Antidepressants for agitation and psychosis in dementia

Abstract

Background

Agitation and psychosis are common among older adults with dementia and are challenging to manage. At the present time, little is known about the efficacy and safety of antidepressant medications when used to treat these symptoms.

Objectives

To assess the safety and efficacy of antidepressants in treating psychosis and agitation in older adults with Alzheimer's disease, vascular, or mixed dementia.

Search methods

We searched the Cochrane Dementia and Cognitive Improvement Group’s Specialized Register which included Cochrane Central Register of Controlled Trials (The Cochrane Library 2009, Issue 3), MEDLINE (January 1950 to October 2009), EMBASE (1980 ‐ October 2009), CINAHL (all dates ‐ October 2009) and PsycINFO (1806 to October 2009).

Selection criteria

Randomized, controlled trials of antidepressants (selective serotonin reuptake inhibitors (SSRIs), tricyclic antidepressants, trazodone, and other antidepressants), compared to either placebo or comparator medications (typical or atypical antipsychotics, anticonvulsants, benzodiazepines, cholinesterase inhibitors, memantine or other medications) for treatment of agitation or psychosis in older adults with dementia.

Data collection and analysis

Two authors independently assessed trial quality and extracted trial data. We collected information on efficacy as measured by dementia neuropsychiatric symptom rating scales and adverse effects. Study authors were contacted for additional information.

Main results

Nine trials including a total of 692 individuals were included in the review. Five studies compared SSRIs to placebo and two studies were combined in a meta‐analysis for the outcome of change in Cohen‐Mansfield Agitation Inventory (CMAI) scores. There was a significant difference between antidepressants and placebo on measures of agitation as reported on the change in CMAI total score (mean difference (MD), ‐0.89, 95% CI, ‐1.22 to ‐0.57) although the results were heavily weighted by one large study. There were no significant differences in change in behavioral symptoms of dementia for SSRIs compared to placebo in the one study that reported on changes in the Neuropsychiatric Inventory and Behavioral Pathology in Dementia scales. One study comparing citalopram to placebo found a significant difference in NPS as measured on the Neurobehavioral Rating Scale (NBRS) after controlling for baseline severity NBRS score although the unadjusted mean difference was not statistically significant (MD ‐ 7.70, 95% CI: ‐16.57 to 1.17). There was no difference in the rates of trial withdrawals due to adverse events for SSRIs compared to placebo for four studies reporting this outcome (relative risk (RR), 1.07, 95% CI: 0.55 to 2.11) or in the number of trial withdrawals due to any cause in the three studies reporting this outcome (RR, 0.91, 95% CI, 0.65 to 1.26). One study compared the SSRI citalopram to the atypical antipsychotic risperidone and found no difference in NBRS scores, trial withdrawals due to any cause or trial withdrawals due to adverse events although the rates of adverse events as measured on the UKU side effect scale total score were lower for citalopram (MD ‐2.82, 95% CI: ‐4.94 to ‐0.70). Three studies compared SSRIs to typical antipsychotics. In meta‐analysis of two studies there was no statistically significant differences in changes in CMAI total scores (MD, 4.66, 95% CI: ‐3.58 to 12.90). There was also no difference in trial withdrawals due to any cause or due to adverse events for SSRIs compared to typical antipsychotics. One study of trazodone compared to placebo did not find any significant difference in change in CMAI total scores (MD, 5.18, 95% CI, ‐2.86 to 13.22) or trial withdrawals due to any cause (RR, 1.06, 95% CI, 0.54 to 2.09). Two studies comparing trazodone to haloperidol also failed to detect any difference in change in CMAI total scores (MD, 3.28, 95% CI, ‐3.28 to 9.85) or trial withdrawals due to any cause (RR, 0.79, 95% CI, 0.43 to 1.46).

Authors' conclusions

Currently there are relatively few studies of antidepressants for the treatment of agitation and psychosis in dementia. The SSRIs sertraline and citalopram were associated with a reduction in symptoms of agitation when compared to placebo in two studies. Both SSRIs and trazodone appear to be tolerated reasonably well when compared to placebo, typical antipsychotics and atypical antipsychotics. Future studies involving more subjects are required to determine if SSRIs, trazodone, or other antidepressants are safe and effective treatments for agitation and psychosis in dementia.

Plain language summary

Antidepressants for agitation and psychosis in dementia

Psychosis and agitation frequently occur in older adults with dementia. Medications are often prescribed to treat these symptoms and antidepressants are increasingly used for these symptoms. We reviewed the evidence for the effectiveness and safety of antidepressants for the treatment of agitation and psychosis in older adults with dementia. We classified antidepressants based on their mechanism of action and included studies that compared antidepressants to treatment with either placebo or other medications frequently used to manage these symptoms. A total of nine studies (including 692 individuals) were identified, four comparing selective serotonin reuptake inhibitors (SSRIs) to placebo, three comparing SSRIs to typical antipsychotics, and one study comparing SSRIs to atypical antipsychotics. One study compared the antidepressant trazodone to placebo, and two compared trazodone to haloperidol. Most of the studies included in the review were relatively small and of uncertain risk of bias due to methodological issues. The SSRIs sertraline and citalopram were associated with a modest reduction in symptoms of agitation and psychosis when compared to placebo in two studies. There were few other statistically significant differences in changes in agitation or psychosis or in most measures of medication tolerability for SSRIs or trazodone when compared to placebo or the antipsychotic haloperidol. We conclude that there is some evidence to support the use of certain antidepressants for agitation and psychosis in dementia and further studies are required to determine the effectiveness and safety of SSRIs and trazodone in managing these symptoms.

Background

Description of the condition

Dementia is common in older adults, affecting 5% of all individuals over age 65 years and 30% of those greater than age 85 (Canadian Study of Health and Aging 1994, Ferri 2005, Plassman 2007). Neuropsychiatric symptoms (NPS) of dementia occur in approximately 80% of all individuals at some point in the course of illness (Selbaek 2007, Zuidema 2007) including 30 to 50% of individuals with dementia who have symptoms of agitation or psychosis (Ballard 1997; Lyketsos 2002; Pitkala 2004; Savva 2009). Agitation and psychosis are important clinical problems as they decrease patient and caregiver quality of life (Hurt 2008) and are a frequent contributor to caregiver's decisions for nursing home admission (Balestreri 2000). Treatment of these symptoms is challenging for clinicians. Although non‐pharmacological interventions are sometimes recommended as first‐line treatment for these symptoms (Sink 2005), there is limited evidence for many non‐pharmacological interventions (Ayalon 2006; Livingston 2005; Jeste 2008; Chung 2002; Vink 2003) and resources for non‐pharmacological interventions may not be accessible to many practitioners.

Description of the intervention

We examined all randomized, double‐blind, controlled trials of antidepressants (including SSRIs, TCAs, trazodone, and other antidepressants) compared to other psychotropic medications (benzodiazepines, antipsychotics, anticonvulsants) or placebo for treatment of agitation and/or psychosis in dementia. This study will expand on a previous review which evaluated the use of trazodone in dementia patients with agitation (Martinón‐Torres 2004).

How the intervention might work

Many antidepressants facilitate serotonergic neurotransmission which is known to be reduced in Alzheimer's disease (AD) (Lanctot 2001). By enhancing serotonin function this may ameliorate some behavioral symptoms of dementia. Norepinephrine is also modulated by many antidepressants and norepinephrine dysregulation is also observed in AD (Herrmann 2004). Antidepressants have broad clinical effects in other psychiatric conditions and are effective in reducing distress and agitation when used to treat mood and anxiety disorders.

Why it is important to do this review

Antidepressants are frequently used to treat neuropsychiatric symptoms (NPS) of dementia in clinical practice. There has been increasing interest in the use of antidepressants for the treatment of NPS of dementia given the established risks and limited benefits associated with antipsychotics which are currently the most frequently used medications for this purpose. A number of studies on this subject have been undertaken recently and to date there have been no comprehensive reviews of antidepressants for agitation and psychosis in dementia.

