Background

Neuroleptic drugs are controversial treatments in dementia, with evidence accumulating that they may hasten clinical decline. Despite these concerns, they are commonly prescribed for elderly and demented patients. Thioridazine, a phenothiazine neuroleptic, has been commonly prescribed because it was thought to produce relatively less frequent motor side effects. The drug has significant sedative effect, and it is thought that this is the main mechanism of action in calming and controlling the patient. However, pharmacologically, it also has marked anticholinergic properties that could potentially have a detrimental effect on cognitive function.

Objectives

To evaluate the efficacy of thioridazine in dementia in terms of: 
1) efficacy in controlling symptoms 
2) cognitive outcome for the patient 
3) safety

Search methods

The CDCIG Specialised register of trials was searched on 13 March 2009 using the terms 'thioridazine' and 'melleril'. This register contains up to date references from major health care databases like MEDLINE and EMBASE as well as records from trials databases in the field of dementia.

Selection criteria

Unconfounded, single‐blind or double‐blind, randomised trials were identified in which treatment with thioridazine was administered for more than one dose and compared to an alternative intervention in patients with dementia of any aetiology. Trials in which allocation to treatment or comparator were not truly random, or in which treatment allocation was not concealed, were reviewed but are not included in the data analysis.

Data collection and analysis

Data were extracted independently by the reviewers (VK, CAK and RJH). For continuous and ordinal variables, the main outcome measures of interest were the final assessment score and the change in score from baseline to the final assessment. The assessment scores were provided by behavioural rating scales, clinical global impression scales, functional assessment scales, psychometric test scores, and frequency and severity of adverse events. Data were pooled where appropriate or possible, and the Peto odds ratio (95%CI) or the weighted mean difference (95%CI) estimated. Where possible, intention to treat data were used.

Main results

The meta‐analysis showed that, compared with placebo, thioridazine reduced anxiety symptoms as evidenced by changes on the Hamilton Anxiety Scale. However, there was no significant effect on clinical global change, and a non‐significant trend for higher adverse effects with thioridazine.

Compared to diazepam, thioridazine was superior in terms of some anxiety symptoms, with similar adverse effects. Global clinical evaluation scales did not favour either treatment.

Compared to chlormethiazole, thioridazine was significantly inferior when assessed on some items of the CAPE and the Crichton Geriatric Behavioural Rating Scales. Thioridazine was also associated with significantly more dizziness.

No superiority for thioridazine was shown in comparisons with etoperidone, loxapine or zuclopenthixol, except to produce fewer side effects than loxapine.

Authors' conclusions

Very limited data are available to support the use of thioridazine in the treatment of dementia. If thioridazine were not currently in widespread clinical use, there would be inadequate evidence to support its introduction.

The only positive effect of thioridazine when compared to placebo is the reduction of anxiety. When compared to placebo, other neuroleptics, and other sedatives, it has equal or higher rates of adverse effects.

Clinicians should be aware that there is no evidence to support the use of thioridazine in dementia, and its use may expose patients to excess side effects.

Keywords: Humans, Antipsychotic Agents, Antipsychotic Agents/adverse effects, Antipsychotic Agents/therapeutic use, Dementia, Dementia/drug therapy, Thioridazine, Thioridazine/adverse effects, Thioridazine/therapeutic use

Criteria for considering studies for this review

Search methods for identification of studies

The trials were identified from a last updated search of the Specialized Register of the Cochrane Dementia and Cognitive Improvement Group on 13 March 2009 using the search terms: thioridazin* or melloril from January 2007 onward. This register contains records from the following major healthcare databasesThe Cochrane Library, MEDLINE, EMBASE, PsycINFO, CINAHL and LILACS, and many ongoing trial databases and other grey literature sources. For more detailed information on what the Group's Specialized Register contains see the Cochrane Dementia and Cognitive Improvement Group methods used in reviews.

The Cochrane Library, MEDLINE, EMBASE, PsycINFO, CINAHL, LILACS and a number of trial registers and grey literature sources were also searched separately on 13 March 2009 for records added to these databases after January 2007. The search strategies used to identify relevant controlled trials on dementia, Alzheimer's disease and mild cognitive impairment for the Group's Specialized Register can be found inAppendix 1.

In addition, Novartis, the pharmaceutical company that developed and markets thioridazine, was approached and asked to release any published or unpublished data that they had on file.

