Low-dose clonidine infusion to improve sleep in postoperative patients in the high-dependency unit. A randomised placebo-controlled single-centre trial

Abstract

Purpose

Dexmedetomidine increases sleep and reduces delirium in postoperative patients, but it is expensive and requires a monitored environment. Clonidine, another $\alpha$2-agonist, is cheaper and is used safely for other purposes in wards. We assessed whether clonidine would improve sleep in postoperative high-dependency unit (HDU) patients.

Methods

The Clonidine at Low dosage postoperatively to Nocturnally Enhance Sleep (CLONES) study was a double-blind, placebo-controlled, parallel-group pilot effectiveness randomised trial involving adult elective surgery HDU patients in a single academic hospital. Patients received clonidine 0.3 μg/kg/h or saline placebo on the night of surgery. The primary outcome was total sleep time measured using a consumer actigraphy/photoplethysmography device.

Results

Of the 83 randomised patients, three had no data available, leaving 80 (39 clonidine, 41 placebo) in the intention-to-treat analysis, modified for missing data. Median patient ages of the groups were similar (61 and 59 years), as were other baseline characteristics. Clonidine patients had a mean of 100.8 (95% confidence interval [CI] 38.2–163.4) minutes (p = 0.002) longer total sleep time (mean 497.2 vs. 396.4 min) and reported better sleep overall. Delirium was only observed in one patient prior to study drug infusion, and none at the end of the study. Safety outcomes were not different. Four clonidine patients had their medication ceased due to bradycardia and hypotension that required no additional treatment.

Conclusion

Among postoperative elective surgical patients admitted to HDU, low-dose non-titrated clonidine, compared to placebo, was associated with longer and subjectively better-quality sleep.

Supplementary Information

The online version contains supplementary material available at 10.1007/s00134-024-07619-w.

Take-home message

In this randomised trial, postoperative high-dependency unit patients who received a low-dose non-titrated clonidine infusion (in comparison to placebo) slept longer, and the next morning reported a better-quality sleep

Introduction

Poor sleep is common after surgery, especially in patients admitted to the intensive care unit (ICU) or high-dependency unit (HDU)[1]. Sleeping poorly is unpleasant and has negative associations, including more postoperative pain, cardiovascular events, and longer hospital stays [2]. Sleep deprivation is also a risk factor for delirium [3, 4], the most common surgical complication in elderly patients [5]. Postoperative delirium is associated with prolonged hospital stay, long-term cognitive decline, other major complications, and higher mortality [6].

Several non-pharmacological approaches improve postoperative sleep, including multi-modal opioid-sparing analgesia, reducing nighttime interventions, and noise- and light-reduction [2]. However, these strategies can be resource intensive and despite attempts at implementation, many postoperative patients still experience poor sleep [7].

Many medication classes have been used to improve sleep in hospitalised patients, including antipsychotics, statins, melatonin, dexamethasone, gabapentin, and benzodiazepines. A recent meta-analysis concluded that none significantly improves sleep compared to placebo [8].

α2 agonists act on the locus coeruleus to cause sedation and anxiolysis, on the spinal cord to reduce pain conduction [9], and appear to hold promise for the treatment of insomnia, inducing sleep with a similar architecture to normal physiological sleep [9]. Two α2 agonists are in common clinical use: dexmedetomidine and clonidine. Elderly postoperative extubated patients randomised to low-dose (0.1 μg/kg/h), non-titrated dexmedetomidine on their first postoperative night had longer sleep, and subjectively improved sleep quality, compared to controls given placebo [10]. The same dexmedetomidine regimen in 700 elderly postoperative ICU patients reduced delirium compared to placebo [11]. Low-dose (25 μg enteral) clonidine increased the duration of rapid eye movement (REM) sleep compared to baseline in healthy volunteers [12]. Sleep latency, duration, and subjective sleep quality were better with clonidine than with the cyclopyrrolone hypnotic zopiclone in a crossover study of 160 patients [13].

Clonidine would have advantages over dexmedetomidine if it were effective in promoting sleep or preventing delirium, including clonidine’s lower cost in many countries, accessibility of oral and transdermal formulations with high bioavailability in addition to the parenteral form, longer half-life, and familiarity outside a critical care environment. No studies have investigated clonidine for the purpose of promoting postoperative sleep or reducing delirium.

Accordingly, we sought to test whether clonidine is effective at improving sleep and preventing delirium in postoperative HDU patients, in a manner similar to that demonstrated for dexmedetomidine. The results of this study were presented at the 2024 Australian and New Zealand College of Anaesthetists Annual Scientific Meeting.

