Intervention Review

You have free access to this content

Withdrawal versus continuation of chronic antipsychotic drugs for behavioural and psychological symptoms in older people with dementia

  1. Tom Declercq1,*,
  2. Mirko Petrovic2,
  3. Majda Azermai3,
  4. Robert Vander Stichele4,
  5. An IM De Sutter1,
  6. Mieke L van Driel5,
  7. Thierry Christiaens1

Editorial Group: Cochrane Dementia and Cognitive Improvement Group

Published Online: 28 MAR 2013

Assessed as up-to-date: 6 DEC 2012

DOI: 10.1002/14651858.CD007726.pub2

How to Cite

Declercq T, Petrovic M, Azermai M, Vander Stichele R, De Sutter AIM, van Driel ML, Christiaens T. Withdrawal versus continuation of chronic antipsychotic drugs for behavioural and psychological symptoms in older people with dementia. Cochrane Database of Systematic Reviews 2013, Issue 3. Art. No.: CD007726. DOI: 10.1002/14651858.CD007726.pub2.

Author Information

  1. 1

    Ghent University, Department of General Practice and Primary Health Care, Ghent, Belgium

  2. 2

    Ghent University Hospital, Department of Geriatrics, Ghent, Belgium

  3. 3

    Ghent University, Heymans Institute of Pharmacology, Ghent, Belgium

  4. 4

    Heymans Institute of Pharmacology, Ghent, Belgium

  5. 5

    The University of Queensland, Discipline of General Practice, School of Medicine, Brisbane, Queensland, Australia

*Tom Declercq, Department of General Practice and Primary Health Care, Ghent University, De Pintelaan 185, UZ-6K3, Ghent, B-9000, Belgium.

Publication History

  1. Publication Status: Edited (no change to conclusions)
  2. Published Online: 28 MAR 2013


Characteristics of included studies [ordered by study ID]
Ballard 2004

MethodsDouble-blind, placebo-controlled study

Participants100 older people (aged > 65 years) care-facility residents in the UK with probable or possible AD, no severe behavioural disturbances, taking neuroleptics for longer than 3 months were included. All participants having severe behavioural symptoms, i.e. individual scores > 7 on 1 of the 12 items of the NPI-scale at the time of evaluation were excluded

InterventionsAbrupt discontinuation of antipsychotics

OutcomesChanges in behavioural and psychiatric symptoms measured by the NPI, quality of life measured by well-being using DCM


Risk of bias

BiasAuthors' judgementSupport for judgement

Random sequence generation (selection bias)Unclear risk"Subjects were then randomized to neuroleptic (N=54) or placebo (N=46)"

Method of sequence generation not described

Allocation concealment (selection bias)Unclear riskMethod of allocation concealment not reported

Blinding (performance bias and detection bias)
All outcomes
Low risk"The study was conducted using a double-blind design. All study neuroleptics were encapsulated by an independent company to maintain blind, and dispensing was coordinated by the pharmacy departments at the 2 centres. Prescriptions were written prior to randomization in a twice-daily regimen, allocating to each participant the closest dose to their pre-existing prescription from the doses encapsulated"

"...the centre coordinator, blinded to neuroleptic status..."

Encapsulation of the administered drugs ensured blinding of participants and doctors/nurses, but blinding of outcome assessors is not described

Incomplete outcome data (attrition bias)
NPI (Neuro Psychiatric Inventory) and DCM (Dementia Care Mapping)
Low risk"All evaluations were undertaken at baseline. The NPI and DCM assessments were also completed at 1- and 3-month follow-up. Study withdrawals and the proportion of people developing marked behavioural symptoms are described and compared between groups using the chi-square test"

"Fourteen patients (26% active treatment, 30% placebo) withdrew from the study in each group. There were only 6 withdrawals in the placebo-treated group (13%) and 5 withdrawals in the active treatment group (9%) because of behavioural deterioration. Other withdrawals were because of physical health problems (active N=3 (6%), placebo N=3 (7%)), protocol violation (active N=2(4%), placebo N=1(2%)) or withdrawal of consent (active N=3(6%), placebo N=2(4%))"

"Eighty-two (82%) of the patients completed at least 1 follow-up evaluation and were included in the primary outcome analysis"

"For all participants who completed at least 1 follow-up assessment, the last evaluation was carried forward"

Selective reporting (reporting bias)Low riskNPI and well-being scale reported, but MMSE only assessed at start of study

Other biasUnclear riskNot clear how participants were recruited

Ballard The DART-AD Trial

MethodsRandomised, blinded, placebo-controlled parallel 2-group treatment discontinuation study

Participants165 institutionalised people in the UK currently prescribed neuroleptics thioridazine, chlorpromazine, haloperidol, trifluoperazine or risperidone for behavioural or psychiatric disturbance in dementia for at least 3 months were included. The following people were excluded: those who were unable to complete primary outcome measures at baseline assessment; if a clinician responsible for care or study clinician considered the person suffered from any physical condition that would make participation in the trial distressing or likely to increase suffering; if the person was currently taking thioridazine and showing a prolonged QTc on electrocardiogram; if the person was likely to be unable to take capsules

InterventionsAbrupt discontinuation of neuroleptics

OutcomesPrimary outcome: total SIB score

NPS were evaluated by the NPI


Risk of bias

BiasAuthors' judgementSupport for judgement

Random sequence generation (selection bias)Low risk"Randomisation was performed centrally at the Centre for Statistics in Medicine in Oxford (CSMO), using dedicated computer software (MINIM)"