Data collection and analysis

SELECTION OF STUDIES 
A single reviewer (VK), eliminated citations considered as irrelevant on the basis of title abstract. In the presence of any suggestion that an article could possibly be relevant, it was retrieved for further assessment. Three reviewers (VK, RH and CK) independently selected the trials for inclusion in the review from the culled citation list. Disagreements were resolved by discussion.

QUALITY ASSESSMENT 
The same three reviewers (VK, RH and CK) assessed the methodological quality of each trial. The quality of the methodology of each selected trial was rated using the Cochrane Collaboration guidelines (Mulrow 1987).

Category A (adequate) is where the report describes allocation of treatment by: 
(i) some form of centralised randomised scheme, such as having to provide details of an enrolled participant to an office by phone to receive the treatment group allocation; 
(ii) some form of randomisation scheme controlled by a pharmacy; 
(iii) numbered or coded containers, such as in a pharmaceutical trial in which capsules from identical‐looking numbered bottles are administrated sequentially to enrolled participants; 
(iv) an on‐site or coded computer system, given that the allocations were in a locked, unreadable file that could be accessed only after inputting the characteristics of an enrolled participant; or 
(v) if assignment envelopes were used, the report should at least specify that they were sequentially numbered, sealed, opaque envelopes; 
(vi) other combinations of described elements of the process that provides assurance of adequate concealment.

Category B (intermediate) is where the report describes allocation of treatment by: 
(i) use of a "list" or "table" to allocate assignments; 
(ii) use of "envelopes" or "sealed envelopes"; 
(iii) stating the study as "randomised" without further detail.

Category C (inadequate) is where the report describes allocation of treatment by: 
(i) alternation; 
(ii) reference to case record numbers, dates of birth, day of week, or any other such approach; 
(iii) any allocation procedure that is entirely transparent before assignment, such as an open list of random numbers or assignments.

Empirical research has shown that lack of adequate allocation concealment is associated with bias. Trials with unclear concealment measures have been shown to yield more pronounced estimates of treatment effects than trials that have taken adequate measures to conceal allocation schedules, but less pronounced than inadequately concealed trials (Schultz 1995). Thus trials were included if they conformed to categories A or B, and those falling into category C were excluded.

The influence of the quality of randomisation on the results was assessed by pooling studies on these characteristics.

Other aspects of trial quality were not assessed by a scoring system although details were noted of blinding, whether intention‐to‐treat analyses were extractable from the published data, and the number of patients lost to follow‐up.

DATA EXTRACTION 
Data were independently extracted by the same two reviewers (RH and CK) and cross‐checked. Any discrepancies were discussed.

Data were sought on every patient with each outcome measure. To allow an intention‐to‐treat analysis, the data was sought irrespective of compliance, whether or not the patient was subsequently deemed ineligible, or otherwise excluded from treatment or follow‐up. Where dichotomous data were missing, the patients were assumed to have suffered the least favourable outcome. If continuous data were not available in the publications, data for an "on‐treatment" analysis are abstracted and indicated as such.

For continuous or ordinal variables, the main outcomes of interest were the final assessment score and the change in score from baseline (ie pre‐randomisation or at randomisation) to the final assessment. For binary outcomes such as dependency, institutionalisation and death, the endpoint itself is of clinical relevance. The baseline assessment score was the first available score, no longer than two months prior to the randomisation.

In studies where a cross‐over design was used, only data from the first treatment period were eligible for inclusion.

Data from studies including a titration period prior to the randomisation phase of the study were treated as follows. The data from these non‐randomised titration periods were not used to assess safety or efficacy since patients were usually not randomised, nor were treatment or dose allocations concealed.

DATA ANALYSIS 
For continuous or ordinal variables (such as psychometric test scores, clinical global impression scales, functional and quality of life scales) the main outcomes of interest were the final assessment score (corrected for baseline) and the change in score from baseline (i.e. pre‐randomisation or at randomisation) to the final assessment. If ordinal scale data appeared to be approximately normally distributed or if the analysis that the investigators performed suggested parametric tests were appropriate, then the outcome measures were treated as continuous data.

For binary outcomes such as institutionalisation, global impression and death, the endpoint itself was of interest and the Peto method of the 'typical odds ratio' was used.

A test for heterogeneity of the treatment effect between the trials was made using a standard chi‐squared statistic. If a test of heterogeneity was negative then a weighted estimate of the typical treatment effect across trials, the 'typical odds ratio' (i.e. the odds of an unfavourable outcome amongst treatment‐allocated patients to the corresponding odds amongst controls) was calculated using Peto's log‐rank test adapted for ordinal data (EBCTCG 1994). If, however, there was evidence of heterogeneity of the treatment effect between trials then either only homogeneous results were pooled, or a random effects model was used (in which case the confidence intervals was broader than those of a fixed effects model).