Methods

Study design and setting

The Clonidine at Low dosage postoperatively to Nocturnally Enhance Sleep (CLONES) study was a 1:1 parallel-group randomised, placebo-controlled, double-blinded pilot effectiveness study comparing intravenous (IV) infusions of clonidine and placebo administered to postoperative non-cardiac surgical patients admitted following extubation to the surgical HDU of a single tertiary/quaternary academic hospital. This HDU has sleep-promotion policies including nocturnal noise and light reduction and minimisation of nursing interventions. No other systematic approach to delirium reduction, such as the A2F bundle [14], was used in these low-risk extubated patients during the study period. Patients were allocated nursing staff at a ratio of at least 1:2, and a specialist intensivist or senior critical care trainee doctor was constantly present. Study patients could not receive supplemental α₂ agonists unless treating doctors judged this in their best interest; if so, study medication was ceased but data collection continued. This would not have been treated as a protocol violation, as clinicians were empowered by the study protocol to use any treatments they believed to be in their patients’ best interests. Such a patient would be included in the primary “intention to treat” analysis, but if the study medication had been given for less than 2 h, the patient would not be included in the “treatment received” analysis. The prospectively registered trial protocol (ACTRN12619000669190/U1111-1229-9703), approved by the health service Human Research Ethics Committee, appears in electronic supplementary material (ESM). A Data Safety Monitoring Committee reviewed all adverse events. There was no interim analysis.

Measures taken to minimise bias

A computer-generated randomisation scheme was uploaded to the randomisation module in the online case report form (REDCap [15]). Enrolled patients were randomised to study treatment group by an unblinded research coordinator who prepared syringes of study medication from the general ICU stock of clonidine and saline. ICU pharmacists were not directly involved in the study but might have been able to guess patient allocation from the rate of utilisation of clonidine on certain days. Allocation concealment was preserved in the recruitment and randomisation process.

Simple stratified randomisation was based upon two variables:

1. Intraoperative or post-anaesthesia care unit administration of IV or oral clonidine up to a total dose of 1 μg/kg (yes/no)

2. Expected frequent awakenings, i.e. one or both of the following: known or suspected obstructive sleep apnoea; two-hourly or more frequent neurological or neurovascular monitoring (yes/no)

To preserve blinding of patients, treating clinicians, and investigators, both clonidine and saline placebo were given by continuous infusion. Infusion equipment marked only with a registry number was prepared away from the bedside by a research coordinator.

Participants

Patients expected to be admitted to the HDU following surgery were informed of the study in the pre-admission clinic or ward at least one day prior to their operation, with verbal informed consent sought to participate. Written informed consent was confirmed on the morning of surgery. When consenting patients arrived in the HDU, study eligibility was re-confirmed, agreement from the treating critical care doctor was obtained, and randomisation occurred. Study drug infusion commenced at 08 PM.

Patients were excluded if they were < 18 years old; pregnant or breastfeeding; had a bradyarrhythmia identified at the time of HDU admission; were admitted to the ICU (not HDU); had been prescribed an α2 agonist for chronic use preoperatively; had premorbid advanced dementia requiring professional nursing care; had previously been enrolled in a clinical trial of a sedative, antipsychotic or anti-delirium medication during this admission; had a known previous adverse reaction to α2 agonists; had end-stage kidney disease or use of dialysis; had comorbidities that would interfere with sleep measurement; were planned for discharge home directly from the HDU; if research coordinators were unavailable for enrolment or completion of study assessments. Intraoperative use of small doses of clonidine as an adjunct to other analgesic medications is common in our hospital. As excluding such patients would have threatened trial viability, this was not an exclusion criterion.

Interventions

Study medication was given by 5 mL/h IV infusion. Patients randomised to clonidine received 0.3 μg/kg/h (up to a maximum dosing weight of 100 kg) with no loading dose. Placebo 0.9% saline was administered at the same rate as the active treatment. Infusions continued until 6 AM on the first postoperative day.

While receiving the study medication, any other element of standard care could be administered. Only supplemental α2 agonists were prohibited. Treating clinicians could interrupt study medication if they had concerns over adverse effects, and they could ask the study Chief Investigator (or designate) to unblind the patient (facilitated by the unblinded research co-ordinator) if they thought this was required for patient safety. Patients could also request study medication be stopped. Unless patients requested their data be deleted, data continued to be recorded, and the patient was retained in the intent-to-treat analysis.

Outcomes

The primary outcome was total sleep time between 8 PM and 6 AM measured using a wrist-worn Fitbit Alta HR (Fitbit, San Francisco, CA, USA) actigraphy and photoplethysmography device.

Secondary effectiveness outcomes included phase-based (REM, Light, Deep) sleep time measured using the Fitbit Alta HR device; time spent awake; subjective sleep quality scored by patients and the bedside nurses using the Richards-Campbell Sleep Questionnaire (RCSQ); incidence of delirium identified using the Confusion Assessment Method for the ICU (CAM-ICU [16]) at the time of HDU admission and 7–9 AM on the first postoperative day; presence of agitation identified using Richmond Agitation Sedation Scale (RASS) [17] at the same time points; use of sedatives/antipsychotics; and nursing reports of delirium or agitation.

