Early abolition of cough reflex predicts mortality in deeply sedated brain-injured patients

Study design and setting

This was a prospective, multicenter, international observational cohort study, approved by the ethics committee of Paris (Ile de France IV), France (Approval number 2014-AO1102-45) and Brescia (NP 1840, 04-11-2014), Italy. Written informed consent was obtained from the patients’ legal representative in France. In Italy, the Ethics Committee waived the requirement for consent because relatives are not regarded as legal representatives of the patient in the absence of a formal designation. The current study was a pilot study preceding the design of a larger prospective multicentre study assessing the prognostic value of brainstem dysfunction in sedated critically ill patients (ClinicalTrials.gov number: NCT02395861). Patients were recruited in four intensive care units, including one neuro ICU and three medical and surgical ICUs. Patients were recruited between December 2011 and February 2015. The Strengthening the Reporting Studies in Epidemiology (STROBE) guidelines were followed thoroughly (Von Elm et al., 2007).

Participants

Adult patients were eligible if they were deeply sedated following a major brain injury: severe traumatic brain injury, subarachnoid or intraparenchymal cerebral hemorrhage, ischemic stroke, or following a complicated neurosurgical or endovascular procedure. Deep sedation was defined by a Richmond Assessment Sedation Scale (RASS) score < −3 (Ely et al., 2003). Patients were included if sedation lasted between 12 and 24 h, and could be neurologically assessed by one of the PI (SK, JA, CR, CV, FR, NH, GM, GS, NL and FB) at working hours.

Patients were excluded if they were affected by a peripheral neurologic disorder involving the cranial nerves (e.g., Guillain Barré syndrome), or had been referred following cardiac arrest. Based on our previous findings in non-brain-injured patients, we planned to include 150 patients in the present study (Sharshar et al., 2011a; Rohaut et al., 2017).

Sedation

Decisions to initiate or to withdraw sedation as well as the level of sedation were made by the physician in charge of the patient. We recorded the time and reason for initiating deep sedation as well as the GCS prior to sedation. Since sedation was administered as part of the treatment of the cerebral insult, no systematic interruptions of sedation or decreasing trials were performed during the first 24 h. Nevertheless, depth of sedation was monitored using the RASS every 4 h. After the first 24 h, the possibility of discontinuing sedation was assessed on a daily basis. Titration of sedation was performed at least twice daily, targeting physician in charge-defined RASS levels. The date of awakening, defined by spontaneous eye opening and visual contact >10 s (i.e., RASS ≥ −1), was recorded.

Sedation was obtained through a continuous infusion of midazolam and/or propofol, in combination with sufentanil or fentanyl. Total cumulative doses of administered drugs at the time of neurologic examination were collected.

Neurologic examination

The detailed procedure pertaining to neurological examination has been previously described (Sharshar et al., 2011a). Briefly, we assessed: (1) depth of sedation (RASS); (2) reactivity, using the motor and eye response components of the GCS; (3) brainstem reflexes, including pupil size (miosis, normal, or mydriasis), pupillary light reflex, corneal reflex, facial muscle movement in response to noxious stimulation of the temporo-mandibular joint, and the cough reflex in response to tracheal suctioning.

Neurologic examination was performed after 24 ± 12 h of continuous sedation (day 1). Retaining the same definition used in our previous studies, corneal and pupillary reflexes were considered abolished only when both right and left side reflex were abolished. The oculocephalic response to lateral passive head rotation was not performed in traumatic brain injured patients. The Full Outline of Unresponsiveness (FOUR) score was recorded (Wijdicks et al., 2005). At the time of neurological examination, ICP was recorded if available. Intracranial hypertension was defined as ICP > 25 mmHg lasting for more than 5 min or by the need of an additional treatment to control the ICP (additional sedative agent, ventricular drainage, craniotomy with hematoma evacuation or decompressive craniectomy).

Baseline clinical and biological and imaging data

Demographic characteristics, body weight, date, time and cause of ICU admission, co-morbidities using McCabe score (McCabe & Jackson, 1962), date of invasive mechanical ventilation (MV) initiation and its duration, ICU length of stay, occurrence of microbiologically documented ventilator-associated pneumonia as well as the date and cause of death were recorded. The Simplified Acute Physiological Score II (SAPS-II) (Le Gall et al., 2005), the Sequential Organ Failure Assessment (SOFA) (Vincent et al., 1996) as well as key interventions and standard biological tests needed to calculate these scores were recorded. Clinical, biological and neuroradiological data were collected as part of the routine care.

