Acute Distress Respiratory Syndrome After Subarachnoid Hemorrhage: Incidence and Impact on the Outcome in a Large Multicenter, Retrospective Cohort

This multicenter retrospective study includes consecutive aSAH patients admitted to the neuro-intensive care unit (NICU) of the "San Gerardo" Hospital in Monza, the NICU of the "Medical University Hospital" of Innsbruck and the medical-surgical ICU of the "Erasme Hospital (ULB)" in Brussels between January 2009 and December 2017 (Fig. 1).

Inclusion criteria were:

Exclusion criteria were:

The aims of this study were:

We used the electronic data systems of each center to collect data regarding the patients' previous medical history, including age, smoking habits, and pre-injury comorbidities (e.g., hypertension and chronic obstructive pulmonary disease history). We assessed the Glasgow Coma Scale score (GCS) [8] on admission and the World Federation of Neurosurgical Societies grading system for aSAH (WFNS) [9] for each patient according to their medical records. We recorded the modified Fisher scale (mFisher) [10], the arteriography or angio-CT findings (such as aneurysm location), the information regarding the aneurysm treatment (if surgical or endovascular procedure), and we noted whether and how the invasive intracranial pressure was monitored (through an external ventricular drain (EVD) or an intraparenchymal catheter). Neurological outcome was assessed through the Glasgow Outcome Score (GOS) [11] or the modified Rankin scale (mRS) [12] at ICU discharge. To homogenize the outcome evaluation, we converted the mRS score to GOS score. Thus, a mRS of 0 or 1 corresponds to a GOS score of 5, a mRS of 2 to a GOS score of 4, mRS 3 or 4 to a GOS of 3, mRS 5 to a GOS of 2 and mRS 6 to a GOS score of 1. We then defined a good outcome as a GOS score of 4 or 5. We defined an unfavorable outcome as a GOS ≤ 3.

We recorded the length of mechanical ventilation in hours and ICU length of stay (LOS) in days and defined prolonged mechanical ventilation as > 7 days.

For each patient, we collected data regarding oxygenation, ventilator settings (such as tidal volume (TV), positive end-expiratory pressure (PEEP)), and arterial blood gas analysis (ABG): pH, partial pressure in oxygen (PaO₂), and partial pressure of carbon dioxide (PaCO₂) at day one, three and seven from ICU admission. We collected the lowest and highest value for each variable at each daily time point, then we calculated the lowest values of the ratio of the PaO₂ over the inhaled fraction of oxygen (PaO₂/FiO₂). We defined a PaO₂ value < 60 mmHg as the cut-off for hypoxemia.

Two trained investigators for each center reviewed all chest X-rays independently, checking for the presence of bilateral infiltrates. In case of disagreement, a third investigator was involved in interpreting the radiological findings.

To evaluate the ARDS incidence in our cohort, we applied the Berlin definition of ARDS that includes:

Regarding the imaging criteria, every case was reviewed in light of echocardiography, electrocardiography when available, or the medical file to rule out the presence of cardiac failure. When the ARDS criteria were fulfilled, we assessed whether the same episode lasted for multiple days during the first week of ICU admission, thus avoiding the risk of a repeated diagnosis.

We described quantitative variables as the median and interquartile range (IQR), and categorical variables as number and percentage. In order to compare the characteristics of patients with or without ARDS at ICU admission, and with favorable (GOS 4–5), or unfavorable (GOS 1–3) primary outcome, we computed differences between medians (and bootstrap confidence intervals) and proportions (and exact confidence intervals with Yates’ continuity correction) for continuous and categorical data, respectively.

We estimated the incidence of ARDS, as defined by the Berlin definition, by Aalen–Johansen estimator accounting for competing events of death and discharge and plotted stacked crude cumulative incidence by time since ICU admission. We then conducted a generalized linear mixed model with logit link in order to adjust for the variables affecting the neurological outcome. We included all variables that presented an association with the outcome with a p value < 0.15 (by Fisher exact and Wilcoxon test for categorical and continuous variables, respectively) excluding collinear variables [13]. In order to account for center effect, we included a random effect of center in the model. The linearity of continuous variables was checked by the use of polynomials and variables not fulfilling the assumption were categorized for ease of interpretation. Due to missing values in predictors (age 1%, high blood pressure 0.8%, EVD 0.8%, Glasgow coma score at admission 1%, WFNS score 0.9%, mFisher 0.9%, Surgery 1%, white blood count at admission 15.9%, highest temperature at day 1 0.7%), both complete case analysis and multiple imputations were performed. We assumed data were missing at random, in particular for white blood count at admission, that presented the highest missing rate, we believed missing was mainly due to underreporting values from peripheral centers addressing the patient and that white blood count at day 3 and 7 would well allow to impute its value. We performed multiple imputation by MICE algorithm with the method of the chained equation. Ten imputed datasets were created using the following variables: age, ARDS occurrence, EVD, admission GCS, WFNS, mFISHER, surgery, WBC at admission and at day 3 and 7, hypoxia, prolonged mechanical ventilation, smoking history, chronic obstructive pulmonary disease, high blood pressure, aspiration pneumonia, intracranial pressure monitoring, use of neuromuscular blockers, the highest temperature at admission, pathological bronchial secretions, bilateral chest infiltrates, mean arterial pressure at admission, procalcitonin at admission, C reactive protein at admission, heart rate at ICU admission, ICU length of stay and outcomes. Results are presented as odds ratios (OR) with a 95% confidence interval (CI).

