Neuroleptic Rotation for Terminal Agitation in Patients with Cancer and Delirium at an Acute Palliative Care Unit: A Single-Centre, Parallel-Group, Double-Blind, Randomised Clinical Trial

Abstract

Background:

The role of neuroleptics for terminal agitated delirium is controversial. We assessed the effect of 3 neuroleptic strategies on refractory agitation in cancer patients with terminal delirium.

Methods:

In this single-centre, double-blind, parallel group randomised trial, patients with advanced cancer, age≥18, admitted to a Palliative Care unit with refractory agitation despite low dose haloperidol were randomised to haloperidol dose escalation 2 mg intravenously every 4 hours; neuroleptic rotation with chlorpromazine 25 mg intravenously every 4 hours; or combined haloperidol 1 mg and chlorpromazine 12.5 mg intravenously every 4 hours until death or discharge. Rescue doses identical to the scheduled doses were administered at inception then hourly as needed. Permuted block randomization (block size=six, 1:1:1) was conducted, stratified by baseline Richmond Agitation Sedation Scale (RASS) scores. Research staff, clinicians, patients and caregivers were blinded. The primary outcome was change in RASS from time 0 to 24 hours. Comparisons among group was conducted with modified intention-to-treat analysis. This completed study is registered with Clinicaltrials.gov, NCT03021486.

Findings:

Between 7/5/2017 and 7/1/2019, 68 patients were enrolled and 45 received the blinded study interventions (escalation n=15, rotation n=16, combination n=14). RASS decreased significantly within 30 minutes and remained low at 24 hours in the escalation group (number analyzed, mean RASS change between 0 and 24 h [95%CI]: n=10, −3·6 [−5,−2·2]), rotation group (n=11, −3·3 [−4·4,−2·2]) and combination group (n=10, −3.0 [−4·6,−1·4]), with no difference among groups (P=0·71). The most common serious toxicity was hypotension (escalation n=6 [40%], rotation n=5 [31%], combination n=3 [21%]); there were no treatment-related deaths.

Interpretation:

Our data provided preliminary evidence that the 3 strategies of neuroleptics may reduce agitation in patients with terminal agitation. These findings should be put in context of the single-centre design, small sample size and lack of a placebo-only group.

Funding:

National Institute of Nursing Research

Introduction

In the last days of life, over 90% of cancer patients experience terminal delirium^{1, 2} with up to 70% developing restlessness, agitation and/or violent behavior.³ A majority of patients with delirium were aware of their confusion and this experience was highly distressing to them, their caregivers, and healthcare professionals.^4–6

Despite treatment of reversible causes and optimal non-pharmacologic measures, many patients continue to experience severe delirium-related symptoms and distress, raising the need for medications. Although neuroleptics have long been prescribed as first line pharmacologic therapy for management of delirium and associated agitation,⁷ their use has recently been called into question with several clinical trials and systematic reviews suggesting that they offer no benefit.^8–13 However, no randomised trials examining neuroleptics have specifically examined agitation as a primary outcome, a key reason for prescribing neuroleptics by clinicians and a particularly important therapeutic goal in the terminal delirium setting.

Compounding this clinical problem is that approximately 50% of delirious patients admitted to palliative care units have refractory agitation despite low dose haloperidol and the second line options are ill-defined. In a small randomised trial, we previously found that the combination of lorazepam and haloperidol resulted in better agitation control and improved perceived comfort compared to haloperidol alone over 8 hours.¹⁴ However, this study only examined a single dose of rescue medication, the study duration was short and it did not assess neuroleptics other than haloperidol. Chlorpromazine represents an appealing option for refractory agitation because it can be given intravenously and its α1 adrenergic blockage effect may reduce agitation.^15–17 Prior to the current study, the only randomised trial that compared haloperidol and chlorpromazine for the treatment of delirium; however, agitation was not a study outcome.¹⁸ For patients with refractory terminal agitation despite low dose haloperidol, haloperidol dose escalation, neuroleptic rotation to chlorpromazine and combination therapy with both haloperidol and chlorpromazine represent 3 potential therapeutic strategies. The goal of these scheduled high dose neuroleptic strategies is to reduce restlessness/agitation, which represent a step before palliative sedation to induce continuous deep sedation. A better understanding of the role of these therapeutic strategies may help clinicians to better manage this highly distressing syndrome. The objective of this double-blind randomised clinical trial was to estimate the effect of haloperidol dose escalation, neuroleptic rotation to chlorpromazine, and combination therapy with both haloperidol and chlorpromazine on agitation intensity in cancer patients with delirium admitted to our Palliative and Supportive Care Unit (PSCU).

