Key points from the evidence

The content of this evidence summary was up-to-date in March 2015. See summaries of product characteristics (SPCs), British national formulary (BNF) or the MHRA or NICE websites for up-to-date information.


Evidence from randomised controlled trials (RCTs) suggests that valproate preparations (including sodium valproate and valproate semisodium) are no more effective than placebo for treating agitation or behavioural disturbances in people with dementia. Adverse effects such as falls, sedation, gait disturbances, tremor, muscular weakness, thrombocytopenia, gastrointestinal disorders and urinary tract infections were more common in people taking valproate preparations than placebo.

Regulatory status: off‑label. This topic was prioritised because there is uncertainty about the balance of risks and benefits of valproate preparations for this indication.


  • No statistically significant difference between valproate preparations and placebo after 6 weeks' treatment for change from baseline on the Cohen Mansfield Agitation Inventory score in people with dementia (meta‑analysis of 3 RCTs, n=216).

  • No statistically significant difference between valproate semisodium and placebo after 24 months' treatment in time to development of clinically significant agitation or psychosis in people with Alzheimer's disease (1 RCT; n=313).


  • Nausea and tremor are very common adverse effects (occurring in 10% or more people) of valproate preparations (Epilim summary of product characteristics).

  • Meta‑analysis of 4 RCTs (n=394) showed about double the rate of any adverse at 6 weeks among participants taking valproate preparations compared with those taking placebo.

Patient factors

  • Dropout rates in the largest RCT were high (around 61% in both groups). Dropout rates due to adverse effects were 16.3% in the valproate semisodium group and 7.5% in the placebo group (1 RCT; n=313).

Resource implications

  • The dosages of sodium valproate or valproate semisodium used in the studies varied.

  • Prices of valproate vary according to the preparation. For example, 100 sodium valproate 500 mg tablets cost £8.52. Valproate semisodium 500 mg gastro‑resistant tablets (Depakote) cost £29.15 for 90.

Introduction and current guidance

According to the NICE guideline on dementia and a best practice guide produced by the Alzheimer's Society and endorsed by the Department of Health, non‑pharmacological interventions tailored to the individual person's preferences, skills and abilities are recommended first line. NICE advises that people with dementia who develop non‑cognitive symptoms or behaviour that challenges should be offered a pharmacological intervention in the first instance only if they are severely distressed or there is an immediate risk of harm to the person or others. Treatment with an antipsychotic drug may be offered after various conditions have been met. An acetylcholinesterase inhibitor or memantine may be offered in some circumstances.

Other drugs have been used off‑label for non‑cognitive symptoms of dementia; however, evidence to support their use is limited. The NICE full guideline on dementia concluded that there was insufficient evidence to support the use of anticonvulsant mood stabilisers, such as sodium valproate, valproate semisodium or carbamazepine, for the treatment of depression or anxiety in people with dementia. This evidence summary reviews the best available evidence for the use of valproate preparations for managing aggression, agitation and behavioural disturbances in dementia. It includes a Cochrane review which included the 3 studies on valproate preparations that were considered by NICE (Porsteinsson et al. 2001, Sival et al. 2002 and Tariot et al. 2001) and 2 additional studies. Another evidence summary considers carbamazepine for these indications.

Full text of introduction and current guidance.

Product overview

Sodium valproate is available in a variety of standard‑release oral preparations which are licensed for the treatment of epilepsy. It is also available in a variety of prolonged and modified‑release preparations; licensed indications for these vary according to the preparation (see individual summaries of product characteristics for details). Valproate semisodium (also known as divalproex sodium; Depakote) is licensed for the treatment of manic episodes in bipolar disorder when lithium is contraindicated or not tolerated.

None of the valproate preparations are licensed for the management of aggression, agitation and behavioural disturbances in dementia, therefore using any of them for this indication is off‑label.

In January 2015, the Medicines and Healthcare Products Regulatory Agency (MHRA) issued new information and strengthened warnings relating to the risk of abnormal pregnancy outcomes associated with valproate preparations.

Full text of product overview.

Evidence review

  • This evidence summary is based on a Cochrane review (Lonergan et al. 2009) and 1 RCT which was published after the Cochrane review (Tariot et al. 2011). An observational study (Meinhold et al. 2005) is also discussed.

