Frail, elderly patients with Alzheimer's disease who are given widely used antipsychotic drugs such as Zyprexa and Risperdal have a higher risk of dying than patients who are given sugar pills, and doctors should be cautious in prescribing the drugs, according to a new analysis of earlier studies.

The report is likely to complicate the limited choices that families face while caring for elderly patients with Alzheimer's. And it lends support to a decision by the Food and Drug Administration earlier this year to require warning labels on the drugs.

"These medications are only modestly effective and have significant side effects associated with them," said Peter V. Rabins, a professor of psychiatry at Johns Hopkins University School of Medicine who reviewed the new analysis. "The drugs should not be used for trivial problems like difficulty sleeping at night and anxiety," he added. "Clinicians first need to really weigh the potential risks and benefits in deciding whether the potential harm from the symptoms is worth the risk."

An editorial by Rabins was published yesterday, along with the new study, in the Journal of the American Medical Association.

About a third of the 4 million to 5 million Americans with the degenerative brain disorder are given antipsychotic drugs, estimated Samuel E. Gandy, chairman of the medical and scientific advisory council of the Alzheimer's Association, a nonprofit research and advocacy group.

The drugs are approved for the treatment of psychoses in adults but have not been specifically approved for use among elderly patients with dementia, so physicians prescribing them for this group are relying on their clinical judgment. Experts said the new study should help doctors better weigh the risks and benefits.

"It should make physicians cautious when putting patients with dementia or any frail patient on medicines of this sort that are basically major tranquilizers," Gandy said. At the same time, he and others noted, the choices are limited when patients fall into the acute grip of agitation and delusions.

The lead investigator, Lon S. Schneider of the University of Southern California, said his meta-analysis of 15 trials involving about 5,000 patients found that the risk of death rose from 2 percent to 3 percent for patients taking the drugs in studies that lasted about 12 weeks. Most of the deaths came from cardiac problems or respiratory disorders. Schneider said the drugs might increase the risk because they sedate patients and confine them to bed for longer periods, making them more likely to develop infections.

Although Schneider, Gandy and Rabins all agreed that the difference in risk was small and that the patients were vulnerable to begin with, Gandy said that, "because these drugs are so widely used, it could translate into a substantial number of individuals" at risk.