added Landreth, suggesting that if the drug can be shown to work in humans it might be best targeted at people in the early stages of the disease.

Clinical trials for humans are being designed and should produce early results in the coming year, researchers said.

Bexarotene was initially made by US-based Ligand Pharmaceuticals under the brand name Targretin.

It gained orphan drug status in the United States -- approval by the US Food and Drug Administration -- in 1999 as a treatment for cutaneous T-cell lymphoma, a rare cancer of the immune system that manifests in the skin and liver.

The Japanese pharmaceutical giant Eisai bought the worldwide rights for it in 2006. Bexarotene is now available in 26 countries in Europe, North America and South America.

Scott Turner, director of the Georgetown University Medical Center's Memory Disorders Program, who was not involved in the research, welcomed the findings.

"This looks very exciting," he said. "This is a brand new way to move forward in human trials of Alzheimer's disease and it works great with mice."

Turner, a neurologist and leading expert in Alzheimer's disease, however cautioned that more study was needed to see if the same results can be seen in humans.

"One obstacle is that the mice may not be a good model of Alzheimer's disease. We have so many things that work in mice and we try them in humans and they just completely fail," he said.

Bexarotene has a good safety profile, though women who are pregnant or may become pregnant are warned to stay away from it because it risks causing fetal defects.

Typical side effects include diarrhea, dizziness, nausea, dry skin and trouble sleeping.

Since the drug is typically given to cancer patients, Landreth said there have been no anecdotal reports of improved memory in humans, since most do not live long enough to reach the stage of Alzheimer's.

"We have clinical consultants or dermatologists who use this all the time but they hadn't thought to look at this so there is very little anecdotal data around."

Trials should begin in the next month or so, Landreth said.

"Perhaps the most important thing is to ask the question: Does this drug work in human beings as it does in mice? Does it get into the brain? And does it have an effect on amyloid levels and increase ApoE levels?

"We need to do that in normal human beings and see if humans are like mice."

Alzheimer's and other forms of dementia afflict 35.6 million people worldwide, with cases forecast to nearly double by 2030, according to Alzheimer's Disease International which puts the annual global costs of the disease at $604 billion.