Computed tomography in the assessment of aneurysmal subarachnoid haemorrhage for clinical outcome: an observational cohort study

Background: Aneurysmal subarachnoid haemorrhage (aSAH) is a life-threatening event with major complications such as delayed cerebral infarction (DCI) or acute hydrocephalus and poor neurological outcome. DCI occurs most frequently 7 days after aSAH and can last for a prolonged period. The ability to predict these complications would allow the neuro-intensivist to identify patients at risk and select the most appropriate unit for hospitalization. Methods: A 3-year single-centre retrospective cohort study was conducted in our neuroscience critical care unit. Initial computed tomography (CT) scans in patients hospitalized for aSAH were blindly assessed using eight grading systems: the Fisher scale, modified Fisher scale, Barrow Neurological Institute scale, Hijdra scale, Intraventricular Haemorrhage (IVH) score, Graeb score, and LeRoux score. We evaluated and compared these radiological scales for the early prediction of DCI, acute hydrocephalus, and poor neurological outcome at 3 months. Results: Of 200 patients with aSAH who survived to day 7 and were included for DCI analysis, 39% cases were complicated with DCI. The Hijdra scale was the best predictor for DCI, with a receiver operating characteristic area under the curve (ROCAUC) of 0.80 (95% confidence interval [CI], 0.74–0.85). The ideal cut-off score for all patients was 20/42, with a sensitivity of 85% (95% CI, 75%–94%) and specificity of 63% (95% CI, 54%–71%). The IVH score was the most effective grading system for predicting acute hydrocephalus, with a ROC AUC of 0.85 (95% CI, 0.79–0.89). In multivariate analysis, the Hijdra scale was the only independent predictor of the occurrence of DCI (hazard ratio, 1.18; 95% CI, 1.10– 1.27). Conclusions: Although these results have yet to be prospectively confirmed, our findings suggest that the Hijdra scale may be a good predictor of DCI and could be useful in daily

clinical practice.

Background
Aneurysmal subarachnoid haemorrhage (aSAH) is a life-threatening event. Severe complications can occur after the aneurysm is secured, such as delayed cerebral ischemia (DCI) or acute hydrocephalus requiring intensive care monitoring for 12-21 days after aSAH [1,2]. The management of these complications in high-volume hospitals with neurosurgical and endovascular services seems to be associated with better outcome [3].
However, specialized hospitals have limited capacity in terms of neurocritical care unit beds. Optimizing resource allocation requires the ability to select patients at high risk of complications after aSAH. In this context, a radiological score that would be predictive of complications would be useful for identifying patients who need intensive care unit (ICU) monitoring.
Since the early 1980s and publication of the Fisher grade, the occurrence of vasospasm and prognosis with aSAH have been recognized as being influenced by the severity of the initial bleeding, which can be evaluated on an early computed tomography (CT) scan [4][5][6][7][8][9][10][11][12]. However, several studies demonstrated low sensitivity and specificity of this scale for predicting DCI [6][7][8][9][10]. Recently, a systematic review assessing the association of radiological scales for grading aSAH with DCI showed that patients with Fisher grade 4 have a significantly lower risk of DCI compared to those with Fisher grade 3 [13]. With current clinical management including nimodipine, hypertensive therapy, and endovascular treatment, the Fisher grade predicts symptomatic vasospasm in only half of cases [14]. For this reason, other radiological scales have been developed to assess the amount of blood present in the subarachnoid spaces [7][8][9][10]. These scales qualitatively divide aSAH into categories, as do the Claassen scale [7] and the modified Fisher grade [9], or involve a semi-quantitative assessment, as do the Hijdra [8] or the Barrow Neurological Institute (BNI) [10] grading scales. Other grading systems, including the Graeb [15] or Leroux scales [16], or the IntraVentricular Haemorrhage (IVH) score [17], use a semi-quantitative method to assess the amount of blood present in the ventricles.
The goal of this study was to evaluate eight radiological scales in grading subarachnoid or ventricular haemorrhage or both for predicting DCI, acute hydrocephalus, and functional outcome at 3 months in a large aSAH population.

Methods
Guidelines for reporting this study were derived from the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) Statement [18].

