CLINICAL QUESTIONS
What are the most clinical and cost effective and safe pharmacological interventions for the prevention of delirium in people in hospital?
What are the most clinical and cost effective and safe pharmacological interventions for the prevention of delirium in people in long-term care?
What are the most clinical and cost effective and safe pharmacological interventions for treating people with delirium in hospital?
What are the most clinical and cost effective and safe pharmacological interventions for treating people with delirium in long-term care?
14.1. Background
A wide variety of pharmacological interventions are available for the prevention and treatment of delirium. The drugs have varying pharmacological actions, and patients may potentially be troubled by a wide spectrum of adverse effects depending on the agent administered.
In making rational treatment choices, healthcare professionals need to carefully weigh up evidence on the anticipated benefits against that of any relevant concerns about the safety and tolerability. There are two important aspects in a review of adverse effects data for drugs in delirium:
Evaluation of comparative data among different drugs can help healthcare professionals arrive at a treatment decision for a particular agent based on whether the safety profile (nature and frequency of adverse effects) is more, or less, acceptable than the other available agents.
Healthcare professionals should be aware of the most important adverse effects that can arise after giving the therapy so that they can take appropriate measures to detect and minimize the risk from adverse effects
In most illnesses, patients are given adverse effects information to guide their choice of treatment and to enable them to seek medical attention for any untoward symptoms. However, patients receiving treatment for delirium may have little say in the matter, and have to rely on the actions of the healthcare professional. As such the onus is on the healthcare professional to make the appropriate decisions and to institute relevant monitoring and precautionary measures.
While some details on adverse effects have been covered in the parallel efficacy reviews of delirium, there is limited information on the specific adverse effects. It is also unclear whether the classes of drugs differ in their safety and tolerability profile.
14.1.1. Objective
To determine what specific adverse effects may arise from drug therapy for prevention or treatment of delirium.
14.2. Selection criteria
The selection criteria described in the general methodology section (section 2.3.1) were used, but some were specific to the evaluation of adverse effects and are reported in the following sections.
14.2.1. Types of studies
We did not apply any specific inclusion criteria based on study design; however, we aimed to exclude:
Published case reports and case series of specific adverse events, as there is a large degree of publication bias stemming from authors’ and editors’ decisions favouring manuscripts covering esoteric or interesting patients. Such cases are unlikely to be representative of the general patient population
Cross-over studies, as it is impossible to discriminate between events that arise as a complication of the first (previous) treatment, or as events resulting from the present therapy (carry-over).
Small studies with fewer than 20 patients exposed to the intervention of interest, as such studies are unlikely to be able to detect any important adverse effects, and may lead to falsely reassuring findings that no safety problems were identified.
14.2.2. Types of participants
Adults (18 years and over)
Patients requiring treatment for delirium or being given treatment to prevent delirium
Not end-of-life patients or patients with primarily psychiatric disorders such as schizophrenia, bipolar disorder or other psychoses.
Following GDG advice and post-hoc evidence from an indirect population was included in order to investigate stroke as an adverse event. The GDG extended the population to include older patients and those with dementia.
14.2.3. Interventions of interest
Typical antipsychotics: haloperidol
Atypical antipsychotics: risperidone, olanzapine, quetiapine, amisulpride
Benzodiazepines: diazepam, flunitrazepam
Cholinesterase inhibitors: donepezil, rivastigmine
5-HT3 antagonists: ondansetron
Duration of intervention: any
14.2.4. Comparators
For controlled studies, we accepted comparisons of any of the above agents versus placebo or no treatment. We also included studies that directly compared two or more agents from the above list of interventions. However, we excluded studies if the relevant intervention was tested against an active comparator that was not on the list of included drugs, as it would then be impossible to draw any valid conclusions on the relative safety profile of the agent of interest (safer or more harmful than an intervention of unknown effect).
14.2.5. Outcomes
All outcomes reported within the categories of ‘adverse effects, side effects, adverse events, complications, safety, or tolerability’.
