The new findings have their pluses and minuses, one outside expert said.

"What is disappointing about the study is there was no difference in the amount of resources caregivers needed whether the person was on donepezil or placebo," said Dr. Gary J. Kennedy, director of geriatric psychiatry at Montefiore Medical Center in New York City. "There were also no differences in behavioral disturbances such as agitation," he said.

However, "the article is consistent with what donepezil is supposed to do, which is improve cognition, so it does help with cognition -- even if that is pretty severely impaired," Kennedy continued.

"This is really just a confirmation of what had already been shown last year [by two other research groups]," added Maria Carrillo, director of medical and scientific relations at the Alzheimer's Association. "It's a good thing in that it showed less of a decline for people who we thought were hopeless. But it does just further confirm that piece."

Aricept, a cholinesterase inhibitor, was approved a decade ago for mild-to-moderate Alzheimer's dementia.

An estimated 21 percent of the 4.5 million Americans with Alzheimer's in 2000 had severe disease. In addition to reduced cognitive abilities and diminished social interaction, individuals in the severe stage of the disease also have problems performing activities of daily living, such as bathing and going to the toilet. The burden on the caregiver increases markedly as the disease progresses, and many patients are institutionalized.

Some experts have argued against treating patients with advanced disease, saying that reversing cognitive and functional decline is not likely, nor is reducing costs associated with Alzheimer's. Others, however, maintain that leveling off the decline and delaying institutionalization may be worthwhile.

This trial, led by Dr. Sandra Black of the University of Toronto, Ontario, Canada, included 343 people with severe Alzheimer's disease treated at clinics worldwide. The patients were randomized to receive either Aricept or a placebo for six months.

Cognitive function either stabilized or improved in 63 percent of people taking Aricept, versus 39 percent of people on the placebo. People taking Aricept also showed improvements in memory, language, attention and recognizing their own name, as well as less of a decline in social interaction and arranging sentences.

The most common side effects were diarrhea, insomnia, nausea, infection and bladder problems, all of which are consistent with what experts already knew about the drug.

"It's rare that the benefits [in cognition] are really robust -- that when a patient walks in the door, I know they're better," Kennedy said. Despite improvements, however, "The drug didn't reduce behavioral disturbances, it didn't make patients more independent, and it didn't make caregivers much less burdened," he said.

Notwithstanding these limitations, it is likely that the drug will now be widely prescribed for patients in all stages of Alzheimer's, Kennedy said. "This drug is, to some extent, counteracting criticisms that medications only work in the early stages," he said. "It's no longer considered ethically supportable to keep patients off drugs."

More information

Learn more about Alzheimer's from the U.S. U.S. National Institute on Aging.