a. Cholinesterase Inhibitors

The efficacy of tacrine in mild to moderate Alzheimer's disease has been extensively studied, although its effects on patients with more severe or very mild Alzheimer's disease or with other forms of dementia have not been assessed. At least five double-blind placebo-controlled trials with parallel-group comparisons including a total of more than 2,000 patients have been reported (131–135). Overall, these clinical trials consistently demonstrated differences between tacrine and placebo. Approximately 30%–40% of patients taking tacrine who completed the trials showed modest improvements in cognitive and functional measures over study periods ranging from 6 to 30 weeks, compared to up to 10% of those taking placebo. Modest improvement in these studies corresponded to maintaining or improving function by an amount typically lost over 6 months in untreated groups of similar patients with Alzheimer's disease. Response appeared to be related to dose, at least in the largest clinical trial (133), in which patients who could tolerate 120–160 mg/day were more likely to respond. However, only approximately 60% of the patients were able to complete the tacrine trials even at moderate doses; 30% of the subjects were dropped from these trials prior to completion because of elevation in hepatic transaminases, and another 10% had to leave because of other adverse effects, mainly cholinergic effects (e.g., nausea and vomiting). The benefits and adverse effects of administration beyond 30 weeks are unknown.

The efficacy of donepezil has been evaluated in more than 15 randomized, double-blind, placebo-controlled trials (136–146). Trial sizes have varied from fewer than 20 subjects to 818 subjects (138), and many trials have been multicentered. Most trials have been 12–24 weeks in duration, although at least two have lasted 1 year (137, 139), and one lasted more than 2 years (136). Studies have generally been conducted with community-dwelling patients, although one trial included nursing home residents with moderate to severe dementia (145). The majority of studies have enrolled patients with mild to moderate Alzheimer's disease (MMSE scores in the 10–26 range), although at least two have focused on patients with moderate to severe dementia (MMSE scores in the 5–17 range) (107, 142), and one focused exclusively on patients with early Alzheimer's disease with MMSE scores in the 21–26 range (146, 480).

Published studies consistently have found a benefit of donepezil over placebo in both cognitive and functional measures, including measures of clinicians' impressions of improvement. The size of the effect of donepezil has likewise been consistent across most studies, with improvement over placebo of about 1 point on the MMSE and 3 points on the ADAS-cog. On cessation of donepezil, improvement was lost over a 3–6-week period in all studies that examined this outcome. In one study in which donepezil treatment was interrupted for 6 weeks and then reinstated at the original dose, patients' cognition and function did not return to the level achieved prior to donepezil discontinuation (166). Most studies comparing 5 mg/day dosing with 10 mg/day found greater benefit with the higher dosage (138, 140), although this result has not been found in some studies comparing these dosages (141). Donepezil is administered once daily.

The question of whether donepezil treatment delays nursing home placement is an important one and has been addressed in two studies. One study found a delay in nursing home placement in patients treated with open-label donepezil for up to 240 weeks (481). However, these findings have been contested on methodological grounds (482). The AD2000 trial conducted in the United Kingdom followed 565 community-dwelling patients randomly assigned to receive donepezil or placebo for more than 2 years (136). Although donepezil-treated patients had better cognitive and activities of daily living scores than the placebo group, there was no difference in the primary endpoint of time to institutionalization. However, the trial was underpowered, had a high dropout rate, and may have been influenced by treatment interruptions (483).

The efficacy of rivastigmine has been evaluated in at least eight randomized, placebo-controlled, double-blind studies of patients with Alzheimer's disease (147–152). The sizes of the studies have varied from 50 subjects (148) to as many as 725 subjects (149). The duration of most trials was 26 weeks or shorter, although one trial lasted 12 months (150). All trials thus far have been conducted with community-dwelling patients with mild to moderate dementia severity (480).

These studies consistently have found a benefit of rivastigmine over placebo on both cognitive and functional measures, including measures of clinicians' impressions of improvement. The size of the effect of rivastigmine has likewise been consistent across most studies, with improvement over placebo of about 1 point on the MMSE and 3 points on the ADAS-cog, a magnitude of effect similar to that of donepezil. The dosage range found to have maximum efficacy is 6–12 mg/day in divided doses.

The efficacy of galantamine has been evaluated in at least eight randomized, placebo-controlled, double-blind studies (153–159). The numbers of subjects have ranged from under 100 to 978 (157). Duration of most trials was one-half year or shorter. All trials thus far have been conducted with community-dwelling patients with mild to moderate Alzheimer's disease (480).

These studies consistently have found a benefit of galantamine over placebo in both cognitive and functional measures, including measures of clinicians' impressions of improvement. The size of the effect of galantamine has likewise been consistent across most studies, with improvement over placebo of about 1 point on the MMSE and 3 points on the ADAS-cog, a magnitude of effect similar to that of donepezil. The dosage range found to have maximum efficacy is 16–24 mg/day in divided doses. An extended release, once-daily dosing form of galantamine has recently been released.

A number of randomized, double-blind, placebo-controlled trials have been conducted in patients with vascular dementia or mixed Alzheimer's disease and vascular dementia. In two 24-week trials conducted with patients with probable or possible vascular dementia (616 subjects and 603 subjects, respectively), donepezil (5 mg/day and 10 mg/day) was compared to placebo (484, 485). Improvements in measures of cognition and function were found in patients given either dose of donepezil, although in one of the trials (484) one of the two primary outcome measures did not demonstrate benefit of donepezil, compared with placebo. The effect size was comparable to that found in Alzheimer's disease trials. Two 6-month trials of galantamine (592 and 786 subjects, respectively) in patients with vascular dementia or mixed Alzheimer's disease and vascular dementia had similar findings (358, 486).

Of concern, in one unpublished trial of donepezil for vascular dementia, there was a significantly higher rate of death in the subjects taking donepezil than in the placebo group (167), raising a significant safety concern that requires further study.

One 20-week, randomized, double-blind, placebo-controlled study with 120 subjects examined the effects on cognition of 6–12 mg/day of rivastigmine in patients with Lewy body dementia (168). The results showed overall cognitive benefits with rivastigmine, compared with placebo, although differences were not statistically significant for all measures. A small 4-week, placebo-controlled, double-blind, double-crossover, randomized trial of donepezil also demonstrated cognitive benefits in subjects with Lewy body dementia (169).