Secondary safety outcomes included episodes of hypotension or bradycardia requiring study medication cessation; incidence and duration of fever [18]; incidence of Serious Adverse Events (SAEs) or Suspected Unexpected Serious Adverse Reactions (SUSARs); HDU mortality; and HDU and hospital length of stay.

Statistics and power analysis

Sample size was calculated based on a mean total sleep time of 407.1 min (standard deviation [SD] 50.1 min) reported in a healthy population of middle-aged adults [19]. To show a 30-min difference in total sleep time between treatment groups (approximately half that reported in the study of dexmedetomidine in extubated postoperative patients [10]) using a two-tailed Student’s t test with two independent groups, α 0.05, power 0.90, 120 patients were required.

The trial was terminated prematurely on 14 September 2020 after only 83 of the planned 120 patients had been randomised. Funding to support patient recruitment had been exhausted, and the coronavirus disease 2019 (COVID-19) pandemic had substantially reduced the number of patients undergoing elective surgery with planned postoperative HDU care. Ongoing uncertainty and redeployment of staff forced cessation of the trial. No data analysis had occurred prior to this decision. The Data Safety and Monitoring Committee and the responsible Human Research Ethics Committees were informed of the requirement to discontinue the trial.

The primary analyses were conducted on an intention-to-treat basis, modified for patients missing all study data. No data imputation was performed. Continuous outcomes were inspected for normality in the entire study group and, where appropriate, reported as either mean or median of differences with 95% confidence intervals (CIs) and compared using either Mann–Whitney tests or Student’s t tests. Categorical outcomes were compared using Fisher exact tests and reported as differences in proportions (95% CIs). The primary outcome was additionally adjusted for the two stratification variables using linear regression. All statistical analyses were performed using Stata version IC 17 (StataCorp) with a two-sided P < 0.05 considered significant. No adjustment was made for multiple comparisons, so the prespecified secondary outcomes presented should be considered exploratory.

No interim effectiveness analysis was planned. A blinded interim safety analysis was allowed in the Data Safety Monitoring Committee plan if thought to be required.

Results

From 20 December 2018 to 14 September 2020, we randomised 83 patients out of 1169 patients scheduled for elective surgery followed by planned HDU admission (Fig. 1). The most common reasons for inability to participate in the trial were unavailability of a research co-ordinator, inability to contact the patient preoperatively, and refusal of consent. Only 9.5% of potentially eligible patients had a relative clinical contraindication to clonidine. Three randomised patients had no sleep data recorded due to study protocol errors, and were excluded from the analysis. Three out of 39 (8%) clonidine patients and one of 41 (2%) placebo patients did not receive the allocated intervention for at least two hours. A sensitivity analysis was conducted by treatment received, excluding these four patients. No patients received open-label α2 agonists, and no clinician requested that a patient allocation be unblinded.

Fig. 1.

Patient flow diagram of the CLONES trial

Study patients had low Acute Physiology and Chronic Health Evaluation (APACHE) II scores (Table 1). Almost all had undergone general anaesthesia, and approximately 30% had received small doses (45–60 μg) of clonidine intraoperatively. Almost all (92–95%) were prescribed only oral analgesia for the postoperative period. Approximately two thirds received opioids, typically at a low dose (median 75–160 μg fentanyl equivalents [20] over the HDU admission), with no significant differences between study groups (Table 1).

Table 1.

