Agar 2017.

Methods	Randomized clinical trial, multi‐site, placebo‐controlled, double‐blind, parallel groups, dose‐titrated Study duration: 72 hours from randomisation, with follow‐up for 6 months Location: Australia
Participants	Total: 249 participants randomised. 2 removed as ineligible. Sample of 247 Setting: 11 inpatient hospice or hospital palliative care services Mean age years (SD): risperidone arm 74.5 years (10.6); haloperidol arm 76.5 years (8.2); placebo arm 73.8 (10.7) years Gender no. female (%): risperidone 25 (31%); haloperidol 33 (41%); placebo 27 (32%) Diagnosis (%): risperidone (93% cancer); haloperidol (83% cancer); placebo (89% cancer) Delirium severity: baseline delirium severity assessed by the MDAS. Median and IQR scores were: 15.1 (5.8) in the risperidone arm; 14.6 (5.0) in the haloperidol arm; and 13.7 (4.8) in the placebo arm. Participants had mild‐ to moderate‐severity delirium. Performance status: Median Australian Karnofsky Performance Status scores were 40 (IQR 30 to 50) in both the risperidone and placebo groups, and 50 (IQR 40 to 50) in the haloperidol group. AKPS scores range from 0 (dead) to 100 (normal). Inclusion criteria: i) Adults receiving hospice or palliative care who required inpatient care from a specialist palliative care team ii) Delirium diagnosed via criteria from the DSM‐IV (Fourth Edition,Text Revision) iii) Memorial Delirium Assessment Scale (MDAS) score of 7 or more iv) Presence of the target symptoms of delirium associated with distress v) Able to speak English Exclusion criteria: Delirium due to substance withdrawal, history of neuroleptic malignant syndrome, regular use of antipsychotic drugs within 48 hours, previous adverse reaction to antipsychotic drugs, extrapyramidal disorders, prolonged QT interval, clinician‐predicted survival of 7 days or fewer, cerebrovascular accident or seizure in the prior 30 days, and pregnancy or breastfeeding, unable to swallow.
Interventions	Intervention 1: risperidone: oral risperidone solution 1 mg/4 ml. n = 82 Intervention 2: haloperidol: oral haloperidol solution 1 mg/4 ml. n = 81 Comparison: placebo: oral placebo solution. n = 84 Participants 65 years or younger in the intervention groups received a 0.5 mg loading dose administered with the first dose of 0.5 mg, then 0.5 mg maintenance doses every 12 hours. Doses could be titrated by 0.25 mg on day 1 and by 0.5 mg thereafter to a maximum dose of 4 mg/d. For participants older than 65 years, the loading, initial, and maximum doses were halved The placebo solution was titrated similarly using matching volumes of solution for each dose level. Doses were increased if the sum of NuDESC scores for items 2, 3, and 4 (delirium symptoms score) was 1 or more at the most recent assessment, conducted every 8 hours. Dose reduction to the prior dose could occur for adverse effects, resolution of delirium (MDAS score of < 7 for 48 hours), or resolution of symptoms (all NuDESC item scores < 1 for 48 hours) Timing of treatment: maintenance dose every 12 hours after the baseline does (0 hours baseline, 12, 24, 36, 48, 60 hours) Treatment duration: 72 hours
Outcomes	Primary outcome: Delirium symptom scores were assessed by the Nu‐DESC. Scores could range from 0 to 6, with higher scores indicating more severe symptoms. The average of the last 2 delirium symptom scores on day 3 were used Secondary outcomes: Delirium severity assessed daily by the MDAS score Lowest delirium symptoms score Daily use of midazolam Extrapyramidal symptoms assessed by the Extrapyramidal Symptom Rating Scale Sedation assessed by the Richmond Agitation‐Sedation Scale Adverse events assessed by the National Cancer Institute Common Terminology Criteria for Adverse Events Survival
Notes	Study funder: Australian Government’s Department of Health under the National Palliative Care Strategy. Individual site funding was supplemented by grant NHMRC 480476 from the National Health and Medical Research Council, Australia Conflicts of interest disclosure: none reported Trial registration: ACTRN12607000562471
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Site randomisation schedules were generated using random number tables at an independent blinded central registry
Allocation concealment (selection bias)	Low risk	Allocation concealment was by sealed opaque envelopes. Site clinical trial pharmacists who opened the treatment schedules to prepare the intervention were not otherwise involved in patient care. Study medication was dispensed in opaque screw‐top bottles, which were identical in terms of volume, colour, and smell and taste of the contents
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Treatment assignment was double‐blinded: both participants and investigators were masked to treatment group for the duration of the study
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Quote: "Treatment allocation will not be disclosed to patient and their proxy, study staff, treating clinicians or investigators"
Incomplete outcome data (attrition bias) All outcomes	High risk	Full details on attrition and exclusions are reported. The proportion of missing data varied across trial arms. Dropout was higher in risperidone group (52%) than in haloperidol group (37%) and the placebo group (35%). Dropout rates could be different because of impact of intervention
Selective reporting (reporting bias)	Low risk	Protocol available. The primary outcome and secondary outcomes were specified in the protocol
Sample size	Unclear risk	Number of participants in both arms between 50 and 100