In a 24-week, randomized, double-blind, placebo-controlled study with 541 subjects, the effects on cognition and function of 3–12 mg/day of rivastigmine were examined in patients with mild to moderate Parkinson's disease dementia (170). Improved cognition and function were found in the rivastigmine group, compared with the placebo group, and rivastigmine was tolerated by this patient population. In a 48-week open-label active treatment (3–12 mg/day of rivastigmine) extension study that included 334 subjects who completed the above-mentioned 24-week study, cognitive and functional benefits appeared to continue over time (171). Patients treated with placebo in the initial study who were then treated with rivastigmine in the open-label study had improvements in cognitive and functional scores similar to those of the patients who received rivastigmine in the initial study.

Two randomized, double-blind, placebo-controlled studies have investigated donepezil for the treatment of mild cognitive impairment, neither of which demonstrated benefit in the primary study outcomes. In one study of 270 subjects, there was no benefit of donepezil on most (but not all) cognitive tests studied, including the primary efficacy measures (172). The second study included 769 subjects with mild cognitive impairment and used a primary endpoint of progression from mild cognitive impairment to meeting the criteria for the diagnosis of possible or probable Alzheimer's disease over a 3-year period (173). Although fewer donepezil-treated subjects had progressed to Alzheimer's disease in the first year of the study, compared to placebo-treated subjects, there was no difference between the groups by the end of 3 years. Donepezil-treated subjects did better than the placebo group on a number of cognitive tests, but this modest difference did not persist beyond 18 months.

There have been two randomized, placebo-controlled, clinical trials of galantamine in subjects with mild cognitive impairment. Each study was of 2 years' duration and included approximately 1,000 patients. Overall, there were no statistically significant benefits for galantamine compared to placebo, either in increasing the time to the onset of dementia or improving cognitive function, activities of daily living, or global assessment ratings. Of concern in these two trials together, 13 subjects who were taking galantamine died, compared to one subject who received placebo. This finding was statistically significant and represents a precaution in the use of galantamine in this patient population. It is noteworthy that the rates of death in these two trials were much lower in both the placebo and the galantamine groups than would have been expected based on previously conducted clinical trials in patients with actual dementia. As in trials of cholinesterase inhibitors for Alzheimer's disease subjects, there were substantial dropouts due to adverse events in the trials for subjects with mild cognitive impairment.

b. Memantine

Memantine, a noncompetitive NMDA antagonist, has been studied extensively in recent years for the treatment of Alzheimer's disease and vascular dementia. Trials have ranged from 6 weeks to 6 months in duration and have primarily included outpatients, although one study included nursing home residents (180). Studies have enrolled patients with moderate to severe dementia (MMSE scores ranging from 3 to 15) as well as mild to moderate dementia (MMSE scores ranging from 10 to 24).

Among randomized placebo-controlled trials that have included patients with mild to moderate Alzheimer's disease, two unpublished trials did not find benefit of memantine over placebo (487), whereas one trial demonstrated cognitive and functional improvement with memantine, compared to placebo (177). A meta-analysis of these three trials showed a statistically significant but very small advantage to memantine over placebo (108). Nonetheless, the FDA has not approved the use of memantine for treatment of mild Alzheimer's disease.

Among studies of patients with moderate to severe Alzheimer's disease, two published trials (with 252 and 404 subjects, respectively) (174, 175) found cognitive and functional improvement with memantine, compared with placebo. In one of those studies, which included only patients who were already taking stable dosages of donepezil, random assignment to treatment with memantine led to improvement in cognition and function, compared to random assignment to the placebo group (175). One unpublished study did not show any benefit of memantine over placebo (487). In a trial that included 166 subjects with severe dementia due to Alzheimer's disease or vascular dementia, cognitive and functional improvement was greater with memantine than with placebo (180).

In summary, there is evidence supporting the use of memantine for moderate to severe Alzheimer's disease, and memantine is approved by the FDA for this use. Data are not yet available to argue for or against the use of memantine beyond 6 months (108, 176).

There have been two large 6-month clinical trials (with 579 and 321 subjects, respectively) of memantine to treat patients with mild to moderate vascular dementia (178, 179). In both trials, cognition and behavior improved, but there was no demonstrated functional improvement and no improvement on clinical global ratings of change (Clinician's Interview-Based Impression of Change Plus). No studies of memantine have been conducted with patients with severe vascular dementia alone. Nonetheless, in one randomized placebo-controlled trial that included 166 patients with severe dementia, of which 51% had vascular dementia and 49% had Alzheimer's disease, there was improvement in cognitive and functional outcome measures for both the Alzheimer's disease and vascular dementia subjects (180).

c. Vitamin E

There has been considerable interest in vitamin E (-tocopherol) as a treatment for Alzheimer's disease and other dementias because of its antioxidant properties and efficacy in Parkinson's disease. Vitamin E has been shown to slow nerve cell damage and delay death in animal models and cell cultures (including damage associated with amyloid deposition) and thus may be relevant to the development and progression of Alzheimer's disease (183, 488–490).

One large clinical trial of vitamin E for treatment of Alzheimer's disease has been conducted (183). This placebo-controlled, double-blind, multicenter trial included 341 subjects with moderate Alzheimer's disease (CDR of 2) who were randomly assigned to receive 1,000 IU b.i.d. of vitamin E alone, 5 mg b.i.d. of selegiline alone, both agents, or placebo. Vitamin E alone, selegiline alone, and the combination each delayed reaching the study endpoints (defined as a poor outcome, namely death, institutionalization, or significant functional decline). The benefit observed was equivalent to a delay of approximately 5–7 months in reaching the composite endpoint. It is noteworthy that no evidence was found for improvement in function or cognition, compared to baseline. Despite the evidence for a better functional outcome in the treatment groups, compared to the placebo group, all groups showed similar rates of cognitive decline during the 2-year study period. There are no studies of the effects of vitamin E in subjects with Alzheimer's disease with mild or severe impairment or in subjects with other dementias. There are also no data concerning the effect of vitamin E in combination with medications other than selegiline.