Psychotropic medications such as antipsychotics and antidepressants are frequently prescribed for agitation and psychosis in dementia (Pitkala 2004; Ballard 2006; Herrmann 2007; Nijk 2009; ). Antipsychotics are the most commonly used medications for this indication and up to 40% of all individuals with dementia receive antipsychotics (Pitkala 2004; Ballard 2006). Recently there have been increasing concerns over the safety and efficacy of antipsychotics in this setting. Antipsychotics have been associated with increased rates of stroke (Sacchetti 2010) and death (Schneider 2005; Gill 2007) when used for dementia. In addition, large randomized, controlled trials of antipsychotics for NPS of dementia have found limited efficacy, and poor tolerability associated with these medications (Schneider 2006). Despite these concerns, at the present time antipsychotics remain the best supported treatment for severe symptoms of agitation or psychosis where such behaviours are refractory to non‐pharmacological interventions or when patient safety is jeopardized (Ballard 2006, Jeste 2008; Lonergan 2002). The effects of other medications used in the treatment of dementia such as cholinesterase inhibitors (Birks 2006) and memantine (McShane 2006) have also demonstrated only modest effects on NPS.

Antidepressants have been suggested as alternative, perhaps safer, medications that may be used instead of antipsychotics or as second‐line treatments for agitation and psychosis (Salzman 2008; Schneider 2001; Lonergan 2004). Surveys of geriatric psychiatrists indicate that many clinicians recommend antidepressants as potential treatments for these symptoms (Greve 2005; Condren 2001). Similar to antipsychotics, antidepressants are frequently prescribed to this population with 30% to 40% of all individuals with dementia receiving antidepressants (Pitkala 2004). One review of pharmacological treatments for NPS of dementia recommended that selective serotonin reuptake inhibitors (SSRIs) could be considered as a second‐line treatment after atypical antipsychotics (Sink 2005). Some randomized trials have found that the SSRI citalopram may have effects similar to antipsychotics in treating NPS of dementia (Pollock 2002a, Pollock 2007a, Schneider 2009) although a review of antipsychotics for the treatment of NPS concluded that the evidence for non‐antipsychotic treatment of these symptoms, including antidepressants, was unclear (Jeste 2008). There is increasing interest in this topic and a number of studies of antidepressants for NPS of dementia have been completed (Pollock 2002a, Pollock 2007a, Schneider 2009) or are underway (NCT00898807;NCT00260624).

Given the known risks associated with antipsychotics, clinicians may presume that antidepressants such as SSRIs are safer alternatives. Antidepressants, however, are not without their own safety concerns. Older antidepressants such as tricyclic antidepressants (TCAs) are associated with anticholinergic side‐effects and cardiac toxicity. Newer and relatively safer antidepressants such as SSRIs are also associated with serious adverse drug events such as gastrointestinal bleeding (van Walraven 2001), hyponatremia (Fabian 2004), falls, and fractures (Richards 2007). These adverse effects may be especially common when antidepressants are used in frail older adults with cognitive impairment.

As the symptoms of agitation and psychosis in dementia frequently change during the course of illness (Steinberg 2004; Savva 2009) many individuals may only require temporary treatment with psychotropic medications. It would be hoped that antidepressants may provide a safe and effective method of reducing symptoms and distress caused by agitation and psychosis in dementia and improve quality of life for patients and caregivers.

At the present time little is known about the efficacy and safety of antidepressants for psychosis and agitation in dementia when compared to placebo or other psychotropic medications.

Objectives

To assess the safety and efficacy of antidepressants in treating psychosis and agitation in adults with AD, vascular dementia or mixed dementia. The effects of different classes of antidepressants (SSRIs, TCAs, trazodone, other antidepressants) will be compared to either placebo or alternative psychotropic medications including antipsychotics, benzodiazepines, anticonvulsants, cholinesterase inhibitors or memantine.

Methods

Criteria for considering studies for this review

Types of studies

All randomized controlled comparison studies of antidepressants compared to either placebo or other psychotropic medications where the primary study outcome was treatment of psychosis, agitation, or other NPS associated with dementia were included in the review.

Types of participants

We included all studies that included individuals diagnosed with AD, vascular, or mixed AD and vascular dementia, dementia with Lewy bodies (DLB), and dementia not otherwise specified in this review. The diagnosis of dementia were made according to standard diagnostic criteria including ICD‐9, ICD‐10, DSM‐III, DSM‐IV, NINCDS‐ADRDA (McKhann 1984), NINDS/AIREN (Roman 1993) or DLB (McKeith 2005).

We included all age groups, all levels of dementia severity, inpatient and outpatient populations, and studies of individuals living in the community and long‐term care. Studies of individuals with frontotemporal dementia, and Parkinson's disease dementia who will likely respond to treatments differently than individuals with AD or vascular dementia were excluded. We also excluded studies treating individuals with dementia and concomitant major depressive disorder and studies treating depressive symptoms without mention of agitation or psychosis.

Types of interventions

We considered antidepressant treatment to be any daily administered oral antidepressant including: selective serotonin reuptake inhibitors (citalopram, escitalopram, paroxetine, fluoxetine, fluvoxamine, sertraline); tricyclic antidepressants (amoxapine, amitriptyline, maprotiline, nortriptyline, desipramine, trimipramine, imipramine, protriptyline, doxepin, clomipramine); trazodone; and other antidepressants (nefazodone, venlafaxine, duloxetine, reboxetine, bupropion, mirtazepine, mianserin, moclobemide, phenelzine, tranylcypromine).

Comparators for antidepressants included: placebo; any antipsychotic (typical or atypical); benzodiazepines; anticonvulsants; or other medications (including cholinesterase inhibitors and memantine).

Studies which implemented pharmacological co‐interventions along with antidepressant interventions were excluded.

Types of outcome measures

The following outcomes measuring agitation, psychosis, and other NPS of dementia were recorded: symptoms of agitation and psychosis; total NPS scores; cognitive impairment; caregiver distress; safety and tolerability outcomes. The outcome measures are outlined below.

Efficacy Outcomes Scales:

• Clinical Global Impression Scale (CGI)

• Neuropsychiatric Inventory (NPI) (Cummings 1994)

• Gottfried, Brane, and Steen Scale (GBS) (Gottfries 1982)

• Cohen Mansfield Agitation Inventory (CMAI) (Cohen‐Mansfield 1989)

• Behavioural Pathology in Alzheimer’s Disease (BEHAVE‐AD) (Sclan 1996)

• Neurobehavioral Rating Scale (NBRS) (Levin 1987)

• Behavioural Rating Scale for Dementia (BRSD) (Tariot 1995)

Total Neuropsychiatric Symptom Scores: 
 The baseline and change in total scores from all scales were obtained when applicable.

Measures of Agitation:

Agitation has been described as "inappropriate verbal, vocal, or motor activity that is not explained by needs or confusion per se." (Cohen‐Mansfield 1986). Changes in symptoms of agitation associated with antidepressant treatment were determined through changes in the total score of the CMAI or by changes on the agitation subscale scores contained in other neuropsychiatric symptom rating scales for dementia including:

• CMAI total score

• NPI ‐ agitation subscale score

• BEHAVE‐AD ‐ agitation subscale score or aggressiveness category score

• BRSD‐ irritability/agitation category score

• NBRS ‐ hyperactivity/agitation subscale item score

Measures of Psychosis:

• NPI ‐ Delusions subscale score

• NPI ‐ Hallucinations subscale score

• BEHAVE‐AD ‐ delusional thoughts and hallucination subscale scores

• BRSD ‐ psychotic category score

Cognitive Impairment:

• Alzheimer's Disease Assessment Scale cognitive subscale (ADAS‐cog) (Rosen 1984)

• Mini‐Mental State Examination (MMSE) (Folstein 1975)

• Clinical Dementia Rating sum of the boxes (CDR) (Berg 1998)

• Severe Impairment Battery (SIB) (Panisset 1994)

Safety and Tolerability Outcomes:

• Trial withdrawals due to adverse events

• Overall rates of trial withdrawal

• Trial withdrawal due to worsening of symptoms.

• Deaths during treatment.

• Rates of specific side‐effects (including but not limited to): somnolence, insomnia, headache, nausea, diarrhea, falls, bleeding, extrapyramidal symptoms, and hyponatremia

Primary outcomes

The primary efficacy outcome was change in symptoms of agitation and psychosis in dementia as measured on the various dementia NPS scales. The primary outcome for adverse effects was trial drop‐out due to adverse events. The timing of outcomes will be the last time measured in the study with no minimum duration.