"The randomisation programme included a minimisation algorithm to ensure balanced allocation of participants across the intervention groups for important prognostic factors"

Allocation concealment (selection bias)Low risk"The statistician carrying out the randomisation had no direct contact with patients and allocation was, therefore, totally independent of patient recruitment"

"The clinician responsible for randomisation of a patient faxed a randomisation form to the CSMO (or sent e-mail) in exceptional circumstances) and provided details appropriate and sufficient for establishing eligibility. If a person was eligible and informed consent/assent had been obtained and baseline assessments had been completed, the patient was randomised by the statistician either to continue taking medication or to discontinue (placebo group). The statistician directly communicated the allocation to the relevant trial pharmacy, ensuring concealment"

Blinding (performance bias and detection bias)
All outcomes
Low riskThe clinicians, those administering the trial medication, the carers, the relatives, the participants themselves, and those assessing the outcomes were all blinded to treatment allocation

Incomplete outcome data (attrition bias)
NPI (Neuro Psychiatric Inventory) and DCM (Dementia Care Mapping)
Low riskTo give a completed data set the imputation method was used "filling in" missing data with plausible values

A sensitivity analysis was used to test the robustness of the SIB result. This analysis was limited to those people for whom the risk of possible floor and ceiling effects was smallest, i.e. SIB baseline cut-off values ≥ 40 but ≤ 90

Selective reporting (reporting bias)Low riskAll intended primary and secondary outcomes are reported in the first and the follow-up study

Other biasLow riskThere are limiting factors to consider the interpretation of this excellent trial

First, recruitment focused on people living in residential care where moderate and severe dementia usually predominates, and the participants generally are older and frailer than their counterparts in other settings. Thus, the results are not easily extrapolated to individuals who are cared for in other community settings

Second, 89% of the participants were taking haloperidol or risperidone, but pharmacological profiles of neuroleptics differ, so that the study might not adequately represent the effects of discontinuation of other neuroleptics

Furthermore, polypharmacy is common in residential care, and the study did not consider other psychotropic prescriptions

Finally, high participant attrition sharply reduced the statistical power and scope for analysis of outcomes at 12 months. Imputation procedures and sensitivity analyses established robustness of estimates, but they cannot account for type II errors (i.e. false-negative interpretation)

Bridges-Parlet 1997

MethodsDouble-blind, baseline-treatment-controlled pilot study

Participants36 nursing home residents in the US with a diagnosis of dementia and a history of physically aggressive behaviour taking for 3 months prior to the study a stable dose of any neuroleptic. People with primary psychiatric diagnoses, mental retardation and terminal illness or other recent acute changes were excluded

InterventionsAbrupt withdrawal or tapering off a neuroleptic when baseline dose exceeded the equivalent of 50 mg of chlorpromazine. The tapering was done by dropping the baseline neuroleptic dose by half during week 1 and then discontinuing the neuroleptic completely at the beginning of week 2

OutcomesPrimary outcome measures: completion of the 4 weeks of study, change in the amount of observed physically aggressive behaviour in each week over the course of the trial

Secondary outcomes were a number of other behaviours, such as verbally aggressive behaviour, walking or amount of time spent sleeping

These behaviours were observed by experienced study personnel, blinded to treatment assignments and using a portable bar-code reader capable of storing several hours of observation

Physically aggressive acts were identified by type, each of which was coded separately by the bar-code system. Specific types of physically aggressive behaviour included hitting, biting, scratching, kicking and pushing. The severity of a particular instance was not recorded, but if multiple instances of striking occurred, the code for the particular type of physically aggressive behaviour was recorded for each minute in which physical contact occurred. The measure of physically aggressive acts that was used as the primary outcome measure was the sum of all bar-code reader-identified instances of all five types of physically aggressive behaviour

Verbal aggressiveness was defined as an instance of speaking in an angry tone of voice, swearing or yelling in anger. Individual instances were recorded in 1-minute increments, so that if a sustained period of swearing or yelling occurred, each minute in which the behaviour occurred was noted

For behaviours such as walking, sitting restrained or sleeping, the observers recorded the person's current activity at least every minute. Based on this complete activity log, we were able to tabulate the total time spent in these activities

The observers also kept hand-written notes of the period of observation as a back-up


Risk of bias

BiasAuthors' judgementSupport for judgement

Random sequence generation (selection bias)Low risk"At the end of week 1, subjects were randomly assigned to either withdrawal or no withdrawal"

"Assignment was based on a predetermined sequence such that three patients were assigned to withdrawal for every two not withdrawn"

Allocation concealment (selection bias)Unclear riskMethod of allocation concealment not described. It may not have been blinded

Participant groups were well matched for age, chlorpromazine-equivalent neuroleptic dose and physically aggressive behaviour at baseline

Blinding (performance bias and detection bias)
All outcomes
Low risk"The patients were directly observed by experienced study personnel who were blinded to treatment assignments"

"Patients in both groups received identical-appearing capsules prepared at the University of Minnesota Hospital Pharmacy. Patients receiving their medication in crushed form, received in the placebo group tablets of vitamin C instead of capsules. The patient receiving intramuscular mesoridazine daily was given intramuscular saline from a nurse not directly involved in the patient's care"

