Antipsychotic Drugs in Dementia: Title and subTitle BreakWhat Should Be Made of the Risks?

The introduction of the first antipsychotic drug, chlorpromazine, into clinical practice more than 50 years ago revolutionalized psychiatry and neurology.¹ The efficacy of this drug and other drugs in the phenothiazine class demonstrated that a disease considered a “mental illness” could respond to a biologically mediated therapy and heralded the introduction of other neuromodulating therapies such as levodopa for Parkinson disease.

However, since their introduction, the phenothiazines and other antipsychotic neuroleptic agents have raised challenging questions about their adverse effects and toxic effects. Skeptics of the effectiveness of antipsychotic drugs in schizophrenia suggested that sedation rather than a direct drug action was causing patients to report fewer symptoms. This question was resolved by one of the first randomized controlled trials (RCTs) with an active placebo (the original streptomycin trial² for tuberculosis had been randomized but the control group was not administered a placebo) in which May³ demonstrated that the effectiveness of chlorpromazine could not be explained by sedation because a phenobarbital comparison group did not show similar benefit.

In the 1960s, questions arose about the induction of persistent and sometimes permanent movement abnormalities caused by neuroleptic drugs. Skepticism about this adverse effect was resolved only in the 1970s with the demonstration that individuals exposed to these drugs had higher rates of movement abnormalities, now termed “tardive dyskinesia” given their delayed onset, than individuals not exposed.^{4 - 5} At the same time, the prescribing of this class of agents broadened considerably beyond schizophrenia. Tourette syndrome,⁶ the chorea of Huntington disease,⁷ and bipolar disorder⁸ are examples of other disorders for which antipsychotic drugs are widely used.

Antipsychotic drugs are also used to treat the psychiatric and behavioral symptoms that affect 60% to 90% of patients with dementia.^{9 - 10} In nursing homes, where a majority of residents have dementia, 25% of individuals receive such drugs.¹¹ Modest efficacy for treating these symptoms was demonstrated by Schneider et al¹² in a 1991 meta-analysis, which found an 18% benefit over placebo.

In the 1990s, the introduction of atypical or second-generation antipsychotic drugs was met with enthusiasm because of lower reported rates of parkinsonian adverse effects and tardive dyskinesia. However, in 2003, Brodaty et al¹³ reported higher but not statistically significant rates of cerebrovascular adverse events and mortality in patients with dementia-associated agitation and psychosis who were treated with risperidone compared with those receiving placebo in a randomized trial. This finding raised a number of questions. Was it a false-positive result (type I error) related to multiple comparisons? If the finding was valid, was it a specific effect of risperidone or did it reflect a more general effect of all drugs in its class? And what might be the mechanism of such untoward outcomes?

The meta-analysis of randomized placebo-controlled trials by Schneider et al¹⁴ in this issue of JAMA offers important new information on several of these questions. First, the authors demonstrated an increased mortality risk among patients with dementia and neuropsychiatric symptoms treated with most of the second-generation agents available in the United States. A risk ratio (RR) of approximately 1.5 (range, 1.3-1.9) was found for all second-generation drugs for which data could be found, although the results reached statistical significance only when all studies were combined. These findings support the validity of the study by Brodaty et al¹³ and expand it to all drugs in the class for which data are available.

A second, equally important finding is that the one first-generation antipsychotic drug, haloperidol, for which comparable data could be located was also associated with increased mortality with an RR of approximately 2.1. This suggests that increased mortality is a risk for all first- and second-generation antipsychotic drugs when they are used to treat the neuropsychiatric symptoms of dementia. Unfortunately, Schneider et al could not access enough data to determine the causes of death; however, as the authors suggest, the importance of this question requires that relevant data be analyzed by an independent party.

