Introduction

Sleep abnormalities are extremely common in critically ill patients [1]. These patients exhibit considerable reduction in rapid eye movement (REM) and slow-wave sleep (SWS) and more frequent arousals and awakenings than normal [2, 3]. Although the total sleep time may be normal or even increased, the quality of sleep is poor. As a result, these patients are considered to be qualitatively sleep deprived. The impaired sleep quality may cause cardiorespiratory, neurological, immunological, and metabolic consequences, leading to increased morbidity [4–9].

Mechanically ventilated critically ill patients often receive sedatives to facilitate care [10]. Propofol, a GABAA agonist, is commonly used in these patients because of its predictable pharmacokinetics [11]. Although this strategy has been advocated to promote sleep [12] and, thus, reverse the detrimental consequences of sleep deprivation, the effects of propofol on sleep quality in critically ill patients are not well established. Studies have shown that prolonged sedation with propofol in rats is associated with a restorative effect similar to sleep [13]. Propofol administration in sleep-deprived animals results in recovery from sleep deprivation that is not different from that obtained with normal sleep [14]. In normal humans as well as in critically ill patients propofol-induced loss of consciousness is accompanied by the appearance of EEG slow waves that resemble the slow waves of non-REM (NREM) sleep [15, 16]. Nevertheless, in all these studies the doses of propofol were much higher than those recommended in critically ill patients [10]. Indirect data in the literature indicate that at lower dose propofol may adversely affect sleep [17]. The aim of this physiological study was to assess the effects of propofol on sleep quality in a group of critically ill patients ventilated on assisted modes. The rate of propofol infusion was adjusted to maintain a sedation level of 3 on the Ramsay scale (response to commands) [18], as recommended for this group of patients [10].

Methods

Patients

The study, performed between October 2009 and October 2011, was approved by the human studies subcommittee and informed consent was obtained from patients and surrogates. Critically ill patients who had been receiving mechanical ventilation for at least 48 h and who were anticipated to be on assisted modes for two consecutive days were studied. At the time of the study, the patients were hemodynamically stable without vasoactive drugs and ventilated on assisted modes of support through cuffed endotracheal or tracheostomy tubes. We think that it was mandatory for proper interpretation of the results to perform sleep studies in patients not receiving any sedative or opioid, because both may affect sleep architecture; this study design made it possible to compare sleep with and without propofol. Thus, patients not requiring sedation or analgesia with opioids were selected. Exclusion criteria were (1) Glasgow coma scale less than 11; (2) acute physiology score portion of the APACHE II greater than 15 [19]; (3) presence of delirium at the time of the study, as defined by the confusion assessment method for ICU; (4) administration of any sedative drugs or opioids over the last 24 h; (5) detectable plasma levels of sedative drugs (i.e., benzodiazepine, propofol) or opioids (i.e., morphine) before the study; (6) history of epilepsy or any other neurological disease that may potentially have significant effect on the quality of sleep; (7) history of sleep apnea; and (8) ongoing sepsis. The mode of support, level of assist, positive end-expiratory pressure (PEEP), and fractional concentration of inspired O2 (FIO2) were determined by the primary physician, who was not involved in the study. Changes either in the mode of support or in the ventilator settings based on the primary physician’s judgment resulted in patient withdrawal from the study. Administration of opioids and/or neuroleptic (antipsychotic) medications during the entire study period was also a reason for patient withdrawal. Where necessary, nonsteroidal anti-inflammatory drugs (NSAIDs) were used for analgesia.

Measurements

Polysomnography was performed on each patient as previously described [1]. Sleep architecture was scored manually using standard criteria [20]. Total sleep fragmentation index (TSFI) was calculated as the sum of arousals and awakenings per hour of sleep [20]. Periodic breathing was identified visually [16]. Respiratory variables were measured on a breath-by-breath basis during NREM sleep and in a similar fashion during wakefulness. Major asynchrony, such as ineffective efforts, double triggering, and autotriggering, was evaluated by inspection of flow–time and pressure–time waveforms [21]. Asynchrony index (expressed in percentage) was calculated as the number of asynchrony events divided by the total respiratory rate computed as the sum of the number of ventilator cycles (triggered or not) and of ineffective efforts [22].