Baseline patient characteristics and process of care measures

	Clonidine n = 39	Placebo n = 41	Difference (clonidine–placebo) (95% CI); level of significance
Age, years, median (IQR)	61 (49–70)	59 (52–67)
Male, n (%)	16 (41)	23 (56.1)
APACHE II Acute Physiology Score at HDU admission, median (IQR)	6 (4–7)	6 (5–8)
APACHE II total score at HDU admission, median (IQR)	10 (8–12)	9 (8–11)
APACHE II death risk at HDU admission, median (IQR)	0.07 (0.04–0.1)	0.07 (0.04–0.11)
APACHE II surgical category, n (%)
Cardiovascular	2 (5)	3 (7)
Neurosurgery incl. neurotrauma	12 (31)	14 (34)
Gastrointestinal	11 (28)	15 (37)
Other	14 (36)	9 (22)
Primary anaesthetic type, n (%)
General	37 (95)	40 (98)
Neuraxial	1 (3)	0 (0)
Other regional	1 (3)	1 (2)
Surgery duration (hh:mm), median (IQR)	03:13 (02:25–03:53)	02:50 (02:13–04:09)
Time of day of HDU admission (hh:mm), median (IQR)	16:00 (14:35–17:25)	15:50 (14:10–17:35)
Expected frequent awakenings in the HDU, n (%)	23 (59)	21 (51)
Clonidine given intraoperatively, n (%)	11 (28)	12 (29)
In patients given clonidine intraoperatively, total dose administered, μg median (IQR)	60 (30–75)	45 (38–60)
Clonidine given in PACU, n (%)	1 (3)	2 (5)
In patients given clonidine in PACU, total dose administered, μg median (IQR)	60 (60–60)	48 (45–50)
Clonidine given either intraoperatively or in PACU, n (%)	11 (28)	14 (34)
Postoperative analgesia type, n (%)
PCA	3 (8)	2 (5)
PCEA	0 (0)	0 (0)
Epidural	1 (3)	0 (0)
Regional	0 (0)	0 (0)
Oral (any medication)	36 (92)	39 (95)
Study drug duration, median (95% CI), hours	9.8 [8.6–10]	10 [9.5–10]	− 0.15 (− 0.48 to 0.18) p = 0.191a
Any opioids given in HDU, number (%) of patients	22/39 (56)	27/41 (66)	− 10% (− 30% to 10%) p = 0.492b
Opioid use in HDU, median (95% CI), oral fentanyl equivalents, μg [20]	75 [0–320]	160 [0–420]	− 85 (− 296 to 126) p = 0.59a

APACHE II Acute Physiology and Chronic Health Evaluation II, CI confidence interval, IQR interquartile range, HDU high-dependency unit, PACU post-anesthesia care unit, PCA patient-controlled analgesia, PCEA patient-controlled epidural analgesia

^aMann–Whitney U. Treatment group results reported as median [IQR]; difference of medians (95% CI), p value

^bFisher’s exact test. Treatment group results reported as frequency/total (percentage)−difference of proportions (95% CI), p value

Primary outcome

Compared with the placebo group, patients in the clonidine group slept significantly longer (mean difference 100.8 (95% CI 38.2–163.4) minutes; p = 0.002, Table 2 and Fig. 2). This difference remained significant when adjusted for the two stratification variables (p = 0.001; data not shown).

Table 2.

Primary and secondary sleep and delirium effectiveness outcomes; modified intention to treat analysis

