Nonprofit, nonpartisan journalism. Supported by readers.


UCare generously supports MinnPost’s Second Opinion coverage; learn why.

Antipsychotic drugs are riskier for older dementia patients than previously thought, study finds

U.S. Government Accountability Office

Physicians often prescribe antipsychotic drugs to older people with dementia to control non-memory related behavioral symptoms, including agitation, aggressiveness, delusions and hallucinations. 

But officials at the U.S. Food and Drug Administration (FDA) have never approved antipsychotic medications — such as haloperidol (Haldol), risperidone (Risperdal), olanzapine (Zyprexa) and quetiapine (Seroquel) — for that purpose. Indeed, these medicines come with a “black box” FDA warning that their use to control behavioral disturbances in people with dementia is associated with an increased risk of premature death.

Many physicians have continued to prescribe the drugs to their patients with dementia anyway, believing that the benefits — primarily protecting patients from harming themselves or others — outweigh the risks.

A study published Wednesday in the journal JAMA Psychiatry, however, presents troubling new evidence that those risks are significantly higher than previously reported.

A VA database

For the study, researchers from the University of Michigan, the University of Southern California and the Veterans Administration analyzed medical data of almost 91,000 U.S. veterans (mostly men) aged 65 and older with diagnosed dementia. Approximately half the patients had been treated for behavioral problems with an antipsychotic, anticonvulsant (valproic acid) or antidepressant medication; the other half didn’t receive any of these drugs.

After crunching the data, the researchers found that veterans who were given antipsychotic medications had a risk of dying significantly higher than those not taking the medications. For example, 20.7 percent of the patients taking haloperidol died within six months — 3.8 percent more than the non-users. Among the patients taking quetiapine, 11.8 percent died within six months — 2.0 percent more than the non-users. The increased risk of death for the olanzapine and risperidone patients fell somewhere between.

These risks are two to four times higher than previously cited in the medical literature, the study’s authors point out.

The new analysis also revealed that the higher the dose of an antipsychotic medication, the greater the risk of premature death.

Another troubling finding was that people prescribed haloperidol — the riskiest of the drugs — were more likely to be unmarried, African-American or living in facilities with fewer beds for patients.

As for the two other types of drugs used to treat behavioral problems in people with dementia  — valproic acid and antidepressants — the researchers found that the increased risk of premature death associated with the anticonvulsants was not statistically significant and the increased risk associated with the antidepressants was only slightly higher than for non-users.

Number needed to harm

Another way of assessing the risk from the antipsychotic medications is with an epidemiological measurement tool known as the “number needed to harm” (NNH). In this study, the NNH reflects the number of older adults with dementia who would have to be taking a particular drug for one of them to die within six months.

For the haloperidol users in the study, the NNH was 26. In other words, for every 26 patients with dementia taking haloperidol, one would be expected to die within half a year. Risperidone was only slightly less risky, with an NNH of 27. For olanzapine the NNH was 40, and for quetiapine it was 50.

By comparison, the NNH in this study for antidepressant use was 166. (There’s little evidence, however, that antidepressants are effective for patients with dementia.)

Time to raise the threshold

Earlier this month, federal investigators with the Government Accountability Office (GAO) issued a report in which they cited evidence of the widespread overuse of psychiatric drugs in older Americans with dementia. 

In 2012, about one-third of older adults with dementia who spent at least 100 days in  long-term nursing home and about 14 percent of those living outside a nursing home were prescribed an antipsychotic medication, the report noted.

“The harms associated with using these drugs in dementia patients are clear, yet clinicians continue to use them,” said Dr. Donovan Maust, the lead author of the JAMA Psychiatry study and a geriatric psychiatrist at the University of Michigan and the VA Center for Clinical Management Research, in a released statement.

 “That’s likely because the symptoms are so distressing,” he added.“These results should raise the threshold for prescribing further.”

Maust says the emphasis on treating behavioral problems in older people with dementia should focus on non-drug strategies first. But, of course, those strategies take more time and are not always fully reimbursed by Medicare or private health insurance.

 “Non-pharmacologic approaches will only succeed if we as a society agree to pay front-line providers for the time needed to ‘do the right thing,’ ” said one of Mast’s co-authors, Dr. Helen Kales, in the released statement.

Unfortunately, the JAMA Psychiatry study is behind a firewall at the journal’s  website, despite the fact that it was funded in large part by taxpayer money through the National Institute of Mental Health and the National Institute on Aging. 

You can read the GAO report on the overprescribing of antipsychotic medications at that agency’s website.

Comments (3)

  1. Submitted by Rachel Kahler on 03/19/2015 - 10:01 am.

    This will sound callous, but…

    We should be focusing on quality of life rather than quantity of life. If antipsychotic drugs provide quality of life over other options, then the concern for “premature” death should be minimal. If, however, antipsychotic drugs reduce quality of life over other options, then they should not be used and coverage by various insurance should be evaluated on ethical grounds. If, on the third hand, the results in quality of life are equal over other treatments, which are not covered by insurance, then cost should be considered as a factor. If there is no benefit whatsoever, they should not be used at all. Period.

