Girard 2010a.

Methods	RCT comparing haloperidol, ziprasidone, and placebo on the number of days alive and without delirium or coma among ICU patients (49% of all patients were delirious at enrolment, and 35% were comatose) Study took place in 6 ICUs (mixed medical and surgical) at 6 tertiary care centres in the USA
Participants	Participants included 101 (N = 35 haloperidol, median age 51 (IQR 35 to 59) years, 20/35 (57% male); N = 30 ziprasidone, median age 54 (IQR 47 to 66), 21/30 (70%) male; N = 36 placebo, median age 56 (IQR 43 to 68), 22/36 (61%) male) mechanically ventilated, critically ill patients with an abnormal level of consciousness or receiving sedative or analgesic medications Study enrolment between February 2005 and July 2007
Interventions	Participants received 5 mg haloperidol (as a solution containing 1 mg/mL), 40 mg ziprasidone (as a solution containing 8 mg/mL), or placebo (as a 5‐mL solution). In patients without gastric access, study drug was given via 0.5‐mL intramuscular injection (to a maximum of 8 doses). If QTc remained < 500 ms, the second dose of study drug was administered 12 hours after the first, and subsequent doses were given every 6 hours until a change in frequency was warranted. If 2 consecutive assessments for delirium/coma were negative, drug frequency was decreased to every 8 hours, and the drug was discontinued if no delirium or coma was noted for 48 hours. Study drug was reduced if patients remained over‐sedated (RASS ≥ 2 levels deeper than target score) despite discontinuation of sedatives. Study drug was restarted or increased in frequency when over‐sedation was resolved. Study drug was restarted (if discontinued prior) or increased to the previously effective dose if delirium recurred. Study drug was discontinued if extrapyramidal symptoms (≥ 3 points on 3 or more categories of the Simpson‐Angus Scale) or QTc prolongation (> 500 ms) occurred and was restarted only if these were resolved All patients stopped study drug on day 14 regardless of clinical status Other treatments including approaches to sedation were determined by the managing ICU team. Daily spontaneous awakening trials were common but were not protocolized An open‐label antipsychotic was strongly discouraged during the trial but could be used if the ICU team considered it necessary for breakthrough agitation Assessment: brain dysfunction was assessed twice daily using CAM‐ICU and RASS Non‐drug strategies: none of the ICUs used formalized non‐pharmacological interventions to prevent or treat delirium
Outcomes	Primary Number of days alive without coma or delirium (over 21 study days) Secondary Daily delirium risk Duration of delirium Duration of coma Number of days alive and breathing without assistance (in 21 study days) (i.e. ventilator‐free days) Time to ICU and hospital discharge All‐cause 21‐day survival
Notes	Funding Study was investigator‐initiated Study drug was provided by Pfizer Inc., which had no role in the design or conduct of the trial Dr Girard received support from National Institutes of Health (HL007123) Hartford Geriatrics Health Outcomes Research Scholars Award Program Vanderbilt Physician Scientist Development Program VA Tennessee Valley Geriatric Research, Education, and Clinical Center (GRECC) Dr Pandharipande received support from VA Clinical Science Research and Development Service (VA Career Development Award) ASCCA‐FAER‐Abbott Physician Scientist Award Vanderbilt Physician Scientist Development Program Dr Ely received support from VA Clinical Science Research and Development Service (VA Merit Review Award) VA Tennessee Valley GRECC National Institutes of Health (AG027472) Registration Study was registered at clinicaltrials.gov under NCT00096863 Study authors were not contacted
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Participants were randomly assigned in a 1:1:1 manner via a computer‐generated, permuted block randomization scheme stratified according to study centre
Allocation concealment (selection bias)	Low risk	A co‐ordinating centre biostatistician designated treatment group assignments on a list that was provided only to the investigational pharmacists at each study centre, who referred to their unique list to determine group assignment after each patient was enrolled
Blinding of participants and personnel (performance bias) All outcomes	Low risk	This was a double‐blind study. Except for the pharmacist, neither study personnel nor participants were aware of treatment group assignment. Participants received 1 of the 3 colourless, odourless, and tasteless study drugs
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Blinded trained study personnel evaluated participants twice daily for acute brain dysfunction, diagnosing delirium with CAM‐ICU
Incomplete outcome data (attrition bias) All outcomes	Low risk	Data were analysed via an intention‐to‐treat principle. 2 participants were excluded after randomization, before study drug was administered, because of ventricular tachycardia. No outcome data could be collected for these 2 participants after their withdrawal
Selective reporting (reporting bias)	Low risk	All outcomes reported in methods were included in results. Data presented for all participants were included in the analysis
Other bias	Low risk	Open‐label antipsychotic administration was strongly discouraged during the trial but could be provided if the clinical team considered it necessary for breakthrough delirium and agitation. Pfizer Inc. had no role in the design or conduct of the trial; in the collection, analysis, or interpretation of data; nor in the preparation, review, approval, or publication strategy of the study manuscript