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

People with dementia often have behavioural problems that can be difficult for carers to manage. Antipsychotic drugs are often prescribed to control symptoms and assist with controlling difficult behaviour. Many people with dementia continue to take these drugs over long periods of time. This review investigates whether withdrawal of long-term antipsychotic treatment is feasible in older people with dementia suffering from behavioural symptoms (often called neuropsychiatric symptoms or NPS). These include agitation, aggression, hallucinations, anxiety, apathy, depression, delusions (beliefs that cannot be true), wandering, repeating of words or sounds, and shouting. Nine studies with 606 participants provided data for the review. Most of the participants were residents in nursing homes, but some were outpatients. The studies differed considerably in participants, methods and outcomes so that is was not possible to combine most of the data numerically.

The evidence suggests that older nursing home residents or outpatients with dementia can be withdrawn from long-term antipsychotics without detrimental effects on their behaviour. Caution is required in older nursing home residents with more severe NPS, as two studies suggest these peoples' symptoms might be worse if their antipsychotic medication is withdrawn. Moreover, one study suggested that older people with dementia and psychosis or agitation and a good response to their antipsychotic treatment for several months could relapse after discontinuation of their antipsychotic medication. We do not know if there are beneficial effects of withdrawal on intellectual processes, quality of life or ability to carry out daily tasks, or if the risk of harmful events is reduced by drug withdrawal. One study suggests that older people with dementia who continue to take antipsychotics might die earlier.

We recommend that programmes that aim to withdraw older nursing home residents from long-term antipsychotics should be incorporated into routine clinical practice, especially if the NPS are not severe. More research is needed to identify people for whom withdrawal is not indicated and risk of relapse should be weighed against the risk of adverse events with long-term antipsychotic treatment.

Authors' conclusions: 

Our findings suggest that many older people with Alzheimer's dementia and NPS can be withdrawn from chronic antipsychotic medication without detrimental effects on their behaviour. It remains uncertain whether withdrawal is beneficial for cognition or psychomotor status, but the results of this review suggest that discontinuation programmes could be incorporated into routine practice. However, two studies of people whose agitation or psychosis had previously responded well to antipsychotic treatment found an increased risk of relapse or shorter time to relapse after discontinuation. Two other studies suggest that people with more severe NPS at baseline could benefit from continuing their antipsychotic medication. In these people, withdrawal might not be recommended.

Read the full abstract...

Antipsychotic agents are often used to treat neuropsychiatric symptoms (NPS) in dementia, although the literature is sceptical about their long-term use for this indication. Their effectiveness is limited and there is concern about adverse effects, including higher mortality with long-term use. When behavioural strategies have failed and drug therapy is instituted, regular attempts to withdraw these drugs are recommended. Physicians, nurses and families of older people with dementia are often reluctant to try to stop antipsychotics, fearing deterioration of NPS. Strategies to reduce antipsychotic use have been proposed, but a systematic review of interventions aimed at withdrawal of antipsychotic agents in older people with dementia has not yet been performed.


To evaluate whether withdrawal of antipsychotic agents is successful in older people with dementia in community or nursing home settings, to list the different strategies for withdrawal of antipsychotic agents in older people with dementia and NPS, and to measure the effects of withdrawal of antipsychotic agents on behaviour.

Search strategy: 

ALOIS, the Specialized Register of the Cochrane Dementia and Cognitive Improvement Group (CDCIG), The Cochrane Library, MEDLINE, EMBASE, PsycINFO, CINAHL, LILACS, clinical trials registries and grey literature sources were searched on 23 November 2012. The search included the following terms: antipsychotic* or neuroleptic* or phenothiazines or butyrophenones or risperidone or olanzapine or haloperidol or prothipendyl or methotrimeprazine or clopenthixol or flupenthixol or clothiapine or metylperon or droperidol or pipamperone or benperidol or bromperidol or fluspirilene or pimozide or penfluridol or sulpiride or veralipride or levosulpiride or sultopride or aripiprazole or clozapine or quetiapine or thioridazine combined with terms such as discontinu* or withdraw* or cessat* or reduce* or reducing or reduct* or taper* or stop*.

ALOIS contains records from all major healthcare databases (The Cochrane Library, MEDLINE, EMBASE, PsycINFO, CINAHL, LILACS), as well as from many clinical trials registries and grey literature sources.

Selection criteria: 

Randomised, placebo-controlled trials comparing an antipsychotic withdrawal strategy to continuation of antipsychotics in people with dementia.

Data collection and analysis: 

Review authors independently assessed trials for inclusion, rated their risk of bias and extracted data.

Main results: 

We included nine trials with 606 randomised participants. Seven trials were conducted in nursing homes, one trial in an outpatient setting and one in both settings. In these trials, different types of antipsychotics prescribed at different doses were withdrawn. Both abrupt and gradual withdrawal schedules were used. The risk of bias of the included studies was generally low regarding blinding and outcome reporting and unclear for randomisation procedures and recruitment of participants.

There was a wide variety of outcome measures. Our primary efficacy outcomes were success of withdrawal (i.e. remaining in study off antipsychotics) and NPS. Eight of nine trials reported no overall significant difference between groups on the primary outcomes, although in one pilot study of people with psychosis and agitation that had responded to haloperidol, time to relapse was significantly shorter in the discontinuation group (Chi2 = 4.1, P value = 0.04). The ninth trial included people with psychosis or agitation who had responded well to risperidone therapy for four to eight months and reported that discontinuation led to an increased risk of relapse, that is, increase in the Neuropsychiatric Inventory (NPI)-core score of 30% or greater (P value = 0.004, hazard ratio (HR) 1.94, 95% confidence interval (CI) 1.09 to 3.45 at four months). The only outcome that could be pooled was the full NPI-score, used in two studies. For this outcome there was no significant difference between people withdrawn from and those continuing on antipsychotics at three months (mean difference (MD) -1.49, 95% CI -5.39 to 2.40). These two studies reported subgroup analyses according to baseline NPI-score (14 or less versus > 14). In one study, those with milder symptoms at baseline were significantly less agitated at three months in the discontinuation group (NPI-agitation, Mann-Whitney U test z = 2.4, P value = 0.018). In both studies, there was evidence of significant behavioural deterioration in people with more severe baseline NPS who were withdrawn from antipsychotics (Chi2 = 6.8; P value = 0.009 for the marked symptom score in one study).

Individual studies did not report significant differences between groups on any other outcome except one trial that found a significant difference in a measure of verbal fluency, favouring discontinuation. Most trials lacked power to detect clinically important differences between groups.

Adverse events were not systematically assessed. In one trial there was a non-significant increase in mortality in people who continued antipsychotic treatment (5% to 8% greater than placebo, depending on the population analysed, measured at 12 months). This trend became significant three years after randomisation, but due to dropout and uncertainty about the use of antipsychotics in this follow-up period this result should be interpreted with caution.