	Clonidine	Placebo	Difference (clonidine–placebo) (95% CI); level of significance	Test
Primary outcome
Total sleep time, mean (95% CI), mins	497.2 (467.2–527.1)	396.4 (341.3–451.5)	100.8 (38.2–163.4); p = 0.002	[a]
Secondary effectiveness outcomes
Sleep
Awakenings, mean (95% CI), number of events	14 (10.6–17.4)	9.5 (7.1–11.8)	4.5 (0.5–8.5); p = 0.027	[a]
Sleep fragmentation index, mean (95% CI), awakenings/h	1.7 (1.3–2.2)	1.6 (1.1–2.1)	0.1 (− 0.5–0.8); p = 0.75	[a]
REM sleep time, mean (95% CI), mins	59.8 (47.4–72.2)	58.4 (49.4–69.3)	− 0.9 (− 19.4–21.7); p = 0.93	[a]
Light sleep time, mean (95% CI), mins	385.2 (350.5–420)	297.5 (245.7–349.3)	87.7 (28.7–146.7); p = 0.005	[a]
Deep sleep time, mean (95% CI), mins	35.8 (24.4–47.2)	42.4 (27.8–56.9)	− 6.5 (− 24.3–11.1); p = 0.46	[a]
Awake time, mean (95% CI), mins	68.8 (58.7–78.9)	55.9 (46–65.9)	12.9 (− 1–26.8); p = 0.07	[a]
Sleep efficiency (fraction of time in bed spent asleep), mean (95% CI), proportion	0.87 (0.85–0.89)	0.84 (0.78–0.89)	0.02 (− 0.03–0.09); p = 0.30	[a]
RCSQ total score Q1–5 (patient), mean (95% CI), mm	62.2 (54.5–69.7)	48.3 (38.9–57.7)	13.9 (1.8–25.9); p = 0.03	[a]
RCSQ Q1 depth score (patient), mean (95% CI), mm	53.6 (44.5–62.6)	39.3 (29.3–49.3)	14.3 (0.9–27.6); p = 0.037	[a]
RCSQ Q2 latency score (patient), mean (95% CI), mm	62.8 (52.6–73)	53.2 (41.1–64.2)	9.6 (− 5.3–25.5); p = 0.20	[a]
RCSQ Q3 awakenings score (patient), mean (95% CI), mm	62.3 (51.7–72.9)	45.9 (35.9–55.9)	16.4 (2.1–30.7); p = 0.02	[a]
RCSQ Q4 return to sleep score (patient), mean (95% CI), mm	72.4 (63.2–81.7)	54.3 (43.7–64.9)	18.1 (4.2–32.1); p = 0.01	[a]
RCSQ Q5 quality score (patient), mean (95% CI), mm	59.5 (49.3–69.7)	48.9 (36.4–60.3)	10.6 (− 4.6–25.8); p = 0.17	[a]
RCSQ Q6 noise score (patient), mean (95% CI), mm	57.3 (47.5–67.1)	61.1 (51.0–71.1)	− 3.8 (− 17.6–10.1); p = 0.59	[a]
RCSQ total score Q1–5 (nurse rated), mean (95% CI), mm	62.6 (55.7–69.6)	53.1 (45.1–61)	9.6 (− 0.9–20.1); p = 0.07	[a]
RCSQ Q1 depth score (nurse rated), mean (95% CI), mm	52.3 (44.2–60.4)	45.6 (36.7–54.6)	6.7 (− 5.2–18.6); p = 0.27	[a]
RCSQ Q2 latency score (nurse rated), mean (95% CI), mm	69.9 (62.3–77.5)	56.9 (47.4–66.4)	13.1 (1–25.2); p = 0.03	[a]
RCSQ Q3 awakenings score (nurse rated), mean (95% CI), mm	60.7 (51.1–70.3)	52.4 (43.9–60.8)	8.3 (− 4.2–20.9); p = 0.19	[a]
RCSQ Q4 return to sleep score (nurse rated), mean (95% CI), mm	71.2 (63.1–79.4)	60.0 (51.2–68.9)	11.2 (-0.7–23.1); p = 0.06	[a]
RCSQ Q5 Quality Score (nurse rated), mean (95% CI), mm	59.1 (50.2–68.1)	49.5 (41.2–57.8)	9.6 (− 2.4–21.6); p = 0.11	[a]
RCSQ Q6 noise score (nurse rated), mean (95% CI), mm	44.1 (35.9–52.2)	46.7 (38.5–54.8)	− 2.6 (− 14.0–8.8); p = 0.65	[a]
Delirium
Delirium at start of study. number (%) of patients	1/39 (3%)	0/41 (0%)	(− 2.3%–8.3%); p = 0.487	[c]
Delirium at end of study. number (%) of patients	0/39 (0%)	0/41 (0%)		[d]
RASS at start of study. median (95% CI) score	0 [0–0]	0 [0–0]	(0–0); p = 0.710	[b]
RASS at end of study. median (95% CI) score	0 [0–0]	0 [0–0]		[d]
Antipsychotics administered PRIOR to study, number (%) of patients	0/39 (0%)	0/41 (0%)		[d]
Antipsychotics administered DURING study, number (%) of patients	0/39 (0%)	1/41 (2%)	(− 6.3%–2.3%); p = 1.000	[c]
Sedatives administered PRIOR to study, number (%) of patients	0/39 (0%)	0/41 (0%)		[d]
Sedatives administered DURING study, number (%) of patients	0/39 (0%)	0/41 (0%)		[d]
Nursing reports of delirium, number (%) of patients	0/39 (0%)	0/41 (0%)		[d]
Nursing reports of agitation, number (%) of patients	0/39 (0%)	0/41 (0%)		[d]

CI confidence interval, RASS Richmond Agitation-Sedation Scale, RCSQ Richards-Campbell Sleep Questionnaire, REM rapid eye movement

Statistical tests used:

[a] Treatment group results reported as mean (95% CI); mean of differences (95% CI); t test with equal variances p value)

[b] Mann–Whitney U. Treatment group results reported as median (IQR), median differences (95% CI); p value

[c] Fisher's exact test. Treatment group results reported as frequency/total (percentage), 95% CI of difference in proportion; result reported as p value

[d] No statistical test performed due to lack of data/variation

Fig. 2.

Primary outcome: total sleep time, the amount of sleep (minutes) that the patient had on the night of surgery as recorded by the Fitbit device

Secondary outcomes

Patients who received clonidine spent longer in light sleep and reported better sleep overall, as quantified by the total Richards-Campbell Sleep Questionnaire (RCSQ) score. Significant differences in elements of this score included patient self-assessment of deeper sleep, fewer awakenings, and ease of returning to sleep, all better with clonidine. Nurses observed a difference in sleep latency, with clonidine patients falling asleep more rapidly. Paradoxically, clonidine patients had more awakenings, perhaps because they were asleep for longer and so had more opportunity.

Delirium was observed in only one patient prior to study drug infusion. By the end of study drug infusion, no patients had delirium.