The presence or absence of infratentorial lesions, including brainstem lesions, was assessed on the first cerebral computed tomography scanner (CT-scan) performed within 24 h following ICU admission, by the senior neuroradiologist and ICU physician in charge.

Outcomes

Primary recorded outcome was ICU mortality. Secondary outcomes were the occurrence of delayed awakening and of delirium following sedation discontinuation. Delayed awakening was defined by RASS < −1 despite discontinuation of sedation for more than 72 h. Delirium was assessed daily using the confusion assessment method for the ICU (CAM-ICU) (Ely et al., 2001). Other secondary outcomes were the duration of MV, the length of stay in the ICU and the occurrence of microbiologically documented ventilator-associated pneumonia. Patients had follow-up until ICU discharge.

Withdrawal of life-sustaining therapies (WLST) was determined according to applicable French and Italian Law. In current practice, no such decision involves the result of early neurological examination in a deeply-sedated patient.

Bias and confounding factors

We sought to mitigate potential confounding factors that might influence neurological examination as well as relevant outcomes, namely mortality and the occurrence of delirium. Neurological examination was performed by senior ICU physicians who were either neurologists or neuro-intensivists, or were specifically trained by a senior neurologist. The investigator was different from the clinician in charge of the patient. Inter-observer agreement for such an examination has been shown to be satisfactory (kappa scores ranged from 0.62 to 1) (Sharshar et al., 2011b). Brainstem reflexes are routinely assessed in neuro-ICU patients. Management of deep sedation was assessed by collecting cumulative doses and duration of sedation as well as daily RASS. The cause of death and its main risk factors were also assessed, including the GCS prior to sedation, SAPS-II and the SOFA score as well as the cause of brain injury.

Statistical analysis

Data are reported as numbers (percentage), mean (standard deviation), or median (inter-quartile range). Groups were compared using the Wilcoxon rank sum test or the chi-square test. Logistic regression was used to explore the associations between the pre-sedation GCS, SAPS-II, RASS scores, sedation doses, cough reflex, corneal reflex, and ICU mortality. These variables were determined a priori according to our previous findings (Rohaut et al., 2017). Since the number of events was limited, we used Firth penalized logistic regression for multivariable models, in order to limit small-sample bias. Discrimination of multivariable models was assessed using the concordance (c) index, which is equivalent to the area under the receiver operating characteristics curve. It varies theoretically between 0.5 and 1.0, a value of 1.0 indicating a perfect discrimination. Missing data were handled by multiple imputation by chained equations, all variables considered for analysis being used in the imputation model. Since about half patients had at least one variable missing (35 patients had missing data for one variable, one had missing data for seven), 50 imputed datasets were generated (White, Royston & Wood, 2011). Each imputed dataset was analyzed separately, and estimates were then pooled to provide point estimates and confidence intervals (CI) (Marshall et al., 2009). P-values < 0.05 were considered as statistically significant. All analyses were carried out using the R statistical software version 3.4.1 (The R Foundation for Statistical Computing, Vienna, Austria).

Patient characteristics

Among 151 consecutive brain injured patients receiving deep sedation within 12 h of admission, a total of 14 were excluded (Fig. 1). Overall, 137 patients were enrolled; their baseline characteristics are presented in Table 1. Cause of brain injury was blunt trauma in 70 (51%) patients, vascular in 40 patients (29%) and miscellaneous (mainly post-surgical or endovascular procedure) in 27 (20%). MV was initiated for neurological reasons (airway management in comatose patients) in 120 (88%) patients. Other reasons for intubation and MV were acute respiratory failure in 6 patients (4%), shock in one patient (1%), post-operative period in 6 patients (4%), and undetermined in 4 patients (3%). The median age was 50 (34–63) years, 80 (59%) patients were male, and the median SAPS II was 46 (36–55).

