Comparing radiographic scores for prediction of complications and outcome of aneurysmal subarachnoid hemorrhage: Which performs best?

Maryam Said, Svenja Odensass, Meltem Gümüs, Jan Rodemerk, Mehdi Chihi, Laurèl Rauschenbach, Thiemo Florin Dinger, Marvin Darkwah Oppong, Philipp Dammann, Karsten Henning Wrede, Ulrich Sure, Ramazan Jabbarli

PMID: 36371646
DOI: 10.1111/ene.15634

Abstract

Background and purpose

Aneurysmal subarachnoid hemorrhage (aSAH) is characterized by high morbidity and mortality proceeding from the initial severity and following complications of aSAH. Various scores have been developed to predict these risks. We aimed to analyze the clinical value of different radiographic scores for prognostication of aSAH outcome.

Methods

Initial computed tomography scans (≤48 h after ictus) of 745 aSAH cases treated between January 2003 and June 2016 were reviewed with regard to Subarachnoid Hemorrhage Early Brain Edema Score (SEBES), and Claassen, Barrow Neurological Institute (BNI), Hijdra, original Graeb and Fisher scale scores. The primary endpoints were development of delayed cerebral ischemia (DCI), in-hospital mortality and unfavorable outcome (modified Rankin Scale score >3) at 6 months after subarachnoid hemorrhage. Secondary endpoints included the different complications that can occur during aSAH. Clinically relevant cutoffs were defined using receiver-operating characteristic curves. The radiographic scores with the highest values for area under the curve (AUC) were included in the final multivariate analysis.

Results

The Hijdra sum score had the most accurate predictive value and independent associations with all primary endpoints: DCI (AUC 0.678, adjusted odds ratio [aOR] 2.83; p < 0.0001); in-hospital mortality (AUC 0.704, aOR 2.83; p < 0.0001) and unfavorable outcome (AUC 0.726, aOR 2.91; p < 0.0001). Multivariate analyses confirmed the independent predictive value of the radiographic scales for risk of decompressive craniectomy (SEBES and Fisher score), cerebral vasospasm (SEBES, BNI score and Fisher score) and shunt dependency (Hijdra ventricle score and Fisher score) after aSAH.

Conclusions

Initial radiographic severity of aSAH was independently associated with occurrence of different complications during aSAH and the final outcome. The Hijdra sum score showed the highest diagnostic accuracy and robust predictive value for early detection of risk of DCI, in-hospital mortality and unfavorable outcome after aSAH.

INTRODUCTION

Aneurysmal subarachnoid hemorrhage (aSAH) is characterized by high morbidity and mortality, proceeding from the initial severity and from complications occurring during the course of the disease [1]. Therefore, timely recognition of the risks of outcome-relevant complications and proper prognostication of the further course of disease are of paramount importance. Early risk stratification and clinical management of aSAH patients tailored to their risk profile might be helpful in outcome improvement.

The amount and pattern of intracranial bleeding after aneurysm rupture have been widely recognized as major contributors to the risk of early and late complications and poor outcome of aSAH [2]. Accordingly, the admission computed tomography (CT) scan has been of interest in many radiographic scores that aim to quantify the burden of intracranial bleeding after aneurysm rupture. Of these scores, the original Fisher scale published in 1980 by Fisher and colleagues [3] is probably the most widely used aSAH severity scale. In recent decades, several modifications of the original Fisher scale have been developed to improve diagnostic accuracy regarding the prediction of secondary complications of aSAH [4–6]. Moreover, other radiographic scores addressing specific components of subarachnoid hemorrhage (SAH) severity, such as the extent of brain edema (Subarachnoid Hemorrhage Early Brain Edema Score [SEBES]) [7] and intraventricular bleeding (Graeb scale [8]), or scores incorporating different SAH aspects (Hijdra sum score [9]), are also of potential clinical relevance.