Methods

Study design

This single-centre, double-blind, double-dummy parallel group randomised clinical trial assigned patients with refractory hyperactive or mixed delirium in a 1:1:1 ratio to receive (1) haloperidol dose escalation, (2) neuroleptic rotation to chlorpromazine, or (3) combination therapy with haloperidol and chlorpromazine until death, discharge or withdrawal. The trial protocol is available as an online Supplement (appendix pp 13–24). The Institutional Review Board at MD Anderson Cancer Centre approved this study (protocol number 2016–0687).

Patients

Patients were recruited from the PSCU at the University of Texas MD Anderson Cancer Centre in Houston, Texas, United States. Inclusion criteria included a diagnosis of advanced cancer, age 18 years or older, a diagnosis of delirium by DSM-V criteria, a history of agitation with Richmond Agitation Sedation Scale (RASS) ≥+1 over the past 24 hours despite being on scheduled haloperidol of 1–8 mg/day or receiving ≥4 mg/day of rescue haloperidol. RASS +1 was selected as a cutoff for enrollment because persistent restlessness with non-purposeful movements in the terminal delirium setting can be highly distressing to caregivers and often triggered the use of rescue medications to maximize comfort. Patients were excluded if they had contraindications to neuroleptics (i.e. Parkinson’s disease, Alzheimer’s dementia, myasthenia gravis, acute narrow angle glaucoma, neuroleptic malignant syndrome, active seizure disorder, documented QTc prolongation, hypersensitivity) or were already on scheduled chlorpromazine within the past 48 hours of study enrollment. Survival was not a pre-specified criterion in this study because of uncertainty in prognostication; patients who fulfilled the above eligibility criteria were expected to have a survival in terms of days based on a similar study.¹⁴ Written surrogate consent was obtained from the medical power of attorney or legal representative.

The PSCU was chosen as the study setting for several reasons. First, agitation is a particularly common and distressing concern among patients with terminal delirium. The ability to manage agitation is of paramount importance. Second, the PSCU represented a controlled environment with standardized care delivered by an interdisciplinary palliative care team experienced in delirium management. In addition to comprehensive symptom management and treatment of any potentially reversible causes such as infections, hypercalcemia, polypharmacy, and opioid neurotoxicity, all patients routinely received non-pharmacologic measures including orientation cues, window light, efforts to minimize unnecessary stimuli, and caregiver education. Third, PSCU physicians were actively engaged in this study by helping to identify potential patients, and PSCU nurses helped with study medication administration and documentation of study outcomes. Both physicians and nurses were blinded to study assignment.

Randomisation and masking

Randomisation occurred immediately after enrollment to ensure the blinded study medications would be ready when needed. Patients were assigned to one of three treatment groups based on a computer-generated randomisation sequence with a permuted block size of 6 and stratified by RASS score at the time of enrollment (RASS ≤+1 vs. +2 vs. +3–4).

Allocation concealment was maintained by using a secured website that was only accessible to the study pharmacist, who then assigned patients to the study intervention and prepared the blinded study medications. Research staff enrolling patients and conducting the study assessments, bedside nurses, attending physicians, patients and caregivers were blinded to the allocation of the study medication and study outcomes throughout the study. The success of masking was assessed by asking the caregivers and research staff to guess the assignment at the end of the study.

Procedures

Immediately after enrollment, the neuroleptic regimen was standardized to open-label haloperidol 2 mg every 6 hours intravenously and 2 mg every 1 hour as needed. We then monitored patients’ RASS every 2 hours. When a patient’s absolute RASS score was ≥+1 and the patient required rescue medication in the bedside nurse’s judgement, we initiated the blinded study medication regimen.

The blinded study medication regimen consisted of a scheduled dose every 4 hours intravenously and rescue dose every 1 hour intravenously as needed. In addition to the scheduled dose, patients received one rescue dose at the time of blinded study medication initiation. The dose of rescue medications was identical to scheduled study medications. All study medications were generic from commercial sources procured by our pharmacy. Using a double-dummy design, each intravenous scheduled or rescue dose included one syringe consisting of either haloperidol or 5% dextrose administered at 5 mL/min and one bag consisting of either chlorpromazine or normal saline infused at 2·5 mL/min. The starting scheduled or rescue dose was 2 mg of haloperidol for the dose escalation group, 25 mg of chlorpromazine for the neuroleptic rotation group, and 1 mg of haloperidol and 12·5 mg of chlorpromazine for the combination group. These medications dosages were designed to be approximately equivalent based on the concept of haloperidol equivalent daily dose, in which 1 mg of haloperidol was comparable to 12·5 mg of chlorpromazine.^{3, 19, 20} Rescue study medication was administered when a patient’s absolute RASS score was +1 or higher and the patient required rescue medication in the bedside nurse’s judgement. Medications in the three groups had identical appearance and volume to ensure proper blinding. The medication regimen continued until death, discharge, or withdrawal and the doses were titrated according to standardized procedures (appendix p 3). Specifically, both scheduled and as needed medication doses were increased by 1 dose level if patients had RASS ≥+2 anytime or required 3 or more rescue doses within a 4 hour period. The use of other medications and withholding of scheduled study medications were permissible as per standard of practice according to the clinical judgement of the attending physician and bedside nurse.

