his large longitudinal observational study revealed that a higher proportion of patients exposed to antipsychotic medications, especially conventional antipsychotics, were admitted to a nursing home or died compared with those who never took these medications. However, in time-dependent statistical models, these associations were no longer present after we adjusted for the symptoms for which the antipsychotic treatments were usually used (hazard ratio=2.2 compared with 1.3). This suggests that the primary correlate of negative outcomes was the psychiatric symptomatology and not the drugs used to treat these symptoms.
This observational study does not support the association between mortality and antipsychotic use that has been reported in institutionalized elderly patients (34). The discrepancy may be a result of the fact that nursing home patients are more clinically heterogeneous and tend to be more cognitively and physically impaired than individuals commonly seen in outpatient clinics. Furthermore, interactions among multiple factors (including psychiatric symptoms) that have yet to be identified may be associated with antipsychotic-associated death in nursing home patients. For example, in a study conducted in Finnish institutionalized patients with dementia, while neither conventional nor atypical antipsychotics were associated with death, the use of restraints doubled the risk of mortality (13).
Similarly, outcome studies from large databases should be interpreted with caution, as they include not only institutionalized and noninstitutionalized patients, but also patients with a lifetime history of psychiatric disorders and exposure to antipsychotics or with neurodegenerative processes that are themselves associated with increased mortality (e.g., Parkinson’s disease) (35, 36). This may explain the lack of association found in some studies conducted in elderly individuals that did not include these types of patients (11, 13). In addition, it seems that the highest reported risk of death occurs with conventional antipsychotics (e.g., haloperidol) and with some atypical antipsychotics (e.g., risperidone), but not with others (e.g., quetiapine) (34, 37). Although our study had enough participants to examine the effects of antipsychotics on mortality and nursing home admission, we lacked statistical power to compare the effects among individual antipsychotic agents.
The results of this study suggest that it is the symptoms and not the medications that predict nursing home admission and death. The study of long-term effects of antipsychotics in patients with Alzheimer’s disease has strengths and limitations. It is difficult in observational studies to determine the duration of exposure to these medications as well as dosages. In addition, brief exposure to medications may be missed, especially if these occur between clinic visits or in the late stages of the disease. Thus, in the present study, if medication initiation and death occurred in the last 6 months between contacts, we may not have detected the use of antipsychotics. While studies based on claims data provide better information about dosages and date of therapy onset, they do not have information about whether the medication was actually used or about previous exposure to antipsychotics (34). Similarly, while randomized controlled trials provide useful short-term information, they would have to last many years in order to capture time to nursing home admission or death. Furthermore, if a patient in a placebo group had increasingly severe symptoms, he or she would have to be placed on medication (i.e., breaking the blind) and then continued in long-term observation. If the medications are effective at treating the symptoms (in the medicated group), any decision to stop the medication incurs the risk that the symptoms will immediately return in some patients, who may then require long-term maintenance therapy. Indeed, in a recent study of risperidone treatment for patients with Alzheimer’s disease with psychosis and agitation, symptom relapse occurred when the antipsychotic was discontinued in a randomly assigned, double-blind fashion (38).
Psychosis was a significant predictor of death, even after adjustment for antipsychotic use. This finding is consistent with a previous observation of a faster cognitive decline (21) and increased mortality (23) in patients with Alzheimer’s disease with psychotic features. This suggests that these patients live with a more aggressive Alzheimer’s disease phenotype, but the study of psychiatric symptoms and mortality in Alzheimer’s disease is complex. The psychotic phenomenon in Alzheimer’s patients does not occur in isolation and tends to develop in a constellation of symptoms that include agitation, aggression, sundowning, and inappropriate behavior across the spectrum of disease severity (29). Thus, the psychotic symptoms may co-occur with other disturbing behavioral symptoms that can lead to death (39). For example, agitation and aggression can lead to falls, which can cause head trauma or fractures, in turn leading to physical immobilization and subsequent increased risk of death. Aggression was also a risk factor for death when we examined the relationship between antipsychotic exposure and mortality (hazard ratio=1.30 (95% CI=1.01–1.67; see Table S4 in the online data supplement).
Extrapyramidal signs were an independent predictor of nursing home admission and death, consistent with previous studies (40, 41). Notably, extrapyramidal signs are frequent in patients with Alzheimer’s disease, and practically all will develop an extrapyramidal sign during the course of the disease (41). The presence of extrapyramidal signs can predispose patients to reduce their mobility, consequently increasing the risk of infections (e.g., pneumonia). The increased need for personal care in patients with extrapyramidal signs may also lead to nursing home admission. In this study, extrapyramidal signs were more frequent in patients taking conventional than atypical antipsychotics, but we did not examine the effect of the severity of the extrapyramidal signs (as distinct from their presence or absence) on mortality.
The study of factors related to mortality in patients with Alzheimer’s disease is complex, with multiple factors converging to increase the risks of nursing home admission and death. In this outpatient-based population with mild to moderate dementia, exposure to antipsychotics was not associated with an increased risk of nursing home admission after the presence of disruptive behaviors was taken into account. Rather, it was the psychotic/agitated phenotype that emerged as a critical factor influencing the natural and treated history of Alzheimer’s disease.