The incidence of shunt-dependent hydrocephalus (SdHCP) after aneurysmal subarachnoid hemorrhage (SAH) ranges from 14% to 47% [1,2,4,12,14,17,20,23]. The risk factors for SdHCP have been widely investigated, and a recent meta-analysis demonstrated that age, Hunt-Hess grade at admission, Fisher grade, acute HCP or external ventricular drainage (EVD) challenge, in-hospital complications, intraventricular hemorrhage (IVH), rebleeding, and posterior circulation aneurysm were significantly correlated with the risk of SdHCP [18]. However, some factors have shown contradictory results, and this discrepancy may be attributed to the heterogeneity of study designs and populations.

Some investigators recently suggested the use of risk scoring systems for predicting SdHCP [4,9,12]. Although these scoring systems are systematic and validated, their parameters are heterogeneous and somewhat complex.

The aim of this study was to identify independent predictive factors for SdHCP in a large series and devise a simple and reliable scoring system for use before aneurysm obliteration to help early identification of patients who require permanent shunt and thus, decrease the length of hospital stay, EVD procedure, and associated complications [9,22].

The study protocol was approved by the Institutional Review Board (EMCIRB 17-109). We reviewed 301 consecutive patients who underwent aneurysm obliteration following SAH from September 2007 to December 2016. The exclusion criteria for this study were previous aneurysm obliteration, previous major cerebral infarction, the presence of a cavum septum pellucidum, midline shift of >10 mm on initial computed tomography (CT), and in-hospital mortality (Fig. 1). We finally recruited 254 patients.

We retrospectively analyzed the following data : age, sex, history of hypertension and diabetes mellitus, Hunt-Hess grade, Fisher grade, aneurysm size and location, type of treatment (clipping or coiling), bicaudate index (BCI) on initial CT scan, IVH, cerebrospinal fluid (CSF) drainage (EVD or lumbar drainage), vasospasm, modified Rankin scale (mRS) score at discharge, and SdHCP. Aneurysm size was categorized as <5 mm, 5 to 10 mm, and >10 mm. The BCI was defined as the ratio of the width of the frontal horns at the level of the caudate nucleus to the distance of the brain at the foramen of Monro7). We considered a BCI of ≥0.2 as acute HCP regardless of patient’s clinical symptoms. Cerebral vasospasm was defined as the highest mean velocity of the middle cerebral artery of >150 cm/s or a Lindegaard ratio of >3 on transcranial Doppler ultrasonography. A favorable outcome was defined as a mRS score of 0 to 2. For interpretation purposes, we dichotomized other variables as follows : age at 50 years (<50 vs. ≥50 years), Hunt-Hess grade at 4 (grade 1–3 vs. 4–5), Fisher grade at 4 (grade 1–3 vs. 4), and aneurysm location as anterior (anterior cerebral artery, internal cerebral artery, and middle cerebral artery) vs. posterior (vertebrobasilar) circulation.

We included 254 patients (180 women, 70.9%) with a mean age of 55.3±12.3 years (range, 22–90). A summary of the patients’ demographic and clinical characteristics is shown in Table 1. SdHCP occurred in 47 patients (18.5%). Age of ≥50 years, diabetes mellitus, higher Hunt-Hess grade (4–5), Fisher grade 4, larger aneurysm, higher BCI, IVH, CSF drainage, and poorer mRS scores were significantly associated with SdHCP in the univariate analysis (Table 1). Acute HCP (BCI of ≥0.2), Fisher grade 4, and age of ≥50 years were significantly associated with the occurrence of SdHCP in the multivariate analysis (Table 2).

In the current study, we found that acute HCP (BCI of ≥0.2), Fisher grade of 4, and an age of ≥50 years were independent risk factors on the development of SdHCP following aneurysmal SAH. Additionally, the AFA score reliably predicted the occurrence of SdHCP before aneurysm treatment.

