Objectives We performed an updated meta-analysis of antipsychotic treatment in patients with delirium, based on a previous meta-analysis published in 2007.
Methods Included in this study were randomised, placebo-controlled or usual care (UC) controlled trials of antipsychotics in adult patients with delirium. Our primary outcome measure was response rate at the study end point. The secondary outcome measures included improvement of severity of delirium, Clinical Global Impression-Severity Scale (CGI-S), time to response (TTR), discontinuation rate and individual adverse effects. The risk ratio (RR), the number-needed-to-treat/harm (NNT/NNH), 95% CIs and standardised mean difference (SMD), were calculated.
Results We identified 15 studies (mean duration: 9.8 days) for the systematic review (total n=949, amisulpride=20, aripiprazole=8, chlorpromazine=13, haloperidol=316, intramuscular olanzapine or haloperidol injection=62, olanzapine=144, placebo=75, quetiapine=125, risperidone=124, UC=30 and ziprasidone=32), 4 of which were conference abstracts and unpublished. When pooled as a group, antipsychotics were superior to placebo/UC in terms of response rate (RR=0.22, NNT=2), delirium severity scales scores (SMD=−1.27), CGI-S scores (SMD=−1.57) and TTR (SMD=−1.22). The pooled antipsychotic group was associated with a higher incidence of dry mouth (RR=13.0, NNH=5) and sedation (RR=4.59, NNH=5) compared with placebo/UC. Pooled second-generation antipsychotics (SGAs) were associated with shorter TTR (SMD=−0.27) and a lower incidence of extrapyramidal symptoms (RR=0.31, NNH=7) compared with haloperidol.
Conclusions Our results suggested that SGAs have a benefit for the treatment of delirium with regard to efficacy and safety compared with haloperidol. However, further study using larger samples is required.
Statistics from Altmetric.com
Delirium is defined by an acute disturbance of alertness, awareness and attention, with a fluctuating course (Diagnostic and Statistical Manual of Mental Disorders, fifth edition: DSM-V), commonly occurring in medically ill patients, with an occurrence rate varying between 11% and 42%.1 Thus, how to manage patients with delirium is crucial in acute hospital settings. However, the pathophysiology of delirium remains poorly understood. One hypothesis suggests that increased sensitivity of the brain to a peripheral inflammatory process precipitately leads to imbalanced neurotransmitters in the brain.2 Reduced cholinergic function, excess release of dopamine and norepinephrine, and altered serotoninergic activity, are considered to play important roles in the development of delirium.3 ,4 Among them, acetylcholine deficiency and dopamine excess particularly contribute to the development of delirium.5 ,6 In the light of the fact that dopamine and acetylcholine are inversely related, dopamine-blocking neuroleptic drugs, such as haloperidol, have been empirically used to manage symptoms of delirium following thorough evaluation and elimination of the underlying aetiologies of delirium.7 ,8 Furthermore, given that serotonin and noradrenaline control dopamine activity that subsequently affects cholinergic pathways, second-generation antipsychotics (SGAs) that modulate these neurotransmitters have been increasingly used in clinical practice.9 A previous systematic review of antipsychotics, however, failed to show compelling evidence for the treatment of delirium, mainly due to the paucity of data from included studies.10 Moreover, one retrospective, observational cohort study in which antipsychotics were administered to 11% of intensive care unit (ICU) patients who exhibited delirium, reports that antipsychotic exposure was associated with increased length of stay in ICU and hospital as well as mortality.11 Whereas the most recently published clinical practice guideline by American College of Critical Care Medicine shows lack of evidence of antipsychotic use for the treatment of delirium in ICU patients,12 no comprehensive data or clinical practice guidelines have been reported regarding antipsychotic use for the treatment of delirium in non-ICU settings. Furthermore, in two randomised controlled trials (RCTs) comparing antipsychotics and placebo in ICU patients with delirium,13 ,14 most of the patients were mechanically intubated, implying that these studies included critically worse non-verbal participants than other studies involving participants in non-ICU settings. A recently conducted prospective observation study comparing SGAs (olanzapine, quetiapine, risperidone) with haloperidol reported that these experimental medications were equally efficacious and safe,15 although no placebo comparison was included in the study. Moreover, several RCTs investigating efficacy and safety of antipsychotics in non-ICU settings have been published since the previous systematic review was published,16 ,17 although these studies lack adequate sample sizes. Moreover, even though the recent meta-analysis of elderly patients with dementia showed that SGAs were associated with higher incidence of death compared with placebo, the effect size was small (number-needed-to-harm (NNH)=87).18 As such, to overcome the limitations of small studies and to cover broader outcome measures such as improvement of delirium, cognitive function, discontinuation rate, incidence of death and individual adverse effects, we performed an updated and more comprehensive meta-analysis of the RCTs of antipsychotic medication for the treatment of delirium. We also performed several subgroup analyses of efficacy outcomes divided by treatment environment (ICU setting vs non-ICU settings) and antipsychotic class (SGAs vs first-generation antipsychotics (FGA)).