  • The Cochrane review included 5 RCTs which compared valproate semisodium or sodium valproate with placebo in people with dementia and agitation. Three of these studies were included in a meta‑analysis of efficacy (n=216). There was no statistically significant difference between valproate preparations and placebo for change from baseline in the Cohen Mansfield Agitation Inventory (CMAI) score after 6 weeks' treatment (mean difference −2.20; 95% confidence interval [CI] −6.38 to 1.99). The CMAI score examines 29 types of agitated behaviour, including pacing, verbal or physical aggression, screaming, and restlessness. The Cochrane review reported that the results from the other 2 studies could not be interpreted due to methodological problems. Overall, the Cochrane review concluded that valproate preparations are ineffective for treating agitation in people with dementia and are associated with an unacceptable rate of adverse events. Based on current evidence they could not recommend valproate treatment for the management of agitation in people with dementia.

  • Tariot et al. (2011) was a double‑blind RCT (n=313) which investigated whether treatment with valproate semisodium could delay or prevent the development of psychiatric signs and symptoms in people with moderate Alzheimer's disease but no agitation or psychosis. After 24 months' treatment, there was no statistically significant difference between the valproate semisodium group and the placebo group for time to development of clinically significant agitation or psychosis (as defined by Neuropsychiatric Inventory (NPI) scores and clinical assessment) [hazard ratio 0.96; p=0.88]. However, the study may not have had sufficient power to detect a difference between the 2 groups.

  • A US observational study (Meinhold et al. 2005) aimed to assess the behavioural, mood and cognitive effects of valproate semisodium (as monotherapy or in combination with benzodiazepines or antipsychotics) in 450 nursing home residents with a history of behavioural problems associated with dementia. Behaviour and mood symptoms were measured using a minimum data set assessment. However, it is unclear if this is a valid and appropriate tool to measure behaviour and mood symptoms. Results after initiation of valproate treatment were inconsistent, with improvements reported for some measures of mood and behaviour but not others. The results of this observational study should be viewed with caution. Where improvements were reported, the differences appeared minimal and the clinical significance of the findings is unclear

  • In the Cochrane review (Lonergan et al. 2009) meta‑analysis of 4 of the included studies (n=394) showed a statistically significant increase in any adverse effects at 6 weeks among participants taking valproate preparations compared with those taking placebo (odds ratio 1.99; 95% CI 1.29 to 3.08). One of the studies included in this meta‑analysis was terminated early due to the high drop‑out rate in the valproate group compared with the placebo group. Nineteen people (22%) in the valproate group and 3 people (4%) in the placebo group dropped out due to adverse effects (p=0.001). The meta‑analysis also found statistically significant increases in adverse effects such as sedation, nausea, vomiting or diarrhoea, urinary tract infections and thrombocytopenia in participants taking valproate compared with placebo.

  • In the study by Tariot et al. (2011) there were statistically significant increases in sedation scores for participants receiving valproate semisodium compared with those receiving placebo (p=0.02). Some other adverse effects such as falls, gait disturbances, tremor, muscular weakness, depressed mood, diarrhoea and constipation occurred more frequently (in the range 17–39%) in the valproate group compared with the placebo group (in the range 8–32%).

Full text of evidence review.

Context and estimated impact for the NHS

The dosages of sodium valproate and valproate semisodium used in the studies varied. One of the studies used a low dose of sodium valproate (240 mg twice a day); whereas another study used a dose of sodium valproate titrated to 1500 mg per day. The studies which evaluated valproate semisodium used a dose titrated to around 800−1000 mg per day or 10−12 mg/kg bodyweight per day. Cost will depend on the preparation used and the dose.

Full text of context and estimated impact for the NHS.

Information for the public

A plain English summary is available on the NICE website. This sets out the main points from the evidence summary in non‑technical language and may be especially helpful for people or their carers' with behavioural disturbances associated with dementia who are thinking about trying valproate preparations.

About this evidence summary

'Evidence summaries: unlicensed or off‑label medicines' summarise the published evidence for selected unlicensed or off‑label medicines that are considered to be of significance to the NHS, where there are no clinically appropriate licensed alternatives. The summaries provide information for clinicians and patients to inform their decision‑making and support the construction and updating of local formularies.

The summaries support decision‑making on the use of an unlicensed or off‑label medicine for an individual patient, where there are good clinical reasons for its use, usually when there is no licensed medicine for the condition requiring treatment, or the licensed medicine is not appropriate for that individual.

The strengths and weaknesses of the relevant evidence are critically reviewed within this summary, but this summary is not NICE guidance.