These findings should have a direct effect on clinical practice. The results do not contraindicate the use of antipsychotic drugs in the treatment of patients with dementia who have psychotic symptoms and agitation; instead, they change the risk-benefit analysis such that antipsychotic drugs should be used only when there is an identifiable risk of harm to the patient or others, when the distress caused by the symptoms is significant, or when alternate therapies have failed and symptom relief would be beneficial. Antipsychotic drugs should not be used when nondrug treatments are available and the risk of harm or significant distress is low. Fortunately, a range of alternative therapies,^{10 ,15} including behavioral interventions and antidepressant drugs, can be effective.¹⁶

The study by Schneider et al raises 2 other issues that require extensive discussion. The authors’ demonstration of an increased mortality risk associated with antipsychotic drugs required detective work that would thwart most investigators. Similar challenges were encountered by those who raised concerns of increased cardiac morbidity associated with the cyclooxygenase 2 inhibitor nonsteroidal anti-inflammatory drugs.^{17 - 18} Because there is a significant likelihood that rare events will not be identified in the 2 RCTs required by the US Food and Drug Administration (FDA) for drug approval, the FDA must put more effort into formal postmarketing surveillance¹⁹ and require that data from all RCTs be made readily available within a specific time period after data collection is completed. In addition, mortality should be an end point in all drug studies premarketing or postmarketing (William B. Greenough III, MD, written communication, 2005), and postmarketing assessment of mortality should be made mandatory for all newly approved drugs.

The study by Schneider et al illustrates the effort it will take, at least in the short run, to acquire such data. These authors were resourceful and persistent in accumulating the unpublished studies that provided much of their data. Ironically, it was only because pharmaceutical companies were seeking an expansion of approved indications for antipsychotic drug use in dementia, and for nonsteroidal anti-inflammatory drugs in cancer prevention, that the risks came to widespread attention.

Another issue raised by Schneider et al is whether short-term data can be extrapolated to longer periods. The authors speculate about the impact of extrapolating from a 1% number of excess deaths at 8 to 12 weeks to a 4% to 5% risk difference over 1 year. They provide little supporting data on which to base such a speculation and they note that it is equally plausible that the risk is primarily expressed during the initial 12-week drug exposure. What is concerning is that no data are available to answer this basic question. This is another reason that better systems for collecting long-term follow-up data on morbidity and mortality must be established by the FDA.

So what should a clinician do when caring for a patient with dementia who develops psychotic symptoms or aggression? First, etiologies other than dementia need to be considered. Delirium, untreated or undertreated medical illnesses, overmedication, environmental triggers, lack of engaging activities, and misinterpretation of disease symptoms are among the potential etiologies of such behaviors and symptoms. Because all have specific therapies, they should be considered when such symptoms first develop. Second, clinicians should consider the risk/benefit ratio for each patient. For example, patients with hallucinations and delusions that are neither distressing nor placing them or others at risk or harm should not be treated with antipsychotic drugs. Third, once antipsychotic drugs have been prescribed, careful assessment and documentation of the need for continued care is necessary. The Omnibus Budget Reconciliation Act (OBRA) regulations of 1987 require such a reassessment in long-term care,²⁰ but the need for medication continuation should be regularly reassessed and justified for all individuals. Given the high rates of dementia in assisted living homes,²¹ similar practices should be instituted in those settings as well.

Finally, Schneider et al raise an important issue that requires further research. Could the increased morbidity and mortality they found also occur in other groups of frail individuals exposed to this or other classes of drugs? This hypothesis is plausible enough that immediate attention should be given to answering it. Unfortunately, the paucity of long-term data will limit the ability to study this question for many classes of drugs in the near future. International efforts are needed to improve long-term monitoring for adverse events and for research into the important questions raised by this meta-analysis.

Just as skepticism about the effectiveness of chlorpromazine led to refinements in the study of treatment efficacy and helped establish the RCT as a standard, the concerns about antipsychotic drugs should lead to a broad reexamination of the types of data required for drug approval and postmarketing surveillance. Recent events underscore the urgency of these issues.