Up to one-third of all nursing home residents, primarily patients with dementia, receive antipsychotic medications for the treatment of symptoms of psychosis, agitation, and aggression (
1). In patients with dementia, the presence of psychosis or agitation predisposes to worse long-term outcomes, including a higher rate of institutionalization and death (
2). Antipsychotic medications that are used to treat these symptoms are associated with a variety of side effects. In particular, a meta-analysis of placebo-controlled trials of antipsychotics in dementia revealed a 60%−70% greater risk of mortality, which led to a Food and Drug Administration (FDA) black box warning in 2005 (
3). A more recent review found a 54% greater mortality risk for antipsychotics compared with placebo in these patients (
4).
The study by Lopez et al. (
5) in this issue of the
Journal examines whether the presence of psychiatric symptoms could account for the negative long-term effects of antipsychotic treatment, particularly institutionalization and death. In a series of 957 outpatients with probable Alzheimer’s disease who were followed for an average of 4.3 years during the period from 1983 to 2005, 241 patients received both conventional and atypical antipsychotics. The presence of psychosis predicted nursing home admission and time to death, both of which occurred at more than double the rate for patients who received conventional antipsychotics than in those who received atypical antipsychotics. However, this effect of conventional antipsychotics disappeared after several key variables were controlled in the analyses, particularly the co-occurrence of psychiatric symptoms. Critically, in this study the risk of death was not increased in patients exposed to antipsychotics, both conventional and atypical. Worse cognitive performance, extrapyramidal signs, heart disease, agitation, and psychosis were associated with an increased likelihood of admission to a nursing home, while increased age, education level, male gender, worse cognitive performance, extrapyramidal signs, and psychosis were associated with a greater risk of death. Time-dependent statistical models were used, which is the state-of-the-art approach when dealing with longitudinal data that include symptoms and conditions that might appear or disappear at different time points during follow-up.
Some surveys conducted in nursing homes also did not find an increase in mortality in patients with dementia who received antipsychotic medications (
6,
7). In contrast, a report from a large Medicare and Medicaid database of 75,445 patients (
8) demonstrated greater mortality risk associated with the conventional antipsychotic haloperidol compared with the reference antipsychotic, risperidone. Quetiapine was associated with a slight decrease in mortality risk compared with risperidone. In that study, the greater mortality risk for haloperidol remained after controlling for the presence of behavioral disturbances. Nursing home surveys and surveillance data from large health care databases are important indicators of widespread clinical practice, but psychiatric symptoms are typically assessed in these studies from serial data compiled by nurses and aides who utilize forms such as the Minimum Data Set that fulfill regulatory requirements (
8). Measures of this type are different from the fine-grained clinical observations that were made in the study by Lopez et al. at a specialty clinic in a major university-based medical center. The intensity of monitoring, and presumably the standard of care, is higher in such a setting compared with most outpatient clinical settings and nursing homes, and the approaches to prescribing psychotropic medications may have also been different. In academic clinical settings, patients with prior psychiatric history and treatment or the presence of comorbid disorders that increase mortality can be systematically excluded from the study sample (
5), but in nursing home surveys and health care databases this level of detailed information is not available for the careful selection of participants.
In the surveillance study from the Medicare and Medicaid database (
8), the increase in mortality related to antipsychotics in nursing homes was largely explained by the use of high dosages of antipsychotics. The use of high dosages of haloperidol nearly doubled the mortality risk compared with the use of low dosages of haloperidol, and high dosages of risperidone were associated with a 35% increase in mortality compared with low dosages of risperidone (
8). Many primary care physicians in nursing homes have been using high dosages that are above the narrow therapeutic windows identified for haloperidol (0.5–2.0 mg/day) (
9) and risperidone (0.5–2.0 mg/day) (
10) for patients with dementia. The atypical antipsychotic olanzapine has also been shown to lead to greater side effects at higher dosages without a concomitant increase in efficacy (
11). Dosage, a key clinical issue, was not analyzed in the study by Lopez et al., although presumably the dosages prescribed were relatively low, and this may account for the lack of an observed effect on mortality.
Lopez et al. conducted their study in outpatients, whereas most surveys have been conducted in nursing homes, where mortality is much higher. Nearly all of the placebo-controlled clinical trials that showed an overall increase in mortality for antipsychotics relative to placebo were conducted in nursing homes (
3). There remains a need to assess the impact of antipsychotics on nursing home admission and mortality in outpatient settings, where the majority of patients with dementia are evaluated and treated. Research in this area is further complicated by the difficulty in identifying the appropriate comparison group for patients who receive antipsychotics. A comparison group of patients who did not receive antipsychotics is also likely to have less comorbidity and medical illness, thereby confounding the interpretation of any differences observed in mortality rates. One approach that has been used is to compare different antipsychotics to a reference antipsychotic, which provides useful information about relative risk (
8).
The FDA warning about increased mortality risk was based on a meta-analysis of placebo-controlled short-term trials of antipsychotics in patients with dementia. Placebo-controlled trials remain the gold standard and cannot be entirely replaced by naturalistic data from surveys, even though the sample sizes in the latter studies are much larger. In some surveys (
6,
7), a possible explanation for the lack of association between antipsychotics and mortality may be that some physicians changed their prescribing behavior after the black box warning about mortality was issued. Physicians may have become less likely to prescribe antipsychotics to nursing home patients with severe medical illness and instead may have restricted their use of antipsychotics to patients with mild to moderate medical illness, thereby lowering mortality.
So what should the psychiatrist do in clinical practice with these patients? The FDA has not approved the use of any antipsychotic to treat psychosis or agitation in patients with dementia, but off-label use is permitted, as occurs frequently in the treatment of many other disorders. Of note, risperidone is approved in Germany for the treatment of behavioral complications in dementia. Federal regulations in the United States require the discontinuation of antipsychotic medications in nursing homes every 3–6 months, and the Center for Medicare and Medicaid Services recently issued an additional regulation (
12) requiring nursing homes to reduce antipsychotic usage by 15%, an arbitrary number. However, there is often a need to continue antipsychotics in many patients with dementia. The Antipsychotic Discontinuation in Alzheimer Disease trial reported that after treatment response to risperidone was maintained for 4 to 8 months, randomized double-blind discontinuation to placebo was associated with a markedly increased risk of relapse relative to continuation risperidone (
13). Clearly, completely avoiding the use of antipsychotics is not feasible in many patients with dementia who develop symptoms of severe agitation and psychosis that require urgent treatment, particularly because no other class of medication has been shown to be consistently superior to placebo in controlled trials (
14,
15). Behavioral strategies have a place, but no study to date has had an adequate control group with a duration of patient-caregiver exposure to clinical staff comparable to that of the intervention group (
16). In patients who need an antipsychotic to treat symptoms of psychosis or agitation, the recommended strategy is to start the medication at a low dosage and to raise the dosage slowly to avoid exceeding the upper limit of the therapeutic window, thereby optimizing the risk-benefit ratio.