The bedside nurse assessed RASS at fixed times: immediately prior to study medication administration (time 0), 0·5, 1, 2 hours, then every 2 hours during the first 24 hours and then every 4 hours until death, discharge or withdrawal. RASS is a validated 10-point numeric rating scale that ranges from −5 to +4 at 8 hours.^{21, 22} Negative numbers indicate deeper levels of sedation (−1=drowsy, −2=light sedation, −3=moderate sedation, −4=deep sedation, and −5=unarousable), 0 indicates that the patient was alert and calm, and positive numbers indicate increasing levels of agitation (+1=restless, +2=agitated, +3=very agitated and +4=combative). The inter-rater agreement of RASS has been assessed in a previous study at our PSCU and found to be 0·79 (95% CI 0·56, 1·00; P<0·01).¹⁴

We documented all neuroleptics and benzodiazepines doses and time of administration during the study period. The need for any rescue medications or dose level increase both indicate failure of the scheduled study medication to adequately reduce restlessness/agitation.

Starting the day after blinded study treatment, we assessed the perceived level of comfort and level of agitation in the patient by asking the blinded caregivers and bedside nurses independently to answer the following questions: “In my opinion, the patient was more comfortable after the study medication” and “In my opinion, the patient was less restless/agitated after the study medication”.¹⁴ Responses were provided on a 5-point Likert scale from “strongly agree” to “strongly disagree”.

Delirium-related distress in nurses and caregivers with the Delirium Experience Questionnaire, which examined both the recalled frequency of 7 delirium symptoms and associated distress in the rater.¹⁴

Delirium severity was assessed daily from study enrollment with the Memorial Delirium Assessment Scale, a 10-item assessment scale validated for assessment of delirium in cancer patients.^{23, 24} Each item was assigned a score between 0 to 3 for a total score between 0–30. Patients who were not responsive at the time of assessment or had a significantly prolonged response time were assigned a score 3 for each item. A total score of 13 or higher indicates delirium.²³

Symptom burden as perceived by caregivers was assessed daily from study enrollment using the Edmonton Symptom Assessment Scale in which the average intensity of 10 symptoms over the past 24 hours were each rated on an 11-point numeric rating scale from 0 (none) to 10 (worst).²⁵ ESAS has been extensively validated in the palliative cancer care setting and can be completed by a proxy.²⁶

Adverse events of study medications were assessed by monitoring vital signs every 8 hours in the first 24 hours and then every 12 hours thereafter. We also graded adverse events daily using the National Cancer Institute Common Toxicity Criteria Adverse Effects (NCI-CTCAE) v4·03.²⁷ The Udvalg for Kliniske Undersogelser (UKU) rating scale was administered at baseline and then on day 3 of blinded medications to assess the severity of 8 neurologic symptoms related to neuroleptics over the past 3 days.²⁷ Each item was assigned a score from 0 (absent) to 3 (most severe). Overall survival from the time of blinded medication administration was documented. Details of these assessments are provided in the Supplement (appendix pp 1,2).

Outcomes

The primary outcome was change in RASS between time 0 (immediately before initiation of blinded treatment) and 24 hours later. Secondary outcomes defined a priori included the proportion of patients with target RASS −2 to 0 within the first 24 hours; the need for any rescue neuroleptics or benzodiazepines and the need for dose level increase during the first 24 hours; perceived level of comfort and agitation on day 1 as reported by caregivers and nurses, in which the responses “strongly agree” and “agree” were combined for reporting; the change in Delirium Experience Questionnaire, Memorial Delirium Assessment Scale, and Edmonton Symptom Assessment System between day 0 and day 1; and adverse effects including change in vital signs between time 0 and 24 hours, CTCAE documentation, change in UKU rating scores between baseline and day 3, and overall survival. The quality of end-of-life care from the bereaved caregivers’ perspective was an outcome but not reported here because of the focus on the acute hospitalization period.

Statistical analysis

The primary analysis was to assess the within-arm change in RASS over the first 24 hours in each study group separately. With 15 patients per group and 13 measurements over time, we calculated that we had 90% power to detect an effect size of 0·2 with an alpha of 2·5%, assuming the correlation between repeated measures was 0·7 on 1-way, repeated-measures ANOVA (G*power 3·1·9·7. Heinrich-Heine-Universität, Düsseldorf, Germany). The trial was stopped after target enrollment was reached. At the end of the study, we calculated that the actual correlation was 0·52 and the average number of measurement was 11·4 (SD 2·8).

