Psychiatric Medication and Abnormal Behavior as Predictors of Progression in Probable Alzheimer Disease

Objective  To examine whether the use of psychiatric medication and the presence of abnormal behaviors affects the progression of Alzheimer disease.

Design  Cross-sectional with longitudinal follow-up and the likelihood of arriving at 4 end points: (1) Mini-Mental State Examination score of 9 or lower; (2) Blessed Dementia Rating Scale score of 15 or higher for activities of daily living; (3) nursing home admission; and (4) death, evaluated using a proportional hazard model with 9 variables: psychosis, insomnia, wandering, aggression, psychomotor agitation, depression, and use of antidepressants, antipsychotic agents, or sedatives/hypnotics.

Setting  Multidisciplinary dementia research clinic.

Patients  We examined baseline and follow-up behavioral symptoms and the use of psychiatric medication in 179 mildly to moderately impaired patients with probable Alzheimer disease participating in a longitudinal study of dementia. Patients were observed from 2.4 to 172 months (mean duration±SD, 49.5±27.4 months).

Results  Nine patients (5%) were taking sedatives/hypnotics; 16 (9%), antipsychotic agents; and 22 (12%), antidepressants at study entry. Patients taking antipsychotic agents had lower Mini-Mental State Examination scores and higher Blessed Dementia Rating Scale scores for activities of daily living than patients not taking any medication. Using proportional hazard analysis with time-dependent covariates for individual psychiatric symptoms and medications, we found that the development of psychosis was associated with functional decline (time to Blessed Dementia Rating Scale score of ≥15), institutionalization, aggression, and agitation with functional decline after adjusting for age at study entry, education, Mini-Mental State Examination scores, and Blessed Dementia Rating Scale scores. Use of antipsychotic medication was associated with functional decline, and sedatives/hypnotics with death. Neither the presence of psychiatric symptoms nor use of medication was associated with rate of cognitive decline (time to Mini-Mental State Examination score of ≤9).

Conclusions  These findings indicate that the use of antipsychotic agents and sedatives can affect the natural course of Alzheimer disease. Psychosis, agitation, and aggression are important predictors of outcome, even when the effects of medication to treat them is taken into account.

PSYCHOTIC SYMPTOMS (eg, delusions, hallucinations), aggressive behavior, psychomotor agitation, wandering, and depression are common in patients with Alzheimer disease (AD)^1-3 and are the source of major problems for families and treating physicians. Furthermore, the presence of psychosis is a major predictor of cognitive and functional decline and institutionalization in AD,^4-7 and agitation and wandering can affect disease progression.^8,9 However, although the literature seems to support the perspective that psychosis and abnormal behavior can affect progression in AD, there are few studies on whether depression is a predictor of progression of AD. After a 1-year follow-up, Lopez et al¹⁰ found that depression did not affect cognitive decline, although Burns et al¹¹ found that patients with AD and depression died earlier than those without depression. By contrast, other studies did not find an association between depression and death in AD.^8-12 Importantly, none of the studies that assessed psychiatric symptoms as predictors of progression in AD have taken into account the medication used to control these symptoms.

Antipsychotic^13-15 and nonantipsychotic^16-26 medications are widely used to treat abnormal behavior in patients with AD. Despite their widespread use, there are only a few medication-related outcome studies conducted in large cohorts, and meta-analyses of controlled trials of abnormal behavior treatment in dementia show that no single antipsychotic or nonantipsychotic agent is better than another.^27,28 Furthermore, antipsychotic medications have the potential to worsen cognitive deficits (through their anticholinergic effect) and functional abilities (through secondary parkinsonism, orthostatic hypotension, and sedation). Because of these adverse effects, these medications have the potential to alter the natural course of AD. However, there is little information about the effects of psychiatric medication on the progression of the disease. Similarly, although antidepressant medication is used to treat both depression and behavior,¹⁷ to our knowledge there are no studies on its effect on the natural course of AD.