This meta-analysis was performed according to the Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) guidelines.19
Search strategy and inclusion criteria of studies
To identify relevant studies, three of the authors (TK, TH and SM) independently searched MEDLINE, EMBASE, CENTRAL, CINAHL and PsycINFO, without language restrictions, from inception up to 6 November 2014, using the following search strategy: (Antipsychotics OR aripiprazole OR amisulpride OR asenapine OR blonanserin OR clozapine OR iloperidone OR lurasidone OR olanzapine OR paliperidone OR risperidone OR sulpiride OR ziprasidone OR zotepine OR chlorpromazine OR clotiapine OR droperidol OR haloperidol OR loxapine OR mosapramine OR mesoridazine OR molindone OR fluphenazine OR flupenthixol OR pimozide OR perospirone OR perphenazine OR prochloperazine OR quetiapine OR tiapride OR thioridazine OR thiothixene OR trifluoperazine OR zuclopenthixol) AND delirium AND (randomized OR random OR randomly). Included in this study were randomised, placebo-controlled or usual care (UC) controlled trials of antipsychotics in patients with delirium (adult patients: age >18 years). We excluded studies where antipsychotic medications were used to prevent the occurrence of delirium. Studies regarding substance induced/withdrawal delirium were also excluded. Three of the authors (TK, TH and SM) independently assessed inclusion/exclusion criteria and selected studies. The references of included articles and review articles in this area were also searched for citations of further relevant published and unpublished research, such as conference abstracts.
Data synthesis and outcome measures
Our primary outcome measure was response rate at the study end point, which was defined by the original studies using severity of delirium scales and global scales (please refer to online supplementary table S1). The secondary outcome measures included improvement of severity of delirium, measured using scores on Delirium Rating Scale (DRS),20 Delirium Rating Scale-Revised-98 (DRS-R-98),21 as well as Memorial Delirium Assessment Scale (MDAS)22 and the delirium severity scales (DSS) on days 2–3 and at the end point, Clinical Global Impression-Severity Scale (CGI-S)23 scores, time to response (TTR), discontinuation rate, death, Mini-Mental Status Examination (MMSE)24 scores, use of additional antipsychotic medication(s) and individual adverse effects. The method of data synthesis in our meta-analysis is shown in online supplementary table S1.
Three of the authors (TK, TH and SM) independently extracted data from included studies. We used intention-to-treat or modified intention-to-treat analysis as far as possible. Where such data were unavailable, data for completer analysis were extracted from each study. When data required for the meta-analysis were missing, we contacted the investigators of each study and requested unpublished data. The following four categorical meta-analyses of RCTs were performed for evaluating each outcome: (1) either pooled or individual antipsychotic(s) versus placebo/UC, (2) FGA versus FGA (there was only one study that compared chlorpromazine with haloperidol), (3) either pooled or individual SGA(s) versus haloperidol (because haloperidol was used as a comparator in all studies included in this categorical meta-analysis) and (4) head-to-head comparisons of SGAs versus SGA.
The meta-analysis was undertaken using Review Manager software (V.5.3 for Windows, Cochrane Collaboration, http://tech.cochrane.org/Revman). The random effects model was chosen for this meta-analysis because of potential heterogeneity across studies. For dichotomous outcomes, the risk ratio (RR) was estimated along with its 95% CI whereas for continuous outcomes standardised mean differences (SMDs) were used, combining the effect size data (Hedges’ g). In this study, when the random effects model in dichotomous outcomes showed significant between-group differences, the number-needed-to-treat (NNT)/NNH was calculated. Then, NNT/NNH values were derived from the risk differences (RD) using the formula NNT/NNH=1/RD, with the 95% CIs of NNT/NNH being the inverse of the upper and lower limits of the 95% CI of RD. In addition, we assessed the methodological quality of the trials with the Cochrane risk-of-bias criteria (Cochrane Collaboration, http://www.cochrane.org/). We also investigated study heterogeneity using the χ2 test of homogeneity (p<0.05) together with the I2 statistics, considering values of 50% or higher to reflect the considerable heterogeneity.25 Although we did not find significant heterogeneity with respect to the efficacy outcome in all meta-analyses, we carried out a subgroup analysis regarding efficacy outcomes divided by ICU or non-ICU setting studies.