Baseline characteristics were summarized by descriptive statistics. The pre-specified primary outcome, change in RASS from immediately before blinded study medication administration (time 0) to 24 hours after, was assessed with one-way, repeated measures analysis of variance. We compared between study groups by using the Wilcoxon rank sum test. We also assessed the change in RASS from time 0 to 30 min, 4 hours and 8 hours.

We computed the changes before and after blinded medication administration and corresponding 95% confidence interval for secondary outcomes. The 2-tailed Wilcoxon Rank Sum test for continuous variables and 2-tailed Fisher’s exact test for categorical variables. All analyses were two-sided tests. For our pre-specified primary outcome analysis, a 2-sided P-value of 0·05 or less was considered to be statistically significant. We did not adjust for multiple comparisons and all secondary findings are considered to be hypothesis-generating.

Several post-hoc analyses were conducted. We examined the proportion of patients with breakthrough restlessness/agitation, defined as RASS ≥+1 anytime from time 0 to 4 h, 8 h or 24 h. We conducted pairwise comparisons of breakthrough restlessness and need for dose escalation by computing the between-group change and 95% CI. Overall survival was calculated from time of blinded study medication administration to last follow-up or death. The Kaplan Meier method was used for time-to-event analysis and the log-rank test was used to compare overall survival between groups. Because many patients died or were discharged during the open-label phase before they received the blinded study medication, we conducted post-hoc modified intention-to-treat analysis including all patients who started the blinded study interventions.

We also conducted worst-case sensitivity analysis post-hoc by assuming that the patients who started but did not complete the study intervention had no change in RASS score from baseline at 24 hours. Missing data were not imputed for secondary outcomes.

The Statistical Analysis System (SAS version 9·4, SAS Institute, Cary, North Carolina) and R 3·6·3 (R Foundation for Statistical Computing, Vienna, Austria) were used for statistical analysis. This study is registered with Clinicaltrials.gov, NCT03021486.

Role of the funding source

This study was funded by the National Institute of Nursing Research (R21NR016736). The funder of this study had no role in study design, data collection, data analysis, data interpretation, or writing of the report. The corresponding author had full access to all of the data and the final responsibility to submit for publication.

Results

Between July 5, 2017 and July 1, 2019, 998 patients admitted to the PSCU were screened (Figure 1). Among the 116 fully eligible patients, 68 (59%) were enrolled and randomised immediately (dose escalation group, n=23; neuroleptic rotation group, n=22; combination group, n=23). After a median observation period of 6 hours (IQR 4–20 hours, range 2–68 hours) on open-label haloperidol, 45 (66%) patients developed restlessness/agitation (RASS ≥+1) and proceeded to the blinded phase. The primary outcome analysis at 24 hours included 10 (67%) of 15 patients treated with dose escalation, 11 (69%) of 16 patients treated with neuroleptic rotation and 10 (71%) of 14 patients treated with combination therapy. Table 1 shows the characteristics of enrolled patients.

Figure 1. Participant Flow Diagram.

At the time of enrollment, patients were randomised to haloperidol dose escalation, neuroleptic rotation or combination therapy. All enrolled patients were immediately started on a standardized regimen with haloperidol 2 mg every 6 hours intravenously and 2 mg every 1 hour as needed for agitation. Because of the fluctuating nature of delirium, we monitored patients’ RASS every 2 hours until it was at least +1 and required rescue medication in the bedside nurse’s judgement. 23/68 (34%) patients did not develop further agitation until death, discharge or dropout while on open label phase, and thus did not require the study medication.

Table 1.