Study protocol

All studies were performed between 10:00 p.m. and 7:00 a.m. in a single room in the ICU with the window blinds closed. Noise, nursing, and other interventions were minimized during the night of the study. Light was decreased to a minimum level that did not interfere with patient’s assessment. Care was taken to ensure similar environmental conditions between the two study nights.

Each patient was studied during two consecutive nights with or without propofol in a random order. On the propofol night, a bolus of 0.01–0.05 mg/kg of propofol was given over 2 min at 10.00 p.m. followed by continuous infusion at a rate that was adjusted to maintain a sedation level of 3 on the Ramsay scale [18]. In order to prevent bias in propofol administration, only two highly experienced physicians were responsible for propofol titration (EK and CA, senior intensivists) and did not have access to EEG. Continuous infusion was stopped at 7.00 a.m. Apart from this time interval the patients did not receive any sedation throughout the study. If there was need for sedation as judged by the primary physician, the patient was withdrawn from the study. On both experimental conditions arterial blood gasses were measured in the beginning and at the end of the study.

Data were analyzed using non-parametric tests. Continuous variables were expressed as medians (interquartile ranges, IQR) and compared using the Wilcoxon test. Categorical variables were compared using the Fisher exact test. A p value less than 0.05 was considered statistically significant.

Results

Thirteen patients were studied. Ten patients were ventilated with pressure support (PS), 2 with proportional assist ventilation with load-adjustable gain factors (PAV+), and one with continuous positive airway pressure (CPAP) (Purritan-Benett 840, Tyco, Gosport, UK). With propofol one patient ventilated with PAV+ exhibited severe respiratory acidosis (PaCO2 80 mmHg, pH <7.30) necessitating a mode change and thus was excluded from further analysis. Baseline characteristics of the remaining 12 patients are shown in Table 1 (see also Table S1 in ESM). Apart from propofol all other medications were similar (in terms of type and dose) between the two study periods. The median propofol infusion rate was 0.86 mg/kg/h (IQR 0.67–1.25). In all but one patient (no. 12) the requirements for sedation remained constant throughout the study night.

Table 1 Patients’ characteristics
Full size table

None of the patients exhibited either atypical (EEG features of SWS and absence of typical stage 2 sleep as indicated by no recognition of K complexes and sleep spindles) or burst suppression EEG pattern [23]. With propofol sleep efficiency was not different from that without propofol (Table 2; Fig. 1). In all patients and on both study nights abnormal sleep architecture was observed consisting of a lack of sequential progression through sleep stages. Independent of the experimental condition SWS was infrequently observed, occurring in two patients without and in four with propofol (Fig. 2). Without propofol 6 out of 12 patients exhibited REM sleep. Propofol infusion virtually abolished the REM sleep (p = 0.04), which was observed only in one patient (p = 0.02), accounting for 3 % of total sleep time (Table 2; Fig. 2). TSFI did not differ between the two experimental conditions (Tables 2 and S2).

Table 2 Sleep architecture without and with propofol
Full size table
Fig. 1
figure 1

Sleep efficiency in each patient with and without propofol

Full size image
Fig. 2
figure 2

Sleep architecture in each patient without (upper panel) and with (lower panel) propofol. REM rapid eye movement; S1, S2 sleep stage 1 and 2, respectively; SWS slow wave sleep. With and without propofol patient no. 11 did not achieve sleep

Full size image

Both during wakefulness and NREM sleep respiratory variables did not differ with or without propofol (Table 3). Without propofol the incidence of major asynchronies was low and remained unaffected by the propofol infusion [0.3 % (0.1–1.1) vs. 0.3 % (0.2–1.2), respectively]. One patient exhibited significant periodic breathing (more than 10 % of NREM sleep) in both experimental conditions. Arterial blood gasses remained relatively stable throughout the night (Table 4) and did not differ between experimental conditions (Table S3 in ESM).