Secondary outcomes

Secondary efficacy outcomes included changes on total scores for dementia NPS scales, changes on the CGI scale, changes in cognitive impairment scores, and caregiver distress. Secondary tolerability outcomes will include: falls, headache, gastrointestinal upset, worsening of dementia, anxiety, headache, bleeding, extrapyramidal symptoms, and hyponatremia.

Search methods for identification of studies

Electronic searches

We searched ALOIS (www.medicine.ox.ac.uk/alois) ‐ the Cochrane Dementia and Cognitive Improvement Group’s Specialized Register. The search terms included: citalopram, escitalopram, paroxetine, fluoxetine, fluvoxamine, sertraline, trazodone, nefazodone, venlafaxine, duloxetine, reboxetine, bupropion, amoxapine, amitriptyline, maprotiline, nortriptyline, desipramine, trimipramine, imipramine, protriptyline, doxepin, clomipramine, mirtazepine, mianserin, moclobemide, phenelzine, tranylcypromine, antidepressant.

ALOIS is maintained by the Trials Search Co‐ordinator and contains dementia and cognitive enhancement studies identified from:  

1. Monthly searches of a number of major healthcare databases: MEDLINE, EMBASE, CINAHL, PsycINFO and Lilacs

2. Monthly searches of a number of trial registers: meta Register of Controlled Trials; Umin Japan Trial Register; WHO portal (which covers ClinicalTrials.gov; ISRCTN; Chinese Clinical trials Register; German Clinical trials register; Iranian Registry of Clinical trials and the Netherlands National Trials Register, plus others)

3. Quarterly search of The Cochrane Library’s Central register of Controlled trials (CENTRAL)

4. Monthly searches of a number of grey literature sources: ISI Web of knowledge Conference Proceedings; Index to Theses; Australasian Digital Theses

Additional separate searches were run in each of the above sources to ensure that the most up‐to‐date results were retrieved. The search strategy used for the retrieval of reports of trials from MEDLINE (via the Ovid SP platform) can be seen in Appendix 1.

Searching other resources

We handsearched references retrieved from studies identified in the electronic searches. Science Citation Index was searched to identify additional relevant studies that had cited included studies. The Tripdatabases (www.tripdatabase.com) and the National Guideline Clearinghouse (www.guideline.gov) were also searched for relevant articles. The clinical trial databases ClinicalTrials.gov (www.clinicaltrials.gov) and pharmaceutical industry trial registers including Astra Zeneca (www.astrazenecaclinicaltrials.com, Bristol‐Myers Squibb (ctr.bms.com/ctd/registry.do), Eli Lilly and Company (www.lillytrials.com), GlaxoSmithKline (ctr.gsk.co.uk/medicinelist.asp), Novartis (www.novartisclinicaltrials.com/webapp/etrials.home.do) Roche (www.roche‐trials.com/registry.html) Wyeth (www.wyeth.com/clinicaltriallistings), Lundbeck (www.lundbecktrials.com), and Forrest (www.forestclinicaltrials.com) were also searched for published and unpublished trials. Study authors and experts in dementia were contacted for additional references to published and unpublished studies.

Data collection and analysis

Selection of studies

Two authors (DS and NA) independently reviewed the search results and the searches were merged using reference management software. Titles and abstracts were first reviewed to identify relevant reports. Full text of potentially relevant reports were retrieved and multiple reports of the same study were merged. The full‐text reports were then examined independently by two authors for inclusion criteria a third author was used to determine study eligibility in cases of disagreement. We contacted study authors for additional information to determine eligibility or information from missing results when necessary.

Data extraction and management

Two authors independently extracted information from each report. Information on study participants including age, gender distribution, baseline cognitive impairment (measured on MMSE or other measure), and baseline symptoms of severity of agitation or psychosis for the antidepressant and comparison groups were extracted for all included studies. Information on the name and dosages of antidepressants and comparison drugs were also obtained. For all outcomes, the mean difference and standard deviations was recorded for continuous outcomes and proportions for categorical outcomes. We recorded all outcomes based on intention‐to‐treat analysis. A standardized paper data extraction form was utilized in this process.

Assessment of risk of bias in included studies

The "Risk of bias" assessment provided by the Cochrane Collaboration was used to evaluate the potential risk of bias for the included studies. Information on the risk of bias for the following domains were recorded for each study: sequence generation; allocation concealment; blinding of participants, personnel and outcome assessors; incomplete outcome data; selective outcome reporting; and, other sources of bias. We rated each domain as having a low, high or unclear risk of bias according to the "Risk of bias" tool. We presented the assessment of risk of bias results graphically using a methodological quality and methodological summary graphs.

Measures of treatment effect

The efficacy outcomes were recorded as mean change in agitation, psychosis, and total dementia symptom rating scale scores for the primary and secondary efficacy outcomes. We also recorded the number of individuals classified as treatment responders where authors' defined response to therapy as a reduction in symptom scores by a prespecified percentage. Tolerability outcomes were recorded as the proportion of individuals experiencing the outcome for each treatment arm. We did not attempt to combine studies into common effect measures such as a standardized mean difference for studies that used different efficacy outcome measures.

Dealing with missing data

Study authors were contacted to provide additional information in the cases of missing data. In the case of missing data, standard deviations for continuous data was calculated when possible from information contained in the mean change and test statistics.

Assessment of heterogeneity

We examined the studies qualitatively for heterogeneity in study populations. The I² and Q statistics were used to assess statistical heterogeneity among studies.

Assessment of reporting biases

We included studies published in any language to minimize the potential for language bias. Publication bias was minimized by including searches unpublished reports and other methods as outlined in the search strategy. Potential publication bias was also examined by visual inspection of funnel plots of the meta‐analyses for the primary efficacy outcomes.

Data synthesis

We calculated weighted mean differences and standard deviations in meta‐analyses using random‐effects models for studies reporting continuous outcomes using the same rating scale. We used random‐effects Mantel‐Haenszel risk ratios and confidence intervals for meta‐analysis of binary outcomes for studies reporting the same outcome. Studies were assessed qualitatively to decide if meta‐analysis was appropriate for the outcome based on similarity of patient populations.

The following comparisons were planned for all efficacy and tolerability outcomes:

Selective Serotonin Reuptake Inhibitors

• SSRIs versus placebo

• SSRIs versus any antipsychotic

• SSRI versus atypical antipsychotics

• SSRI versus typical antipsychotics

• SSRI versus other medications

Tricyclic Antidepressants

• TCAs versus placebo

• TCAs versus any antipsychotic

• TCAs versus atypical antipsychotics

• TCAs versus typical antipsychotics

• TCAs versus other medications

Trazodone:

• Trazodone versus placebo

• Trazodone versus any antipsychotic

• Trazodone versus atypical antipsychotics

• Trazodone versus typical antipsychotics

• Trazodone versus other medications

Other Antidepressants:

• Other antidepressants versus placebo

• Other antidepressants versus any antipsychotic

• Other antidepressants versus atypical antipsychotics

• Other antidepressants versus typical antipsychotics

• Other antidepressants versus other medications

Subgroup analysis and investigation of heterogeneity

If significant heterogeneity was identified, we first attempted to stratify studies into more homogeneous subgroups. We planned to assess the risk of bias on outcomes by conducting subgroup analyses of higher and lower risk of bias of studies. We also planned to assess duration of treatment and medication dosage in a subgroup analyses. Additional subgroup analyses were planned for baseline severity of cognitive impairment and baseline severity of behavioral symptoms. We also analysed treatment setting (long‐term care or community residence) in a subgroup analysis.We planned to include a subgroup analysis for antidepressants that are currently marketed if any of the included medications have been withdrawn from most major markets (e.g. nefazodone). We also planned subgroup analyses of studies which included individuals with DLB as this population is more sensitive to adverse events related to antipsychotics (McKeith 2005).

Sensitivity analysis

We planned to investigate the effect of individual studies on the summary measures of effect obtained from meta‐analysis.