Patient characteristics at time of enrolment

	Open-label haloperidol only (n=23)*	Received masked treatment*
		Escalation group (n=15)	Rotation group (n=16)	Combination group (n=14)
Age, years	63 (55–74)	65 (61–75)	63 (59–70)	60 (57–74)
Sex
Female	11 (48%)	5 (33%)	6 (37%)	8 (57%)
Male	12 (57%)	10 (67%)	10 (63%)	6 (43%)
Race
White	19 (83%)	10 (67%)	13 (81%)	10 (71%)
Black	2 (9%)	2 (13%)	2 (12%)	2 (14%)
Hispanic	1 (4%)	1 (7%)	1 (6%)	2 (14%)
Other	1 (4%)	2 (13%)	0	0
Education
Some high school or less	0	1 (7%)	0	3 (21%)
Completed high school	6 (26%)	5 (33%)	7 (44%)	4 (29%)
Some college	6 (26%)	4 (27%)	3 (19%)	3 (21%)
Completed college	8 (35%)	4 (27%)	5 (31%)	2 (14%)
Advanced degree	3 (13%)	1 (7%)	1 (6%)	2 (14%)
Cancer type
Breast	1 (4%)	3 (20%)	1 (6%)	1 (7%)
Gastrointestinal	6 (26%)	5 (33%)	7 (44%)	2 (14%)
Genitourinary	5 (22%)	1 (7%)	1 (6%)	2 (14%)
Gynaecological	0	0	0	1 (7%)
Haematological	3 (13%)	0 (0%)	3 (19%)	2 (14%)
Respiratory	4 (17%)	4 (27%)	3 (19%)	3 (21%)
Other	4 (17%)	2 (13%)	1 (6%)	3 (21%)
Metastatic cancer or advanced stage disease	23 (100%)	15 (100%)	16 (100%)	14 (100%)
Karnofsky Performance Status†
10%	1 (4%)	1 (7%)	1 (6%)	0
20%	18 (78%)	5 (33%)	6 (37%)	10 (71%)
30%	4 (17%)	9 (60%)	9 (56%)	4 (29%)
Reason or reasons for acute palliative care unit admission‡
Delirium	20 (87%)	12 (80%)	11 (69%)	6 (43%)
Pain	21 (91%)	13 (87%)	15 (94%)	14 (100%)
Dyspnoea	13 (57%)	11 (73%)	9 (56%)	9 (64%)
Others	4 (17%)	2 (13%)	3 (19%)	4 (29%)
Memorial Delirium Assessment Scale§	26 (20–29)	25 (19–27)	24 (21–29)	28 (19–29)
Medication administered in the 24 h before enrolment
Haloperidol scheduled	23 (100%)	14 (93%)	15 (94%)	14 (100%)
Haloperidol as needed	19 (83%)	15 (100%)	15 (94%)	14 (100%)
Chlorpromazine scheduled	0	0	0	0
Chlorpromazine as needed	5 (22%)	4 (27%)	7 (44%)	7 (50%)
Benzodiazepine scheduled	0	1 (7%)	0	1 (7%)
Benzodiazepine as needed	1 (4%)	1 (7%)	2 (13%)	3 (21%)
Haloperidol doses in the 24 h before enrolment
Scheduled, mg	5 (3–6)	4 (2–6)	5 (2–6)	5 (4–7)
Rescue, mg	2 (1–6)	4 (3–5)	3 (2–6)	3 (2–4)
Number of breakthrough doses	2 (1–3)	3 (2–3)	3 (2–3)	2 (1–2)

Data are median (IQR) or n (%).

^*Unless otherwise specified.

^†A validated assessment of performance status that ranges from 0% (dead) to 100% (normal, no complaints).

^‡Some patients had multiple reasons for admission.

^§A 10-item, clinician-rated assessment scale validated for assessment of delirium in cancer patients23, 24 that examines the level of consciousness, disorientation, memory, recall, attention, disorganised thinking, perceptual disturbance, delusions, psychomotor activity and sleep, assigning a score between 0 to 3, for a total score of between 0 and 30. A score of 13 or higher indicates delirium.

As shown in Figure 2, RASS significantly decreased in all 3 groups within 30 minutes. The mean change in RASS between 0 and 24 hours was −3·6 (95% CI −5, −2·2) in the escalation group, −3·3 (95% CI −4·4,−2·2) in the rotation group and −3.0 (95% CI −4·6,−1·4) in the combination group (Table 2), with no significant difference among the 3 groups (Table 2, P=0·71).

Figure 2. Change in Richmond Agitation Sedation Scale over the first 24 Hours after Study Medication Administration.

The mean RASS scores (data markers) are plotted for each study group starting at time 0 (i.e. immediately before blinded study medication administration) over the next 24 hours, with error bars indicating standard error.

Table 2.

Changes in RASS score

	Escalation group (n=15)	Rotation group (n=16)	Combination group (n=14)	p value*
Primary outcome
RASS change between baseline and 24 h	10, −3·6 (−5·0 to −2·2)	11, −3·3 (−4·4 to −2·2)	10, −3.0 (−4·6 to −1·4)	0·71
Secondary RASS outcomes
RASS scores
RASS change between baseline and 30 min	15, −2·6 (−3·6 to −1·6)	16, −2·4 (−3·0 to −1·8)	14, −2·1 (−3·0 to −1·3)	0·86
RASS change between baseline and 4 h	15, −2·5 (−3·8 to −1·3)	16, −2·9 (−3·6 to −2·3)	14, −2·8 (−3·9 to −1·7)	0·78
RASS change between baseline and 8 h	13, −3·1 (−4·3 to −1·9)	16, −2·7 (−3·5 to −1·9)	14, −3·1 (−4·4 to −1·7)	0·72
RASS score between 0 and −2 at 30 min	15, 8 (53%)	16, 10 (63%)	14, 8 (57%)	0·87
RASS score between 0 and −2 at 4 h	15, 8 (53%)	16, 5 (31%)	14, 7 (50%)	0·41
RASS score between 0 and −2 at 8 h	13, 2 (15%)	16, 7 (44%)	14, 4 (29%)	0·25
RASS score between 0 and −2 at 24 h	10, 2 (20%)	11, 3 (27%)	10, 5 (50%)	0·32
Breakthrough restlessness
RASS ≥1 anytime during the first 4 h	15, 11 (73%)	16, 3 (19%)	14, 7 (50%)	0·009
RASS ≥1 anytime during the first 8 h	15, 12 (80%)	16, 4 (25%)	14, 7 (44%)	0·01
RASS ≥1 anytime during the first 24 h	15, 12 (80%)	16, 7 (50%)	14, 9 (64%)	0·12