Unfortunately, only few studies compared these scores with regard to their diagnostic accuracy [10–13]. Studies analyzing multiple radiographic scales within a representative aSAH cohort are still lacking. Furthermore, the literature on the predictive value of these scores regarding adverse events other than vasospasm/delayed cerebral ischemia (DCI) and poor outcome is sparse.

In this study, we aimed to investigate the predictive value of the various radiographic scores for SAH for the occurrence of certain complications and outcome measures after aSAH. Additionally, a special emphasis was placed on identifying the most sensitive score for each endpoint and the clinically relevant cutoff point for its predictive value, independent of other factors.

MATERIALS AND METHODS

Patient population

For our retrospective analyses, we included all eligible consecutive cases treated for aSAH at our institution between January 2003 and June 2016. All patients were aged 18 years or older and required available native pre-treatment CT scans <48 h after ictus, enabling assessment of the selected radiographic scores. Our study was approved by the local ethics committee (Ethik-Kommission, Medizinische Fakultät der Universität Duisburg-Essen, registration number: 15-6331-BO). The study is registered in the German trial registry (DRKS, unique identifier: RKS00008749).

Subarachnoid hemorrhage management

Patients were admitted to our neurosurgical intensive care unit and underwent digital subtraction angiography to confirm the ruptured intracranial aneurysm. After interdisciplinary consensus was reached with the Department of Neuroradiology, treatment of the aneurysm was performed <24 h after admission, either by microsurgical clipping or endovascular coiling. Conservative treatment consisted of the administration of nimodipine for 21 days, and maintaining normovolemia and mean arterial pressure >70 mmHg. Acute hydrocephalus was treated with an external ventricular drain. Continuous intracranial pressure (ICP) monitoring was then consecutively performed. In case of pathologically increased ICP >20 mmHg refractory to conservative treatment consisting of forced cerebrospinal fluid (CSF) drainage, deep sedation and osmotic diuresis, a decompressive craniectomy (DC) was performed. If signs of pathologically increased ICP were present on admission, DC was performed at that time. In patients with persistent hydrocephalus, the external CSF drain was replaced by a permanent ventriculoperitoneal shunt.

Transcranial Doppler sonography (TCD) of the intracranial vessels was performed daily in the acute phase after SAH (21 days) to monitor for vasospasms. An increase of flow velocities in the middle cerebral artery >120 cm/s was deemed to be suspicious for cerebral vasospasm. Accordingly, in unconscious patients with TCD vasospasm, and/or in case of occurrence of neurological deterioration not attributable to any other complication (e.g., rebleeding or hydrocephalus), a repeat digital subtraction angiography was conducted for the angiographic confirmation of vasospasm, and intra-arterial spasmolysis was performed when required.

Computed tomography scans were performed at admission, with every neurological deterioration, after treatment of the aneurysm, during weaning of the external CSF drain, and before and after placement of a ventriculoperitoneal shunt.

Data management

The first available pre-treatment CT scan was reviewed by the first author (M.S.), blinded at this time to any clinical information. The SEBES, and Claassen, Barrow Neurological Institute (BNI), Graeb, original Fisher and Hijdra ventricle, cistern and sum scores for all patients were thus assessed, as described in the original papers [3, 6–9, 14]. The follow-up CT scans were reviewed for the occurrence of cerebral infarction(s) by the senior author (R.J.), as previously described [15]. Data on patient demographics, initial clinical aSAH characteristics, as well as complications and outcome measures, were collected from the institutional prospective aneurysm database.

The recorded adverse events during initial hospital stay after aSAH included: aneurysm rebleeding before treatment, need for DC, occurrence of vasospasm in TCD, angiographic vasospasm requiring treatment (intra-arterial spasmolysis), ventriculoperitoneal shunt placement due to chronic hydrocephalus, development of DCI on the follow-up CT scans up to 6 weeks after aSAH, defined as new hypodensities not seen on the pre- and early post-treatment imaging, occurrence of systemic infections, and epilepsy during aSAH treatment.

Clinical severity was assessed using the World Federation of Neurosurgical Societies (WFNS) score [16]. Functional outcome was assessed at 6 months after aSAH using the modified Rankin Scale (mRS) [17]. An mRS score >3 at 6 months after aSAH was defined as unfavorable outcome.