There were no significant differences in the incidence of any of the measured safety outcomes (Table 3). No SAEs or SUSARs (as defined by the Australian Therapeutics Goods Administration and in the study protocol) were reported. Median HDU and hospital lengths of stay were no different. Four clonidine patients (10%) had their medication ceased due to bradycardia and hypotension. Although not statistically higher than the placebo group, this is notable given the known adverse effect profile of clonidine. It might be that a statistically significant difference in incidence was not observed because of the low study number combined with the low incidence of these effects. In all four cases, the bradycardia and hypotension resolved without the need for additional treatment.

Table 3.

Safety outcomes

	Clonidine	Placebo	Difference (clonidine–placebo) (95% CI); level of significance	Test
Incidence of fever > 38 °C, number (%) of patients	0/39 (0)	1/41 (2)	(− 6.3–2.3); p = 1.000	[a]
Time patient was febrile > 38 °C, mean (95% CI) hh:mm	–	02:00 [02:00–02:00]	–	[b]
Transfer to ICU due to deterioration, number (%) of patients	0/39 (0)	1/41 (2)	(− 6.3–2.3); p = 1.000	[a]
HDU length of stay, median (95% CI), days	0.9 [0.8–1]	0.9 [0.8–1]	− 0.02 (− 0.09–0.06); p = 0.89	[c]
Hospital length of stay, median (95% CI), days	3.4 [2.2–6.4]	2.4 [2.2–4.3]	0.95 (− 0.76–2.68); p = 0.46	[c]
Any SAEs or SUSARs, number (%) of patients	0/39 (0)	0/41 (0)		[d]
Study drug ceased for hypotension, number (%) of patients	4/39 (10)	3/41 (7)	(− 9–15); p = 0.709	[a]
Study drug ceased for bradycardia, number (%) of patients	4/39 (10)	0/41 (0)	(0.5–19); p = 0.052	[a]

CI confidence interval, ICU intensive care unit, HDU high-dependency unit, SAE serious adverse event, SUSAR suspected unexpected serious adverse reaction

Statistical tests used:

[a] Fisher's exact test. Treatment group results reported as frequency/total (percentage), 95% CI of difference in proportion; result reported as p value

[b] No statistical test performed due to lack of data/variation

[c] Mann–Whitney U. Treatment group results reported as median (IQR), median differences (95% CI); p value

[d] No statistical test performed due to lack of data/variation

When all analyses were repeated excluding the one patient who did not receive study medication because they refused an intravenous cannula and the three patients whose clinicians terminated study medication after < 2 h due to lack of equipoise (all due to bradycardia)—i.e. a “treatment received as intended” analysis—there were no qualitative differences in any of the between-group comparisons.

Discussion

In this trial involving elective surgery patients planned to be admitted to an HDU for postoperative monitoring, clonidine increased the time spent asleep on the first postoperative night by approximately 100 min. Patients who received clonidine reported having had a better night’s sleep, and were observed by their bedside nurse to return to sleep more quickly after awakenings. No effect on delirium was able to be observed in this low-risk patient cohort. The only reported adverse effects from clonidine were bradycardia and hypotension, which were uncommon and required only cessation of the study medication.

The results of this study are consistent with several trials of dexmedetomidine following surgery. In a similar design to our trial, a study of low-dose (0.1 μg/kg/h), non-titrated dexmedetomidine in 700 elective surgical patients ≥ 65 years admitted to a Beijing ICU [11] found a reduction in postoperative delirium incidence from 23% to 9%. While the mean APACHE II score on admission, 10.2–10.6, was similar to our study, just over 50% of those patients were intubated at study enrolment, and approximately 75% used patient-controlled opioids, perhaps explaining the higher incidence of delirium. Duration of sleep was not measured, but subjective sleep quality was significantly better in the dexmedetomidine group on postoperative days 1–3, consistent with our more detailed sleep results. In a similar study, of 61 extubated ICU postoperative patients, those randomised to 0.1 μg/kg/h dexmedetomidine (compared to placebo) on the first postoperative night had longer deep sleep, increased sleep efficiency, and better subjective sleep quality [10]. A 100-patient trial of predominantly mechanically ventilated ICU patients without delirium at enrolment found low-dose (0.2 μg/kg/h) nocturnal dexmedetomidine, compared to placebo, resulted in less than half the incidence of delirium although no difference in subjective sleep quality [21]. Others have found a nocturnal postoperative dexmedetomidine bolus [22] or infusion [23] compared to placebo-reduced delirium [22, 23] following cardiac surgery, but neither of these trials assessed sleep as an outcome. No trial has assessed clonidine as we have done in this study.