In this study, we examined the predictive value of psychiatric medication and of abnormal behavior on progression of the disease in a large, carefully selected group of patients with probable AD.²⁹ Furthermore, this longitudinal study provides an opportunity to examine physicians' treatment of the most common behavioral problems.

We examined the clinical characteristics of 179 patients with probable AD²⁹ enrolled in a longitudinal study of dementia at the University of Pittsburgh from 1983 to 1988. Each patient received extensive neurologic, psychiatric, and neuropsychological examinations. Details of this cohort, as well as inclusion and exclusion criteria, have been published previously.¹ Patients with clinical or radiological evidence of cerebrovascular disease were not included in the study. This included patients with a Hachinski Ischemic Scale score³⁰ higher than 4 or with a single ischemic lesion found in at least 2 consecutive cuts on a computed tomographic scan of the brain. Patients who had a small ischemic lesion in a critical area affecting either cognitive function or motor function were not included in the study. In addition, patients with poor control of hypertension (eg, inadequate medical control of the medication or inadequate patient compliance) or secondary forms of hypertension with poor therapeutic response were excluded, as were patients with severe and uncontrolled ischemic heart disease (eg, signs and symptoms of congestive heart failure, cardiac dysrhythmia) or diabetes mellitus. Similarly, patients who met criteria for Parkinson disease at the baseline examination or had developed extrapyramidal signs before the onset of dementia were excluded from the cohort.

Subjects were examined annually, and the follow-up period ranged from 0.2 to 14.16 years (mean follow-up period±SD, 4.16±2.38 years). Five patients were observed for less than 1 year, 13 for 1 to 2 years, 73 for 3 to 4 years, 54 for 5 to 6 years, 14 for 7 to 8 years, and 10 for 7 years or more. Seventy-five patients had the diagnosis of definite AD.²⁶

Psychiatric evaluations were conducted annually by psychiatrists using a semistructured interview with patients and primary caregivers.³¹ None of these patients had psychiatric disorders before the onset of dementia. Details of the clinical criteria for individual psychiatric symptoms have been published previously.³²

The medications used by the patients were recorded on a summary form that included dosages, date of the initial prescription, and date of discontinuation of medication. The selection, administration, and response evaluation of each treatment were conducted by the referring physician. For the purpose of this study, psychiatric medications were classified as antidepressants, antipsychotic agents, and sedatives/hypnotics/anxiolitics (SHAs). Five patients taking diphenhydramine hydrochloride were not entered into the survival analysis. None of these subjects was treated with cholinesterase inhibitors³³ or anticonvulsants as mood stabilizers.^23,24

Functional capacity was assessed with the Blessed Dementia Rating Scale (BDRS) for activities of daily living.³⁴ We used a BDRS score of 15 or higher as the end point, which represents a moderate to severe functional impairment.

Global cognitive function was assessed using the Mini-Mental State Examination (MMSE).³⁵ We used an MMSE score of 9 or lower as the end point in the analysis because we considered it representative of moderate to severe AD.

We considered institutionalization to have occurred when the patient started to receive health care at home (eg, visiting nurses), or when he or she was institutionalized, regardless of the level of care (eg, personal care facility, health care facility).

Student t tests were used to analyze demographic and neuropsychiatric data. Proportional hazard models were used to determine if there were differences in the time from study entry to death, institutionalization, BDRS scores of 15 or higher, and MMSE scores of 9 or lower between patients with AD with and without psychiatric symptoms, and as a function of medication use and type. Proportional hazard models were also used to individually assess the effect of the presence of individual symptoms or medications on these same outcomes.

Because the development of psychiatric symptoms and use of medication is more frequent as the disease progresses,^36,37 the effects of individual symptoms and medications were assessed as time-dependent covariates. This allowed us to determine the effect of medications and symptoms if they were present after baseline examination.