Study design and population
This was a single-centre retrospective cohort study of consecutive patients with an aSAH admitted during a 33-month period (January 1, 2013 to July 30, 2016) at our neuroscience critical care unit (NCCU). Access to health information was approved by an ethics committee (Comité d'éthique pour la recherche en Anesthésie-Réanimation -IRB 000102542019081), which waived the requirement for individual consent according to French law at the time of the study [19]. Inclusion criteria were age older than 18 years, an available head CT scan demonstrating aSAH prior to any neurosurgical intervention (external ventricular drainage (EVD), aneurysm clipping, or endovascular treatment), and confirmed ruptured aneurysm on subsequent digital subtraction angiography. Exclusion criteria were the presence of non-aneurysmal vascular malformations and of intracranial artefacts (prior embolization or aneurysm clipping). Patients with SAH from other causes such as head injury, arteriovenous malformation, or arterial dissection or without aneurysms confirmed on CT or angiography were excluded. Patients for whom an initial CT could not be retrieved, with incomplete CT, or with an initial CT obtained more than 24 hours after bleeding were also excluded.

Clinical management
Our aSAH management policy has been described previously in detail [1]. Briefly, all patients were managed in a dedicated NCCU according to a standardized aSAH protocol in accordance with published European guidelines [20]: administration of intravenous nimodipine and ventricular drainage in cases of hydrocephalus allowing continuous monitoring of intracranial pressure. All aneurysms were secured within 24 hours after admission with endovascular coil embolization or surgery. All patients were followed with transcranial Doppler sonography. Those with neurological symptoms deemed suspicious for vasospasm underwent CT angiography, followed by conventional angiography in case of moderate or severe vasospasm.

CT grading
For each patient, two independent clinicians blinded to clinical data reviewed the initial 32-slice CT scans. Each head CT was graded according to eight grading systems: the  [16], and IVH score [17]. Each grading system is detailed in Additional file 1 (Table S1), and an illustrative example of patient evaluation is depicted in Fig. 1.
Data collection, DCI definition, and outcome assessment Data corresponding to clinical characteristics such as age, sex, the Glasgow Coma Scale [21], and the World Federation of Neurosurgical Societies (WFNS) scale [22] were recorded at admission. The Simplified Acute Physiology Score 2 (SAPS II) [23] was calculated within 24 hours after NCCU admission.
Based on the latest recommendations [24,25], DCI was defined as follows: development of focal neurologic signs; reduction by at least 2 points on the Glasgow Coma Scale that lasts for at least one hour and is associated with angiographic cerebral vasospasm, detected either with CT angiography or digital subtraction angiography; or a new cerebral infarction detected on a CT scan, either within 6 weeks after an aSAH or before discharge, after excluding a procedure-related infarction.
Acute hydrocephalus was defined as the need to place an EVD within the first 72 hours.

Statistical analysis
Means and standard deviations were calculated for continuous variables with normal distributions, and medians and interquartile ranges for non-continuous variables. For categorical variables, numbers and percentages were used. Comparison between continuous variables from two groups was assessed by an unpaired two-sample t-test (normally distributed) or a Wilcoxon-Mann-Whitney U test (no assumption for distribution).
Differences between categorical variables were assessed by Fisher's exact test. A receiver operating characteristic (ROC) curve was plotted to determine the ROC area under the curve (ROC AUC ) and the optimal cut-off value of grading scales that best predicted DCI, early hydrocephalus requiring EVD, poor outcome, and mortality. The ROC AUC of each scale was compared with those with higher ROC AUC values, using the method described by DeLong et al [27]. The interobserver variability of the eight scales was assessed. A weighted Cohen kappa coefficient (κ) was calculated for each pair per scale used with κ < 0.2, κ = 0.21 to 0.4, κ = 0.41 to 0.6, κ = 0.61 to 0.8, and κ > 0.8 corresponding to poor, fair, moderate, strong, and near-complete agreement, respectively [28]. We performed adjusted analyses via ordinal logistic or linear regression modelling, as appropriate. All statistics were carried out using JMP (version 14.0, SAS, Cary, NC, USA) except for the ROC analyses and kappa coefficient calculation, which were performed using MedCalc (version 9.2, MedCalc Software, Ostend, Belgium). Statistical significance was assumed at p values of 0.05 and below.