Results

Description of studies

Results of the search

A total of 1807 citations were retrieved from the initial search of electronic databases. Of these, the titles and abstracts of 393 citations were reviewed in further detail. The number of full length articles retrieved and reviewed was 78. After review of articles a total of nine studies met inclusion criteria.

Included studies

Nine studies (Finkel 2004; Nyth 1990; Olafsson 1992; Gaber 2001; Pollock 2002; Auchus 1997; Pollock 2007; Sultzer 1997; Teri 2000) with a total of 692 individuals met inclusion criteria. A full description of the characteristics of included studies are described in Characteristics of included studies.

There were five studies comparing the SSRIs to placebo. Finkel 2004 examined the effects of sertraline (25 ‐ 200 mg/d) compared with placebo in a twelve week trial involving 244 individuals with Alzheimer's dementia who failed to respond to 8 weeks of treatment with open‐label donepezil. Nyth 1990 compared the effects of citalopram (20 to 30 mg/d) to placebo in a 4 week trial of 98 individuals with dementia. This study involved an initial randomized, placebo‐controlled phase for 4 weeks followed by open‐label treatment and a second placebo‐controlled randomized trial involving withdrawal of medication. Information from the initial 4 week placebo‐controlled treatment study were used for the analysis. The third study Olafsson 1992 compared fluvoxamine (50 to 150 mg/d) to placebo in a 6 week trial involving 46 individuals with dementia. The remaining placebo‐controlled trials by Auchus 1997 and Pollock 2002 also included a typical antipsychotic comparison group and are described below.

There were three studies evaluating the effects of SSRIs compared to the typical antipsychotics. Auchus 1997 compared the SSRI fluoxetine (20 mg/d) to haloperidol (3 mg/d) and placebo in 15 older adults with AD over 6 weeks of treatment. Gaber 2001 evaluated the effect of sertraline (25 to 50 mg/d) to haloperidol (1 to 2 mg/d) in 23 individuals with Alzheimer's, vascular or mixed dementia in a 10 week trial. Pollock 2002 compared citalopram (20 mg/d) to perphenazine (0.1mg/kg/d) and placebo in individuals with AD, vascular, mixed, DLB or unspecified dementia in a 17 day trial. The proportion of individuals with possible or probable DLB in the Pollock 2002 trial was 15/31 (48%), 8/33 (24%), and 9/21 (43%) in the citalopram, perphenazine, and placebo groups respectively.

Pollock 2007 compared the effects of citalopram (20 ‐ 40 mg/d) to risperidone (1 ‐ 2 mg/d) in individuals with AD, vascular, mixed, DLB or unspecified dementia in a 12 week trial. The number of participants with dementia with Lewy bodies in the citalopram arm was 14/53 (26.4%) and 17/50 (34%) in the risperidone arm (personal communication, Dr. Bruce Pollock).

Two studies evaluated the effects of trazodone. Sultzer 1997 compared trazodone (50 ‐ 250 mg/d) to haloperidol (1 ‐ 5 mg/d) among 28 individuals with dementia. The second study Teri 2000 compared trazodone (50 to 300 mg/d) to haloperidol (0.5 to 3 mg/d), placebo, or behavioral management in 148 individuals with Alzheimer's dementia.

Excluded studies

There was also insufficient information available for data extraction from another unpublished study Schneider 2009 to include in the review.

Risk of bias in included studies

The potential risk of bias within included studies is summarized in the methodological quality graph Figure 1 and methodological quality summary Figure 2. None of the included studies were of low risk of bias on all methodological quality items and for many studies the potential risk of bias was unclear from information reported in the study publication.

1.

Methodological quality graph: review authors' judgements about each methodological quality item presented as percentages across all included studies.

2.

Methodological quality summary: review authors' judgements about each methodological quality item for each included study.

Effects of interventions

SSRIs compared to Placebo

Five studies evaluated the effects of SSRIs when compared to placebo for symptoms of agitation and psychosis in dementia (Auchus 1997; Finkel 2004; Nyth 1990; Olafsson 1992; Pollock 2002).

Change in CMAI Total Scores:

Two studies reported the effects of SSRIs on change in CMAI total scores (Auchus 1997; Finkel 2004) Analysis 1.1. The mean difference in CMAI total score for SSRIs compared to placebo in the study Auchus 1997 was 2.80 (95% CI ‐5.84 to 11.44) while the mean difference in change in CMAI total score for the study Finkel 2004 was ‐0.90 (95% CI ‐1.20 to ‐ 0.57). In meta‐analysis there was a significant reduction in CMAI total score for SSRI compared to placebo (MD ‐0.89, 95% CI, ‐1.22 to ‐0.57, Z=5.32, p<0.00001) Figure 3.

1.1. Analysis.

Comparison 1 SSRI versus Placebo, Outcome 1 CMAI Change in Total Score.

3.

Forest plot of comparison: 1 SSRI versus Placebo, outcome: 1.1 CMAI Change in Total Score.

Change in NPI Total Scores:

One study (Finkel 2004) examined change in NPI total scores for SSRIs compared to placebo. The change in NPI total score for SSRIs was ‐4.7 (SD 17.6) compared to ‐6.5 (SD 12) for placebo with no statistically significant difference in the mean change in NPI total scores for the two groups (MD 1.80, 95% CI ‐2.01 to 5.61, Z=0.93, p=0.35) Analysis 1.2.

1.2. Analysis.

Comparison 1 SSRI versus Placebo, Outcome 2 NPI Change in Total Score.

Change in BEHAVE‐AD Total Score:

The change in BEHAVE‐AD total scores was reported in one study (Finkel 2004). The change in BEHAVE‐AD total scores was ‐1.5 (SD 5.5) for SSRIs compared to ‐0.8 (SD 4.4) for placebo with no statistically significant difference in the mean change in BEHAVE‐AD total scores (MD ‐0.70, 95 CI, ‐1.95 to 0.55, Z=1.09, p=0.27) Analysis 1.3.

1.3. Analysis.

Comparison 1 SSRI versus Placebo, Outcome 3 BEHAVE‐AD Change in Total Score.

Change in NBRS Total Score:

The change in NBRS total scores was reported in one study (Pollock 2002). The change in NBRS total score was ‐10 (13.62) for citalopram compared to ‐2.3 (17.46) for placebo. The difference in change in NBRS was statistically significant after adjusting for baseline score although the unadjusted differences NBRS change scores were not statistically significant (MD ‐7.7, 95 CI, ‐16.57 to 1.17, Z=1.70, p=0.09) Analysis 1.4.

1.4. Analysis.

Comparison 1 SSRI versus Placebo, Outcome 4 NBRS Change in Total Score.

Trial Withdrawals due to Adverse Events

Four studies reported on the proportion of trial participants withdrawn from trials due to adverse events associated with study medication for SSRIs compared to placebo. The proportion of individuals with trial withdrawals due to adverse events was 24/200 (12%) for SSRI and 21/199 (11%) for placebo with no significant difference in the RR for trial withdrawals due to adverse events for SSRIs compared to placebo (relative risk (RR), 1.07, 95% CI, 0.55 to 2.11, Z=0.20, p=0.84) Analysis 1.5Figure 4.

1.5. Analysis.

Comparison 1 SSRI versus Placebo, Outcome 5 Trial Withdrawal Due to Adverse Events.

4.

Forest plot of comparison: 1 SSRI versus Placebo, outcome: 1.2 Trial Withdrawal Due to Adverse Events.

Trial Withdrawals due to Any Cause

Three studies reported on the risk of trial withdrawals due to any cause for SSRIs compared to placebo (Finkel 2004; Olafsson 1992; Pollock 2002). The proportion of individuals with trial withdrawal due to adverse events was 45/177 (25%) for SSRIs and 44/166 (27%) for placebo with no significant difference in the proportion of individuals who withdrew from studies due to any cause (RR, 0.91, 95% CI 0.65 to 1.26, Z=0.59, p=0.56) Analysis 1.6.

1.6. Analysis.

Comparison 1 SSRI versus Placebo, Outcome 6 Trial Withdrawal Due to Any Cause.

SSRIs compared to Atypical Antipsychotics

Change in NBRS Total Scores:

One study reported the effect of citalopram to risperidone on change on the NBRS total score (Pollock 2007). The mean change change in NBRS score was for citalopram ‐1.26 (SD 4.58) and ‐0.73 (4.91) for risperidone with no significant difference noted in the change in NBRS scores between the two groups (MD ‐0.53, 95 CI, ‐2.37 to 1.31, Z=0.57, p=0.57) Analysis 2.1.