The computerised search yielded 336 references after duplicates were removed (figure 1). We excluded 308 studies based on title and abstract review. Thirteen additional articles were excluded after full-text review because one of them was a review article; two articles pertained to observational studies; six studies were RCTs comparing the efficacy of prophylactic use of antipsychotics with placebo; two studies were RCTs of adjunct therapy on antipsychotics in patients with delirium: bright light adjuvant therapy on risperidone and rivastigmine adjuvant therapy on haloperidol; and two studies were RCTs comparing antipsychotics with dexamedetomidine and ondansetron. In total, we included 15 studies; 11 had been published13 ,14 ,16 ,17 ,26–32 and the remaining four studies were conference abstracts and remained unpublished33–36 (figure 1 and online supplementary table S2). Although one Chinese study28 did not report whether the patients in the control group received placebo, the patients in the control group of that study did not receive any psychotropic drugs. Since the investigators who conducted the trial did not respond to our request for clarification of data, we determined that the control group in the study could be considered as UC. The following results are a brief summary of included studies, and the details of each study are described in online supplementary table S2; mean duration: 9.8 days, total n=949: amisulpride=20, aripiprazole=8, chlorpromazine=13, haloperidol=316, intramuscular olanzapine or haloperidol injection=62 (the conference abstract did not report number of patients randomised to each treatment group), olanzapine=144, placebo=75, quetiapine=125, risperidone=124, UC=30 or ziprasidone=32 (see online supplementary table S2). Four studies13 ,14 ,32 ,36 were conducted in ICU settings, one study27 included ICU patients and non-ICU patients and the rest of the studies were completed in non-ICU settings. Although three studies17 ,27 ,28 used completer analysis, we used these data in our meta-analysis to increase sample size as much as possible. Most of the studies were of high methodological quality based on the Cochrane risk-of-bias criteria, and many were double-blind RCTs and mentioned the required details of the study design (see online supplementary figures S1 and S2).
Antipsychotics versus placebo/UC
When pooled as a group, antipsychotics were superior to placebo/UC in terms of response rate (RR=0.22, 95% CI 0.15 to 0.34, p<0.00001, I2=0%, NNT=2, 95% CI 2 to 3, N=3, n=240; figure 2), DSS scores (SMD=−1.27, 95% CI −2.44 to −0.11, p=0.03, I2=93%, N=3, n=245; figure 2), CGI-S scores (SMD=−1.57, 95% CI −1.91 to −1.23, p<0.00001, I2=0%, N=2, n=204; see online supplementary figure S3) and TTR (SMD=−1.22, 95% CI −1.55 to −0.90, p<0.00001, I2=0%, N=2, n=204; see online supplementary figure S4; table 1). Individually, haloperidol and olanzapine had benefit on DSS scores (SMD=−1.64, 95% CI −2.13 to −1.15 and SMD=−2.14, 95% CI −2.66 to −1.61, respectively), response rate (RR=0.18, 95% CI 0.09 to 0.35, NNT=2, 95% CI 1 to 3 and RR=0.25, 95% CI 0.15 to 0.44, NNT=2, 95% CI 1 to 3, respectively), CGI-S (SMD=−1.62, 95% CI −2.11 to −1.14 and SMD=−1.52, 95% CI −2.00 to −1.04, respectively) and TTR (SMD=−1.11, 95% CI −1.56 to −0.65 and SMD=−1.35, 95% CI −1.81 to −0.88, respectively); however, these results were derived from only one UC-controlled study (table 1). Although we did not detect significant heterogeneity in terms of the meta-analysis of response rate, we performed a subgroup analysis divided by ICU or non-ICU setting studies (see online supplementary table S3). When we performed a subgroup analysis using only ICU study, the pooled group was marginally superior to placebo in the outcome (RR=0.25, 95% CI 0.06 to 1.02, p=0.05, I2=not applicable, N=1, n=36). On the contrary, when we performed another subgroup analysis using non-ICU setting studies, pooled antipsychotics were significantly superior to UC in response rate (RR=0.22, 95% CI 0.15 to 0.34, p<0.00001, I2=0%, NNT=2, N=2, n=204). Other efficacy outcomes did not include ICU setting studies. Neither pooled nor individual antipsychotics differed from placebo/UC regarding discontinuation rate (see online supplementary figure S5) and incidence of death, extrapyramidal symptoms, akathisia and QTc prolongation (see online supplementary table S4). Pooled antipsychotics were associated with higher incidence of dry mouth (RR=13.0, 95% CI 1.82 to 93.10, p=0.01, I2=0%, NNH=5, 95% CI 3 to 9, N=2, n=204) and sedation (RR=4.59, 95% CI 1.36 to 15.50, p=0.01, I2=2%, NNH=5, 95% CI 4 to 7, N=3, n=240) compared with placebo/UC (table 1). Individually, haloperidol was associated with higher incidence of dystonia compared with UC (RR=19.32, 95% CI 1.21 to 307.84, NNH=3, 95% CI 2 to 5) and olanzapine was associated with higher incidence of dry mouth compared with UC (RR=16.4, 95% CI 1.02 to 262.58, NNH=4, 95% CI 3 to 6; table 1). Visual inspection of the funnel plots for primary outcome in both treatment groups suggested no publication bias (data not shown).