Proportional hazard models, developed using a stepwise selection procedure, were used to determine which psychiatric symptom or medication was independently associated with the outcome measures, after adjusting for possible confounding variables such as education, age, and sex. For baseline BDRS and MMSE scores, indicator variables were created to show if the patient had a score greater than or equal to the median value (eg, MMSE score, ≤19; BDRS score, ≥6). Indicator variables were also created indicating if the patients were 70 years or older at study entry or had an education level of 12 years or higher. These indicator variables were included in the stepwise selection procedures.

Once a model was selected, the proportionality of the hazards was assessed using a time-dependent variable for each factor in the model. No significant deviations from proportionality were noted. For those patients not reaching an outcome (eg, death, institutionalization, BDRS score of ≥15, MMSE score of ≤9), the data were censored using the date of the last contact.

A total of 41 patients (23%) were taking at least 1 psychiatric medication at study entry; 22 (12%) took antidepressants; 16 (9%), antipsychotic agents; and 11 (6%), SHAs. Table 1 shows the demographic and neuropsychiatric characteristics of the patients. Patients taking antidepressants were younger (t₁₇₈=−3.24, P=.001), and had higher Hamilton Depression Rating Scale scores³⁸ (t₁₇₈=2.29, P=.02) than patients who were not taking any medication. Patients taking antipsychotic medication had lower MMSE scores (t₁₇₈=2.09, P=.02) and higher BDRS scores for activities of daily living (t₁₇₈=−1.90, P=.05) than patients who were not taking any psychiatric medication. There was no statistical difference between patients taking and not taking psychiatric medication in disease duration, education, Hachinski Ischemic Scale score, or New York University Scale for Parkinsonism³⁹ scores.

Thirty-three patients (18%) had episodes of verbal or physical aggression; 42, (23%) psychomotor agitation; 13, (7%) wandering; 52, (29%) insomnia; 24, (13%) major depression; and 46 (26%), psychotic symptoms at baseline examination (Table 2). There were more depressed patients using medication than not using medication (χ²₁=36.3, P<.001). During follow-up, 33 patients (18%) were noted to be aggressive; 69 (38.5%) experienced psychomotor agitation; 34 (19%), wandering; 67 (37%), insomnia; 16 (9%), depression; and 68 (38%), psychosis. There were more patients treated than not treated who had aggression ( χ²₁=35.1, P<.001), psychomotor agitation (χ²₁=42.3, P<.001), and depression (χ²₁=17.7, P<.001).

Since patients can have more than 1 psychiatric symptom and can be taking more than 1 medication, Table 3 shows how clinicians targeted each symptom. Antipsychotic agents and SHAs, alone or in combination, were used to treat aggression, agitation, wandering, and psychosis. Antidepressants were used to treat depression, insomnia, and nocturnal agitation.

At baseline examination, 7 (17%) of the 41 treated patients were taking 2 or more drugs; 3 patients (7%) were treated with both antidepressants and antipsychotic agents; 3 (7%), with SHAs and antidepressants; and 2 (5%), with 3 types of medication. Table 4 shows the type of medication used by the patients. During follow-up, 61 patients (34%) took psychiatric medications; 21 (12%) took antidepressants; 40 (22%), antipsychotic agents; and 23 (13%), SHAs (see Table 4). Twenty-four patients (15%) were treated with more than 1 psychiatric medication; 6 (3%), with antidepressants and SHAs; 7 (4%), with antidepressants and antipsychotic agents; and 12 (8%), with antipsychotic agents and SHAs. Seven patients (4%) used more than 1 antipsychotic agent through the course of the disease; 2 (1%) used more than 1 antidepressant; and 2 (1%) used more than 1 SHA.