Data are n, mean (95% CI); or n, number of observations (%). The number of patients with data available for each analysis is shown. The number of patients with missing data varied because of attrition (eg, death), the specific timing of study assessments (eg, day 1 vs day 3) and the availability of caregivers or bedside nurses. RASS=Richmond Agitation Sedation Scale.

^*Change in study outcomes before and after medication administration between groups was assessed using two-tailed Wilcoxon Rank Sum test for continuous variables and two-tailed Fisher’s exact test for categorical variables.

Secondary analyses showed no significant differences among the 3 groups in the proportion of patients who achieved RASS 0 to −2 within the first 24 hours (Table 2). The rotation group had fewer patients who required dose level increase (appendix p 4). In contrast, the combination group was significantly more likely to require rescue benzodiazepines (appendix p 4).

A majority (60–75%) of patients were perceived by blinded caregivers and nurses to be more comfortable comparing the day before to the day after treatment (appendix p 5). Delirium Experience Questionnaire revealed significant within-group reduction in distress related to psychomotor agitation in all 3 treatment groups as assessed by nurses but not caregivers (appendix p 6). No significant change in MDAS was detected in any group (appendix p 7). There was a significant within-group reduction in caregivers’ rating of patients’ ESAS pain, fatigue, nausea and anxiety in the neuroleptic rotation group and sleep in all 3 groups (appendix p 7).

Patients in the rotation group and combination group had a significant within-group reduction in respiratory rate, systolic and diastolic blood pressure by 24 hours (appendix p 8). CTCAE reporting revealed that hypotension was the most common severe adverse event, occurring in 6 (40%) patients in the escalation group, 5 (31%) in the rotation group and 3 (21%) in the combination group (Table 3). No patients required dose level reduction. One (2%) patient discontinued treatment secondary to potential treatment adverse effect (combination group, Grade 3 akathisia). No patients had worsening of any of the 8 neuroleptic symptoms documented in the UKU questionnaire between day 0 and day 3 (appendix p 9). Fourteen (93%) patients in the escalation group, 16 (100%) patients in the rotation group and 14 (100%) patients in the combination group died within 30 days of the study. There were no treatment-related deaths; all patients died of progressive cancer in this setting. Among the patients who received the blinded study medication, the median overall survival was 62·5 h (95% CI 35·8 h to 74·3 h) with a median follow-up time of 84 h (IQR 35 h, 144 h) and no significant difference among the 3 groups in post-hoc analysis (appendix p 11).

Table 3.

Adverse events

	Escalation group (n=15)				Rotation group (n=16)				Combination group (n=14)
	Grade 1–2	Grade 3	Grade 4	Grade 5	Grade 1–2	Grade 3	Grade 4	Grade 5	Grade 1–2	Grade 3	Grade 4	Grade 5
Akathisia	0	0	0	0	0	0	0	0	0	1 (7)	0	0
Apnoea	0	0	0	0	0	0	0	0	0	1 (7)	0	0
Death, NOS*	0	0	0	14 (93%)	0	0	0	16 (100%)	0	0	0	14 (100%)
Hypotension	0	1 (8%)	1 (8%)	4 (31%)	0	0	2 (13%)	3 (19%)	0	0	1 (7%)	2 (14%)
Hypoxia	0	2 (15%)	0	0	0	0	0	0	0	0	0	0
Sinus tachycardia	0	0	0	0	1 (6%)	0	0	0	2 (14%)	0	0	0

Data are n (%). Grade 1–2 adverse events occurring in at least 10% of patients in any group and the corresponding grade 3–5 events, or grade 3–5 events occurring in at least 10% of patients in any group and the corresponding grade 1–2 events. NOS=not otherwise specified. Adverse events assessed by Common Terminology Criteria for Adverse Events, version 4.03.

^*All deaths within 30 days of study were documented.

After 24 hours of treatment, 8 (36%) blinded caregivers and 7 (32%) nurses correctly guessed the identity of treatment, suggesting that blinding was likely intact.