In one clinical trial, the effects of donepezil, vitamin E, and placebo were compared in patients with mild cognitive impairment (173). These patients were followed for 3 years. Overall, no differences were found between vitamin E (2,000 IU/day) and placebo on measures of cognition, level of function, or progression to dementia.

Although vitamin E has been widely used clinically and in numerous clinical trials for a variety of indications and has been considered to have low toxicity, more recent evidence suggests that vitamin E may be associated with a small but significantly increased risk for morbidity and possibly even mortality. At high doses, it may worsen blood coagulation defects in patients with vitamin K deficiency (184). Of greater concern is evidence linking vitamin E usage in clinical trials to increased mortality in a dose-dependent fashion. A meta-analysis of 11 clinical trials of vitamin E in a mixed population that included some individuals with significant cardiac disease found a very small but statistically significant increase in mortality in trials using doses greater than 400 IU/day and a dose-dependent increase in mortality with doses above 150 IU/day (181). In addition, a carefully conducted large, randomized clinical trial of vitamin E (400 IU/day) for the prevention of heart disease or cancer in patients with diabetes mellitus and/or vascular disease had an unanticipated finding that vitamin E was associated with an increased risk of heart failure (182).

d. Other Agents

A number of medications marketed for other indications have been proposed for the treatment of dementia on the basis of epidemiological data or pilot studies. Aspirin and other NSAIDs have been proposed because of epidemiological data suggesting that they protect against the development of dementia (185–189) and because of hypotheses regarding the involvement of inflammatory mechanisms in the pathogenesis of Alzheimer's disease (491). In a single small treatment trial for patients with Alzheimer's disease, patients receiving 100–150 mg/day of indomethacin experienced less decline over 6 months than did a matched control group (186). However, a number of larger studies with other NSAIDs (rofecoxib, naproxen, and diclofenac) did not show benefit over placebo in slowing the cognitive decline in Alzheimer's disease (190–192). Moreover, several studies have suggested cardiac toxicity, especially with more selective cyclooxygenase-2 inhibitors. Finally, use of aspirin and NSAIDs can be associated with other significant adverse events such as gastrointestinal bleeding and impairment of renal function. Thus, NSAIDs are not recommended for the treatment of Alzheimer's disease.

Hormone replacement therapy is known to affect cognitive function (492) and was shown to be beneficial in the treatment of dementia in at least two case series (493, 494). It was also associated with later onset and/or decreased risk of cognitive decline in at least two observational studies of postmenopausal women (495, 496). In contrast, in the prospective Women's Health Initiative Memory Study (WHIMS), increased rates of conversion to Alzheimer's disease were found in women age 65 years or older who were randomly assigned to receive an estrogen/progestin combination compared with those who received placebo (193). Results of a number of other prospective trials also showed no benefit of estrogen over placebo in the treatment of cognitive symptoms of Alzheimer's disease (194, 195). Therefore, hormone replacement therapy is not recommended for use in the treatment of cognitive symptoms of Alzheimer's disease.

There is also interest in the hormone melatonin and in botanical agents such as ginkgo biloba, which are available without a prescription. There are no data supporting the use of most of these agents in Alzheimer's disease. Numerous clinical trials of ginkgo biloba have been conducted. Most of these have been small trials with considerable methodological problems (497). In one randomized placebo-controlled trial that included 309 subjects, 1-year efficacy of ginkgo was found in subjects with Alzheimer's disease or vascular dementia, but there were significant methodological problems with the trial, including a very high dropout rate (498). In a large, 26-week, randomized, double-blind, placebo-controlled trial that included 513 subjects with mild to moderate Alzheimer's disease (MMSE score of 10–24), the effects of 120 mg/day of ginkgo, 240 mg/day of ginkgo, and placebo were compared (196). There was no advantage to ginkgo over placebo for cognitive function, but these results are qualified by the fact that there was very little cognitive decline in the placebo group as well. At this point, the preponderance of the evidence does not support the routine use of ginkgo for the treatment of dementia (197, 198).

The chelating agent desferrioxamine has also been studied as a possible treatment for Alzheimer's disease on the basis of hypotheses regarding heavy metals in the pathogenesis of the disease. In one small single-blind trial, there was some evidence of a decrease in cognitive decline over 2 years (499). One study of another chelating agent failed to confirm this finding (500). Because chelating agents are quite toxic and support for them is so weak, they are not recommended for the treatment of dementia. Newer agents that chelate copper and zinc are in clinical development as potential Alzheimer's disease therapies because they may lower beta-amyloid burden (501).

As reviewed in a Cochrane meta-analysis that included negative trials, the irreversible MAO-B inhibitor selegiline has been studied in a large number of randomized controlled clinical trials, with mixed results (199). One large trial comparing selegiline with vitamin E and placebo demonstrated some benefit of selegiline over placebo (183), but numerous other trials have not shown clinically meaningful benefit. Although use of selegiline at dosages of 5–10 mg/day is generally safe and well tolerated, few clinicians actually use this medication in clinical practice for the treatment of cognitive decline in dementia. Selegiline use is considered contraindicated in combination with meperidine, SSRIs, or tricyclic antidepressants. Use of the selegiline patch for treatment of dementia has not been studied.

A mixture of ergoloid mesylates is currently marketed under the trade name Hydergine for the treatment of nonspecific cognitive impairment. It has been available for at least 40 years and has been studied in at least 150 clinical trials, seven of which were double-blind, placebo-controlled, randomized trials with a parallel-group design involving a total of 297 patients with diagnoses consistent with Alzheimer's disease. Studies have been conducted in patients with vascular dementia as well. Although a number of trials have produced weakly positive results, these have generally been on isolated cognitive measures or on measures of psychiatric symptoms, and the overall evidence suggests little or no effect (200). Side effects are usually mild and affect the gastrointestinal system. Dosages used in the studies ranged from 3 to 9 mg/day.