14.2.6. Assessment of Validity of Adverse Effects Data
The methods for assessing validity were based on recommendations of chapter 14 of the Cochrane Handbook of Systematic Reviews. This focuses on two major factors:
In view of this, the following parameters were recorded:
What methods (if any) did the studies stipulate for the specific assessment of adverse effects?
Did the investigators prespecify any possible adverse events that they were particularly looking out for?
What categories of adverse effects were reported?
14.3. Identification of studies
Articles that had already been retrieved for the efficacy reviews were considered and reference lists were checked to identify specific articles on adverse effects.
A total of 170 full text articles were screened, with 16 studies fulfilling the inclusion criteria.
However, we had to make further exclusions due to no adverse effects data being extractable. Three eligible studies failed to mention anything about adverse effects and were not evaluated any further. (Hu 2006: olanzapine, haloperidol and control; Liu 2004: risperidone; Moretti 2004: rivastigmine).
Adverse effects data were extracted from 13 included papers (Aizawa 2002; Bayindir 2000; Breitbart 2002; Kalisvaart 2005; Kaneko 1999; Kim 2001; Lee 2004; Liptzin 2005; Miyaji 2007; Pae 2004; Parellada 2004; Prakanratta 2007; Skrobik 2004).
Following GDG advice, indirect evidence was obtained from three further studies (Douglas 2008; Gill 2005; Hermann 2004).
14.3.1. Study Design
The following types of studies were included in the adverse effects analysis:
14.3.2. Population
The studies looked at a wide range of participants, but for the most part were in patients undergoing surgery or admission to intensive care. Three of the studies (Douglas 2008; Gill 2005; Hermann 2004) reported on stroke adverse events associated with antipsychotics in older patients, who were likely to be at risk of delirium. These studies were on an indirect population.
14.3.3. Intervention and Comparisons
There was a diverse range of interventions, and associated comparator agents across the trials.
14.3.4. Assessment and Reporting of Adverse Effects
A diverse range of methods were used, with the most well-defined ones being scales for assessing extrapyramidal signs and symptoms. It is not clear though how useful such scales are in postoperative or intensive care patients, in contrast to ambulant psychiatric patients.
14.4. Results
The interventions, comparators and populations were extremely varied, as was the reporting of adverse effects outcomes. Descriptive summaries are given in Appendix D.
14.4.1. Direct comparison of active agents
Five studies (Gill 2005; Herrmann 2004; Lee 2004; Miyaji 2007; Skrobik 2004) reported direct comparisons between two antipsychotic agents.
Extrapyramidal adverse effects were the main focus of three studies (Lee 2004; Miyaji 2007; Skrobik 2004), with one study (Skrobik 2004) describing specific efforts to “carefully record” such events. Two studies reported specifically on stroke as an adverse event (Gill 2005; Herrmann 2004). One study was in older adults (mean age 81.7 years) (Herrmann 2004) and one study was in older adults with dementia (mean age 82.6 years) (Gill 2005).
No extrapyramidal events were found in the Lee (2004) study, but both Miyaji (2007) and Skrobik (2004) studies described a higher incidence of extrapyramidal effects with haloperidol as compared to quetiapine, and olanzapine respectively. However the Miyaji (2007) study was retrospective while Skrobik (2004) was quasi-randomised, and neither study had any blinding and are thus subject to bias from investigators who may favour the new atypical antipsychotics when recording the extrapyramidal effects.
While the ascertainment of mortality is less subjective, the baseline differences in populations receiving the interventions in the Miyaji (2007) study makes it difficult to draw any reliable conclusions, simply because the more severely ill patients may have received parenteral haloperidol.
Two studies carried out multivariate analyses (Gill 2005; Herrmann 2004). The Gill (2005) study did not take into account confounders such as smoking history, presence and severity of hypertension, lipid status and specific valvular heart conditions. Similarly the Herrman (2004) study did not take into consideration smoking or obesity. The most commonly prescribed antipsychotic was risperidone in both studies (Gill 2005: 76%; Herrmann 2004: 61%)
The Gill (2005) study reported that in older patients with dementia there is no significant difference in the effects of atypical antipsychotics compared with typical antipsychotics.