    It’s interesting that this column has focused on life quality and end of life conversations, yet still highlights “problems” like these without looking at the quality of life issues associated with them. Based on this particular article, it would seem to me that the only concern is risk of death. Like it or not, we all die. Whether or not death is “premature” in these situations might need to be discussed.

  2. Submitted by Thomas Jaeger on 03/19/2015 - 11:24 am.

    Never easy to treat Dementia

    I was able to review this study methods. While the results are important, a number of limitations need to be kept in mind. In the end, it certainly adds to the body of evidence that identifies risk of anti-psychotics, but I don’t believe that we can accept the results implicitly.

    The authors did an admirable job at what they set out to do, but the study- a retropsective cohort study- has inherent limitations. First, they used administrative data, which though much more reliable and accessibile isn’t the same as directly observing a pateint’s care. For instance, though they tried to measure other risk factors for death, you just don’t know if the data was uniformly accurate in relation to medical or behavioral variables. Second, though they tried to adjust for a number of confounders, no observational study can truly control for all factors that might contribute to an individuals death. I have no doubt that these drugs have risk, but we can’t eliminate the uncertainty of whether the use of antipsychotics isn’t associated with something else that contributes to their risk of death. For example, they couldn’t accurately measure or account for the severity of dementia or of behavioral symptoms, and much of dementia is due to vascular disease that is itself often fatal. And there is no attempt (nor would it have been possible) to account for the fact that these pateints were cared for at many different locations, and the caregiver’s skills or barriers to care could be all over the map.

    Because of these and other limitations, we need to remember that the number needed to harm reported is an estimate, not a directly measured effect of the antipsychotics.

    The biggest mistake we could make from this study is to conclude that anti-epileptic drugs or anti-depressants should be the preferred drugs. The hypothesis and methods of the study simply cannot make this comparison, as the other drug classes are used for different problems. Might they be less risky? Sure, but they probably have little or no efficacy for severe agitation.

    Carng for patients with Dementia is one of the largest public health issues of our generation. It is extremely challenging, and expensive. The pool of caregivers trained to anticipate and respond to their patient’s needs without medication simply does not exist. We need to be cautious when presribing anti-psychotics for behavioral problems associated with Dementia. But there are still situations where there is no good alternative. I

    The study by Maust and colleagues is the type of study needed to identify potential harm, and the estimate of magnitude of risk is valid to a point. IMHO we shouldn’t use this data to make over-reaching pronouncements about never using this or that agent in times of desperate need.

  3. Submitted by Tom Hicks on 09/16/2015 - 12:50 am.

    Antipsychotic Risk from ONLY the Raw Data

    Please try and slam me. I want to learn what this debate is all about.

    (Qualifier: I am only looking at the raw data and am not adhering to statistical detail such as CI, P values and range. I notice that the author of this article has not included them either.)

    I crunched some numbers from ONLY the raw data reported in the study referenced at bottom.

    Risperidone 180 day trial data shows a 1.64 FOLD INCREASE IN MORTALITY
    Of 6,338 pairs, 883 patients taking the drug died and 538 patients not taking the drug died.
    Therefore, 883/538 = approx.1.64 FOLD increase in the risk of death OR 64% BUT BUT BUT only 5.44% more of the patients in the group that received the drug died during the specific time period of 180 days . (883-538) / 6,338 = approx. 5.44%.

    Showing a percentage like 5.44% can only make sense when the trial period is also given AND then it is still confusing. I believe this has misled many physicians to assume that there is only this slight % increase in mortality risk. ( 5.44% vs 64%. ) In most of the articles they publish the risk is reporting as “a small increase in the risk of mortality”.

    Example: Assuming that we use only the raw data to infer mortality risk for other time periods.
    When 8.49% of the nonusers are dead, 13.93% of the users will be dead (5.44% increase)
    When 50% of the nonusers are dead, 82% of the users will be dead (32% increase)
    When 61% of the nonusers are dead, 100% of the users will be dead (39% increase)

    The reporting of the mortality risk statistics in these articles is misleading and a great many articles report the “slight percentage increase” without even reporting the trial time period at all.

    Of interest, the other two antipsychotics drugs reported in this study show lower but similar risk (approx. 1.42 and 1.44 FOLD using raw data). The similarity seems to be significant and possibly amazing that such a patter is shown.
    Also, the data agrees well with the Study used for the FDA Black Box WARNING 4/11/2005
    “These studies enrolled a total of 5106 patients, and several analyses have demonstrated an approximately 1.6-1.7 fold increase in mortality in these studies.”

    Notice that the FDA WARNING says “FOLD increase”. It sounds like the FDA has made things more understandable and used the data to infer a 60%-70% increase in the risk of death for any time frame rather than presenting the confusing percent increase in the number of patients that died during a specific time frame.

    I found the raw data here:
    Antipsychotics, Other Psychotropics,
    and the Risk of Death in Patients With Dementia
    Number Needed to Harm
    JAMA Psychiatry Published online March 18, 2015

Leave a Reply