Neurological features of the patients

Main neurological features are presented in Table 2. The median GCS prior to sedation was 7 (4–11). At the time of inclusion, all patients were deeply sedated (RASS < −3), with 90 (66%) patients exhibiting RASS-5. Deep sedation was obtained using one (n = 114, 83%), two (n = 21, 15%) or three (n = 2, 2%) hypnotic agents. One hundred twenty-two (89%) patients received midazolam, 35 (26%) propofol and 5 (4%) sodium thiopental. Sufentanil was administered to 107 (78%) patients. At inclusion, median FOUR score was 4 (2–5). The most frequently preserved brainstem reflex was the corneal reflex, present in 103 patients (79%), and grimacing to pain was the most frequently abolished reflex (present in 32 patients, 30%). The cough reflex was present in 100 patients (74%). The proportion of patients with RASS-5 was greater in patients with abolished cough reflex (80% vs. 60%, p = 0.039). Similarly, the cumulative doses administrated at time of clinical exam of midazolam and opioid agents were greater in patients with abolished cough reflex (1.2 vs. 1.9 mg/Kg, p = 0.005 and, 5.2 vs. 8.5 µg/Kg, p = 0.009 respectively) (Table 3).

Brain imaging was performed for 133 patients (97%) by a CT scan, of whom 45 (34%) patients exhibited infra-tentorial lesions including brain stem lesions.

ICU mortality

Thirty patients (22%) died in the ICU (Table 1). Causes of death were multi-organ failure in 6 patients (20%), brain death in 5 patients (17%) and, withdrawal of life sustaining therapies (WLST) in 16 patients (53%). Cause of death was not reported for 3 patients (10%). WLST-related death occurred at day 8 (6–17) (ranging from day 3 to day 48 after admission). SAPSII (OR 1.06 per unit, p = 0.001) was significantly greater in non-survivors (Table 4). At admission, GCS were significantly lower (OR 0.80 per unit, p = 0.004) in non-survivors. The proportion of patients with RASS-5 (p = 0.32) did not statistically differ between the two groups. The corneal (OR 2.69, p = 0.034) and cough reflexes (OR 5.12, p = 0.0003) were more frequently abolished in non-survivors. After adjustment to pre-sedation GCS, SAPS-II, and RASS (OR: 5.19, 95% CI [1.92–14.1], p = 0.001) or midazolam and sufentanyl doses (OR: 8.89, 95% CI [2.64–30.0], p = 0.0004), an abolished cough reflex was associated with ICU mortality.

Other outcomes

Median duration of sedation was 5 (3–10) days while median duration of MV was 13 (8–24) days. Median time to awakening was 4 (1–12) days following discontinuation of sedation. Delayed awakening was observed in 67 patients (51%) and delirium in 66 patients (59%). Eighty-six patients (64%) developed ventilator associated pneumonia. Among the 103 patients in whom ICP was monitored, 50 patients (49%) suffered from at least one episode of elevated ICP.

Limits

Our study has several limits. First, the causes of brain injuries are multiple. Nevertheless, except for grimacing to pain and pupillary light reflexes, the neurological patterns were similar among the different subpopulations (Table 2). Our results suggest that the association between cough reflex abolition and ICU mortality is independent from the cause of brain injury. Second, the optimal way to assess the depth of sedation remains a matter of debate. In accordance with our previous findings, we chose to assess the depth of sedation using a specifically designed clinical scale (the RASS; Table 4, model 1). However, since doses of sedation were significantly different within patients with preserved and abolished cough reflex, we also decided to include this parameter in our multivariate analysis (Table 4, model 2). Continuous EEG monitoring and/or determining plasma levels of sedatives could have provided more detail regarding the level of sedation, however these assessments are not routinely undertaken. Based on our results, and without measurement of plasmatic sedation levels, assessing the cause of cough reflex inhibition is not feasible. It would be interesting to determine the prognostic value of the cough reflex against EEG, and to assess whether its abolition depends on circulating levels of sedatives. Third, as the main cause of death was WLST, our study could have been exposed to the self-fulfilling prophecy bias. However, the proportion of WLST in our study was comparable to other reports (Leblanc et al., 2018). In addition, it is not a common clinical practice to base WLST solely on day 1 neurological examination. In our study, no WLST-related death occurred before day 3. Nevertheless, a neurological examination at day 1, 2, and 3 could have been more valuable than solely at day 1, and this must be further evaluated. It must also be noted that since the reason for ICU discharge was not prospectively assessed, we cannot rule out that patients have been discharged from ICU for palliation on the ward and have died on the ward. Only ICU mortality at day 28 was addressed. Fourth, our data does not allow any exploration of the causality of brainstem dysfunction. In addition to the proposed mechanisms (i.e., primary injury, oversedation or neuroinflammation), the causal link between brainstem failure and multi-organ failure (which was responsible for 20% of the mortality in our study) has to be explored. Understanding to what extend brainstem dysfunction could be a cause or a consequence of multi-organ failure would need further investigation.