Completion of the 4-week programme was the primary outcome. Nursing staff was involved in decision to discontinue the programme. They were blinded for the treatment allocated, thus outcome assessment may have been adequately blinded

Incomplete outcome data (attrition bias)
NPI (Neuro Psychiatric Inventory) and DCM (Dementia Care Mapping)
Low risk"Of the 22 patients who were withdrawn, 20 (91%) completed the 4-week double-blinded withdrawal"

"Two patients were restarted on medication on the recommendation of the nursing staff; only one went back on a neuroleptic"

"Of the 14 patients not withdrawn, all completed the 4-week trial"

"Of the 576 observation periods there were seven in which the bar-code reader failed. Handwritten back-up notes were used for physically aggressive behaviour frequency"

Selective reporting (reporting bias)Low riskAll intended outcomes reported

Other biasUnclear riskThere may have been selective recruitment limiting the generalisability of the results

"Five nursing homes referred patients to the study. Nursing supervisors identified physically aggressive patients with dementia who were currently being treated with a neuroleptic. We relied on the judgements of the nursing supervisors regarding the prior presence of physically aggressive behaviour. Other than the criteria we gave the supervisors, we did not know what criteria the nursing supervisors used to choose potential study subjects. It was beyond the resources of our project to collect data on all patients whom the nursing supervisors considered as eligible but either than rejected, or who were denied permission to participate by the family. It is possible that our study patients differed in some systematic way from nonparticipating patients. Our participants may have been more suitable for withdrawal than other patients on neuroleptics"

Cohen-Mansfield 1999

MethodsDouble-blind crossover study

Participants58 nursing home residents in the US older than 70 years taking, for at least 4 weeks, haloperidol, thioridazine or lorazepam for agitation were included. Because the diagnosis of dementia was not mentioned in the paper, we emailed the first author to ask her whether the participants included suffered from dementia and her answer was positive (referring to the MMSE scores). The following people were excluded: people with concomitant administration of other antipsychotic or antianxiety drugs other than low-dose trazodone hydrochloride for sleep, life expectancy of < 3 months, psychiatric diagnosis of a major affective disorder of schizophrenia, acute infection within 10 days before entering the nursing home, expectancy of leaving the nursing home within 3 months, and uncontrolled hyperglycaemia or hypoglycaemia

InterventionsWithdrawal of antipsychotic and lorazepam use by tapering to placebo during a 3-week period. After 7 weeks of taking either placebo or medication, the group taking placebo was titrated back to the initial dose and the group taking medication was tapered to placebo

OutcomesPrimary outcomes were overall psychiatric symptoms measured by the BPRS and agitation measured by the CMAI

The BPRS assessed the following symptoms: somatic concern, anxiety, emotional withdrawal, conceptual disorganisation, guilt feelings, tension, mannerisms and posturing, grandiosity, depressive mood, hostility, suspiciousness, hallucinatory behaviour, motor retardation, uncooperativeness, unusual thought content and blunted affect. Each symptom was rated on a severity scale ranging from 1 (not present) to 7 (extremely severe). The BPRS was assessed by daytime and evening nursing staff and by the unit's social worker

The CMAI consists of 29 agitated behaviours, each rated on a 7-point scale of frequency (1 indicates the participant never engages in that specific behaviour, and 7 indicates the participant manifests that behaviour an average of several times an hour). The period rated was the week before administration of the CMAI, and three subtypes of agitation were used: aggressive behaviour, physically non-aggressive behaviour and verbally agitated behaviour. The CMAI was independently completed by daytime and evening shift nursing staff members most familiar with the resident, and by social workers

Secondary outcomes were functioning, adverse effects and global impression scales

NotesCopy of our mail to Prof. Dr. Cohen-Mansfield and her answer:

"Dear Dr. Declercq,


I am quite sure they all had dementia, but I do not remember whether it was an inclusion criterion.  The MMSE scores also fit with a dementia diagnosis.


Jiska Cohen-Manfield, PhD, ABPP

From: Tom Declercq []
Sent: Tuesday, April 21, 2009 9:34 AM
To: Jiska Cohen-Mansfield
Subject: Cochrane review on withdrawal of antipsychotics

Dear Colleague, Professor,


 As first author of an ongoing Cochrane review on withdrawal of antipsychotics, I am very interested in the RCT you conducted on withdrawal of haloperidol, thioridazine and lorazepam in the nursing home (Arch Intern Med 1999). Can you please let me know if the selected patients, residents of nursing homes, had dementia as an inclusion criterion?


Thank you




Tom Declercq MD

Department of Primary Care

Ghent University


Risk of bias

BiasAuthors' judgementSupport for judgement

Random sequence generation (selection bias)Unclear risk"......half the residents were randomly assigned to have their medication dose tapered during a 3-week period, followed by receipt of a placebo (the other half continued their usual medication dosage)"

"Residents were randomly assigned to the placebo versus medication group and stratified both by level of cognitive function and by psychotropic medication"

Method of sequence generation is not described

Allocation concealment (selection bias)Unclear riskNot described

Blinding (performance bias and detection bias)
All outcomes
Low risk"Study medications (usual medication and placebo) were administered as identical liquids to ensure blindness by the care team. Only the dispensing pharmacist, who was not an employee of the nursing home, knew which medication was administered. The care team, residents, family caregivers, and research team were blinded to which group a participant was assigned"