Seventy-four (41%) of the 179 patients were institutionalized; 101 (56%) died; 55 (31%) reached a BDRS score of 15 or higher; and 76 (42%) reached an MMSE score of 9 or lower. Cox proportional hazard models were used to examine the univariate effect of medication and individual psychiatric symptoms (occurring at any time during follow-up) on the outcome variables (Table 5). There was a significant difference in the distribution of the time from study entry to institutionalization (relative risk [RR]=2.10, P=.01) and BDRS scores of 15 or higher (RR=2.04, P=.03) between patients with AD with and without psychosis. Of the psychiatric medications, antipsychotic agents were associated with time to BDRS scores of 15 or higher (RR=2.02, P=.03) and with institutionalization (RR=2.14, P=.02); SHAs were associated with time to death (RR=1.96, P=.05).

Multivariate proportional hazard models were used to determine whether there were significant independent associations between medication use, psychiatric symptoms, and outcome while controlling for age at study entry, education, sex, and functional and cognitive severity of the disease at study entry (Table 6). Psychosis, aggression, and agitation were all independently associated with BDRS scores of 15 or higher; psychosis was also associated with time to institutionalization. Antipsychotic agent use was associated with BDRS scores of 15 or higher, and use of SHAs with time to death. Of the demographic variables, age was associated with time to death, and MMSE scores of 19 or lower with both cognitive and functional end points.

The prevalence of specific types of psychiatric symptoms was consistent with previous studies^40-43; the number of patients with psychomotor agitation, insomnia, psychosis, and wandering increased during follow-up, while the number with depression and aggressive behavior decreased as the disease progressed.³⁶ Psychotic symptoms were significant predictors of institutionalization and functional decline, but not of death or cognitive decline.^5,7 However, other studies have found that psychotic symptoms also predict cognitive decline.^4,6,7 This discrepancy could be attributed to sample size, length of follow-up, and other methodological differences. Previously, we reported⁷ that the presence of psychosis at study entry predicted cognitive and functional decline in patients with mild to moderate AD. In the present study, we used a proportional hazard model that took into account the presence of psychosis at baseline and at follow-up. Perhaps those patients who experience psychosis early in the course of the disease will have more rapid cognitive and functional decline^44,45 than those who develop these symptoms later.

Although depression has been associated with increased risk of death in patients with AD¹¹ and in nondemented elderly individuals,⁴⁶ studies like ours, using proportional hazard models,¹² do not find such an association. Agitation and aggression have been associated with institutionalization in demented individuals^47-49 and patients with AD.⁵⁰ However, to our knowledge there are few longitudinal studies that have assessed the effect of behavioral problems on progression of the disease. Walsh et al⁸ found that wandering and behavioral problems were associated with death in patients with AD, and Moritz et al⁹ found that selected behavioral symptoms (eg, agitation, wandering, hallucinations) predicted death. Our results are difficult to compare with those studies because we excluded at study entry patients with comorbid conditions that could affect time to death (ie, cerebrovascular disease). Further, the mean age of those samples was older than ours.

The rates of medication use in the present study were similar to those reported by Mendez et al,⁴¹ who found that antidepressants were used by 7.4% of the patients, antipsychotic agents by 11.5%, and sedatives/hypnotics by 7.4%. Generally, use of psychiatric medications increases during follow-up,^36,41 but this is due to rising use of antipsychotic agents and sedatives, not antidepressants.³⁶

In spite of concerns about anticholinergic effects of tricyclic agents,^51-53 which may exacerbate cognitive deficits in patients with AD, the use of antidepressants was not associated with any of the end points of interest. Further, current clinical practice intuitively favors the use of newer antidepressants (eg, selective serotonin reuptake inhibitors) that have fewer adverse effects than tricyclic agents. This seems to be supported by a small study that found that fluoxetine is better tolerated than amitriptyline by patients with AD and depression.⁵⁴ However, to our knowledge there are no placebo-controlled studies that have evaluated the efficacy of the tricyclic agents vs selective serotonin reuptake inhibitors.