2.1. Analysis.

Comparison 2 SSRI versus Atypical Antipsychotic, Outcome 1 NBRS Change in Total Score.

NBRS Psychosis Subscale:

Pollock 2007 also reported the change in the psychosis subscale of the NBRS scale and found no significant difference in the change in psychosis subscales for citalopram compared to risperidone (MD 0.26, 95 CI, ‐1.51 to 2.03, Z=0.29, p=0.77) Analysis 2.2.

2.2. Analysis.

Comparison 2 SSRI versus Atypical Antipsychotic, Outcome 2 NBRS Change in Psychosis Subscale.

Trial Withdrawals due to Adverse Events:

The proportion of individuals with trial withdrawal due to adverse events was 4/53 (8%) for citalopram and 9/50 (18%) for risperidone in Pollock 2007 with no statistically significant difference in the proportion of individuals who withdrew due to adverse events for citalopram when compared to risperidone (RR 0.88, 95 CI, 0.63 to 1.24, Z=0.73, p=0.46) Analysis 2.4.

2.4. Analysis.

Comparison 2 SSRI versus Atypical Antipsychotic, Outcome 4 Trial Withdrawal Due to Any Cause.

Trial Withdrawal due to Any Cause:

The proportion of individuals with trial withdrawals due to any cause was 28/53 (53%) for citalopram and 30/50 (60%) for risperidone in Pollock 2007 with no statistically significant difference in the proportion of individuals who withdrew from studies due to any cause for citalopram when compared to risperidone (RR 0.37, 95 CI, 0.11 to 1.30, Z=1.55, p=0.12) Analysis 2.3.

2.3. Analysis.

Comparison 2 SSRI versus Atypical Antipsychotic, Outcome 3 Trial withdrawal Due to Adverse Events.

SSRIs compared to Typical Antipsychotics

Change in CMAI Total Scores:

Two studies reported the effect of SSRIs compared to the typical antipsychotic haloperidol on measures of agitation and psychosis (Auchus 1997; Gaber 2001) both of which evaluated agitation using the CMAI. The mean difference in CMAI total score for SSRI compared to haloperidol in the study Auchus 1997 was 3.80 (95% CI ‐6.18 to 13.78) and the mean difference in CMAI total scores in the study Gaber 2001 was 6.49 (95% CI, ‐8.11 to 21.09). There was no significant difference in the change in CMAI total scores for SSRIs compared to typical antipsychotics in meta‐analysis (MD 4.66, 95% CI, ‐3.58 to 12.90, Z=1.11, p=0.27) Analysis 3.1Figure 5.

3.1. Analysis.

Comparison 3 SSRI versus Typical Antipsychotic, Outcome 1 CMAI Change in Total Score.

5.

Forest plot of comparison: 1 SSRI versus Haloperidol, outcome: 1.1 CMAI Total Score.

Change in NBRS Total Scores:

The mean change in NBRS total scores for citalopram was ‐10 (SD 13.62) and ‐7.2 (SD 17.05) for perphenazine with no significant difference between the citalopram and perphenazine (MD ‐2.80, SD ‐10.34 to 4.74, Z=0.73, p=0.47) Analysis 3.2 in the one study reporting this outcome (Pollock 2002).

3.2. Analysis.

Comparison 3 SSRI versus Typical Antipsychotic, Outcome 2 NBRS Change in Total Score.

Trial Withdrawals due to Adverse Events:

The proportion of individuals with trial withdrawal due to adverse events was 0/5 (0%) for fluoxetine compared to 2/5 (40%) for haloperidol in Auchus 1997 with no significant difference in the risk of withdrawal due to adverse events for fluoxetine when compared to haloperidol (RR, 0.13, 95% CI, 0.00 to 3.52, Z=1.22, p=0.22) Analysis 3.3.

3.3. Analysis.

Comparison 3 SSRI versus Typical Antipsychotic, Outcome 3 Trial Withdrawal Due to Adverse Events.

Trial Withdrawal due to Any Cause:

The proportion of individuals with trial withdrawal due to any cause was 16/31 (52%) for citalopram and 18/31 (58%) for perphenazine in Pollock 2002 with no significant difference in the relative risk of trial withdrawals for citalopram compared to perphenazine (RR, 0.89, 95% CI, 0.33 to 2.37, Z=0.23, p=0.81) Analysis 3.4.

3.4. Analysis.

Comparison 3 SSRI versus Typical Antipsychotic, Outcome 4 Trial Withdrawals Due to Any Cause.

Trazodone compared to Placebo

Change in CMAI Total Scores:

One study reported on the change in CMAI scores for trazodone compared to placebo (Teri 2000). The change in CMAI total scores for the trazodone group was ‐0.76 (SD 16.76) while the change in CMAI total scores in the placebo group was ‐5.35 (SD 18.5). There was no significant difference in the change in CMAI scores for trazodone compared to placebo (MD, 5.18, 95% CI, ‐2.86 to 13.22, Z=1.26, p=0.21) Analysis 4.1.

4.1. Analysis.

Comparison 4 Trazodone versus Placebo, Outcome 1 CMAI Change in Total Score.

Trial Withdrawals Due to Any Cause

The proportion of individuals with trial withdrawal due to any cause was 12/37 (32%) for trazodone and 11/36 (31%) for placebo in Teri 2000 with no significant difference in the risk of trial withdrawal due to any cause for trazodone compared to placebo (RR, 1.06, 95% CI, 0.54 to 2.09, Z=0.17, p=0.86) Analysis 4.2.

4.2. Analysis.

Comparison 4 Trazodone versus Placebo, Outcome 2 Trial Withdrawal Due to Any Cause.

Trazodone compared to Typical Antipsychotics

Change in CMAI Total Scores:

Two studies reported on the change CMAI scores associated with trazodone compared to haloperidol (Sultzer 1997; Teri 2000) with no significant difference in the change in CMAI score for trazodone when compared to haloperidol (MD 3.28, 95% CI, ‐3.28 to 15.74, Z=0.98, p=0.33) Analysis 5.1Figure 6.

5.1. Analysis.

Comparison 5 Trazodone versus Typical Antipsychotic, Outcome 1 CMAI Change in Total Score.

6.

Forest plot of comparison: 4 Trazodone versus Haloperidol, outcome: 4.1 CMAI Change in Total Score.

Clinical Global Impression

Two studies reported on numbers of individuals who were assessed to be much or very much improved according to the CGI for trazodone when compared to haloperidol (Sultzer 1997,Teri 2000). There was no statistically significant difference in the relative risk (RR) of improvement for trazodone compared to haloperidol in meta‐analysis of these two studies (RR 1.25, 95% CI, 0.82 to 2.34, Z=1.05, p=1.00) Analysis 5.2Figure 7.

5.2. Analysis.

Comparison 5 Trazodone versus Typical Antipsychotic, Outcome 2 CGI ‐ much or very much improved.

7.

Forest plot of comparison: 1 Trazodone versus Haloperidol, outcome: 1.1 CGI ‐ much or very much improved.

Trial Withdrawal Due to Any Cause

The proportion of individuals with trial withdrawal due to any cause was 12/37 (32%) for trazodone and 14/34 (41%) for haloperidol in Teri 2000 with no significant difference in the risk of trial withdrawal due to any cause for trazodone compared to haloperidol (RR, 0.79, 95% CI, 0.43 to 1.46, Z=0.76, p=0.45).

Subgroup Analysis

Subgroup analysis of the outcome of trial withdrawals due to any cause for SSRIs compared to placebo was undertaken for studies that included patients with DLB (Pollock 2002) and studies that did not include patients with DLB (Finkel 2004,Olafsson 1992) and no significant difference in the rate of trial withdrawals due to any cause for SSRIs compared to placebo in these two subgroups. There was no information available from other studies to undertake any of the additional planned subgroup analyses.