FGA versus FGA
There were no significant differences in DSS scores on day 2 and at the end point, or on MMSE scores, incidence of death, or dystonia and dyskinesia, between chlorpromazine and haloperidol (see online supplementary table S5). Visual inspection of the funnel plots for primary outcome in both treatment groups suggested no publication bias (data not shown).
SGA versus FGA
Neither pooled SGAs nor individual SGAs outperformed haloperidol regarding DSS scores on days 2–3 and at the end point, regarding response rate on days 2–3 and at the end point (figure 3), and regarding CGI-S scores at the end point (table 2). However, pooled SGAs were associated with shorter TTR (SMD=−0.27, 95% CI −0.54 to −0.01, p=0.04, N=3, n=222; figure 3) and lower incidence of extrapyramidal symptoms (RR=0.31, 95% CI 0.12 to 0.84, p=0.02, NNH=7, 95% CI 4 to 100, N=4, n=296) compared with haloperidol (table 2). Individually, olanzapine was superior to haloperidol in TTR (SMD=−0.35, 95% CI −0.68 to −0.03, p=0.03) and incidence of dystonia (RR=0.08, 95% CI 0.02 to 0.35, p=0.0006, NNH=3, 95% CI 2 to 6; figure 3 and table 2). In the subgroup analysis (see online supplementary table S3), pooled SGAs were superior to haloperidol in DSS scores at the end point in ICU setting studies (SMD=−0.52, 95% CI −0.85 to −0.19, p=0.002, N=2, n=144) but not in the non-ICU setting. On the contrary, pooled SGAs were superior to haloperidol in TTR in non-ICU setting studies (SMD=−0.31, 95% CI −0.59 to −0.02, p=0.03, N=2, n=198) but not in ICU setting studies. Neither pooled nor individual SGAs differed from haloperidol regarding discontinuation rate (see online supplementary figure S6) and individual adverse events (see online supplementary table S6). Visual inspection of the funnel plots for primary outcome in both treatment groups suggested no publication bias (data not shown).
SGAs versus SGA
Amisulpride was similar to quetiapine regarding DSS scores at the end point, TTR, discontinuation rate, and incidence of death and sedation (see online supplementary table S7). Between olanzapine and risperidone, there were no significant differences in DSS scores on day 3 and at the end point, response rate on day 3 and at the end point, MMSE scores, discontinuation rate, or on 12 individual adverse effects including death and extrapyramidal symptoms (see online supplementary table S7). Visual inspection of the funnel plots for primary outcome in both treatment groups suggested no publication bias (data not shown).
We performed an updated meta-analysis of antipsychotic treatment in patients with delirium based on one previous meta-analysis (3 studies were included in that meta-analysis).10 Antipsychotic medications were superior to placebo/UC in efficacy outcomes. However, it requires scrupulous attention to interpret these results because the controlled groups included UC28 and most of all analyses included data of completer analysis.28 There were no significant differences in most of the side effects, including death, between antipsychotics and placebo/UC. Regarding SGAs versus haloperidol, SGAs were associated with shorter TTR (SMD=−0.27, 95% CI −0.54 to −0.01, p=0.04, N=3, n=222) and lower incidence of extrapyramidal symptoms (RR=0.31, 95% CI 0.12 to 0.84, p=0.02, NNH=7, 95% CI 4 to 100, N=4, n=296) compared with haloperidol. Again, however, these results included data of studies with completer analysis and UC.28 Moreover, SGAs were not inferior to haloperidol in outcomes pertaining to other side effects. Emerging evidence suggested that close attention is required when prescribing antipsychotics in the elderly. For example, death occurred in 3.5% of patients with dementia randomised to SGAs compared with 2.3% of patients randomised to placebo, and the pooled OR for death was 1.54 (95% CI 1.06 to 2.23; NNH=87).37 Moreover, antipsychotic use in the elderly can raise the risk of incidences of pneumonia, stroke and falls.18 ,38 Given our results confirming safety of SGAs in comparison to that of haloperidol in the treatment for patients with delirium, we suggest that the clinicians choose SGAs prior to the trial of haloperidol for the treatment of delirium. Owing to the limited number of studies included in our study, however, we were unable to explore which SGAs should be used for the treatment of delirium.