Because treatments for psychosis and behavioral disturbances overlap to a considerable extent, and because investigations often include subjects with both groups of symptoms, they are grouped together in this discussion.

a. Antipsychotics

First-generation antipsychotic medications (also known as "conventional" or "typical" antipsychotic agents) have been extensively studied in the treatment of psychosis and agitation in individuals with dementia. For example, a 1990 review (210) identified seven double-blind, placebo-controlled, randomized, parallel-group clinical trials including 252 patients studied over periods of 3–8 weeks (203–209). Despite some methodological flaws, notably small numbers of subjects and a lack of diagnostic specificity, these studies, when taken together, constitute reasonable evidence for a modest improvement in target symptoms in some patients treated with first-generation antipsychotic medications. A meta-analysis of these seven trials (210), using clinician assessment of improvement in a variety of behavioral symptoms as the primary outcome, showed improvement in 59% of the subjects taking antipsychotics and 41% of those taking placebo. The studies used a wide variety of dosages (ranging from 66 to 267 mg/day in chlorpromazine equivalents), and efficacy for behavioral symptoms was not correlated with standardized dose. Adverse effects were common, but specific rates are not available. Dropout rates were also high, whether associated with side effects or poor efficacy.

In a more recent meta-analysis of studies of first-generation antipsychotics in patients with dementia, clinically significant improvement was found in 64% of patients treated with active medication versus 38% of patients treated with placebo, translating to an effect size of 26% (211). No differences in efficacy were seen among the different agents used. Significant side effects were found in 25% more of the patients who received active treatment, compared with those who received placebo. This meta-analysis did not include a number of more recent studies (212–216); however, the data from these newer studies generally conformed to the results of the meta-analyses (210, 211). In another review of antipsychotics used in dementia (217), which also included several trials of second-generation antipsychotic agents, mean improvement rates were 61% with antipsychotics and 35% with placebo, yielding a treatment effect of 26%. The modest treatment effects estimated by these meta-analyses and reviews are generally consistent with results from placebo-controlled trials of second-generation antipsychotics.

Second-generation antipsychotic agents have also been studied in well-controlled trials in patients with dementia. Three placebo-controlled trials of risperidone have been conducted among nursing home residents with severe dementia complicated by agitation and/or psychosis (212, 218). In the first trial, a fixed-dose study that included 625 patients, the response rates of 45% and 50% for 1 mg/day and 2 mg/day, respectively, were significantly different than placebo (33%) (the 0.5-mg/day dose was not more effective than placebo). Parkinsonism was seen in 21% of patients who received 2 mg/day, and it was concluded that 1 mg/day represented the most effective dose. In the second trial, a flexible-dose study that included 344 patients, response rates were 47%, 63%, and 54%, respectively, for placebo, haloperidol (mean dosage 1.2 mg/day), and risperidone (mean dosage 1.1 mg/day); these response rates did not differ statistically. Secondary outcomes related to aggression decreased in the two active treatment groups, compared with the placebo group, but cognitive and functional outcomes were not different across treatment groups. In the third trial, the effects of flexibly dosed risperidone were compared to those of placebo in 337 nursing home patients with severe dementia who required treatment for aggression; a significant difference in response rates of 37% for placebo and 63% for risperidone (mean dosage 1 mg/day) was found (219).

Three clinical trials have been conducted to study the effects of oral olanzapine for the treatment of persisting agitation and/or psychosis in patients with severe dementia. The first trial, reported only as an abstract, compared flexibly dosed olanzapine (mean dosage of about 2.3 mg/day) with placebo in 238 patients, and no difference in efficacy or tolerability was found. This negative result may in part be explained by the mean dosage's being too low (502). In the second trial, in which participants consisted of 206 nursing home residents, response rates of 66%, 57%, and 43% were found with fixed dosages of 5, 10, and 15 mg/day of olanzapine, respectively, versus 36% for placebo (the response rates with 5 mg/day and 10 mg/day were significantly different from those with placebo) (220). Some patients from this trial received follow-up open-label, flexible-dose treatment for 18 weeks; the results were consistent with the findings of the original report (221). The third trial included 652 residents of nursing homes or continuing care hospitals who met the operational criteria for psychosis (222). Patients were assigned to treatment with olanzapine at doses of 1, 2.5, 5, or 7.5 mg/day or placebo; no drug-placebo differences were seen in measures of psychosis or other behavioral features.

Meehan et al. (223) studied acute treatment of agitation with intramuscular olanzapine in a group of 272 inpatients or nursing home residents with Alzheimer's disease and/or vascular dementia. Olanzapine at doses of 2.5 and 5.0 mg was found to be superior to placebo in treating agitation at 2 hours; the response rates were 62%, 66.7%, and 37.3%, respectively. The response rate for those treated with intramuscular lorazepam was 72.1%; all response rates for participants who received active treatment were different from those for participants who received placebo group but not from each other. Adverse events were not significantly different between groups.

Evidence for the use of quetiapine is limited to findings from three open-label trials that suggested possible benefits for agitation (503–505). A 10-week, multicenter, placebo-controlled trial of flexibly dosed quetiapine versus haloperidol was conducted in a group of elderly nursing home patients with operationally defined psychosis, criteria for which were implemented before the development of the recently proposed clinical criteria for the psychosis of Alzheimer's disease (201). Results from the subgroup of 284 patients with Alzheimer's disease were analyzed separately. In these subjects, the mean daily dosage of haloperidol was 2 mg/day at endpoint, whereas that of quetiapine was about 120 mg/day. Neither of the treatment groups differed with respect to reduction in measures of psychosis, the primary outcome of the trial. One secondary measure of agitation showed improvement with both haloperidol and quetiapine treatment but not placebo. These studies led to a second placebo-controlled trial of 100 mg/day of quetiapine, achieved by day 4, or 200 mg/day, achieved by day 8, in which the participants were 333 nursing home residents with dementia (227). A dose of quetiapine at 200 mg/day was superior to placebo on numerous outcomes, with less benefit seen at 100 mg/day. A more critical review of the data will be possible once the trial results are published. A placebo-controlled trial of quetiapine in a group of 93 subjects with Alzheimer's disease and agitation failed to show any benefit over placebo for the medication (506).