In this study, clonidine had no observable analgesic effect. Clonidine is well recognised as an effective analgesic adjunct for postoperative pain, with dose-dependent opioid sparing effects optimal at 3 μg/kg bolus followed by 3 μg/kg/h infusion [24]. Our infusion of 0.3 μg/kg/h without a preceding bolus is therefore likely to have been insufficient to produce maximal analgesia, but avoided the hypotension of higher doses [24]. Numerically fewer patients in the clonidine group required opioids (56% vs 66%), and the median does was substantially lower (75 mcg vs 160 mcg fentanyl equivalents); the absence of a statistically significant effect might alternatively have been imprecision due to small sample size.

The dose selected for our study was based on both safety and extrapolation of likely effectiveness. Australian guidelines for clonidine as an anaesthesia adjunct in place when the study was designed [25] recommended intravenous loading with 1.5–5 µg/kg (maximum 600 µg) followed by infusion at 0.3 µg/kg/h. This appears based on a 32-patient study of clonidine vs. placebo [26], in which a six-hour postoperative infusion of 0.3 µg/kg/h produced a mean reduction in blood pressure of 13 mmHg and heart rate 2 bpm. In a 70-patient comparison of intravenous clonidine and dexmedetomidine for sedation in predominantly surgical ICU patients [27], median equi-sedative infusion doses required for a RASS of -3 or -4 were 0.4 µg/kg/h for dexmedetomidine and 1.4 µg/kg/h for clonidine. Assuming a linear conversion, scaling the 0.4 µg/kg/h dexmedetomidine dose down to 0.1 µg/kg/h (as used in [11]) would result in an equivalent clonidine infusion rate of 0.35 µg/kg/h. Loading doses of intravenous clonidine cause bradycardia and hypotension [24, 28]. Consequently, we avoided using a loading dose.

We used a consumer-targeted wrist sleep tracking device to assess the study primary outcome. Polysomnography is the gold standard for sleep measurement, but is intrusive, labour intensive, costly, and is impractical in large numbers of patients [29]. Bispectral index (BIS) monitoring is also intrusive and prone to patient dislodgement. Wrist actigraphy is less intrusive and expensive, but although it detects sleep with good sensitivity, it overestimates sleep in immobile ICU patients [30]. Actigraphy is also unable to distinguish sleep stages. Consumer-targeted sleep tracking devices, such as the FitBit Alta HR, combine wrist actigraphy with reflective photoplethysmography to measure heart rate. Using the two modalities, these devices report time spent in each sleep stage (wake, N1 and N2, N3, and REM) and sleep statistics such as total sleep time and number of awakenings with moderate-to-good reliability in healthy volunteers when compared to polysomnography [19]. In a 507-patient study, a FitBit device accurately measured total sleep time in non-intubated ICU patients [31]. Notwithstanding the validation provided by these two studies, the use of heart rate in the FitBit Alta HR sleep detection algorithm is potentially problematic in our trial, as the heart rate effects of clonidine might have interfered with the detection of total sleep time or identification of sleep stages. We did not record detailed heart rate data, so cannot compare this between trial groups. This possibility limits the validity of our findings, mitigated only by the congruent benefit on sleep recorded by other methods.

Subjective sleep quality can also be assessed using questionnaires. At least 13 questionnaires have been used in ICU research [32], but the RCSQ is the only scale validated in an ICU population [33]. The self-reported RCSQ correlated closely with polysomnography in 70 ICU patients [33]. It is unclear whether nurses’ assessment of sleep quality using the RCSQ is a reliable substitute for patient self-assessment [34, 35]. Our study found a clearer effect of clonidine in patient-reported scores, with the exception of the nurses’ observation of the time taken to fall asleep—perhaps the most difficult element for a patient to assess.

We studied clonidine as an alternative to dexmedetomidine for several reasons. Clonidine is substantially cheaper than dexmedetomidine in our hospital. Dexmedetomidine is only available in intravenous form, requiring intravenous access, an infusion pump, and monitoring by nursing staff. If an equivalent dose of clonidine could be administered enterally or transcutaneously, discomfort and cost would be reduced. Dexmedetomidine’s potential cardiovascular side effects require monitored critical-care environments, while enteral or transcutaneous clonidine could be administered in surgical wards. Dexmedetomidine’s short elimination half-life (two hours) allows for easy titration when used as a sedative [36], but if used for low-dose prophylaxis, titratability is less useful. Clonidine’s longer elimination half-life (12–16 h) means any protective effects against postoperative delirium or improvement in sleep—if they exist—might last longer.

These results have potentially important implications for practice in the ward context as well as the HDU. Clonidine is used safely on hospital wards for hypertension, opioid withdrawal, as an analgesic adjunct [37]. Formulation as a transcutaneous patch [38] potentially expands postoperative utility in patients unable to swallow. In patients > 65 years, 75 μg clonidine orally every 3 h produced plasma concentrations at the high end of the therapeutic range. No safety concerns were identified [39]. The oral bioavailability of clonidine approaches 100% [40]. In an 80 kg adult, a dose of 75 μg per os q3hr is therefore approximately comparable to the 0.3 μg/kg/h IV used in our trial. For an average 80 kg adult, a 300 mg clonidine patch would equate to an infusion of 0.16 µg/kg/h [41].