Study endpoints and statistical analyses

The objective of our study was to clarify the predictive value of each of the above-mentioned radiological scores for the complications and functional outcome of aSAH. Our primary endpoints were: development of DCI in the follow-up CT scans, in-hospital mortality, and unfavorable outcome at 6 months after aSAH. Secondary endpoints included the above-mentioned adverse events occurring during the initial hospital stay.

The associations between the radiographic scales and the study endpoints were first evaluated in univariate analyses using Student’s t-test for normally distributed data and the Mann–Whitney U-test for non-normally distributed data. For significant results, a receiver-operating characteristic (ROC) curve analysis was performed to compare diagnostic accuracy according to the area under the curve (AUC). For the scales with the best diagnostic performance, the clinically relevant cutoffs were defined based on the ROC curve analyses.

To confirm the independent predictive value of the radiographic scales for the primary and secondary study endpoints, the scales with best performance were included in the final multivariate binary regression analysis adjusted for age, initial clinical (WFNS grade) and baseline radiographic severity (original Fisher scale as obligatory reference scale), as well as treatment modality. The multivariate analysis was performed in two ways: using the variables as original (continuous) data and in a dichotomized manner. For the latter, patient age was dichotomized at the cohorts’ median age (55 years) the radiographic scales − according to the ROC-based cutoffs, and the WFNS scale was dichotomized as high (4–5) and low grade (1–3). Data analysis was performed using SPSS statistical software (version 25, SPSS Inc., IBM). Correlations with a p value of ≤0.05 were considered statistically significant.

RESULTS

We included a total of 745 aSAH patients in the final cohort with available initial CT scans <48 h after ictus, enabling assessment of the above-mentioned radiographic scales. The baseline characteristics of these patients are detailed in Table 1. TABLE 1. Baseline characteristics of the subarachnoid hemorrhage population

Primary endpoints

Delayed cerebral ischemia risk

In the univariate analysis (Table 2), all radiographic scores showed significant correlations with occurrence of DCI in the follow-up CT scans (p < 0.0001 for all scores; Figure 1a). According to the ROC curves, the Hijdra sum score provided the largest AUC (AUC 0.678, p < 0.0001; Figure 2), with a clinically relevant cutoff at ≥15 points. Multivariate analysis proved that this predictive value of the Hijdra sum score was independent of age, worse clinical presentation, baseline radiographic severity and treatment modality (continuous assessment: adjusted odds ratio [aOR] 1.08 per point increase, 95% confidence interval [CI] 1.05–1.11, p < 0.0001; Hijdra sum score ≥ 15 points: aOR 2.83, 95% CI 1.89–4.23, p < 0.0001 [Table 3]). TABLE 2. Univariate analysis of the radiographic scores for events-to-predict

TABLE 3. Multivariate analysis of the radiographic scores for every event-to-predict

In-hospital mortality

Similarly to DCI risk, all scores proved to be significantly related to risk of in-hospital mortality in the univariate analysis as well (SEBES, p = 0.002, all other scores, p < 0.0001; Figure 1b). The ROC curve for the Hijdra sum score proved again to have the largest AUC, at 0.704 (p < 0.0001; Figure 2), with clinical significance of ≥15 points. Multivariate analysis confirmed the independent association with the endpoint (continuous assessment: aOR 1.05 per point increase, 95% CI 1.02–1.08, p = 0.001; Hijdra sum score ≥ 15 points: 2.83, 95% CI 1.80–4.47, p < 0.0001 [Table 3]).

Unfavorable outcome at 6 months

The relationship between the radiographic scales and unfavorable outcome at 6 months after aSAH was synchronous with the other two major endpoints. Univariate analysis was significant for all scores (p < 0.0001; Figure 1c), with significant correlation for every score according to the ROC curves (p < 0.0001 for all; Figure 2). The Hijdra sum score proved again to have the largest AUC (0.726) and, after dichotomization at ≥15 points, was independently predictive of long-term functional outcome (continuous assessment: aOR 1.05 per point increase, 95% CI 1.02–1.08, p < 0.0001; Hijdra sum score ≥ 15 points: aOR 2.91, 95% CI 1.96–4.32, p < 0.0001 [Table 3]).