Three placebo-controlled studies of aripiprazole have been published or presented in abstract form. In all three studies, the primary outcomes were not reached, but significance on individual outcomes was shown. In a placebo-controlled trial that included 208 outpatients who met the clinical criteria for psychosis of Alzheimer's disease (201), there was no difference between flexibly dosed aripiprazole (mean dosage of 10 mg/day) and placebo on the primary outcome variable of psychosis, although post hoc analyses showed benefits at some time points (224). The second study, with findings presented in abstract form, was a placebo-controlled study of fixed-dose aripiprazole conducted with 587 nursing home residents with dementia and psychotic features (507). A significant effect was found only for the 10-mg/day dose on the primary outcome, a measure of psychosis, with 50% of patients who received placebo and approximately 68% who received 10 mg/day of aripiprazole considered to have responded clinically. The third study, with findings presented in abstract form, was a placebo-controlled flexible-dose study conducted with 256 nursing home residents with dementia and psychotic features (508). The results showed no drug-placebo difference in the primary outcome with a mean aripiprazole dosage of 8.6 mg/day. A fuller appreciation of these studies will be possible once the results are published.

Clozapine has been found to be useful in controlling psychotic symptoms in patients with Parkinson's disease (233) and dementia with Lewy bodies (234) and may also be useful for patients with Alzheimer's disease who are sensitive to the extrapyramidal effects of first-generation antipsychotic agents (509). Currently no specific data are available on the use of ziprasidone in the treatment of elderly patients.

A recent meta-analysis that included the randomized controlled trials described in preceding paragraphs showed that benefits tended to be greater for symptoms of agitation than for psychosis (225). These data also demonstrate a significant placebo response, a finding that underscores the importance of nonpharmacological interventions for relief of these signs and symptoms. The nonspecific aspects of clinical trial participation (e.g., increased attention from staff), as well as the frequent clinical evaluations that occur during a clinical trial, may contribute to the improvement seen in patients in the placebo arm of a trial. Finally, much of the data discussed earlier are from studies that have not been published yet, so they will need to be reevaluated.

The available studies comparing antipsychotics to one another are of limited power but suggest no clinically significant differences in efficacy (40, 210, 212, 215, 225). The first direct comparison of second-generation antipsychotics with placebo in patients with Alzheimer's dementia has been undertaken as part of the NIMH-funded CATIE-AD (228). In this trial 421 outpatients with Alzheimer's disease and psychosis and/or aggression were randomly assigned to treatment with olanzapine, quetiapine, risperidone, or placebo with dosages adjusted as needed and followed for as long as 36 weeks. There were no differences among treatments in the main outcome, time to all-cause discontinuation, with initial treatments maintained for about 8 weeks. Time to discontinuation due to lack of efficacy, however, favored olanzapine and risperidone, both of which were maintained for about 24 weeks, whereas quetiapine and placebo were maintained for approximately 9 weeks. Time to discontinuation due to adverse events or intolerability favored placebo, with discontinuation in 24% of patients who received olanzapine, 16% of patients who received quetiapine, 18% of patients who received risperidone, and 5% of patients who received placebo. Although there were no differences in improvement as rated with the Clinical Global Impression of Change (olanzapine 32%, quetiapine 26%, risperidone 29%, placebo 21%), some symptom ratings favored the drugs over the first 12 weeks. Adverse effects offset advantages in efficacy of second-generation antipsychotic drugs for treatment of psychosis, aggression, or agitation in patients with Alzheimer's disease.

It is important to note that there are limited data on the efficacy of antipsychotic medications for patients with dementia beyond 8–12 weeks of follow-up, although these medications are often used for much longer periods of time in clinical practice. Extensive clinical experience has suggested that they are sometimes helpful for longer periods of time. Several studies of the effects of withdrawal of treatment have suggested that a substantial proportion of patients can be withdrawn from treatment successfully after a period of time (40, 229).

Antipsychotic agents are associated with a risk of serious complications that must be considered in weighing the risks and benefits of antipsychotic treatment.

Tardive dyskinesia, whose incidence increases with increasing dose and duration of treatment and which occurs more commonly in women, is also more common in individuals with dementia and in elderly patients in general. The risk may be as high as 30% for elderly patients with significant exposure to first-generation antipsychotic agents (510–512). This risk may be considerably lower with the use of second-generation antipsychotics. For example, Jeste et al. (513) reported a cumulative incidence of tardive dyskinesia of 2.6% in 330 patients with dementia treated openly with risperidone (mean dosage of about 1 mg/day) for a median of 273 days. This figure is much lower than that reported for older people treated with first-generation antipsychotics (511) and is consistent with data reported in a recent prospective longitudinal study of risperidone and haloperidol in older subjects with mixed psychiatric disorders (514). There are, however, no placebo-controlled studies addressing this issue and no studies employing withdrawal maneuvers. Clozapine is the agent least associated with tardive dyskinesia, although the rare occurrence of clozapine-induced tardive dyskinesia has been reported (515).

Neuroleptic malignant syndrome is a rare but potentially lethal adverse effect of antipsychotic medications. It occurs less frequently with second-generation antipsychotic agents than with first-generation agents, but it has been reported with both types (516–520). The core features of this syndrome are muscle rigidity (leading to elevated serum creatinine phosphokinase levels), fever, leukocytosis, tremor, delirium, autonomic instability, and diaphoresis. Older age and dementia may increase the risk of neuroleptic malignant syndrome.

Clozapine is associated with risk of agranulocytosis (about 1%), which is more common in elderly patients than in younger patients (509), and regular monitoring of blood counts is required. Other antipsychotics may rarely be associated with this adverse event, but its incidence is so infrequent that routine monitoring of blood counts for this syndrome is not required for patients who take other antipsychotics.

Metabolic abnormalities caused by second-generation antipsychotic medications are not well studied in individuals with dementia but may be more common with use of these agents in older individuals (258). These metabolic abnormalities include hyperlipidemia, weight gain, and diabetes mellitus (228).