The null hypothesis of no effect was tested for the following comparisons:

1. thioridazine versus placebo 
2. thioridazine versus no treatment 
3. thioridazine versus alternative pharmacological intervention 
4. thioridazine versus behavioural intervention

Description of studies

Twelve trials were found that were suitable for inclusion in this review, ten were reported between 1973 and 1986, one in 1993 and one in 1996. Four of these had suitable designs but the information reported was too brief to allow analysis. The most common reason for excluding trials was because of an open label design. The eight trials that were analysed were similar in design. They were all double‐blind, randomised, controlled trials of 3 to 8 weeks duration. Two trials had no washout period and the rest had 1 to 2 week washout periods when no neuroleptic drugs were taken. 

Sample sizes ranged from 30 subjects to 610. The mean age of subjects was in the range 72.5 to 80.0 years and one trial only specified subjects were over 65 years of age. Six trials did not specify the type of dementia, one specified Alzheimer's Disease and one Vascular Dementia.

Three trials compared thioridazine with diazepam and one of these had a placebo group. Of the remaining trials one compared thioridazine with loxapine and placebo, one with loxapine, one with zuclopenthixol, one with etoperidone and one with chlormethiazole. The use of different controls constrained the pooling of results. In the trials comparing thioridazine to diazepam and thioridazine to loxapine some items could be pooled. Mostly data were inadequately reported making pooled analyses impossible. 

The dose of thioridazine in the different trials ranged from 10mg to 250mg, however, the given mean doses ranged from 32.9 to 95 mg. The mean dosages of the controls were diazepam 7.2 to 12 mg, chlormethiazole 723.2 mg, loxapine 10.5 to 17.5 mg, zuclopenthixol 6.8 mg and etoperidone 100 mg. (Table 1;Table 2)

The following outcome measures were recorded:

1. Behavioral Rating Scales: 
CGBRS ‐Crichton Geriatric Behavioural Rating Scale 
CAPE ‐ Clifton Assessment Procedures for the Elderly ‐ Behaviour Rating Scale 
NOSIE ‐ Nurses Observation Scale for Inpatient Evaluation 
mNOSIE ‐ Modified Nurses Observation Scale for Inpatient Evaluation 
BEHAVE‐AD ‐ Behavioural Pathology in Alzheimer's Disease Scale 
VIRS ‐ Video Interview Rating Scale

2. Clinical global assessment scales: 
BPRS ‐ Brief Psychiatric Rating Scale 
SCAG ‐ Sandoz Clinical Assessment‐Geriatric 
mHARS ‐ Modified Hamilton Anxiety Rating Scale 
CPRS ‐ Comprehensive Psychopathological Rating Scale 
GBSRS ‐ Gottfries, Brane and Steen Rating Scale 
CGE ‐ Clinical global evaluation of severity or change 
CGI ‐ Clinical Global Impression 
VTSRS ‐ Verdun Target Symptom Rating Scale 
PGRS ‐ Plutchik Geriatric Rating Scale

3. Cognitive Scales: 
CAPE ‐ Clifton Assessment Procedures for the Elderly ‐ Cognitive Assessment Scale

4. Adverse events and side effects: 
UKU ‐ Udvalget for Kliniske Undersogelser Side Effect Rating Scale 
CGE ‐ Clinical evaluation of drug tolerance

5. Functional performance

ADL ‐ Activities of Daily Living scale

No studies recorded the following outcomes with specific measures: 
1. Institutionalisation 
2. Death

Risk of bias in included studies

All eight included trials were double blind and controlled. None of them described randomisation procedures utilised. Two trials did not have a washout period which may be confounding because both trials were of a short duration i.e. 3 and 4 weeks. Only one trial specified a diagnostic system used to diagnose dementia. The results from the six trials that did not specify dementia type may be masking an effect of different responses in different forms of dementia. Four trials commented on ethics and consent. One study used video recordings to improve inter rater reliability. Three trials specified in which groups drop outs occurred, and intention to treat analyses could be performed. Three trials discussed drop outs but did not specify in which group they occurred and two trials did not discuss drop outs at all.

All the trials were of short duration, six lasted 3 to 4 weeks and two 8 weeks.