Discussion

Summary of main results

We found a relatively small number of studies evaluating the effects of antidepressants on symptoms of agitation and psychosis in dementia with only nine studies meeting inclusion criteria. Many of the studies included in the review had relatively small sample sizes and were also at potential risk of bias from various aspects related to study design although information pertinent to assessment of risk of bias was unclear in many studies. Given the small number of studies, and variation in the methodology used to measure symptoms of agitation and psychosis, there were relatively few outcomes that could be included in meta‐analysis. Two studies of found some benefit of citalopram and sertraline for treating NPS of dementia when compared to placebo. For most other outcomes measuring agitation and psychosis there were no statistically significant difference for SSRIs compared to typical or atypical antipsychotics. There were also no statistical differences in measure of drug tolerability such as the proportion of individuals withdrawn from studies due to any cause or for adverse events noted for SSRIs compared to typical or atypical antipsychotics except for citalopram which had an improved side‐effect profile when compared to perphenazine as measured on the UKU side‐effect scale. There was no significant difference noted for trazodone when compared to placebo or typical antipsychotic comparators on measures of efficacy and tolerability.

Quality of the evidence

The overall quality of studies included in this review is questionable given that many of the studies did not provide a detailed description of study methods to allow us to determine the potential risk of bias. As such, our assessment of study quality reported that most items assessed by the risk of bias tool were of unclear risk of bias. However, given the relatively few studies of antidepressant treatment for psychosis and agitation in dementia we included all studies regardless of potential risk of bias to reflect the best evidence currently available on this topic. Also, given the relatively small number of studies and diversity of outcomes reported in the studies we were unable to conduct the planned subgroup analyses based on potential of risk of bias to evaluate the effects of potential bias on observed outcomes. One potential limitation of the studies included in this review is the relatively high rates of trial withdrawal observed in many studies which may limit the interpretability of the results. Also, the reporting of adverse events in many studies was limited and it is difficult to determine if the frequency of adverse events were low or present and not reported.

Potential biases in the review process

We included both English and non‐English publications to minimize the likelihood of publication bias according to publication language. Our search for unpublished trials also helped to minimize the potential for publication bias. Due to the small number of studies we were unable undertake analysis of funnel plots to assess for potential publication bias.

Agreements and disagreements with other studies or reviews

To date, there have been no previous reviews on antidepressants for the treatment of agitation or psychosis in dementia. Previous reviews on the treatment of NPS of dementia have suggested that antidepressants may be considered before antipsychotics in the treatment of this condition presumably due to a decreased rate of adverse events associated with antidepressants when compared to other medications such as atypical antipsychotics (Salzman 2008; Sink 2005). Although atypical antipsychotics are the most extensively studied medication class for treatment of agitation and psychosis in dementia, these medications are known to be associated with an increased risk of both overall mortality (Schneider 2005; Gill 2007) and stroke (Sacchetti 2010). From the current direct comparison studies of SSRIs and trazodone to placebo or typical antipsychotics we conclude that currently there is some limited evidence to support the use of antidepressants as a safe and effective alternative to other pharmacological or non‐pharmacological interventions for agitation and psychosis in dementia.

Authors' conclusions

Implications for practice.

Currently, there are few high‐quality studies examining the efficacy and safety of antidepressants for the treatment of agitation and psychosis in dementia. At the present time, there is limited evidence to support the use of antidepressants for this indication. Antidepressants such as citalopram, sertraline, and trazodone may improve symptoms of agitation and psychosis for some individuals with dementia and given that the tolerability and safety of these medications appears to be similar to placebo and certain antipsychotics, these medications may be considered as a potential treatment for these symptoms. The decision for clinicians to utilize antidepressants for the treatment of agitation and psychosis in dementia should be guided by the availability of alternate treatments, an assessment of potential benefits and harms potentially associated with the treatment, and an assessment of patient and caregiver preference. Future studies are required to determine the safety and efficacy SSRIs or trazodone for the treatment of agitation and psychosis in dementia.

Implications for research.

A better understanding of the effects of antidepressants on agitation and psychosis is needed before concluding that antidepressants are safe and effective treatments for this condition. Results from one completed (Schneider 2009) and two ongoing (NCT00898807; NCT00260624) randomized, controlled trials could provide important additional information on the safety and efficacy of antidepressants for this condition. Given the heterogeneity of study designs, inclusion criteria, and the variety of outcome measures employed in studies of interventions for agitation and psychosis in dementia, utilization of standardized outcome measures and inclusion criteria would facilitate comparisons of outcomes across studies and future meta‐analyses. Future studies are also required to determine if certain patient populations or symptoms are most likely to respond to treatment with these medications.

Appendices

Appendix 1. MEDLINE search strategy

Source

Search strategy

Medline (Ovid SP)

1. (anti‐depres* or antidepres*).mp. 2. Antidepressive Agents/ 3. citalopram.mp. or Citalopram/ 4. escitalopram.mp. 5. paroxetine.mp. or Paroxetine/ 6. fluoxetine.mp. or Fluoxetine/ 7. fluvoxamine.mp. or Fluvoxamine/ 8. sertraline.mp. or Sertraline/ 9. trazodone.mp. or Trazodone/ 10. nefazodone.mp. 11. venlafaxine.mp. 12. duloxetine.mp. 13. reboxetine.mp. 14. bupropion.mp. or Bupropion/ 15. amoxapine.mp. or Amoxapine/ 16. amitriptyline.mp. or Amitriptyline/ 17. nortriptyline.mp. or Nortriptyline/ 18. desipramine.mp. or Desipramine/ 19. trimipramine.mp. or Trimipramine/ 20. imipramine.mp. or Imipramine/ 21. protriptyline.mp. or Protriptyline/ 22. doxepin.mp. or Doxepin/ 23. clomipramine.mp. or Clomipramine/ 24. mirtazapine.mp. 25. mianserin.mp. or Mianserin/ 26. moclobemide.mp. or Moclobemide/ 27. phenelzine.mp. or Phenelzine/ 28. tranylcypromine.mp. or Tranylcypromine/ 29. (SSRI* or "selective serotonin reuptake inhibitor" or TCA*).mp. 30. or/1‐29 31. Dementia/ 32. Dementia, Multi‐Infarct/ 33. Dementia, Vascular/ 34. Alzheimer Disease/ 35. Lewy Body Disease/ 36. Delirium/ 37. Huntington Disease/ 38. "Pick Disease of the Brain"/ 39. Kluver‐Bucy Syndrome/ 40. Wernicke Encephalopathy/ 41. Creutzfeldt‐Jakob Syndrome/ 42. Delirium, Dementia, Amnestic, Cognitive Disorders/ 43. dement*.mp. 44. Alzheimer*.mp. 45. (lewy* and bod*).mp. 46. deliri*.mp. 47. ((cognit* or memory* or mental*) and (declin* or impair* or los* or deteriorat*)).mp. 48. (chronic and cerebrovascular).mp. 49. ("organic brain disease" or "organic brain syndrome").mp. 50. "supra nuclear palsy".mp. 51. ("normal pressure hydrocephalus" and "shunt*").mp. 52. "benign senescent forgetfulness".mp. 53. (cerebr* and deteriorat*).mp. 54. (cerebral* and insufficient*).mp. 55. (confusion* or confused).mp. 56. (pick* and disease).mp. 57. (creutzfeldt or jcd or cjd).mp. 58. huntington*.mp. 59. binswanger*.mp. 60. korsako*.mp. 61. (mci or "subjective memory complaint" or "episodic memory").mp. 62. "pre‐clinical ad".mp. 63. or/31‐62 64. 63 and 30 65. randomized controlled trial.pt. 66. controlled clinical trial.pt. 67. random*.ab. 68. placebo.ab. 69. drug therapy.fs. 70. trial.ab. 71. or/65‐70 72. (animals not (humans and animals)).sh. 73. 71 not 72 74. 64 and 73

Data and analyses

Comparison 1. SSRI versus Placebo.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 CMAI Change in Total Score	2	250	Mean Difference (IV, Random, 95% CI)	‐0.89 [‐1.22, ‐0.57]
2 NPI Change in Total Score	1	240	Mean Difference (IV, Random, 95% CI)	1.80 [‐2.01, 5.61]
3 BEHAVE‐AD Change in Total Score	1	240	Mean Difference (IV, Random, 95% CI)	‐0.7 [‐1.95, 0.55]
4 NBRS Change in Total Score	1	52	Mean Difference (IV, Random, 95% CI)	‐7.7 [‐16.57, 1.17]
5 Trial Withdrawal Due to Adverse Events	4	399	Risk Ratio (M‐H, Random, 95% CI)	1.07 [0.55, 2.11]
6 Trial Withdrawal Due to Any Cause	3	343	Risk Ratio (M‐H, Random, 95% CI)	0.91 [0.65, 1.26]
7 Occurence of Insomnia	1	244	Risk Ratio (M‐H, Random, 95% CI)	1.11 [0.56, 2.17]
8 Occurence of Diarrhea	2	342	Risk Ratio (M‐H, Random, 95% CI)	2.37 [1.35, 4.15]
9 Occurence of Anorexia	1	244	Risk Ratio (M‐H, Random, 95% CI)	1.23 [0.65, 2.30]

1.7. Analysis.

Comparison 1 SSRI versus Placebo, Outcome 7 Occurence of Insomnia.