Worst-case sensitivity analysis of the primary outcome post-hoc revealed similar findings with a significant mean change in RASS between 0 and 24 hours in each of the 3 groups (rotation −2·3 [95% CI −3·3, −1·2], escalation −2·4 [95% CI −3·7, −1·1] and combination −2·1 [95% CI −3·5, −0·8]) and no significant difference among the 3 groups.

In post-hoc analysis, significantly fewer patients in the rotation group had breakthrough restlessness (RASS ≥+1) in the first 4 hours and in the first 8 hours relative to the escalation and combination groups (Table 2, appendix p 10). Pairwise comparison showed that the rotation group had significantly fewer patients with breakthrough restlessness (RASS ≥+1) than the escalation group in the first 4 hours (rotation vs. escalation −54·6% [95% CI −84·0%, −25·2%]; rotation vs. combination −31·3% [95% CI −63·7%, 1·2%]) and in the first 8 hours (rotation vs. escalation −55·0% [95% CI −84·3%, −25·7%]; rotation vs. combination −25·0% [95% CI −58·7%, 8·7%]). The rotation group also had fewer patients who required dose level increase than the combination group (rotation vs. escalation −20·4% [95% CI −45·7%, 4·9%]; rotation vs. combination −43·8% [95% CI −72·5%, −15·0%]).

Discussion

In this preliminary trial of hospitalized cancer patients with delirium and refractory agitation, the strategies of neuroleptic rotation, haloperidol dose escalation and combination therapy given at approximately equivalent doses were found to be associated with a significant reduction in RASS at 24 hours and increased comfort as perceived by caregivers and bedside nurses. Although the mean RASS scores did not differ significantly, patients in the rotation group had fewer episodes of breakthrough restlessness, required fewer rescue doses, and were less likely to require dose level increase. Taken together, our study supports the use of haloperidol and/or chlorpromazine in the management of terminal restlessness and provides preliminary evidence for neuroleptic rotation in patients with refractory restlessness.

Clinical trials on agitated delirium are highly complex logistically but possible. This double-blind, double dummy trial examines a clinically relevant primary outcome (i.e. refractory restlessness/agitation) in a unique population (i.e. terminal delirium). Our study also used high doses of neuroleptics (e.g. scheduled haloperidol starting at 12 mg/day and up to 48 mg/day at dose level 4). We used relatively high doses of neuroleptics in this study because patients continued to experience restlessness/agitation despite multiple measures including repeated trial of lower doses of haloperidol (≤8 mg/day) and the time sensitivity of the situation. We found that repeated administration of high dose haloperidol and/or chlorpromazine was associated with a rapid reduction of restlessness assessed by RASS, in addition to overall impression of agitation control and improved comfort by both caregivers and bedside nurses. Unlike deep palliative sedation, few patients in this study were deeply sedated with RASS of −5.

The primary outcome was the magnitude of change in RASS. In the terminal agitation setting, a greater reduction in RASS was associated with increased perceived comfort.²⁸ However, the magnitude of change only provides a partial picture and deep sedation (RASS −5) may not always be desirable because families often want to see if they could still communicate with the patient, if at all possible.²⁹ Thus, secondary outcomes such as proportion of patients with breakthrough restlessness/agitation and the proportion of patients who required rescue medications or dose level escalation represent complementary, practical and clinically meaningful measures. Based on our findings, future trials may consider using breakthrough restlessness and/or need for dose escalation as primary outcomes.

Without a placebo-only control group, it is not possible to tell if the reduction in RASS scores observed in this study was because of the pharmacologic effect of high dose neuroleptics or simply due to the fluctuating nature of agitated delirium, regression to mean, and/or effectiveness of non-pharmacologic measures. A placebo-only group was considered but ultimately not included in this study because of ethical considerations. Specifically, patients were in the last days of life with very little time to maximize comfort and that all patients who required blinded medications had already suffered from multiple days of restlessness/agitation despite standard non-pharmacologic measures and a trial of both standard and open-label haloperidol.

The very rapid reduction of RASS within 30 minutes of medication administration was consistent with a pharmacologic effect with intravenous neuroleptics. Our previous double-blind randomised clinical trial examining a single dose of haloperidol with or without lorazepam conducted in the same PSCU with almost identical eligibility criteria also led to a rapid reduction in RASS.¹⁴ Although not a direct comparison, this haloperidol alone group could provide some useful historical data to put the current findings in context. Among 29 patients who received a one-time 2 mg dose of intravenous haloperidol, the mean RASS score reduction was −2·3 (95% CI −2·9, −1·6) points at 8 hours and only 37% of patients were perceived to be more comfortable by caregivers the next day.¹⁴ In contrast, the haloperidol escalation group in the current study involved an initial 4 mg dose of intravenous haloperidol followed by repeated dosing with a chance for up-titration. The RASS score reduction in this group was −3·1 (95% CI −4·3, −1·9) at 8 hours and 62% were perceived to be more comfortable by caregivers the next day, with similar effects observed in the other two study groups. Taken together, these observations suggest that neuroleptics likely played a role in RASS reduction on top of non-pharmacologic measures or natural fluctuations of delirium, and that a more intensive dosing schedule may be associated with better control of terminal restless/agitation. However, the preliminary nature of our data precludes any definitive comparisons and further research is warranted. Future studies are also needed to compare lorazepam and chlorpromazine and to identify the dosing schedule to achieve optimal control of agitation.