We recognise several limitations to our study. We screened 1169 patients to randomise only 83, which might indicate inclusion bias. However, this low rate was primarily due to unavailability of a research co-ordinator after extended business hours (346), inability to contact the patient preoperatively (326), and refusal of consent (136), not patient ineligibility (195 with a condition that contraindicated clonidine administration). While we cannot be sure that patients excluded for administrative or consent reasons were not systematically different to those who were included, we have no reason to believe this was the case. The age, sex and operative category demographics of the included patients match those of our overall HDU population. We regret that we did not have sufficient funding to allow patient recruitment after 8 PM or when completion of study assessments would be required on weekends, but this was anticipated at the time of protocol design and prospectively included in both the protocol and trial registration documentation. Only 83 of the planned 120 patients (69%) were randomised. Unplanned early termination of clinical trials has been argued to exaggerate effect size, although in reality this effect is very small [42]. Adjustment for the two major potential confounders identified prior to study commencement had no effect on the results. Despite their small size, our study groups appear relatively well balanced at baseline; however, we did not collect data on several other patient characteristics that might affect sleep, such as body mass index or comorbidities. Our study has many of the characteristics of single-centre trials we have previously identified as potentially misleading [43]: intercurrent care in our HDU might interact with clonidine and so limit external validity; our results might be specific to patients with this low baseline risk; knowledge that a patient was enrolled might have led clinicians to pay them more attention; and more subtle clonidine cardiovascular effects than could be recorded in our monitoring might nonetheless have unblinded treating clinicians. We did not record detailed cardiovascular parameters in trial patients other than the occurrence of bradycardia and hypotension, so we cannot state whether one trial group was mildly but systematically different to the other. Although there was a difference in the primary outcome and several congruent secondary outcomes, the study was underpowered to detect differences in important end points including ICU length of stay, and the patients’ baseline risk of delirium was too low for this outcome to be affected.

Nonetheless, our study has many strengths. Its double-blinded design reduced bias. Patients were treated in an HDU that paid attention to non-pharmacological aspects of sleep optimisation, so the medication effect was in addition to, not instead of, good clinical care. The magnitude of effect on the primary outcome is biologically plausible and consistent with earlier trials of α2 agonists in similar contexts. The outcome differences are patient-centred, especially the superiority in patient-reported sleep scores. The trial’s broad eligibility criteria suggest this intervention might safely translate to postoperative patients in less intensely monitored environments.

Conclusion

Amongst postoperative elective surgical patients admitted to an HDU, infusion of low-dose non-titrated clonidine, compared to placebo, in the context of non-pharmacological measures to enhance sleep, resulted in approximately 100 min of subjectively better-quality sleep. Nurses observed patients receiving clonidine fell asleep more quickly. While the duration of additional sleep might have been affected by the dependence of the measuring device on heart rate, congruent patient- and nurse- reported secondary outcomes support the validity of this primary outcome. This study justifies equipoise for a multicentre pragmatic randomised controlled trial of low-dose postoperative clonidine, ideally informed by preliminary validation of any sleep monitoring device (such as photoplethysmography) in the context of negatively chronotropic drugs. Exploration of oral or transcutaneous clonidine in postoperative ward patients might also be warranted, although the additional cardiovascular risk in a less monitored environment should be considered.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

MCR and DL conceived the study and wrote the trial protocol. DL, EH, JMP and MCR were awarded project grant funding to support the trial. Patients were recruited and data collected by DL, EH, AP, AH, ML-S, JS, CF, AL, B-YM, TS, KH and CAW. Data analysis was performed by DL and MCR. The first draft of the manuscript was written by MCR. All authors read and approved the final manuscript.

Funding

Open Access funding enabled and organized by CAUL and its Member Institutions. The CLONES study was funded by a Queensland Health Junior Doctor Research Fellowship awarded to Dr Liu and a Royal Brisbane and Women’s Hospital Foundation Project Grant.

Data availability

As specified in the trial registration, for five years following this publication, individual participant data underlying published results will be available from the corresponding author to appropriately qualified researchers who provide a methodologically sound proposal, only to achieve the aims in the approved proposal.

Declarations

Conflicts of interest

The authors declare no conflicts of interest.

Ethical standard

The trial was approved by the Human Research Ethics Committees of the Prince Charles Hospital, Brisbane (ID HREC/2018/QPCH/44026) and University of Queensland (2018002382/44026) and received local site governance approval by the Royal Brisbane and Women’s Hospital. The study was performed in accordance with the ethical standards laid down in the 1964 Declaration of Helsinki and its later amendments.