Antipsychotic agents are successfully used to treat abnormal behavior in AD,⁵⁵ despite their serious adverse effects. Sedation, extrapyramidal signs, gait abnormalities, and orthostatic hypotension are associated with the use of antipsychotic agents in nondemented⁵⁶ and demented subjects.¹⁵ The most prescribed antipsychotic drugs were haloperidol and thioridazine, which is consistent with the standard therapy in nondemented psychotic patients. However, we found that the antipsychotic drugs that cause fewer extrapyramidal signs such as thiothixene and perphenazine were prescribed less often than those that cause more extrapyramidal signs.

It seems counterintuitive that the use of antipsychotic agents and psychotic symptoms independently predict functional decline. However, antipsychotic medication was used to treat agitation, aggression, insomnia, or a combination of behavioral disorders, not necessarily psychosis. The faster functional decline in patients treated with antipsychotic drugs could be related to increased sedation or to the motor abnormalities that often occur in patients taking these drugs.

Sedatives are among the oldest drugs to treat behavior in general, but there are limited data on their use for dementia.¹⁹ Benzodiazepines can produce amnesia⁵⁷ and increase the risk of falls and death in nondemented elderly individuals^58-60; short-term use of short-acting agents (eg, lorazepam) is recommended in the treatment of anxiety and sleep disturbances in AD.⁶¹ Our findings concur with previous observations of increased risk of death in nondemented elderly individuals treated with benzodiazepines. Therefore, the use of SHAs, including benzodiazepines, in the treatment of abnormal behavior in AD should be minimized. Although we found that some physicians favored the use of diphenhydramine, an antihistaminic, to treat nocturnal agitation and insomnia, there are no data supporting the use of this medication in AD.

One of the problems in the treatment of behavioral symptoms in AD is the lack of consensus regarding who should be treated; most of the patients with behavioral problems were not being treated at study entry, except those with depression. As the disease progressed, more patients began treatment, especially those with agitation or aggression; most of the patients with depression continued their treatment. Of the patients with psychotic symptoms, fewer than half received any medication. However, some of these patients may not have benefited from drug treatment⁶¹ because, for example, delusional episodes do not always cause anxiety or agitation, and hallucinations can be pleasant. By contrast, other patients can have anxiety and aggression associated with psychosis,^44,62,63 and these may require treatment.

The other problem in the treatment of behavioral symptoms in AD is which medication(s) should be used to treat these behaviors. Patients can have a single behavioral abnormality or a combination of symptoms. Antidepressants were not only used to treat depression but also insomnia and episodes of nocturnal agitation. Antipsychotic agents and SHAs were used to treat agitation, wandering, aggression, psychosis, and insomnia. This suggests that the choice of treatment may be guided by the most disturbing symptom, but there was not much agreement on the type of medication. For instance, patients with insomnia and nocturnal agitation could be treated with SHAs, antidepressants, or antipsychotic agents. Furthermore, patients may not respond to the initial therapy and so their regimens are frequently changed from one type of medication to another. Indeed, patients with agitation and aggression were the most difficult to control at follow-up. They required the use of more than 1 drug, and the treatment of these symptoms had a higher rate of failure. Eight patients (7 taking antipsychotic agents and 2 taking SHAs) required a change in their medication regimen, either because of no response to treatment or the development of adverse effects.

These results indicate that psychiatric symptoms, especially psychosis, agitation, and aggression, are associated with more rapid progression in AD. The data also illustrate the limited knowledge and problems that physicians have trying to find the best ways to use these medications to treat patients with AD. Although newer antidepressants and antipsychotic agents are emerging as the first line of treatment of psychiatric problems in AD, to our knowledge there are no studies that have compared "traditional" therapies with newer medications. Controlled trials are urgently needed to justify their frequent use.⁶¹

Accepted for publication September 24, 1998.

This study was supported by grants AG 03705 and AG 05133 from the National Institute on Aging, Bethesda, Md. Dr Becker is the recipient of Research Scientist Development Award (level II) K02-MH01077, Bethesda.

Corresponding author: Oscar L. Lopez, MD, Neuropsychology Research Program, 3600 Forbes Ave, Suite 502, Pittsburgh, PA 15213.