The Herrmann (2004) study reported results separately for olanzapine and risperidone compared with typical antipsychotics. There was no significant effect of olanzapine [RR 1.1 (95% CI 0.4 to 2.3)] or risperidone [RR 1.4 (95% CI 0.7 to 2.8)] on the incidence of stroke. A head to head comparison (risperidone versus olanzapine) showed no difference in effect [RR 1.3 (95% CI 0.8 to 2.2); ().
14.4.2. Results of specific classes of interventions versus no treatment or placebo
Typical and atypical antipsychotics
One retrospective cohort study (Douglas 2008) was an intra-patient study comparing periods of antipsychotic use and non-use in older adults (indirect population). Median age when first exposed to any antipsychotic drug was 80 years. The study reported on the risk of stroke in patients presenting with first ever stroke (at least 12 months after initial registration on the UK general practice database). The most commonly prescribed atypical antipsychotic was risperidone (81%), followed by olanzapine (18%), amisulpride and quetiapine (4% in each group).
Exposure to any of the antipsychotics was a significant risk factor for stroke [RR 1.73 (95% CI 1.60 to 1.87)]. When typical and atypical antipsychotics were analysed separately, a significant effect was observed [typical antipsychotics: RR 1.28 (95% CI 1.18 to 1.40); atypical antipsychotics: RR 2.32 (95% CI 1.73 to 3.11)]. ().
Haloperidol
There were two included RCTs, both covering the use of haloperidol in postoperative patients. (Kalisvaart 2005, Kaneko 1999)
Both trials reported on active measures to detect adverse effects, with frequent clinical assessments. Haloperidol use in this setting appeared to be relatively safe with no excess of withdrawals from adverse events compared to control, and no extrapyramidal effects.
Atypical antipsychotics
For risperidone, we identified one RCT (Prakanratta 2007) and one observational study (Parellada 2004). There were two open uncontrolled trials of olanzapine (Breitbart 2002, Kim 2001), and one of quetiapine (Pae 2004).
Both the risperidone studies looked for specific adverse effects but did not show any clear trend for harm.
One olanzapine study (Breitbart 2002) used clinical methods to evaluate adverse effects, and this showed sedation to be a problem necessitating dosage reduction.
The remaining two studies (Kim 2001, Pae 2004) did not mention any specific monitoring for adverse effects, and data were sparse.
Cholinesterase inhibitors
One study (Liptzin 2005) which was a randomised double-blind controlled trial of donepezil was identified. Despite methodological strengths elsewhere, this study did not describe any specific monitoring of adverse effects, and did not provide numerical data, even though there was a statement about equivalent rates of adverse effects between drug and placebo. Moreover, adherence to treatment was poor, and as such, no conclusions can be drawn on the relative safety of donepezil.
5-HT3 antagonists
One study (Bayindir 2000) which was an open-label uncontrolled evaluation of ondansetron in postoperative patients was identified. The authors did not state what, if any monitoring was used for detecting adverse effects, and it is difficult to have any confidence in their conclusions that the therapy was safe, without any apparent side effects.