There is also evidence of an increased risk of cerebrovascular adverse events with at least three of the second-generation antipsychotics (aripiprazole, olanzapine, and risperidone) when used in the treatment of patients with dementia. In October 2002, Health Canada issued a letter to health care professionals stating that risperidone use may be associated with cerebrovascular events in elderly patients with dementia (521). In April 2003, the FDA issued a similar warning regarding risk of cerebrovascular events with risperidone in patients with dementia and noted that risperidone had not yet been shown to be safe or effective in treating dementia-related psychosis (522). Pooled data from four placebo-controlled trials suggested a rate of "cerebrovascular events" (variably defined) of 4% in patients treated with risperidone, compared with about 2% in those treated with placebo. The available data suggest that the risk is greater among those with pre-existing risk factors for cerebrovascular disease.

In January 2004, Eli Lilly and Company issued a warning to prescribers that there was an increased incidence of cerebrovascular adverse events in patients with dementia who were treated with olanzapine (1.3%) versus placebo (0.4%), as well as increased mortality (3.5% vs. 1.5%) (523). In February 2005, Bristol-Myers Squibb issued a warning to prescribers that there was an increased risk of cerebrovascular adverse events in patients with dementia who were treated with aripiprazole (1.3% vs. 0.6% in those given placebo).

At present no warning has been issued regarding quetiapine and increased risk of cerebrovascular events. Schneider et al. (225, 524) reported an event rate of 0.8% for quetiapine and 1.9% for placebo among placebo-controlled studies of quetiapine in the dementia population. They also noted that the 95% confidence intervals (CIs) for the relative risks for risperidone, olanzapine, and aripiprazole all indicated increased risk for such events, whereas the CI for quetiapine could not distinguish among increased risk, decreased risk, or no risk. It is possible that patients with existing cerebrovascular disease or other risk factors might be most susceptible to these events. However, this possibility has yet to be definitively addressed with studies designed to assess side effects or efficacy as a function of baseline medical condition.

Finally, a recent meta-analysis of randomized controlled clinical trials of second-generation antipsychotics in patients with dementia performed by Schneider et al. (525) compared mortality in the treatment and placebo groups during the clinical trial period. In total, the meta-analysis included 15 trials, three of aripiprazole, five of olanzapine, three of quetiapine, and five of risperidone. The subjects were nursing home residents in 11 of the trials and outpatients in the remaining four trials; a total of 3,353 subjects received medication and 1,757 received placebo. When data from all the trials were pooled, the odds ratio for death in subjects who received second-generation antipsychotics was 1.54, compared to the placebo group, with a 95% CI of 1.06–2.23. Although for no individual medication or individual trial was the odds ratio of death in the medication group statistically different from that for the placebo group, for all medications and for 12 of the 15 trials the odds ratio favored placebo over the medication.

This finding has been underscored by the "black box" warning applied by the FDA on April 11, 2005, to all of the second-generation antipsychotics, stating that analysis of 17 placebo-controlled trials of aripiprazole, olanzapine, quetiapine, or risperidone in patients with dementia showed a death rate of about 4.5% in those receiving active treatment versus about 2.6% in those receiving placebo. Causes of death were varied, with the most common being cardiovascular (heart failure, sudden death) or infectious (pneumonia) in nature (231).

There is also evidence that first-generation antipsychotics are similarly associated with increased mortality among patients who take them and that this increased mortality may exceed that found with second-generation antipsychotics. In a large retrospective review of 22,890 patients over age 65 years in Pennsylvania who received either first- or second-generation antipsychotic agents between 1994 and 2003, Wang et al. (230) found a 1.37 relative risk (CI, 1.27–1.49) of death among users of first-generation agents versus users of second-generation agents. The risk was highest with higher doses and closer to the initiation of treatment. The increased risk was independent of the presence of dementia and of residence (nursing home versus community). The magnitude of this difference was such that the authors concluded that for every 100 patients treated with first-generation antipsychotics instead of second-generation agents, there would be seven additional deaths. At this time, there is no FDA "black box" warning on first-generation antipsychotics.

Clinicians facing the challenge of treating patients with significant psychosis or behavioral disturbances must weigh the risk of not treating these complications of dementia against the risks of active treatment described in this section. Clinicians must take into account the evidence supporting the efficacy of the agent in question, the morbidity and risk associated with the target symptoms, the patient's general medical condition, and the evidence of risk and benefit of any alternative treatment being considered.

In addition to their association with the serious side effects described in Section V.B.2.a.2.a, antipsychotic medications are associated with numerous more common but milder side effects.

First-generation antipsychotic agents have a broad range of common side effects that vary with medication potency, although any side effect can be seen with any agent. Reviews regarding first-generation agents have cited side effects including akathisia, parkinsonism, sedation, peripheral and central anticholinergic effects, postural hypotension, cardiac conduction defects, and falls (211). Most of these data come from short-term controlled trials; evidence regarding long-term safety is generally lacking. Data available from other studies in the elderly population, however, indicate that caution is warranted. For instance, rates of tardive dyskinesia are five- to sixfold greater in older than in younger populations after long-term treatment with first-generation agents (511). For practical purposes, side effects often guide selection of these agents when used in patients with dementia. High-potency agents (e.g., haloperidol, fluphenazine) are most strongly associated with akathisia (which can worsen the target symptoms) and parkinsonian symptoms. Low-potency agents (e.g., thioridazine, chlorpromazine) are associated with sedation (which can lead to worsening cognition or falls), central anticholinergic effects (e.g., confusion, delirium), postural hypotension (which can also lead to falls), and a variety of peripheral anticholinergic effects (e.g., dry mouth, constipation, bladder dysfunction, tachycardia). When individuals with dementia have co-occurring extrapyramidal disorders (as in dementia with Lewy bodies), extraordinary sensitivity to first-generation antipsychotic agents may be seen (526).