1.8. Analysis.

Comparison 1 SSRI versus Placebo, Outcome 8 Occurence of Diarrhea.

1.9. Analysis.

Comparison 1 SSRI versus Placebo, Outcome 9 Occurence of Anorexia.

Comparison 2. SSRI versus Atypical Antipsychotic.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 NBRS Change in Total Score	1	103	Mean Difference (IV, Random, 95% CI)	‐0.53 [‐2.37, 1.31]
2 NBRS Change in Psychosis Subscale	1	103	Mean Difference (IV, Fixed, 95% CI)	0.26 [‐1.51, 2.03]
3 Trial withdrawal Due to Adverse Events	1	103	Risk Ratio (M‐H, Random, 95% CI)	0.42 [0.14, 1.28]
4 Trial Withdrawal Due to Any Cause	1	103	Risk Ratio (M‐H, Random, 95% CI)	0.88 [0.63, 1.24]
5 UKU Side Effect Scale Total Score	1	103	Mean Difference (IV, Random, 95% CI)	‐2.82 [‐4.94, ‐0.70]

2.5. Analysis.

Comparison 2 SSRI versus Atypical Antipsychotic, Outcome 5 UKU Side Effect Scale Total Score.

Comparison 3. SSRI versus Typical Antipsychotic.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 CMAI Change in Total Score	2	33	Mean Difference (IV, Random, 95% CI)	4.66 [‐3.58, 12.90]
2 NBRS Change in Total Score	1	64	Mean Difference (IV, Fixed, 95% CI)	‐2.8 [‐10.34, 4.74]
3 Trial Withdrawal Due to Adverse Events	1	10	Risk Ratio (M‐H, Random, 95% CI)	0.2 [0.01, 3.35]
4 Trial Withdrawals Due to Any Cause	1	64	Risk Ratio (M‐H, Random, 95% CI)	0.95 [0.60, 1.50]

Comparison 4. Trazodone versus Placebo.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 CMAI Change in Total Score	1	73	Mean Difference (IV, Random, 95% CI)	5.18 [‐2.86, 13.22]
2 Trial Withdrawal Due to Any Cause	1	73	Risk Ratio (M‐H, Random, 95% CI)	1.06 [0.54, 2.09]
3 Occurrence of Somnolence	1	73	Risk Ratio (M‐H, Random, 95% CI)	1.17 [0.39, 3.49]
4 Occurence of Fatigue	1	74	Risk Ratio (M‐H, Random, 95% CI)	0.67 [0.20, 2.17]
5 Occurence of Tremor	1	73	Risk Ratio (M‐H, Random, 95% CI)	0.78 [0.23, 2.67]
6 Occurence of Parkinsonian Gait	1	73	Risk Ratio (M‐H, Random, 95% CI)	1.30 [0.31, 5.39]

4.3. Analysis.

Comparison 4 Trazodone versus Placebo, Outcome 3 Occurrence of Somnolence.

4.4. Analysis.

Comparison 4 Trazodone versus Placebo, Outcome 4 Occurence of Fatigue.

4.5. Analysis.

Comparison 4 Trazodone versus Placebo, Outcome 5 Occurence of Tremor.

4.6. Analysis.

Comparison 4 Trazodone versus Placebo, Outcome 6 Occurence of Parkinsonian Gait.

Comparison 5. Trazodone versus Typical Antipsychotic.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 CMAI Change in Total Score	2	99	Mean Difference (IV, Random, 95% CI)	3.28 [‐3.28, 9.85]
2 CGI ‐ much or very much improved	2	99	Risk Ratio (M‐H, Random, 95% CI)	1.25 [0.82, 1.90]
3 Trial Withdrawal Due to Any Cause	1	71	Risk Ratio (M‐H, Random, 95% CI)	0.79 [0.43, 1.46]
4 Occurence of Somnolence	1	71	Risk Ratio (M‐H, Random, 95% CI)	0.55 [0.22, 1.35]
5 Occurence of Fatigue	1	71	Risk Ratio (M‐H, Random, 95% CI)	0.19 [0.07, 0.51]
6 Occurence of Tremor	1	71	Risk Ratio (M‐H, Random, 95% CI)	0.41 [0.14, 1.20]
7 Occurence of Parkinsonian Gait	1	71	Risk Ratio (M‐H, Random, 95% CI)	0.53 [0.17, 1.64]

5.3. Analysis.

Comparison 5 Trazodone versus Typical Antipsychotic, Outcome 3 Trial Withdrawal Due to Any Cause.

5.4. Analysis.

Comparison 5 Trazodone versus Typical Antipsychotic, Outcome 4 Occurence of Somnolence.

5.5. Analysis.

Comparison 5 Trazodone versus Typical Antipsychotic, Outcome 5 Occurence of Fatigue.

5.6. Analysis.

Comparison 5 Trazodone versus Typical Antipsychotic, Outcome 6 Occurence of Tremor.

5.7. Analysis.

Comparison 5 Trazodone versus Typical Antipsychotic, Outcome 7 Occurence of Parkinsonian Gait.

Characteristics of studies

Characteristics of included studies [ordered by study ID]

Auchus 1997.

Methods	Randomized controlled trial comparing fluoxetine, haloperidol, and placebo Two week washout period followed by 6 week treatment phase
Participants	Participants with possible or probable Alzheimer's dementia CMAI scores >24 Total N=15
Interventions	Fluoxetine 20 mg daily (N=5) Haloperidol 3 mg daily (N=5) Placebo (N=5)
Outcomes	CMAI BEHAVE‐AD sections C,D,E Number of adverse events
Notes	Outpatients
Risk of bias
Bias	Authors' judgement	Support for judgement
Adequate sequence generation?	Unclear risk	"before undergoing random, double‐blind, assignment"
Incomplete outcome data addressed? All outcomes	High risk
Free of selective reporting?	Low risk
Free of other bias?	Low risk

Finkel 2004.

Methods	All individuals completed 8 weeks of open‐label treatment with donepezil prior to randomization Participants continuing to meet inclusion criteria following open‐label donepezil phase were randomized to either sertraline to placebo for 12 weeks
Participants	Probable or possible Alzheimer's dementia NPI total score > 5 and total severity score > 1 on at least one domain MMSE between 8 ‐ 23 Hachinski score < 5 CDR Score < 3
Interventions	Sertraline 25 ‐ 200 mg daily (N=124) Placebo (N=120)
Outcomes	NPI CGI BEHAVE‐AD CMAI
Notes	Outpatients
Risk of bias
Bias	Authors' judgement	Support for judgement
Adequate sequence generation?	Unclear risk	"randomized"
Blinding? All outcomes	Unclear risk	"double‐blind"
Free of other bias?	Unclear risk	Study sponsored by manufacturer of donepezil and sertraline

Gaber 2001.

Methods	Randomized controlled trial comparing sertraline to haloperidol 10 weeks duration
Participants	Participants with probable Alzheimer's, vascular or mixed dementia Total N=23
Interventions	Sertraline 25 ‐ 50 mg daily (N=13) Haloperidol 1 ‐ 2 mg daily (N=10)
Outcomes	CMAI
Notes	Institutionalized inpatients CMAI values were estimated from a graph
Risk of bias
Bias	Authors' judgement	Support for judgement
Adequate sequence generation?	Unclear risk	"randomized"
Blinding? All outcomes	Unclear risk	"double‐blind"

Nyth 1990.

