Impact of sleep alterations on weaning duration of mechanically ventilated patients: how much is bad?

From the authors:

Sleep is markedly altered in intensive care unit (ICU) patients under mechanical ventilation [1–4]. Sleep in these patients is fragmented by numerous arousals and awakenings and mainly comprises light sleep. The deep sleep allowing physical restoration and the rapid eye movement (REM) sleep allowing neurobehavioral restoration may completely disappear. In some ICU patients, normal sleep architecture may disappear and be replaced by electroencephalogram aspects suggesting atypical sleep [5–7]. Whereas it has been shown that sleep deprivation could alter physical and cognitive functions in animals [8], the potential deleterious effects of such sleep disturbances in critically ill patients are unknown.

In a recent issue of the European Respiratory Journal, we reported that sleep alterations, especially atypical sleep or the absence of REM sleep stage, may be associated with poor outcomes in patients under mechanical ventilation with weaning difficulties [9]. Patients with atypical sleep or no REM sleep on polysomnography had prolonged duration of mechanical ventilation as compared to patients with normal sleep. In their correspondence, A.M. Esquinas and colleagues justifiably raise a number of concerns, and it is a great pleasure to reply and to provide supplementary findings to support our results.

It is true that multiple factors may disrupt sleep in the ICU, particularly noise, light, patient care activities, patient–ventilator asynchronies or central apnoeas, which can provoke numerous arousals and awakenings [1–4]. In ICU patients it has been shown that the sleep fragmentation index, i.e. the number of arousals and awakenings per hour of sleep, is very close to that in patients with severe obstructive sleep apnoea syndrome (>30 events·h⁻¹) [1–4]. In our study, sleep fragmentation was high but did not differ between patients with atypical sleep and those with normal sleep (median 33 events·h⁻¹ (interquartile range (IQR) 25–40 events·h⁻¹) versus 31 events·h⁻¹ (IQR 8–46 events·h⁻¹); p=0.57). Although this result was not indicated in our article, it makes it clear that the environment was similar in the two groups.

Another major concern is that patients with atypical sleep could be more likely to have pre-existing sleep disorders, alkalosis or higher levels of bicarbonates, which may induce sleep alterations. In our overall population (45 patients), only two patients had known sleep apnoea syndrome and 12 (27%) had underlying chronic obstructive pulmonary disease (COPD). Although the proportion of COPD patients was similar in the two groups (25% among those with atypical sleep versus 28% among those with normal sleep; p>0.99), we cannot exclude the possibility that patients with atypical sleep may have had sleep difficulties before ICU admission. By contrast, the level of bicarbonates on the day of polysomnography did not differ between patients with atypical sleep and those with normal sleep (29±6 versus 30±6 mmol·L⁻¹; p=0.41). Likewise, the proportion of patients with alkalosis (pH >7.42), which is a major factor for central apnoeas and subsequent arousals [10], was similar in the two groups (45% in patients with atypical sleep (nine out of 20 patients) versus 42% in those with normal sleep (10 out of 24 patients); p>0.99). It must also be emphasised that all patients were conscious, responsive to voice and non-sedated, so that paroxysmal activity was unlikely.

Lastly, and this is probably the most important limitation of our study, patients who exhibited atypical sleep on polysomnography had spent more time in the ICU and had received more sedation prior to polysomnography than patients with normal sleep. Consequently, atypical sleep may be rendered more likely by sleep deprivation due to a more prolonged ICU stay, or may simply be the result of acute brain dysfunction due to more marked impregnation with sedative drugs rather than the result of sleep disturbances. We fully agree that we clearly cannot claim that sleep alterations are necessarily the cause of prolonged weaning and delayed extubation.

Given the baseline differences between patients with atypical sleep and those with normal sleep, we performed a propensity score-matched analysis according to their ICU length of stay and cumulative dose of sedation (midazolam) received prior to polysomnography (R software, MatchIt package; www.r-project.org). Baseline characteristics and outcomes of the 36 patients included in the propensity score-matched cohort (80% of the initial population) are displayed in table 1. After matching by ICU length of stay and cumulative dose of sedation prior to polysomnography, the duration of weaning remained significantly longer in patients with atypical sleep than in patients with normal sleep. Using multivariate analysis in this matched cohort, ICU-acquired weakness and atypical sleep were the two independent factors associated with prolonged weaning from mechanical ventilation, with an adjusted odds ratio for atypical sleep of 16.5 (95% CI 2.7–174; p=0.006).

In view of these findings, we therefore believe that a strong relationship might exist between sleep and weaning from mechanical ventilation in the ICU and that sleep alterations may be an underexplored aspect of weaning difficulties.