Secondary endpoints

In the univariate analysis for the association between the radiographic scales and the complications of SAH (Table 2), aneurysm rebleeding showed significant correlations with Claassen, BNI, Hijdra sum, original Fisher and Graeb scores. The ROC curve of the Fisher scale showed the highest AUC value (0.612, p = 0.011), with a clinically significant cutoff at grade ≥3. In the multivariate analysis, this relationship did not prove to be independent when corrected for the remaining confounders (continuous assessment: aOR 1.24 per point increase, 95% CI 0.59–2.59, p = 0.57; Fisher grade 3–4: aOR 1.39, 95% CI 0.30–6.40, p = 0.672 [Table 3]).

The need for a DC was best predicted by the SEBES (AUC 0.694, p < 0.0001, cutoff at ≥3). In the multivariate analysis, the SEBES (continuous assessment: aOR 1.78 per point increase, 95% CI 1.44–2.18, p < 0.0001; SEBES 3–4: aOR 3.05, 95% CI 1.93–4.81, p < 0.0001) and Fisher (continuous assessment: aOR 1.94 per point increase, 95% CI 1.25–3.00, p = 0.003, Fisher grade 3–4: aOR 6.26, 95% CI 2.02–19.44, p = 0.002) scales were confirmed as independent predictors.

The SEBES, and BNI and Fisher scores showed significant correlations with occurrence of vasospasms on the daily performed TCD (p < 0.0001, p = 0.05 and p = 0.02, respectively) in the univariate analysis. The ROC curves for SEBES (p < 0.001), BNI score (p = 0.03) and Fisher score (p = 0.05) proved significant, with SEBES providing the largest AUC (0.654). Multivariate analysis confirmed the independent predictive value of SEBES (continuous assessment: aOR 1.31 per point increase, 95% CI 1.13–1.51, p < 0.0001, SEBES 3–4: aOR 2.27, 95% CI 1.57–3.26, p < 0.0001) and Fisher grade (continuous assessment: aOR 1.35 per point increase, 95% CI 1.02–1.80, p = 0.036) regarding the development of vasospasm on TCD.

Angiographic vasospasms that require intra-arterial spasmolysis showed significant correlations with all scores except the Hijdra ventricle score and Graeb score in the univariate analysis. The ROC curves showed similar significance, with relatively poor performance for all scores. Of these scores, BNI score had the highest AUC of 0.593 (p < 0.0001). Independent associations of BNI score (continuous assessment: aOR 1.45 per point increase, 95% CI 1.18–1.79, p < 0.0001; BNI score 3–5: aOR 2.26, 95% CI 1.50–3.41, p < 0.0001) and Fisher grade (continuous assessment: aOR 1.83 per point increase, 95% CI 1.30–2.58, p < 0.0001; BNI score 3–5: aOR 3.36, 95% CI 1.35–8.37, p = 0.009) were shown in the multivariate analysis.

According to the univariate analyses and the ROC curves, all radiological scores had a certain predictive value for shunt dependency. The Hijdra ventricle score proved to be most sensitive (AUC 0.680, p < 0.0001) and was predictive independent of baseline characteristics (continuous assessment: aOR 1.11 per point increase, 95% CI 1.05–1.18, p = 0.001; Hijdra ventricle score ≥ 4: aOR 2.18, 95% CI 1.46–3.24, p < 0.0001). The Fisher grade also proved independently significant (continuous assessment: aOR 1.48 per-point-increase, 95% CI 1.07–2.06, p = 0.018; Fisher grade 3–4: aOR 3.70, 95% CI 1.52–9.00, p = 0.004).