Incomplete outcome data (attrition bias)
NPI (Neuro Psychiatric Inventory) and DCM (Dementia Care Mapping)
Low risk"Twenty-three participants discontinued participation in the study before completion for the following reasons: death or dying (3), hospitalization (1), not eating or weight loss (3), increased agitation (9), lethargy (2), withdrawal of consent (4), facial asymmetry (1) and fall (3); some had multiple reasons. For 12 participants, discontinuation occurred during the original drug dosage, for 9 while taking placebo, and for 2 during titration from drug to placebo. Most discontinuations (20 of 23) occurred in the first part of the study, before the crossover stage"

"Participants who discontinued the study were similar in demographic characteristics to those who stayed. Although their levels of agitation at baseline were higher than those who stayed in the study, these differences did not reach statistical significance"

Selective reporting (reporting bias)Low riskAll outcome data reported

Other biasUnclear riskSeveral different analyses were used to assess the robustness of the result. However, it is not clear if an intention-to-treat analysis was used.

By not making difference in outcome reporting between discontinuation of antipsychotics, namely haloperidol and thioridazine, versus discontinuation of lorazepam, a benzodiazepine, it is impossible to retain robust conclusions from this withdrawal study

Devanand 2011

MethodsA 6-month, randomised, double-blind, placebo-controlled discontinuation trial (phase B) following response to haloperidol open treatment during 20 weeks (phase A)

Participants78 outpatients aged 50-95 years with AD and psychosis or agitation were assessed for eligibility in phase A of the trial

Exclusion criteria were: acute unstable medical condition, delirium, alcohol or substance abuse or dependence during the prior year, clinical evidence of stroke, other dementias including vascular or Lewy body or frontotemporal dementia, multiple sclerosis, Parkinson's disease, tardive dyskinesia, diagnosis of a psychotic disorder antedating the onset of dementia, antipsychotic medication usage during the 4 weeks prior to study entry, and contraindication to the use of haloperidol

44 participants were included in phase A

22 responders of phase A were eligible for randomisation in phase B (21 entered Phase B and 20 had at least one follow-up visit)

InterventionsPhase A: open treatment (20 weeks): 44 outpatients with AD and psychosis or agitation receiving psychotropic medication had a 1-week washout prior to entering phase A. During phase A flexible doses of haloperidol 0.5-5 mg daily were individually titrated to maximise therapeutic response and minimise side effects, especially extrapyramidal side effects. Visits occurred at 0, 2, 4, 8, 12, 16 and 20 weeks

Phase B: discontinuation trial (24 weeks): 20 phase A responders were double-blind randomised to a continuation versus placebo (i.e. discontinuation) group. For people randomised to placebo, there was a 2-week double-blind sequential placebo substitution tapering period to placebo. Time points of assessment during phase B were 0 (same as end of phase A), 2, 4, 8, 12, 16, 20 and 24 weeks (i.e. 6 months)

OutcomesPhase A: the 3 most prominent target of psychosis or agitation were identified, scored on a 7-point scale (0 = absent to 6 = extreme) and tracked during the study. Criteria for response (primary outcome) were minimum 50% reduction from baseline in the sum score of these 3 target symptoms, a sum score ≤ 6 on these 3 items (range 0-18), and minimal or greater improvement on the CGI-C score (rated only for symptoms of psychosis or agitation)

Phase B: relapse (primary outcome) was assessed at any single time point during phase B. Criteria for relapse were minimum 50% worsening from the sum score of the 3 target symptoms at the end of phase A, a sum score ≥ 6 on these 3 items (range 0-18), and minimal or greater worsening on the CGI-C score (rated for psychosis and agitation)

Secondary outcomes were: somatic side effects assessed by the TESS, extrapyramidal signs assessed by the UPDRS, and tardive dyskinesia assessed by the Rockland TD scale. Cognition was assessed by the MMSE and impairment in ADL was assessed by the modified BFAS at baseline, end-phase A and end-phase B


Risk of bias

BiasAuthors' judgementSupport for judgement

Random sequence generation (selection bias)Unclear riskNot described

Allocation concealment (selection bias)Unclear risk"Responders by end-Phase A were eligible for Phase B, a 24-week, random assignment (1:1 assignment of haloperidol and placebo), double-blind, trial of continuation haloperidol (same dose as end-Phase A) versus switch to placebo"

No description of the blinding of random allocation

Blinding (performance bias and detection bias)
All outcomes
Low riskParticipant and carer blinding: "Haloperidol and placebo were made up in identical looking opaque white capsules"

Outcome assessors seem to be adequately blinded: "The blind is maintained after study exit to avoid biasing raters. A code-break is authorized only if needed in cases of overdose or medical emergency"

Incomplete outcome data (attrition bias)
NPI (Neuro Psychiatric Inventory) and DCM (Dementia Care Mapping)
Low riskAttrition at end of phase A fully accounted for: "There were 15 Phase A non-completers (34%), with all early terminations attributed either to lack of efficacy (n=9) or side effects (n=6)"

Attrition at end of phase B accounted for and ITT included: "Twenty of the 21 patients randomized in Phase B to continuation haloperidol or placebo had at least one follow-up visit after randomization and were included in the Phase B analysis. Among patients who did not relapse, reasons for early study termination prior to 24 weeks in Phase B were side effects (n=2), moving out of the area (n=1), medical illness (n=1) and noncompliance (n=1). All data from these patients were included in the intent-to-treat, last observation carried forward, analyses"