The data were generally reported in a way that made analysis difficult or impossible. The most common ways outcome measures were presented were as means or means of change without specifying standard deviations or exact P values. This meant much of the information could not be analysed further. Adverse reactions were poorly reported and usually descriptively.

Effects of interventions

Out of 52 studies reviewed, only 12 trials fulfilled the inclusion criteria of which 8 had sufficient information for further analysis. Only two trials compared thioridazine with placebo, and in both of these there was also another comparator drug. Studies were generally of short duration and included small numbers of subjects. The inadequacy of the published results leaves little compatible information that can be synthesised across trials. Very few usable data were extractable from the included studies, with the wide variety of outcome measures used constraining pooling of the data. A very small number of trials contributed to those results that were possible to pool. The maximum number of trials contributing to a comparison is two, and often results from only one trial are presented.

Outcomes were assessed in the domains of behaviour, clinical global change, psychometric change and adverse events. One study assessed functional ability and none assessed or used institutionalisation or death as an outcome.

The analysis showed that, compared with placebo, thioridazine reduced anxiety symptoms as evidenced by changes on the Hamilton Anxiety Scale for items anxious mood/tension/fear/insomnia (OR 4.91; 95% CI: 3.21 to 7.50) and items intellect/agitation/depressed mood/behaviour at interview (OR 3.64; 95% CI: 2.39 to 5.54). However, there was no significant effect on clinical global change (OR 1.58; 95% CI: 0.42 to 5.96), and a non‐significant trend for adverse effects with thioridazine (OR 0.41; 95% CI 0.09 to 1.86). No other placebo controlled data were analysable.

Compared with diazepam, thioridazine was significantly better in terms of anxiety symptoms only in one study that combined items intellect/agitation/depressed mood/behaviour at interview (OR 1.80; 95% CI: 1.04 to 3.10). The significant finding favouring thioridazine on the item fears is due to no events occurring in the control group and thus the confidence interval is infinity (OR 8.55; 95% CI: 1.39 to 52.52). For clinical global evaluation data were pooled for two studies and thioridazine was significantly better than diazepam only on the nurses' evaluation of severity (OR 2.87; 95% CI: 1.30 to 6.32). Rates of adverse effects were similar (OR 0.84; 95% CI: 0.25 to 2.82).

Compared with chlormethiazole, thioridazine fared significantly worse when assessed on the Crichton Geriatric Behavioural Rating Scales for items confusion/alertness (OR 0.31; 95% CI: 0.11 to 0.89) and continence (OR 0.23; 95% CI: 0.07 to 0.76). On the nocturnal confusion item of the Clifton Assessment Procedures for the Elderly thioridazine was also significantly worse than chlormethiazole (OR 0.24; 95% CI: 0.09 to 0.69). Thioridazine was also associated with significantly more dizziness (OR 0.18; 95% CI: 0.04 to 0.80).

No superiority for thioridazine was shown in comparisons with etoperidone, loxapine or zuclopenthixol on any measure, except to produce fewer adverse effects than loxapine (OR 2.81; 95% CI: 1.09 to 7.23).

Adverse events were not reported in any systematic manner in any of the trials. In only three trials were drop outs reported in such as way that an intention to treat analysis could be performed. No deaths were reported in any of the studies, but they were generally of short duration with no included follow‐up after the trial period. Most studies provided sparse or no information regarding ECG changes (Table 3) ; however, one study reported prolongation of the QT in one patient in the thioridazine group.

Implications for practice.

Implications for research.

Appendix 1. Search Report March 2009

Comparison 1. Thioridazine vs Diazepam.

Comparison 2. Thioridazine vs Placebo.

Comparison 3. Thioridazine vs Etoperidone.

Comparison 4. Thioridazine vs Loxapine.

Comparison 5. Thioridazine vs Chlormethiazole.

Comparison 6. Thioridazine vs Zuclopenthixol.

Characteristics of included studies [ordered by study ID]

Characteristics of excluded studies [ordered by study ID]

Internal sources

• East London and the City Mental Health NHS Trust, UK.

• St Bartholomew's Hospital Medical School, UK.

• Division of Neurosciences, Imperial College of Medicine, UK.

• Institute of Neurology, London, UK.

• The National Hospital for Neurology & Neurosurgery, London, UK.

External sources

• Alzheimer's Society (Research fellowship ‐ Richard Harvey), UK.

References to studies included in this review

References to studies excluded from this review

Additional references