The studies included in the meta-analysis had differences in patient characteristics (race and ethnicity, study duration and aetiology of delirium), which could generate heterogeneity when combining data for systematic review and meta-analyses. However, we did not detect significant heterogeneity regarding the primary outcome in any of the categorical meta-analyses. Despite these results, given that delirium is considered to be influenced by several environmental factors such as ICU stay, we performed the subgroup analyses in terms of efficacy outcomes. In antipsychotics versus placebo/UC comparisons, while antipsychotics were marginally associated with producing more responders compared with placebo in the ICU setting, antipsychotics were superior to UC in response rate in the non-ICU setting. As the risk ratio of these two subgroup analyses was similar, we considered that antipsychotics were extremely useful in treating delirium. In SGAs versus haloperidol comparisons, SGAs were superior to haloperidol only in TTR of efficacy outcomes. In subgroup analysis, SGAs were superior to haloperidol in TTR in the non-ICU setting but not in the ICU setting. The study conducted by Hu et al, which used completer analysis as well as UC control, likely contributed to the result of the non-ICU setting subgroup analysis. On the contrary, SGAs were superior to haloperidol in DSS at the end point in the ICU setting but not in the non-ICU setting. Two studies conducted in ICU settings and included in subgroup analysis of our study were homogeneous with the same DSS scale (MDAS) and employed the same method using non-completer analysis. Therefore, although only two studies were included in this subgroup analysis, we recommend that clinicians choose SGAs prior to a trial of haloperidol for the treatment of ICU patients with delirium.
It may be argued that our systematic review and meta-analyses are biased and may lack validity owing to the inclusion of unpublished conference abstracts. Given that the study quality of this grey literature is difficult to assess due to its incompleteness, this critique may be to the point. To investigate if intervention effectiveness is influenced by the exclusion of grey literature in meta-analyses, McAuley et al39 extracted a random sample of 135 meta-analyses from an existing database (MEDLINE), in which they identified that grey literature was included in 33% of the meta-analyses. With exclusion of the grey literature, their result revealed that meta-analyses could generate exaggerated effectiveness of interventions used in original randomised controlled trials. Given the interpretation of their study, inclusion of four unpublished conference abstracts in our study is considered to be appropriate.
Our results suggested that antipsychotic medications were superior to PLA/UC in efficacy outcomes. Moreover, SGAs are more beneficial for the treatment of delirium regarding efficacy and safety outcomes compared with haloperidol. However, because the studies included in the meta-analysis were small, further study using larger samples is required.
The authors thank Dr Djokic and Dr Kim for providing, from their studies, additional, unpublished data relevant to this meta-analysis.
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
- Data supplement 1 - Online supplement
TK and TH contributed equally.
Contributors TK and TH had full access to all of the data in the study, and take responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design, analysis and interpretation of the data, acquisition of the data and statistical analysis were performed by TK and TH. The manuscript was written by TK, TH, SM and NI. NI supervised the review.
Funding TK has received speaker's honoraria from Abbott, Astellas, Daiichi Sankyo, Dainippon Sumitomo, Eisai, Eli Lilly, GlaxoSmithKline, Janssen, Yoshitomi, Otsuka, Meiji, Shionogi, Tanabe-Mitsubishi, Tsumura, Novartis and Pfizer. TH has received speaker's honoraria from Dainippon Sumitomo, Otuka and Novartis. SM has received speaker's honoraria from Eisai, Janssen, Novartis, Daiichi Sankyo, Ono, Eli Lilly, Takeda and Otsuka. NI has received speaker's honoraria from Astellas, Dainippon Sumitomo, Eli Lilly, GlaxoSmithKline, Janssen, Yoshitomi, Otsuka, Meiji, Shionogi, Novartis and Pfizer, and has research grants from GlaxoSmithKline and Otsuka.
Competing interests None declared.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.