Selective reporting (reporting bias)Unclear riskThis pilot study report mentions the following primary outcome: "Criteria for relapse were minimum 50% worsening from the sum score of three target symptoms at end Phase A, a sum score 6 on these three items (range 0–18), and minimal or greater worsening on the CGI-C score (rated for psychosis/agitation)"

The protocol (Devanand 2012a) for the subsequent study (Devanand ADAD-Trial) mentions NPI as primary outcome measure: "Primary Efficacy Measure is NPI sum score of delusions, hallucinations, and agitation/aggression, which are three of 12 domains in the NPI". It is not clear why the primary outcome definition was changed for the subsequent study

Other biasUnclear riskThe discontinuation trial only included people who responded to haloperidol. Non-responders after first 12 weeks (phase A) were excluded from the discontinuation phase (B)

Study supported by NIH grant but authors have financial links with several pharmaceutical companies

Devanand ADAD 2012

Methods6-month, randomised, double-blind, placebo-controlled discontinuation trial (phase B) following response to flexible dose risperidone open treatment for 16 weeks (phase A)

Participants"Patients were recruited from memory clinics (including Alzheimer’s research centers), geriatric psychiatry clinics, and clinics at Veterans Affairs medical centers, as well as through physician referrals and advertising. Patients were eligible for participation in the study if they were outpatients or residents of assisted-living facilities or nursing homes, were 50 to 95 years of age, and met the criteria for dementia of the Diagnostic and Statistical Manual of Mental Disorders, 4th edition (DSM-IV), and the criteria for probable Alzheimer’s disease of the National Institute of Neurological and Communicative Disorders and Stroke–Alzheimer’s Disease and Related Disorders Association"

"In addition, eligible patients had a score on the Neuropsychiatric Inventory (NPI) of 4 or more at both screening and baseline on the delusions or hallucinations subscale (psychosis score) or the agitation–aggression subscale (agitation score) (with scores on all NPI subscales ranging from 0 to 12 and higher scores indicating more pronounced symptoms) and a score of 5 to 26 on the Mini–Mental State Examination (MMSE, with scores ranging from 0 to 30 and higher scores indicating better cognition) in the case of outpatients and a score of 2 to 26 in the case of nursing home residents (with the lower range reflecting the greater severity of dementia in nursing homes)"

"Exclusion criteria were a history of stroke, transient ischemic attack, or uncontrolled atrial fibrillation"

253 people screened; 180 received risperidone; 110 randomised in phase B

Group 1 (continue risperidone 32 people at start, 13 received risperidone at wk 16, 10 completed 48 wks without relapse)

Group 2 (continue risperidone for 16 wks and then placebo, 38 people at start, 27 received placebo at wk 16, 14 completed 48 wks without relapse)

Group 3 (start placebo in phase B, 40 people, 13 received placebo at wk 16, 10 completed 48 wks without relapse)

Interventions"Patients with Alzheimer's disease and psychosis or agitation–aggression received open-label treatment with risperidone for 16 weeks. Those who had a response to risperidone therapy were then randomly assigned, in a double-blind fashion, to one of three regimens: continued risperidone therapy for 32 weeks (group 1), risperidone therapy for 16 weeks followed by placebo for 16 weeks (group 2), or placebo for 32 weeks (group 3)"

1-week washout period for psychotropic medication

"If washout was not feasible ... stable doses of selective serotonin-reuptake inhibitors or low-dose trazodone or of sedatives or hypnotic agents were permitted"

"Lorazepam, at a dose of 1 mg or less per day, was permitted if needed"

OutcomesPrimary outcomes: time to relapse of psychosis ("the primary end point was the time to relapse during weeks 0 to 16 of phase B") or agitation ("the time to relapse during weeks 17 to 32 of phase B")

"In phase A, patients were considered to have had a response if they had a reduction of 30% or more from baseline on the NPI core score (the sum of the subscale scores for agitation–aggression, hallucinations, and delusions) and a score of 1 (very much improved) or 2 (much improved) on the Clinical Global Impression of Change (CGI-C) scale (which ranges from 1 to 7, with higher scores indicating less improvement) for overall psychosis or agitation. In phase B, patients were considered to have had a relapse if they had an increase in the NPI core score of 30% or more, or a 5-point increase from the score at the end of phase A, and a score of 6 (much worse) or 7 (very much worse) on the CGI-C. At any phase B study visit, if the criteria for relapse were met on the basis of scores on the NPI and CGI-C during the preceding 2 weeks, end-of-study procedures were completed, and the patient exited the study to receive open label treatment"

Secondary outcomes: "...assessments of extrapyramidal signs, with the use of the Simpson–Angus scale (which ranges from 0 to 40, with higher scores indicating more extrapyramidal signs); tardive dyskinesia, with the use of the Abnormal Involuntary Movement Scale (AIMS; which ranges from 0 to 35, with higher scores indicating more severe symptoms); general somatic symptoms developing during treatment, as assessed with the use of the Treatment Emergent Symptoms Scale (TESS; which ranges from 0 to 26, with higher scores indicating more somatic symptoms); cognitive status, as assessed with the use of the MMSE and the Alzheimer's Disease Assessment Scale (ADAS)–cognitive score (which ranges from 0 to 70, with higher scores indicating worse cognition); and physical function, as assessed with the use of the Physical Self-Maintenance Scale (PSMS; which ranges from 1 to 30, with higher scores indicating worse functioning)"