Diagnostic performance was poor for predicting systemic infections during hospitalization using the radiographic scales at admission, with the Fisher grade providing the largest AUC (0.593, p < 0.0001). This relationship was also independent in the multivariate analysis for the dichotomized scale assessment (Fisher grade 3–4: aOR 2.68, 95% CI 1.36–5.30, p = 0.004). Finally, none of the radiographic scales showed any clinical value for the prediction of epilepsy in the course of aSAH.

DISCUSSION

In the present study, we showed how the initial amount of intracranial blood after aneurysm rupture is of importance for the further course of disease and its final outcome. According to our analyses, the Hijdra sum score has the best predictive value with regard to risks of cerebral infarction and poor outcome after aSAH. How this initial amount of blood has such a significant influence on the later events following the initial bleeding remains to be definitively explained but the available literature provides certain insights into this process.

The occurrence of complications after the aneurysm rupture has a multifactorial etiology, depending on the timing of these adverse clinical events. In particular, early brain injury developing directly after the bleeding event is the consequence of a cascade of processes causing endothelial damage and cell death, leading to a disruption in the blood–brain barrier by increasing its permeability, thus allowing the development of vasogenic brain edema [18]. The sudden increase in ICP after aneurysm rupture leads to reduced cerebral perfusion pressure, impairs autoregulation and can eventually cause transient or persistent ischemia [19]. In accordance with the Monro-Kellie doctrine, the peak ICP value at the time of aneurysm rupture corresponds to the amount of blood entering the subarachnoid space [20]. This relationship between the burden of intracranial bleeding and the extent of early brain injury explains the significant associations between the radiographic scores and initial clinical presentation of aSAH patients [21].

Among other complications considerably contributing to SAH outcome, cerebral vasospasm has also been strongly linked with the severity of SAH [22]. Angiographic evidence of cerebral vasoconstriction at the site of major subarachnoid blood clots was reported more than 40 years ago [23]. Although the exact mechanism remains unclear to date, the breakdown products of red blood cells are thought to lead to neuro-inflammation which can trigger vasospasm [24]. It is widely accepted that these vasoconstrictions play a crucial role in the genesis of DCI, which in its turn has repeatedly been proven to be a major contributor of morbidity and mortality after aSAH [25, 26]. Further, the development of post-hemorrhagic hydrocephalus has also been connected to the amount and pattern of SAH after aneurysm rupture. Garcia et al. found threefold higher blood volumes in aSAH patients requiring a permanent CSF diversion [27]. Its pathophysiological mechanism is believed to rest on the inflammatory reaction triggered by the blood breakdown products as well as the mechanical obstruction caused by blood cells [28].

For a long time, efforts have been made to quantify the amount of blood in aSAH patients with different scores and to extract a clinical value from these measurements. In 1980, the Fisher score was introduced as the first radiographic score that showed a significant correlation with cerebral vasospasm and could predict the clinical outcome of aSAH [3]. Since that time, multiple other radiographic scores have been developed for different measurable points in the head CT scans of this patient population. The Graeb score followed in 1982 as a semi-quantitative tool for measuring the extent of an intraventricular hemorrhage (IVH) [8]. The Hijdra sum score, from 1990, allows more extensive measurement of the intracranial blood, taking into account blood in the ventricles and cisterns [9]. The amount of subarachnoid and intraventricular blood was found to be predictive of occurrence of DCI, rebleed and outcome in a cohort of 176 SAH patients by the same authors [9, 29]. A recent systematic review confirmed the significant association of the Hijdra and (modified) Fisher scales with DCI risk [30]. To complement the original Fisher scale, Claassen et al. (2001) presented a novel rating scale for the prediction of DCI by means of ventricular and subarachnoid blood [6]. The BNI score was introduced in 2018 for predicting symptomatic vasospasm according to the thickness of the subarachnoid blood measured in the CT scan [14]. Most recently, the SEBES has been introduced as a tool to predict both early and delayed cerebral edema and poor outcome [7, 31]. An independent predictive value for the Fisher grade regarding long-term outcome after SAH was reported [32]. The cisternal Hijdra score has been found to have a prognostic value for functional outcome in SAH patients with Fisher Grade 3 [33]. Kole et al. validated a model based on the Hijdra sum score for the prediction of the presence of an aneurysm in patients with angiogram-negative SAH [34]. Additionally, better inter-observer agreement has been reported for the Hijdra sum score than for the original Fisher score [35].