We found no difference among the 3 groups in our primary outcome of mean RASS score over 24 hours; however, between-group comparison was only a pre-planned secondary objective and for hypothesis-generation only. At this time, all 3 strategies appeared relatively effective for management of agitation and there is inadequate evidence to suggest that one strategy was superior over another. However, secondary outcome analyses suggested that rotation to chlorpromazine may be associated with better control of agitation. In the setting of terminal delirium, time is of the essence and steady control of restlessness was critical. Larger studies are needed to further examine the benefit of chlorpromazine.

Interestingly, we did not observe a synergistic effort nor gradient effect in the combination group; however, this was only a secondary outcome and warrants further research.

We did not detect significant extrapyramidal adverse effects related to neuroleptic use despite the high doses. Both rotation and combination group had a significant reduction in blood pressure; this was not unexpected and had been reported in the literature. Although a few deaths were attributed related to study treatment, assigning attribution was challenging in this setting given that patients were actively dying and hypotension was also commonly seen in the last days of life.³⁰ Importantly, the survival of approximately 3 days was similar to a previous study in which a single dose of haloperidol was given.¹⁴

This study has several limitations. First, our study was conducted at a comprehensive cancer centre by an academic palliative care team and with a selective patient population. Thus, the findings may not be generalizable to other populations. Second, we selected RASS +1 as a cutoff for study medication administration and used neuroleptics at high doses. Although these practices are acceptable for refractory restlessness/agitation in the terminal delirium setting, they may not be applicable to other patient populations with a different prognosis. Third, we only assessed intravenous neuroleptics but not subcutaneous administration. Fourth, the small sample size limits between-group comparisons. Fifth, we did not include a placebo-only group despite the double-dummy design because of the distressing nature of refractory agitation. Sixth, there were many secondary outcomes and these findings were considered hypothesis-generating only.

This study highlights that high dose neuroleptics may still have a role in the management in delirium by providing rapid reduction of restlessness and agitation in the last days of life. It provides preliminary data for neuroleptic rotation as a promising treatment strategy that warrants further investigation.

Research in context.

Evidence before this study

This randomised clinical trial was designed to examine the effect of haloperidol and chlorpromazine on terminal agitation. Prior to this, systematic reviews on the role of neuroleptics in delirium among hospitalized patients concluded that neuroleptics did not alter the duration of delirium, delirium severity, delirium symptoms, nor length of hospital stay. However, few clinical trials have specifically recruited patients with agitated delirium and examined agitation as a primary outcome, the main reason to prescribe neuroleptics. We searched PubMed using the terms “delirium” and “chlorpromazine” and restricted results to those of prospective randomised clinical trials from database inception to May 8, 2020. There was only one study comparing chlorpromazine, haloperidol and lorazepam in patients with acquired immune deficiency syndrome who developed delirium. However, the level of agitation was not a study outcome in that study. Furthermore, no randomised trials have studied the concept of neuroleptic rotation.

There is also a paucity of studies in the terminal delirium setting in which persistent restlessness is particularly common and extremely distressing. Before this study, there was only one randomised clinical trial in the terminal agitation setting that compared the effect of a single dose of haloperidol with or without lorazepam on refractory restlessness. This relative lack of evidence contributes to significant confusion regarding the role of neuroleptics in patients with terminal agitation.

Added value of this study

Studies involving patients in the last days of life are challenging and few have been completed. To our knowledge, this is the first randomized clinical trial to examine the effect of haloperidol and chlorpromazine on patients with delirium focusing on agitation as a primary outcome and the first randomized trial to test the concept of neuroleptic rotation. Our study showed that haloperidol and chlorpromazine given at high doses may reduce agitation rapidly and improve perceived comfort level. Despite the high doses of neuroleptics, extrapyramidal adverse effects were rarely observed in this setting. Neuroleptic rotation appears to be a promising strategy that warrants further investigation.

Implications of all the available evidence

Although the literature suggests that neuroleptics have a limited effect on the duration of delirium and delirium severity, our study highlights that they may have a pharmacologic effect on reducing agitation. This beneficial effect is of clinical significance, particularly in the terminal agitation setting in which palliation is of paramount importance.