Adverse events

NotesOutcome symptoms slightly unevenly distributed in randomised groups in phase B: 9% agitation-aggression in group 1 (continue risperidone) vs 19% in group 2 (switch to placebo after 16 wks) and 18% in group 3 (placebo throughout phase B)

High rates of discontinuation of risperidone (38% in phase A; 68% in group 1 and 29% in group 2)

Risk of bias

BiasAuthors' judgementSupport for judgement

Random sequence generation (selection bias)Low risk"The study statistician prepared a randomized permuted-blocks procedure, with blocks of 3 or 6, to balance the group assignment in each of four (2 × 2) strata, with stratification within each site according to the presence or absence of psychosis at baseline and residence (assisted-living facility or nursing home vs. home)"

Allocation concealment (selection bias)Low risk"Patients who had a response entered phase B of the study and were randomly assigned..."

"The central pharmacy of the New York State Psychiatric Institute maintained the assignment code, and clinicians and raters remained unaware of the group assignments of all patients during the entire study"

Blinding (performance bias and detection bias)
All outcomes
Low risk"...double-blind fashion..."

"...all tablets identical in appearance"

"Immediately before the end of phase A, the pharmacy dispensed prepackaged blister packs of risperidone or placebo tablets that were identical in appearance for patients eligible for randomization in phase B. The number of tablets the patient was receiving daily at the end of phase A was the number he or she received throughout phase B"

Incomplete outcome data (attrition bias)
NPI (Neuro Psychiatric Inventory) and DCM (Dementia Care Mapping)
Low riskAll randomised participants accounted for in the flowchart

"The dropout rates did not differ significantly among the randomized groups (Fig. 1)"

Selective reporting (reporting bias)Low riskAll outcomes reported

Other biasUnclear riskFunding sources:

"...Johnson & Johnson, donated the risperidone tablets and matching placebo but had no role in the conduct of the study or the analysis or reporting of the data"

"Supported by grants from the National Institutes of Health (R01 AG021488 and R01 AG17761) and the Department of Veterans Affairs"

Conflict of interest: not reported in the paper, but included in the Supplementary Appendix available online. The first author received grants from several pharmaceutical companies (inside and outside the submitted work)

Findlay 1989

MethodsRandomised double-blind placebo-controlled trial

Participants36 older people (aged ≥ 65 years) female inpatients in a long-stay psychogeriatric ward in the UK with senile dementia, Alzheimer type, receiving a stable dose of between 10 and 100 mg of thioridazine per day for at least 2 months. Males, multi-infarct dementia and antipsychotic agents other than thioridazine were excluded

InterventionsWithdrawal of thioridazine by tapering to half of the daily dose in the first week and to placebo over the next week; after 4 weeks, all participants were restored to half their original dose of thioridazine with any subsequent alterations being made by their regular medical attendant on an empirical base

OutcomesPrimary outcomes are cognitive function measured by the CAS and cognitive and behavioural dysfunction as measured by the SCAGS and the LPRS

Secondary outcomes are systolic BP and heart rate (HR)


Risk of bias

BiasAuthors' judgementSupport for judgement

Random sequence generation (selection bias)Unclear risk"...matching active and placebo (liquid) formulations of thioridazine were used, each subject being entered separately and allocated by a random code to the active or placebo group in a double-blind manner"

Method of sequence generation not described

Allocation concealment (selection bias)Unclear risk"...allocated by a random code to the active or placebo group in a double-blind manner"

How blinding was ensured is not described. The randomisation process was not completely successful

"The starting difference in Cognitive Assessment Scale scores between active-continued and placebo-substituted groups represents an artefact of the randomization process"

Blinding (performance bias and detection bias)
All outcomes
Low risk"During the first week patients in the "placebo" group received placebo substitution for half of their daily dose of thioridazine and over the next week a total substitution. Similar mock substitutions with thioridazine were given to the "active" group, so that initial medication was continued but the trial remained double-blind. After four weeks all patients were restored to half their original dose of thioridazine with any subsequent alterations being made by their regular medical attendant on an empirical basis"

Incomplete outcome data (attrition bias)
NPI (Neuro Psychiatric Inventory) and DCM (Dementia Care Mapping)
Unclear riskNo information

Selective reporting (reporting bias)Unclear riskPrimary outcome not described; it is unclear if a selection of measured outcomes was reported

Other biasUnclear riskThe randomisation procedure unfortunately resulted in a baseline imbalance in 1 of the 3 cognitive /behavioural rating scales, i.e. the CAS. It is unclear if this has had an impact on outcomes

It is not clear how participants were recruited

Ruths The BEDNURS Study

MethodsRandomised, placebo-controlled, double-blinded multicentre 4-week study

Participants55 nursing home residents with dementia in Norway taking haloperidol, risperidone or olanzapine for nonpsychotic symptoms for at least 3 months before the study as standing medication in stable doses; older people, i.e. aged ≥ 65 years; dementia diagnosis according to the clinical criteria of International Classification of Diseases, 10th Revision; and residence in the facility for at least 3 months before inclusion were included. Participants with antipsychotic use for a primary diagnosis of major psychotic disorder, mental retardation, terminal illness with life expectancy judged to be shorter than 3 months and recent major changes in health status were excluded