Table 3 Respiratory variables during wakefulness (n = 12) and NREM sleep (n = 11)
Full size table
Table 4 Difference in blood gasses between the end and beginning of each study night
Full size table

Discussion

The main findings of our study were that in critically ill patients mechanically ventilated on assisted modes, propofol infusion titrated to achieve the recommended sedation depth suppressed the REM sleep stage and further worsened the poor sleep quality of these patients. These adverse effects of propofol on sleep cannot be explained either by increasing patient–ventilator dyssynchrony or by deteriorating gas exchange, which remained relatively stable during the two study nights.

The Ramsay sedation scale was used for titration of propofol administration [18]. Although this scale has not been validated properly and lacks a sufficient measure of agitation, it is very simple and used extensively in ICUs worldwide. In addition, this scale has excellent inter-rater reliability and demonstrates strong correlation with validated sedation scales such as the Richmond agitation sedation scale [24, 25]. Furthermore in our study in order to avoid errors in sedation level, the titration of propofol was performed by two senior intensivists (EK, CA). The fact that in all but one patient the requirements for sedation remained constant throughout the study night indicates adequate goal-directed sedation. In addition the physicians responsible for titration did not have access to EEG and this guarantees that the titration of propofol followed the usual clinical practice (i.e., use of sedation scale) and was not influenced by EEG.

Contrary to general belief propofol administration at the doses studied did not significantly increase the sleep efficiency. On the other hand in three patients sleep efficiency was considerably lower with propofol than that without. In addition with and without propofol one patient did not achieve sleep at all. Paradoxical excitation, a well-described phenomenon observed with relatively low doses of propofol and consisting of an increase in beta activity on the EEG (13–25 Hz) and purposeless movements [26, 27], may underlie the low sleep efficiency, although none of the studied patients exhibited the clinical picture of this phenomenon. Nevertheless, the inability of propofol at doses studied to increase the sleep efficiency does not guarantee that this sedation depth is associated with sleep pattern in EEG.

Our study reconfirmed previous studies showing that critically ill patients exhibit disorganized and poor quality sleep as evidenced by the lack of sequential progression through sleep stages and low percentages of SWS and REM sleep [2, 3, 17]. Propofol infusion at doses titrated to achieve the recommended sedation depth in critically ill patients not only failed to improve the sleep architecture but worsened the sleep quality because the drug virtually eliminated REM sleep. One could argue that the REM suppression we observed with propofol was due to relatively short sleep duration; with propofol REM may be delayed and would have occurred later if patients had been observed for a greater duration. Although we cannot entirely exclude this possibility, it is highly unlikely because propofol exerts its pharmacological effects within minutes [11, 27]. In addition there is no evidence in the literature indicating that the effects of propofol on EEG are time-dependent [27].

The clinical implication of propofol-induced suppression of REM sleep in critically ill patients is not known. Notwithstanding that the role of REM sleep in humans has not been clarified [28, 29], studies have shown that REM suppression causes insomnia, and might affect memory formation [30–33]. These effects may predispose to delirium and post-traumatic stress disorder (PTSD). Indeed studies in critically patients have shown that propofol administration (as well as other GABA agonists) is an independent risk factor for delirium [34, 35]. In addition indirect evidence in the literature indicates that the inability of critically ill patients to recall memories may be linked to PTSD [36]. Therefore, propofol infusion by depressing REM sleep might be associated with short-term (delirium) and long-term (PTSD) effects in critically ill patients.