Risperidone treatment of patients with dementia is associated with a low to moderate risk of dose-related parkinsonism. In the large trial by Katz et al. (218), the rates of parkinsonism were 21%, 13%, and 7% among patients who received 2 mg/day of risperidone, 1 mg/day of risperidone, and placebo, respectively. In the trial of De Deyn et al. (212), the incidence of parkinsonism was 15% for subjects who received risperidone at a mean endpoint dosage of 1.1 mg/day and 11% for subjects who received placebo. Brodaty et al. (219) found an incidence of parkinsonism of 23% among subjects who received risperidone (mean dosage of 0.95 mg/day) and 16% among subjects who received placebo. These results are similar to those from a smaller study by Chan et al. (215) in which risperidone and haloperidol were compared and are also consistent with the results of a large meta-analysis of randomized controlled trials of risperidone, which found an overall odds ratio for extrapyramidal signs and symptoms of 1.80 (CI, 1.35–2.42) for risperidone versus placebo (225). These studies also showed a greater risk of sedation. In the trial of De Deyn et al. (212), sedation affected 12% of patients taking risperidone but only 4.4% of those who received placebo, a finding also confirmed by the meta-analysis (odds ratio, 2.43; CI, 1.78–3.32). In the meta-analysis, a higher rate of peripheral edema was found in patients treated with risperidone, compared to those who received placebo (225). Risperidone may also cause an abnormal gait in some patients (225).

For olanzapine, side effect information is primarily available from the one randomized controlled trial that demonstrated clear clinical efficacy (220); in one other clinical trial, the doses used were not sufficient to help or harm or to provide meaningful information about side effects (502), and in the other trial specific characteristics were not described for the side effects that occurred (222). In the one trial with side effect information, sedation (at rates of 25%–36%) and abnormal gait (at rates of 14%–20%) were observed at all dosages used (5–15 mg/day). Results of the meta-analysis of randomized controlled trials of olanzapine showed substantially increased risk with the medication, compared to placebo, of sedation (odds ratio, 4.00; CI, 2.27–7.04) and urinary tract infections or incontinence (odds ratio, 6.69; CI, 1.27–35.10) (225). Gait abnormalities also developed in more subjects taking olanzapine than in those who received placebo (225).

In the one trial comparing quetiapine to haloperidol, tolerability of quetiapine was superior, with comparable effects on a simple measure of agitation (527). Several measures of parkinsonism showed worsening with haloperidol but no difference between quetiapine and placebo (527). Sedation was seen in 4.1% of patients who received placebo versus 25.3% and 36.2% of those who received quetiapine and haloperidol, respectively. In the subsequent trial in which the most efficacious dosage of quetiapine was 200 mg/day, sedation occurred in 17.6% of patients taking 200 mg/day, compared with 11.3% of those taking 100 mg/day and 6.5% of those who received placebo (227). A recent meta-analysis of randomized controlled trials of quetiapine similarly found an odds ratio for sedation of 3.90 (CI, 1.41–10.78) for quetiapine versus placebo (225).

Aripiprazole treatment of patients with dementia is commonly associated with sedation, occurring in 5%–15% of those treated versus 1% of those receiving placebo. In the one published randomized controlled trial, 8% of aripiprazole-treated subjects but only 1% of placebo-treated subjects experienced sedation (224). This finding was confirmed in a recent meta-analysis that focused on adverse events (225). Regarding ziprasidone in the treatment of patients with dementia, there are insufficient data to make assertions regarding safety and tolerability.

Most short-term side effects can be minimized by using the lowest effective dose. This principle is particularly important in order to minimize sedation and akathisia, both of which can actually worsen target symptoms (528). It may also be helpful to select an agent with the side effect profile most suited to a given patient. Anticholinergic agents may be effective in the treatment of parkinsonian side effects, but the high risk of associated cognitive decline, delirium, and other anticholinergic effects suggests that they should be used only with extreme caution for elderly patients both with and without dementia.

b. Benzodiazepines

The use of benzodiazepines in the treatment of behavioral symptoms in dementia has been studied in at least eight randomized clinical trials. In six studies including a total of 873 subjects, benzodiazepines were compared to antipsychotics administered orally (238–243). In two studies, benzodiazepines were compared to placebo (529, 530). Most of these studies were limited by the presence of poorly specified diagnoses, a mixture of target symptoms, limited outcome measures, and, in most cases, high doses of long-acting agents. Nonetheless, they demonstrated that benzodiazepines perform better than placebo but not as well as antipsychotics in reducing behavior problems. These results are supported by a more recent randomized controlled trial of comparing intramuscular lorazepam with intramuscular olanzapine, which showed equal efficacy of lorazepam and olanzapine at 2 hours but inferior efficacy of lorazepam at 24 hours (223). There are no data concerning the efficacy of benzodiazepines after 8 weeks or whether one benzodiazepine is more effective than another.

c. Anticonvulsants

Use of carbamazepine has support from several case series (248), a small open trial (249), a double-blind nonrandomized trial (250a), and two double-blind randomized trials (250b, 250c) that showed modest benefit for agitation at low doses, with low side-effect rates over a short treatment period. One of these trials was followed by an open-label extension that further supported efficacy, safety, and tolerability (251). One small randomized crossover trial showed nonsignificant decreases in behavioral measures (252).

Although several favorable case reports and open trials have been reported for the anticonvulsant valproate (531–533), four placebo-controlled trials have not demonstrated efficacy. In a 6-week, randomized, placebo-controlled trial that included 172 nursing home residents with Alzheimer's disease and secondary mania, subjects treated with divalproex sodium had no more improvement in symptoms of mania than did subjects treated with placebo (253). In another randomized placebo-controlled study conducted with 42 patients with Alzheimer's disease and behavioral disturbances, valproic acid (dosage of 480 mg/day) was not more effective than placebo in reducing overall agitation, although in secondary analyses certain individual symptoms were improved (254). However, because these results were derived from an analysis of secondary outcomes, they are not sufficient to define practice. In another 6-week, randomized, placebo-controlled trial that included 56 nursing home residents with dementia and behavioral disturbance, results were suggestive of benefit with divalproex sodium; 68% of the treatment group had improvement in agitation, compared with 52% of the control group (256). The largest study, which prospectively addressed agitation as the primary outcome in 153 nursing home residents with Alzheimer's disease, found no difference between divalproex sodium (mean dosage of 800 mg/day) and placebo in either the primary outcome (agitation) or secondary outcome measures (255).

There are no controlled trials of newer anticonvulsants such as lamotrigine, gabapentin, and topiramate. The few case reports and case series that suggest benefit with gabapentin (e.g., reference 534) do not provide sufficient evidence to recommend their use.

d. Other Agents

A number of other agents have been proposed for the treatment of agitation in patients with dementia (reviewed in references 210, 535, 536). Efficacy data for these agents generally come from case reports or small open trials, often of mixed populations.