InterventionsAbrupt discontinuation of antipsychotic medication

OutcomesBehavioural and Psychological Symptoms measured by the NPI-Q

The NPI-Q covers 12 symptoms: delusions, hallucinations, agitation/aggression, depression, anxiety, euphoria, apathy, disinhibition, irritability, aberrant motor behaviour (restlessness, e.g. purposeless wandering and inappropriate activity), sleep problems and eating disorders. Information on participants' symptoms was obtained by interview with the primary nurse informant. Individual symptoms were scored as 0 (absent), 1 (mild), 2 (moderate), or 3 (severe), providing an NPI-Q sum score rating from 0 to 36. 3 separate ratings were conducted for all the participants: in the second baseline week, in week 1 and in week 4 of the intervention period. These ratings included symptoms occurring during the 7-day period before assessment

Sleep/wake activity was recorded continuously during baseline and intervention (i.e. over 6 weeks) using an Actiwatch portable recorder. The Actiwatch is a small wrist-worn device containing an accelerometer that is optimised for highly effective sleep-week inference from wrist activity. Actigraphically measured wrist activity is a feasible and reliable method for sleep/wake evaluation in nursing home residents. The following actigraphic parameters were calculated: total sleep time, total wake time, sleep efficiency (proportion of sleep during night window, i.e. 11 p.m. to 7 a.m.), daytime activity and night-time activity. The ratio of day-to-night-time activity was calculated and expressed as a light/dark ratio. Mean 24-hours activity and peak times of activity were calculated. Analyses of sleep/wake activity were based on 3 x 7-day records comprising the second baseline week and week 1 and week 4 of the intervention, corresponding with the timing for behavioural ratings


Risk of bias

BiasAuthors' judgementSupport for judgement

Random sequence generation (selection bias)Low riskParticipants were assigned to antipsychotic drug discontinuation (intervention group) or no discontinuation (reference group) by means of computer generated, random, permuted blocks of 4

Allocation concealment (selection bias)Low risk"All study medications were provided by an independent pharmacy to maintain blindness, and dispensed at the same dose frequency as baseline treatment"

Blinding (performance bias and detection bias)
All outcomes
Low risk"In the intervention group, patients received inert placebo capsules consisting of lactose, whereas reference group patients received identically looking capsules containing continued antipsychotic drug treatment at current dose"

"Sealed envelopes, containing details of study medication for each patient, were available for the nursing home physicians in case of serious health events"

It is not clear how outcome assessment was blinded

Incomplete outcome data (attrition bias)
NPI (Neuro Psychiatric Inventory) and DCM (Dementia Care Mapping)
Unclear riskAll 55 participants were included in the analysis

"Seven patients completed the study prematurely, due to unblinding for randomisation code, behavioural deterioration, restless legs or delirium"

Selective reporting (reporting bias)Low riskAll intended outcomes reported

Other biasUnclear riskIt is not clear if the participating nursing homes were different from non-participating

van Reekum 2002

MethodsRandomised, double-blind, placebo-controlled trial

Participants34 residents in nursing homes and from geriatric chronic care floors of an academic health science centre in Canada, having any form of dementia, receiving antipsychotics for 6 months or longer. The following people were excluded: people with a history of antipsychotic discontinuation having failed within the past 6 months, a history of schizophrenia, antipsychotic use for nausea, diagnosis of delirium (DSM-IV criteria), a global rating scale of 3 on the BEHAVE-AD rating scale at the time of the screening, 1 week prior to the start of the study or within the 2 weeks of the pre-trial period

InterventionsWithdrawal by tapering: dose reduction with original medication halved for the first week of the dose reduction period and the remaining dose halved for the second week

OutcomesBehavioural outcome measures included the BEHAVE-AD, the NPI and the ROAS

Cognitive functioning was assessed using the MMSE and the MDRS. The MMSE was completed on a monthly basis and the MDRS was completed at baseline and at the end of the study

Functional level was assessed using the BDS, recording ADL and motivational behaviour sub-scale

Extrapyramidal signs were assessed using the ESRS. The functional and extrapyramidal sign measures were completed on the same schedule as the MMSE

A CGI was also completed for all visits. The CGI quantified the clinical impression of severity of behavioural disturbance on a 7-point, verbally anchored scale. The degree of change from baseline was also ranked on a similar scale

All outcome data were obtained by a trained research assistant upon interview of the prime nurse or the subject as appropriate for the instrument


Risk of bias

BiasAuthors' judgementSupport for judgement

Random sequence generation (selection bias)Low risk"A random number table was used to allocate subjects to receive either continued antipsychotic treatment at the current dose or to receive identical placebo"

Allocation concealment (selection bias)Unclear riskNot described

Blinding (performance bias and detection bias)
All outcomes
Low risk"During all study periods, medications, including placebo, were placed into identical capsules to maintain blindness"

Incomplete outcome data (attrition bias)
NPI (Neuro Psychiatric Inventory) and DCM (Dementia Care Mapping)
Low risk"The total number of subjects who were withdrawn from the study early was 16, with 10 in the placebo group and 6 in the active treatment group The difference in the rate of early study withdrawal was not statistically significant. Subjects were withdrawn due to medical illness (3), death (3), extrapyramidal symptoms (3), and exacerbations in of behavioural problems (7, of 4 in the placebo and 3 in the active group)"

Selective reporting (reporting bias)High riskNPS assessed by NPI, aggression assessed by the ROAS, extrapyramidal signs assessed by the ESRS, cognitive functioning assessed by MMSE and functional outcome assessed by the BDS were not reported in data