Methods	Randomized, controlled trial with comparing citalopram to placebo, subsequent open label treatment and withdrawal phases One week washout period followed by double‐blind treatment phase for 4 weeks
Participants	DSM III criteria for primary degenerative or multi‐infarct dementia Excluded individuals with severe dementia defined as individuals being unable to assist with dressing or eating Total N=98
Interventions	Citalopram 20 mg with dose increase to 30 mg at week 3 if no significant treatment effect Placebo
Outcomes	Clinical Global Impression scale Gottfrie, Brane, and Steen scale UKU side‐effect rating scale
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Adequate sequence generation?	Unclear risk	"randomly assigned"
Blinding? All outcomes	Unclear risk	"double‐blind"

Olafsson 1992.

Methods	Randomized controlled trial comparing fluvoxamine to placebo 2 week washout followed by 6 weeks of treatment
Participants	Participants with DSM III primary degenerative or multi‐infarct dementia Total N=46
Interventions	Fluvoxamine 50 ‐ 150 mg daily (N=22) Placebo (N=24)
Outcomes	Gottfrie, Brane, and Steen Scale
Notes	Inpatient population
Risk of bias
Bias	Authors' judgement	Support for judgement
Adequate sequence generation?	Unclear risk	"randomly allocated"
Blinding? All outcomes	Unclear risk	"double‐blind"
Incomplete outcome data addressed? All outcomes	High risk
Free of other bias?	Low risk

Pollock 2002.

Methods	Randomized controlled trial comparing citalopram, perphenazine, and placebo 3 ‐ 5 day washout period followed by 17 days treatment
Participants	Participants with DSM‐IV and NINCDS‐ADRDA criteria for Alzheimer's, vascular, mixed or dementia not otherwise specified Total N=85
Interventions	Citalopram 10 mg daily for 3 days increased increased to 20 mg daily for 14 days (N=31) Perphenazine 0.05 mg/kg/day for 3 days increased to 0.1 mg/kg/day for 14 days (N=33) Placebo (N=21)
Outcomes	NBRS Udvalg for Kliniske Undersogelser (UKU) Side Effect Rating Scale MMSE
Notes	Inpatient population 32 patients met criteria for possible or probable dementia with Lewy bodies
Risk of bias
Bias	Authors' judgement	Support for judgement
Adequate sequence generation?	Unclear risk	"randomized"
Blinding? All outcomes	Unclear risk	"double‐blind"
Free of selective reporting?	Low risk
Free of other bias?	Low risk

Pollock 2007.

Methods	Randomized controlled trial comparing citalopram to risperidone 12 weeks duration
Participants	DSM IV criteria for Alzheimer's disease, vascular dementia, mixed dementia or dementia with Lewy bodies Total N=103
Interventions	Citalopram 10 mg daily for 3 days, increased to 20 mg daily, further dose increases permitted to maximum of 40 mg daily (N=53) Risperidone 0.5 mg daily for 3 days, increased to 1 mg daily, further dose increased permitted to maximum of 2 mg daily (N=50)
Outcomes	NBRS NPI Cornell scale for depression in dementia UKU side effect scale MMSE Severe Impairment Battery
Notes	Inpatient population
Risk of bias
Bias	Authors' judgement	Support for judgement
Adequate sequence generation?	Low risk	Stratified according to psychosis "generated by study statisticians"
Allocation concealment?	Low risk	"Only the research pharamacist knew the treatment assignments"
Blinding? All outcomes	Unclear risk	"double‐blind"
Incomplete outcome data addressed? All outcomes	Low risk
Free of selective reporting?	Low risk
Free of other bias?	Low risk

Sultzer 1997.

Methods	Randomized controlled trial comparing trazodone to haloperidol One week observation phase prior to randomization Three week dose titration phase followed by 6 weeks of extended treatment
Participants	All participants had significant agitation classified as CMAI > 41 or single CMAI item > 6 Total N=28
Interventions	Trazodone 50 ‐ 100 mg daily increased to maximum of 250 mg (N=14) Haloperidol 1 ‐ 2 mg daily, increased to maximum of 5 mg (N=14)
Outcomes	CMAI CGI Overt Aggression Scale (OAS) Adverse Effects Checklist Barnes Akathisia Scale
Notes	Inpatient population
Risk of bias
Bias	Authors' judgement	Support for judgement
Adequate sequence generation?	Unclear risk	"assigned randomly"
Blinding? All outcomes	Low risk	"Patients and all staff were blind to medication type"
Incomplete outcome data addressed? All outcomes	Low risk
Free of other bias?	Low risk

Teri 2000.

Methods	Randomized controlled trial comparing trazodone, haloperidol, placebo and behavioral management 16 week duration of trial
Participants	Participants with possible or probable Alzheimer's dementia Participants had a two week history of agitation All psychotropic medications were discontinued in the 2 weeks prior to randomization Total N=148
Interventions	Trazodone 50 ‐ 300 mg daily (N=37) Haloperidol 0.5 ‐ 3 mg daily (N=34) Placebo (N=36) Behavioral Management (N=41)
Outcomes	CGI BRSD (CERAD) CMAI Adverse Events
Notes	Outpatients
Risk of bias
Bias	Authors' judgement	Support for judgement
Adequate sequence generation?	Unclear risk	"randomly allocated"
Blinding? All outcomes	Unclear risk	"To insure that interviewers remained blind to treatment assignment, caregivers did not discuss any aspect of their treatment with the interviewer"
Incomplete outcome data addressed? All outcomes	Low risk

Characteristics of excluded studies [ordered by study ID]

Study	Reason for exclusion
Auer 1996	Insufficient information on patient population
Baumann 2001	All participants received perphenazine
Bentham 1990	No antidepressants comparator
Burke 1997	Retrospective study
Burke 1997a	Retrospective study
Cakir 2008	Open label trial
Dehlin 1985	Antidepressant withdrawn from use due to safety concerns
Devanand 2003	Only included participants with depression
Geldmacher 1994	Not randomized
Kim 2000	Case reports
Lawlor 1991	No antidepressant comparator
Lawlor 1994	Cross‐over study
Levkovitz 2001	All individuals received perphenazine prior to treatment with fluvoxamine
Lyketsos 2000	Only included patients with dementia and depression
Mervis 1991	Included individuals with organic personality disorder
Passeri 1987a	Included patients with depressed mood
Passeri 1993	No antidepressant comparator
Ramadan 1999	Not a randomized study
Reifler 1989	Included participants with and without depression
Scharre 2003	Open label study
Schneider 2009	Abstract from meeting with insufficient information to include in review
Siddique 2009	Retrospective study
Timmons 2004	Information on inclusion and exclusion criteria not available in abstract

Differences between protocol and review

In the published protocol studies which included individuals with dementia with Lewy bodies were excluded. In the final review studies that included a proportion of individuals with dementia with Lewy bodies were included in the review and the effects of DLB patients on outcomes were assessed in subgroup analyses.

Contributions of authors

Dallas Seitz and Nikesh Adunuri were involved in all aspects of the study design including writing the review protocol, review of citations and abstracts, data extraction and meta‐analysis. Sudeep Gill assisted with the title registration, protocol development, and writing the final review. Nathan Herrmann, Andrea Gruneir, and Paula Rochon assisted in developing research questions, and writing the study protocol, and drafting the final review.

Sources of support

Internal sources

• Alzheimer Society of Canada, Canada.

  Postdoctoral Fellowship Award ‐ Dr.Seitz

• Canadian Institutes of Health Research, Canada.

  Postdoctoral Fellowship Award ‐ Dr. Gruneir

• This was was supported by an Interdisciplinary Capacity Enhancement Grant (HOA‐80075) from the Canadian Institutes of Health Research (CIHR) Institute of Gender and Health and CIHR Institute of Aging, Canada.

External sources

• No sources of support supplied

Declarations of interest

Dr. Herrmann has received research support and/or speaker's honouraria from Lundbeck, Janssen‐Ortho, Pfizer, Novartis, Wyeth, and Eli Lilly. The remaining authors have no conflicts of interest to declare.

New