It has been shown in critically ill patients that patient–ventilator dyssynchrony adversely affects sleep quality [37]. Propofol administration, mainly by decreasing respiratory drive, could influence patient–ventilator dyssynchrony [21], and thus sleep quality. In our study, however, this was not the case. Propofol, at the doses studied, did not decrease respiratory drive, as indicated by the constant PaCO2 and respiratory variables, while the incidence of major patient–ventilator asynchronies remained unaffected. It follows that neither derangement of arterial blood gasses nor increased patient–ventilator dyssynchrony may explain the observed effects of propofol on sleep. Although pontothalamic and hypothalamic pathways might be involved in REM suppression [28, 29], the design of our study does not permit further clarification of the responsible mechanism.

The inability of propofol to improve sleep quality seems to contradict results of studies in rat models showing that propofol may mimic the restorative effects of normal sleep [13, 14]. Although the apparent discrepancy between these results and ours could be explained by species differences and the critical illness, we believe that the different dosing regimen mainly underlies the discrepancy. Tung et al. [13, 14] used doses of propofol several fold higher than those used in the current study. In humans high doses of propofol are associated with loss of consciousness and appearance of delta activity in EEG closely resembling SWS [15]. Collectively these studies and the current one indicate that the effects of propofol on sleep might be dose-dependent.

It is common practice to sedate difficult to wean patients during the night in order to obtain the restorative effects of sleep and thus facilitate the weaning process. Our findings do not support this practice. It is unknown whether an increase in dose in humans can cause restorative effects of sleep, as animal studies indicate [13, 14]. However, higher doses may depress respiratory drive leading in patients ventilated on assisted modes to hypoventilation and inefficient gas exchange.

Our study demonstrated that the amount of total sleep fragmentation was comparable to that reported in normal subjects in an ICU environment [38]. On the other hand some studies reported considerable sleep fragmentation in critically ill patients [2, 39]. We believe that in our patients the relative normal magnitude of sleep disruption may be due to selection criteria, the minimization of noise, nursing, and other interventions, and the use of a single room to study the patients. Nevertheless, in our patients the magnitude of sleep fragmentation was comparable to that observed by Gabor et al. [38] and Alexopoulou et al. [17] who selected critically ill patients for sleep studies using quite similar patient selection criteria to those in our study.

This study has several limitations. Firstly, in order to facilitate sleep classification using the standard criteria [20] a highly selected group of critically ill patients was studied. These criteria have not been developed for critically ill patients in whom atypical sleep pattern such as EEG features of SWS and absence of typical stage 2 sleep (i.e., absence of K complexes and sleep spindles) may be observed [23]. In our study patients with conditions known to be associated with this pattern (i.e., ongoing sepsis) were excluded [2, 3]. Thus, the results of this study should be applied with caution to a general population of critically ill patients in whom acute critical illness may cause encephalopathy and atypical sleep patterns [23]. Secondly, because of the study design, patients not requiring sedation were studied. In patients needing sedatives for various reasons (i.e., agitation, patient–ventilator dyssynchrony) the results might be different and this is also certainly a limitation. Thirdly, because of the strict selection criteria we applied, the number of patients studied over the 2-year period was small. Nevertheless, the effect of propofol on sleep was rather consistent between patients, partly overcoming this limitation. Fourthly, on both study nights sleep duration was approximately 4 h. This time interval may not reflect the sleep architecture adequately well, particularly in critically ill patients in whom the circadian rhythm is diminished because of disturbances in melatonin secretion [1]. However, long-term (24 h) polysomnography studies in critically ill patients have shown that sleep architecture and quality did not differ between day and night, indicating that night studies are sufficient to characterize the sleep pattern in these patients [2, 3]. We think that it is unlikely that long-term sleep studies may modify the results, at least qualitatively.

In conclusion this physiological study shows that propofol administration to achieve the recommended level of sedation in critically ill patients suppresses the REM sleep stage and further worsens the already impaired sleep quality of these patients. These results do not support the use of propofol to facilitate sleep in critically ill patients ventilated on assisted modes.