Table 14.1GRADE evidence summary - Adverse Events
View in own window
| Typical antipsychotic vs placebo [prevention] |
|---|
| Outcome/Author (year) | Meta-analysis details | Summary statistics | Comments | GRADE details | GRADE Comments |
|---|
Adverse event (extrapyramidal)
(Kalisvaart 2005) | 1 trial;
430 patients; from RCT | RR=1 | Study reported no extrapyramidal events |
Study quality: no serious limitation Directness: Direct Imprecision: Number of events < 300 Inconsistency: consistent Reporting bias: Adequate
| Placebo comparison
Averse events data-prevention trial |
| GRADE evidence rating: Low |
|
Adverse event (sedation)
(Kalisvaart 2005) | 1 trial;
430 patients; from RCT | RR=1 | Study reported no sedation events in either group |
Study quality: no serious limitation Directness: Direct Imprecision: Number of events < 300 Inconsistency: consistent Reporting bias: Adequate
| Placebo comparison
Averse events data-prevention trial |
| GRADE evidence rating: Low |
|
Adverse event (tachycardia)
(Kaneko 1999) | 1 trial;
78 patients; from RCT | RR=3.15 (95%CI 0.13, 75.12) | No significant difference |
Study quality: serious limitation-method of assessment of delirium unclear Directness: Direct Imprecision: CI crosses appreciable harm/benefit Inconsistency: consistent Reporting bias: Adequate
| Placebo comparison
Averse events data-prevention trial |
| GRADE evidence rating: Very low |
| Atypical antipsychotic vs placebo [prevention] |
|---|
| Outcome/Author (year) | Meta-analysis details | Summary statistics | Comments | GRADE details | GRADE Comments |
|---|
Adverse event (cardiovascular instability)
(Prakanratta 2007) | 1 trial;
126 patients; from RCT | RR=0.76 (95%CI 0.18, 3.27) | Not significant |
| Placebo comparison
Adverse events data-prevention trial |
| GRADE evidence rating: Low |
| Atypical antipsychotic 1 vs atypical antipsychotic 2 [prevention] |
|---|
| Outcome/Author (year) | Meta-analysis details | Summary statistics | Comments | GRADE details | GRADE Comments |
|---|
Adverse events
(Lee 2005) | 1 trial;
31 patients; from RCT | RR=1 | No significant adverse events reported such as acute dystonia dyskinesia |
Study quality: serious limitation-incomplete follow up Directness: Direct Imprecision: Number of patients < 400 Inconsistency: consistent Reporting bias: Adequate
| Adverse events data-prevention trial Very small study; 25% missing data in 1 arm |
| GRADE evidence rating: Low |
|
Adverse events (sedation)
(Breitbart 2002) | 1 trial;
79 patients; from Cohort | Proportion (%)30 | High proportion of patients with sedation events |
| Olanzapine; Hospitalised cancer patients; clinical examination for adverse events |
| GRADE evidence rating: Low |
| Cross review typical antipsychotic vs atypical antipsychotic 2 [treatment] |
|---|
| Outcome/Author (year) | Meta-analysis details | Summary statistics | Comments | GRADE details | GRADE Comments |
|---|
Adverse event (extrapyramidal)
(Skrobik 2004) | 1 trial;
73 patients; from Quasi RCT | RR=8.2 (95%CI 0.48, 140.09) | No significant difference |
| Haloperidol vs olanzapine quasi randomised Adverse events data-treatment trial Adverse events carefully recorded; Not blinded. |
| GRADE evidence rating: Very low |
Overall summary
Results for stroke as an adverse effect for different types of antipyschotics in an indirect population is presented in .
summary of results for stroke as an adverse effect.
14.5. Health economic evidence
No relevant health economic papers were identified.
14.6. Clinical evidence statements
14.7. From evidence to recommendations
The GDG recognized the paucity of the reported adverse effects data is a major limitation. Most of the investigators appear to have focused on extrapyramidal effects, and omitted to consider or discuss the possibility of other adverse events. Another important limitation is that patients with delirium are unable to accurately describe of any untoward symptoms, and thus adverse events may have been missed by the clinicians. The heterogeneous data on haloperidol are of interest here, as this may possibly reflect susceptibility to bias in the unblinded studies that found an excess of extrapyramidal symptoms, when compared to newer atypical agents. The data on extrapyramidal effects and mortality should be judged cautiously, given that higher quality RCTs with thorough adverse effects monitoring have failed to replicate these findings.
All three studies (Douglas 2004; Gill 2005; Herrmann 2004) reporting on the incidence of stroke and antipsychotic use attempted to take into account known confounders, but each had limitations; the Gill (2005) study may have been higher quality because it was prospective but was solely in patients with dementia and the results may therefore not be generalisable.
The indirect evidence from the severe adverse events (stroke) of pharmacological interventions was incorporated into the health economic model. The GDG also took into consideration other direct evidence on adverse events (such as extrapyramidal symptoms) when making recommendations.
The GDG weighed up the benefits and harms when making their recommendations on pharmacological treatment (see section 13.8 and GRADE tables, section 13.4).