Other biasUnclear riskThis study has less potential for referral bias than the Bridges-Parlet (1997) and Cohen-Mansfield (1999) study

Characteristics of excluded studies [ordered by study ID]

StudyReason for exclusion

Horwitz 1995This publication is not a clinical trial, but a naturalistic study

McLennan 1992This trial reports a primary outcome (prolactin response to withdrawal of thioridazine assessed in the Findlay 1989 cohort), which has no relationship with NPS we are interested in

Rule 2003This reference probably refers to the registration of a discontinuation study which has not been published until now (no more information found on this reference)

Wessels 2010This trial was not a discontinuation trial

Westbury 2011This trial was a follow-up study and not a discontinuation trial

Characteristics of ongoing studies [ordered by study ID]

Trial name or titleDementia Antipsychotics and Antidepressants Discontinuation Study DESEP

MethodsRandomised double-blind placebo-controlled trial

ParticipantsNursing homes residents with Alzheimer dementia or vascular dementia given antipsychotics or antidepressants for 3 months or more

InterventionsRisperidone, escitalopram, citalopram, sertraline, paroxetine

OutcomesChanges in Neuropsychiatric Inventory, changes in Cornell's Depression Scale, changes in Unified Parkinson's Disease Rating Scale sub-scale

Starting dateAugust 2008



Comparison 1. Discontinuation versus continuation of antipsychotic medication: continuous data, analysis method mean difference

Outcome or subgroup titleNo. of studiesNo. of participantsStatistical methodEffect size

 1 Behavioural assessment by using NPI measuring NPS at 3 months (Ballard 2004 and Ballard DART-AD) (forest plot 1)2194Mean Difference (IV, Fixed, 95% CI)-1.49 [-5.39, 2.40]

Table 1. Antipsychotic drug classes

PHENOTHIAZINES with aliphatic side chain

PHENOTHIAZINES with piperazine structure

FENOTHIAZINES with piperidine structure


INDOLE derivatives

THIOXANTHENE derivatives





Table 2. Antipsychotic drugs with defined daily doses

PHENOTHIAZINES with aliphatic side-chain

N05AA01 Chlorpromazine 0.3 g per os

N05AA02 Levomepromazine 0.3 g per os

N05AA03 Promazine 0.3 g per os

N05AA04 Acepromazine 0.1 g per os

N05AA05 Triflupromazine 0.1 g per os

N05AA06 Cyamemazine

N05AA07 Chlorproethazine

PHENOTHIAZINES with piperazine structure

N05AB01 Dixyrazine 50 mg per os

N05AB02 Fluphenazine 10 mg per os

N05AB03 Perphenazine 30 mg per os

N05AB04 Prochlorperazine 0.1 g per os

N05AB05 Thiopropazate 60 mg per os

N05AB06 Trifluoperazine 20 mg per os

N05AB07 Acetophenazine 50 mg per os

N05AB08 Thioproperazine 20 mg per os

N05AB09 Butaperazine 10 mg per os

N05AB10 Perazine 0.1 g per os

N05AB20 Homophenazine

FENOTHIAZINES with piperidine structure

N05AC01 Periciazine 50 mg per os

N05AC02 Thioridazine 0.3 g per os

N05AC03 Mesoridazine 0.2 g per os

N05AC04 Pipotiazine 10 mg per os


N05AD01 Haloperidol 8 mg per os

N05AD02 Trifluperidol 2 mg per os

N05AD03 Melperone* 0.3 g per os

N05AD04 Moperon 20 mg per os

N05AD05 Pipamperone 0.2 g per os

N05AD06 Bromperidol 10 mg per os

N05AD07 Benperidol 1.5 mg per os

N05AD08 Droperidol

N05AD09 Fluanisone

N05AE INDOLE derivatives

N05AE01 Oxypertine 0.12 g per os

N05AE02 Molindone 50 mg per os

N05AE03 Sertindole* 16 mg per os

N05AE04 Ziprasidone* 80 mg per os

THIOXANTHENE derivatives

N05AF01 Flupentixol 6 mg per os

N05AF02 Clopenthixol 0.1 g per os

N05AF03 Chlorprothixene 0.3 g per os

N05AF04 Tiotixene 30 mg per os

N05AF05 Zuclopenthixol 30 mg per os


N05AG01 Fluspirilene

N05AG02 Pimozide 4 mg per os

N05AG03 Penfluridol 6 mg per os


N05AH01 Loxapine 0.1 g per os

N05AH02 Clozapine* 0.3 g per os

N05AH03 Olanzapine* 10 mg per os

N05AH04 Quetiapine* 0.4 g per os


N05AL01 Sulpiride 0.8 g per os

N05AL02 Sultopride 1.2 g per os

N05AL03 Tiapride 0.4 g per os

N05AL04 Remoxipride 0.3 g per os

N05AL05 Amisulpride* 0.4 g per os

N05AL06 Veralipride

N05AL07 Levosulpiride 0.4 g per os


N05AX07 Prothipendyl 0.24 g per os

N05AX08 Risperidone* 5 mg per os

N05AX09 Clotiapine 80 mg per os

N05AX10 Mosapramine*

N05AX11 Zotepine* 0.2 g per os

N05AX12 Aripiprazole* 15 mg per os

N05AX13 Paliperidone*

*atypical antipsychotics

 * Atypical antipsychotic agents.