All tests were two-sided at a 0.05 significance level; we performed the statistical analysis using R, software version 3.5.2 [14].

The cumulative incidence of ARDS was 2.2% (95%CI 1.2; 3.2%) at day one, 3.2%(95%CI 2.2;4.4%) at day three, and 3.6% (95%CI 2.4;4.9%) at day seven (Fig. 2). Individual ARDS criteria are depicted in the supplementary table. 19 (61.3%) out of the 31 patients presenting an ARDS within 7 days after ICU admission had transthoracic echocardiography, 8/19 (42.1%) patients presented an abnormal ejection fraction. In these patients, the pulmonary opacities were considered incompletely explained by the occurrence of cardiac diastolic failure.

Patients who presented ARDS were similar for demographic characteristics (including hypertension and smoke) but presented with a more severe aSAH, as for neurological status at admission, according to the GCS, and neuro-radiological findings (mFisher) (Table 1). We did not find any difference in aneurysms' location, but patients with ARDS underwent more frequently invasive procedures such as aneurysm clipping, EVD positioning, and ICP catheter insertion (Table 1).

In the whole cohort, the lowest PaO₂/FiO₂ ratio in the first week was 230 [168, 305], 234 [173, 311] in patients without ARDS and 150 [118, 179] in patients who developed ARDS (p < 0.0001, Table 2). Patients without ARDS were ventilated with lower TV (420 mL [400, 460] vs. 460 mL [410, 500], p = 0.099, Table 2), and neuromuscular blockade was less frequently used (8 (26.7%) vs. 121 (15.2%), p = 0.15, Table 2).

Overall, 342 patients (40.0%) presented a poor neurological outcome at ICU discharge (Fig. 2), 23 (74.2%) among patients with ARDS, and 319 (38.7%), among others (p = 0.005). Patients with ARDS had a higher duration of mechanical ventilation (6 [2, 15] vs. 2 days [0, 11], p = 0.002), but a similar LOS (9 days [2, 17] vs. 11 days [4, 21], p = 0.198, Tables 2, 3).

Several variables were associated with poor ICU outcome (Table 3): older age, history of high blood pressure, rebleeding, worse neurological status at admission (assessed by motor and total GCS, as well as WFNS score), modified Fisher grade, the need for surgical intervention, or insertion of an ICP catheter or EVD, the occurrence of aspiration pneumonia, the temperature at admission, ARDS diagnosis, radiological opacities, the use of neuromuscular blockers, tidal volume and PEEP settings, a high white blood cell (WBC) count, higher PaO₂ values, higher duration of intubation and LOS.

To perform the multivariable regression analysis, we excluded the collinear variables such as radiological opacities, the use of neuromuscular blockers, tidal volume and PEEP settings. We, therefore, included: age, high blood pressure, admission GCS, WFNS, mFisher, need for ICP or EVD, surgical intervention, WBC at admission, highest temperature, aspiration pneumonia, hypoxemia, ARDS and prolonged mechanical ventilation. After multiple imputation, the multivariable model evidenced that age > 65 years (OR = 3.52, 95%CI = 2.24–5.53, p < 0.001), prolonged duration of mechanical ventilation (OR = 4.75, 95% CI = 3.18–7.09, p < 0.001), GCS < 8 at admission (OR = 0.47, 95%CI = 0.23–0.93, p = 0.033), a Fisher score of 3 or above (OR = 1.76, 95%CI = 1.04–2.95, p = 0.033), neurosurgery intervention (OR = 1.70, 95%CI = 1.16–2.50. p = 0.0065) and ARDS (OR = 3.00, 95%CI = 1.16–7.72, p = 0.023) were associated with an unfavorable outcome at discharge from ICU (Table 4).

Several limitations need to be mentioned in this study.

ARDS incidence was evaluated accordingly to the daily worst PaO₂/FiO₂ value. Thus, we might have even overestimated the incidence of ARDS, as daily lowest PaO₂/FiO₂ could be the result of temporary episodes of hypoxemia, such as those occurring during an atelectasis episode, during the transports to the radiologic suite, or during an angiographic procedure. Also, the participation of cardiogenic edema to ARDS was ruled out according to medical files, transthoracic and EKG interpretation in the medical files.

In our study, we did not investigate the etiology of ARDS. ARDS after aSAH can be multifactorial. Different mechanisms have been described, such as aspiration, neurogenic pulmonary edema, or brain immune system interaction. Surprisingly, the incidence of aspiration pneumonia and the presence of pathological pulmonary secretions did not differ in patients with and without ARDS.

Moreover, we focused on the early outcome after aSAH, at ICU discharge. We considered the time in ICU by a Cox model on ICU mortality and found consistent results (the hazard of ICU mortality in patients with ARDS was higher than patients without ARDS: hazard ratio 2.41, 95%CI: 1.39; 4.18). Many other factors would have influenced the long-term neurological outcome in aSAH, not only ARDS; as e.g., the aneurysm treatment, complications and the rehabilitation phase.

Due to the retrospective observational design of this study, we reported an association but not causality.

The retrospective data collection resulted also in the lack of precise and complete data regarding ventilation and other confounders that might have influenced the patients' outcome.