Patient demographics
During the study period, of 371 consecutive patients with SAH, 270 suffered acute aSAH and 230 met all inclusion criteria (Fig. 2). Of these, 200 patients survived for more than 7 days and were included in the analysis for factors related to DCI. Patient characteristics are presented in Table 1. A total of 47% of patients were classified as grade 3-5 on the WFNS scale. The in-hospital mortality rate was 24%. DCI was documented in 78 patients (39%) and was related to poor outcome: 42% of patients presenting with DCI suffered a poor outcome (GOSE 1-4), whereas 26% of patients without DCI did so (p < 0.001). Poor neurological condition (WFNS 3-5) at admission to NCCU was statistically associated with poor neurological outcome (p < 0.001) and occurrence of DCI (p = 0.01).  Figure 4 shows the occurrence of DCI in the subpopulation of low-severity aSAH (WFNS 1-2) and high-severity aSAH (WFNS 3-5). In the WFNS 1-2 subpopulation, the Hijdra scale was also the bestperforming scale to predict DCI, with a ROC AUC of 0.82 (95% CI, 0.73-0.89) (Additional file 1 - Fig. S1). The IVH score performed best for predicting early hydrocephalus requiring EVD, with a ROC AUC of 0.85 (95% CI, 0.79-0.89) (Fig. 3D). On univariate analysis, the Hijdra scale and IVH score were of significant prognostic value for the presence of DCI and of early hydrocephalus requiring EVD (Additional file 1 - Table S3). The Fisher scale did not predict any of the three outcome measures.

Discussion
Many studies have compared clinical grading scales such as the Hunt and Hess scale, WFNS, and Glasgow Coma Scale for predicting unfavourable outcome in aSAH [29,30]. To our knowledge, however, this study is the first to compare eight radiological scales, grading subarachnoid or ventricular haemorrhage or both for the prediction of DCI, acute hydrocephalus, and functional outcome at 3 months in a large aSAH population. We identified a 39% rate of DCI in our aSAH population, in agreement with studies using a modern definition of DCI [31]. As previously reported, the Fisher scale failed to predict vasospasm or DCI occurrence, with both poor sensitivity and poor specificity [6,7,9,10].
We confirmed that the Fisher scale could not predict any of the three outcome measures.
We note that in our ICU population, 91% of all patients were classified Fisher grade 3-4, which may have confounded statistical analysis.
The Hijdra scale was the most effective scale for predicting DCI, with an ideal cut-off of 20/42 and excellent interobserver agreement. This scale has been studied mostly for its association with functional outcome and has been found to be superior to the Fisher scale [12, 32]. Dupont et al [33] also reported that a Hijdra score ≥ 23 was strongly associated with the occurrence of vasospasm, and our findings show a strong association of Hijdra score with DCI. The design of the grading system may explain these results. Because DCI development and outcome correlate with the amount of blood on CT scan [4,5,7,9,33,34], grading every cistern and every ventricle likely assesses the overall bleeding with greater precision than other scales. Despite its apparent complexity, the Hijdra scale had a good reliability, and we found an excellent interobserver agreement for it [12, 35,36].
Of interest, both sensitivity and specificity were increased in a WFNS 1-2 subpopulation.
Associated with clinically predictive factors such as smoking, history of diabetes, and hypertension [31], the Hijdra scale could help physicians to better predict DCI occurrence and determine the most appropriate hospitalization unit for these aSAH patients.
Acute hydrocephalus is a frequent complication after brain aneurysm rupture [37][38][39]. In our ICU population, 53% patients needed an EVD. We found that the IVH score best predicted acute hydrocephalus requiring EVD. Moreover, this score stood out from the Graeb and Leroux scores by allowing for reliable estimation of intraventricular bleeding volume [17]. In a 2012 study, Hwang et al found that the IVH score could reliably predict poor neurological outcome, which was associated with an estimated IVH volume > 6 mL [40]. Recently, the interval to blood clearance in the basal cisterns and peripheral subarachnoid spaces has been associated with shunt dependency. In this study, patients with a shorter interval of blood clearance required a shorter duration of shunt use than other patients [41]. Thus, accurate evaluation of blood volume on CT scans using quantitative scales or automated computer blood quantification could be useful for aSAH management [42].
SAPS II was the best predictor of poor outcome at 3 and 6 months in our study. This score was generated from a cohort of medical and surgical ICU patients and was not intended to assess mortality in neurological patients. This expected finding illustrates the importance of clinical variables included in the SAPS II scale, particularly age and initial Glasgow Coma Scale score [23]. Indeed, the initial neurological assessment by means of the   Illustration of the eight CT-scan based grading systems from patient 6 hours after subarachnoid haemorrhage from anterior communicating artery aneurysm.   Occurrence of DCI in low-severity aSAH (WFNS 1-2, white bar) and high-severity aSAH (WFNS 3-5, black bar) for the total population and grouped by Hijdra score <10, 10-30, and >30. * p < 0.05, WFNS1-2 vs WFNS 3-5.

Supplementary Files
This is a list of supplementary files associated with the primary manuscript. Click to download.