For example, data on trazodone are primarily from case reports, case series (260, 261), and a few small trials (262) in which decreased irritability, anxiety, restlessness, and affective disturbance was reported in a total of 13 patients. In a small double-blind, randomized clinical trial that was not placebo controlled, improvement in agitation with trazodone was comparable to that seen with haloperidol (263). However, in the largest and only placebo-controlled trial, which included 37 patients receiving trazodone, no benefit of trazodone, compared to placebo, was found (214). A small randomized, placebo-controlled, double-blind study of trazodone in patients with frontotemporal dementia showed benefit over placebo (264).

Preliminary data suggest that SSRIs may be useful in the treatment of agitation (262, 270). These preliminary reports are supported by two case series and a small controlled trial showing benefit of a variety of SSRIs for reduction of agitation in patients with frontotemporal dementia (537–539), although one trial of paroxetine in patients with frontotemporal dementia did not demonstrate improvement in symptoms but did show worsening of cognition (540). A rigorous placebo-controlled trial examined the short-term benefit of citalopram versus perphenazine in inpatients with agitation or psychotic features (271). Double-blind treatment lasted no more than 3 weeks. Although both perphenazine and citalopram, compared to placebo, produced clinical improvement on general measures of behavior, improvement in agitation and aggression specifically was seen only with citalopram. Patients taking citalopram had few side effects.

The effects of buspirone for treatment of agitation or anxiety in elderly patients with dementia have been reported in a number of case reports (265–267) and assessed in two open trials (268, 269). These limited data are insufficient to establish efficacy.

A 6-week, randomized, double-blind, placebo-controlled trial of propranolol that included 31 subjects with Alzheimer's disease and behavioral disturbance who resided in nursing homes showed benefit of the medication, compared with placebo, for certain symptoms, although it was noted that use of beta-blockers was contraindicated for many subjects who would otherwise have been eligible for the study (277). The mean dose was 106 mg/day. The benefits noted in the 6-week trial were lost to a great extent over the ensuing 6 months of open-label treatment.

a. Antidepressants

Over the last 15 years, eight placebo-controlled studies have examined the efficacy of antidepressants in patients with dementia. Four trials were conducted with SSRIs, one with citalopram, one with fluoxetine, and two with sertraline (289–292). Three trials assessed cyclic antidepressants (imipramine, maprotiline, and clomipramine) (293–295), and one trial examined an MAOI (moclobemide) (296). The available evidence is mixed for the efficacy of these medications for the treatment of depression in patients with dementia, with some trials demonstrating superiority over placebo and others failing to show differences in efficacy. Among the explanations for the variability in trial results are differences in patient selection criteria and differences in the sensitivity of the rating scales used in the various trials (541). Studies that have used the most restrictive criteria for depression generally reported better response to active treatment, compared with placebo. Selective serotonin reuptake inhibitors appear to be the most promising for treating depression in patients with dementia, with sertraline having superior efficacy, compared with placebo (292) and citalopram improving affective symptoms in one study of patients with dementia (289). The cyclic antidepressants, however, were either no more effective than placebo or produced significant side effects. Head-to-head comparisons of SSRIs to cyclic antidepressants and a SSRI (sertraline) to a serotonin-norepinephrine reuptake inhibitor (venlafaxine) showed equal efficacy in treating depression but better tolerability of SSRIs over cyclic antidepressants (297–299).

Although clinical trials support the efficacy of antidepressants in the treatment of depressed elderly patients without dementia (285), extrapolating these data to patients with co-occurring dementia should be done cautiously. The reader is referred to APA's Practice Guideline for the Treatment of Patients With Major Depressive Disorder , 2nd edition (284) for a summary of this literature.

Data concerning the treatment of other affective symptoms such as apathy are much sparser. There is minimal evidence that dopaminergic agents, such as psychostimulants (d-amphetamine, methylphenidate), amantadine, bromocriptine, and bupropion, are helpful in the treatment of severe apathy, but case reports have suggested that efficacy studies are warranted (301, 302). Psychostimulants have also received some support for the treatment of depression in elderly individuals with severe general medical disorders (303–305).

b. Electroconvulsive Therapy

The data supporting the efficacy and safety of ECT in the treatment of depression in dementia are limited to the results of one small retrospective chart review (306). Several larger, prospective studies have supported the efficacy of ECT in the acute treatment of geriatric depression (307, 308).

The available data do not suggest a specific course of action for treating sleep disturbances in patients with dementia. Several small trials of bright light therapy exist, but they have not shown effectiveness in improving sleep problems and their associated behavioral and mood disturbances (315, 318–322). In one small study, improvement in MMSE scores (3 points) was reported with bright light therapy (320). In another small study, patients with vascular dementia, but not patients with Alzheimer's disease, had a decrease in nighttime activity with bright light therapy (318). However, four randomized controlled trials of bright light therapy have failed to show increases in nocturnal sleep time or decreases in nighttime activity, sleep latency time, agitation, or depression (315, 318, 319, 321, 322, 542).

Behavioral interventions and pharmacological agents are often utilized to address the sleep problems of patients with dementia. In one study, nocturnal sleep was improved by targeting daytime activities of nursing home residents during periods when they were most likely to nap (543). In another study, caregiver education and improved sleep hygiene were found to improve sleep quality in patients with dementia (316). A sleep hygiene behavioral intervention for patients with dementia was evaluated in a small trial (317). In this study, training for caregivers in how to implement proper sleep hygiene principles (e.g., consistent rising times, minimizing daytime napping, daily exercise) resulted in improved sleep, better maintenance of a consistent bedtime and rising time, fewer naps during the day, and more physical activity in the form of walking (317). Likewise, only a few pharmacological agents used to treat sleep problems have been rigorously studied in this population. Although 3 mg of melatonin at bedtime was found to prolong sleep and decrease nighttime activity in a small sample (544), a randomized controlled trial that included 157 individuals showed no benefit of 10 mg of melatonin or 2.5 mg of slow-release melatonin, compared with placebo (545). Galantamine, an acetylcholinesterase inhibitor, did not improve sleep quality in patients with dementia (546).