Patients with aneurysmal SAH aged >50 years had a 4-fold higher rate of SdHCP compared with younger patients (odds ratio, 3.938; 95% CI, 1.375–11.275). This finding is consistent with those of recent studies [1,2,5,13,14,17,22,23]. Although each study used a different cut-off value for age, a recent meta-analysis found an increased risk of SdHCP in patients aged >50 years [21]. The precise mechanism is not fully understood, but aging may accelerate an increase in the CSF compartment due to brain atrophy and a reduction of CSF turnover due to accumulation of brain metabolites [10,16]. SdHCP was more frequently occurred in patients with a Fisher grade of 4 in our cohort. Many studies have shown that a higher Fisher grade (3 and 4) was a strong predictive factor for shunt dependency [1,12,22,23]. This finding can be explained by the fact that blood clots in the ventricular system or subarachnoid space disturb CSF circulation and impair CSF absorption at the arachnoid villi. Additionally, recent experiments have shown that thrombin and transforming growth factor-β1 released after SAH induce ventricular dilatation, ependymal cell damage, and subarachnoid meningeal fibrosis and play role in the development of HCP [8,11]. In our subgroup analysis, the presence of intracerebral hemorrhage was the only significant factor in the patients aged <50 years (data not shown). This result implies that the pathogenetic mechanism of SdHCP differs somewhat depending on age. The BCI is a practically useful tool for assessing linear ventricle size [6]. Although the BCI varies with age, many investigators have adapted this index to quantitatively assess HCP [3-5,15]. Accordingly, we defined HCP as a BCI of ≥0.2 and excluded patients with a cavum septum pellucidum to reduce the false-positive rate. Previous studies have demonstrated that patients with SdHCP show poorer clinical outcomes [1,4,17,23]. Consistent with these findings, SdHCP was significantly associated with unfavorable clinical outcomes in our series (40.4% vs. 8.2%, p=0.000).

We propose use of the herein-described preoperative risk scoring system based on acute HCP (1 point), Fisher grade of 4 (1 point), and an age of ≥50 years (1 point). This AFA scoring system reliably predicted SdHCP. We validated the AFA scoring system in another cohort (175 patients with aneurysmal SAH with the same inclusion criteria from January 2001 to August 2007). The AFA scoring system showed significant predictability in this cohort (AUC, 0.895; 95% CI, 0.847– 0.943). Some investigators recently devised a scoring system to predict SdHCP after aneurysmal SAH [4,9,12]. The CHESS score is somewhat complex, and the definition of acute HCP is subjective [9]. The diagnosis of acute HCP was made according to generally acknowledged clinical and radiological criteria. However, it may be difficult to distinguish clinical deterioration due to SAH from that due to acute HCP. Moreover, there is no explanation of how early cerebral infarction is associated with SdHCP. Whether a posterior location of a ruptured aneurysm is a risk factor for SdHCP remains controversial. We found no significant association between the location of the ruptured aneurysm and SdHCP. Motiei-Langroudi et al. [12] suggested a scoring system consisting of EVD insertion, Hunt- Hess grade, and modified Fisher grade. However, the indication for EVD insertion could differ among institutions or attending neurosurgeons. In the SDASH scoring system, the authors used the Barrow Neurological Institute (BNI) to predict SdHCP [4,19]. The BNI was originally devised to predict cerebral vasospasm, like the Fisher grade. Although a previous study reported that the BNI was associated with the development of SdHCP, the BNI has not been used universally.

Our study has certain limitations apart from those inherent to retrospective studies. This was a single-center study, and the results must therefore be validated externally to prove their applicability in a prospectively collected cohort. The decision to perform a shunt procedure is not uniform among institutions and clinicians, possibly resulting in selection bias. In the method to assign a point to each predictor, we simply assigned 1 point to each predictor based on the OR, not the corresponding regression coefficient. This may lead to different overall scores, and thus decrease discriminative ability of the scoring system. Therefore, we validated the scoring system through the receiver operating characteristic curve analysis to overcome this drawback. Additionally, some patients could not undergo a shunt procedure because of their poor clinical condition or loss to follow-up.

The AFA risk scoring system is a simple and useful tool to assess the risk of SdHCP based on demographic and clinical data before aneurysm treatment. We have internally validated its usefulness. Additional external validation will be needed before this risk scoring system is universally adopted.