It can be concluded that many efforts have been made to provide a sensitive tool for the prediction of poor outcome in this generally severely affected young patient population. However, clear comparisons among the different radiographic scores that have been introduced are still lacking. We found one study comparing the Claassen score with the original Fisher score for the prediction of DCI. In that study, both scores were significantly correlated to DCI, with the Claassen score having no additional value [11]. A more recent study, based on a cohort of 271 aSAH patients, found the modified Fisher and Claassen scales to be superior predictors of complications after SAH compared to the original Fisher scale [13]. Another study compared the BNI and modified Fisher scores with several clinical and combined scores in a total of 423 aSAH patients. The authors found clinical scores to be superior in predicting cerebral infarction and poor outcome [36].

In a comparison among several scores measuring the extent of IVH, the Graeb and LeRoux scores were found to predict outcome with similarly good accuracy in patients with primary intracerebral hemorrhage (ICH) [37]. This cohort sample was relatively small with only 73 patients with IVH. In a smaller sample of 43 IVH patients, Bisson et al. found the Graeb score as well as the modified Graeb score to be reliable tools for poor outcome prediction [38]. However, the fact that the Graeb and LeRoux scores were not exclusively intended for aSAH patients must be taken into account. The different pathophysiological pathways leading to spontaneous ICH and IVH versus aSAH may explain the inconsistent correlations with adverse events found in the latter population.

To date, it is not clear which complications during the course of aSAH could be predicted by the above-mentioned scores. Moreover, the literature on the superiority of any of these scores over the others with regard to predicting a certain adverse event is sparse. In the present study, we compared for the first time the predictive value of these different radiographic scores within the same patient cohort. Our large, representative patient sample enabled the detection of the scores with the most accurate prediction of adverse clinical events and poor outcome after aSAH.

In our study, we found that these scores can estimate many clinically relevant complications and outcomes reasonably well. The Hijdra scores, in particular, performed excellently for our primary endpoints. We propose that, in future, the scores with confirmed predictive value should be routinely calculated for aSAH patients to improve the selection process and timing of preventive measures against impending DCI, such as early initiation of induced hypertension and additional diagnostic confirmation (with digital subtraction angiography or perfusion CT scan) and endovascular treatment of cerebral vasospasm, if present. aSAH patients with limited clinical assessability (unconscious patients) in particular might profit from such a proactive diagnostic and therapeutic approach. The radiographic scores evaluated in the present study might add to the accuracy of outcome prediction after aSAH. To address the possible outcome effect, further prospective trials incorporating the proposed radiographic scales in the aSAH management protocols are required.

The major limitations of our study are related to its retrospective single-center design. In particular, aSAH patients with delayed hospital admission and/or missing initial CT scan (performed within 48 h after ictus) were excluded from our final cohort. However, the excluded cohort consisted of mostly mild clinical cases with favorable WFNS grades at admission and lower burden of complications during aSAH (Table S1). Therefore, our study results are representative for aSAH patients at risk of adverse clinical events and poor outcome, and therefore, for individuals who would most profit from the prediction of complications and outcome of aSAH.

In conclusion, this is the first study to compare the diagnostic value of different radiographic scales within a large representative patient cohort and to specify the scales most suitable for the independent and early prediction of certain adverse events and poor outcome after aSAH. The Hijdra sum score was found to have the overall highest diagnostic accuracy and prognostic value for the early detection of risk of DCI, in-hospital mortality and poor outcome in this patient population. Further research should be conducted to validate the predictive value of the Hijdra and other scores, and eventually aid the further amelioration of care for aSAH patients.

ACKNOWLEDGEMENT

Open Access funding enabled and organized by Projekt DEAL.

CONFLICT OF INTEREST

None declared.