Serum glucose as an important predictor of delayed cerebral ischemia in angiogram-negative subarachnoid hemorrhage

Despite benign overall course, angiogram-negative subarachnoid hemorrhage (AN-SAH) still companies with risk of delayed cerebral ischemia (DCI). Serum glucose was previously found to be related to DCI occurrence in aneurysmal subarachnoid hemorrhage (aSAH), but this has not been conrmed in AN-SAH. The aim of this study was to clarify the signicance of serum glucose in DCI prediction in AN-SAH patients.

Despite repeated cerebral angiography, the source of bleeding cannot be identi ed in approximately 15% of patients with spontaneous subarachnoid hemorrhage (SAH). This subgroup has been classi ed as angiogram-negative SAH (AN-SAH) [1][2][3]. According to different bleeding patterns shown on computed tomography (CT), AN-SAH is divided into perimesencephalic AN-SAH (PAN-SAH) and nonperimesencephalic AN-SAH (NPAH-SAH) [4]. It was previously believed that AN-SAH has a different natural history and better outcome compared to aneurysmal SAH (aSAH). However, recent studies have shown that the clinical course and outcome of NPAN-SAH are similar to that of aSAH [5]. Although the overall course is relatively benign, AN-SAH still has a certain incidence of delayed cerebral ischemia (DCI) and poor outcome, particularly in NPAN-SAH patients [5]. Therefore, it is important to nd early risk factors that predict DCI and unfavorable outcome for patients with AN-SAH.
However, few studies have focused on the risk factors of DCI in patients with AN-SAH. Previous studies found that hyperglycemia at admission was associated with increased short-term mortality in aSAH patients without diabetes [9]. Moreover, another study found a correlation between early serum glucose levels and the occurrence of DCI and poor outcome in patients with aSAH [6]. In addition, serum glucose variability during hospitalization was also con rmed to be related to mortality, delayed cerebral infarction and neurological outcome after aSAH [10][11][12]. However, there was no study exploring the clinical value and predictive role of serum glucose levels in AN-SAH.
In this study, we analyzed our institutional data to illustrate the relationship between serum glucose and risk of DCI occurrence, and we investigated whether early serum glucose could be a predictor of DCI in AN-SAH.

Study population
In this study, consecutive patients with AN-SAH admitted to the Second A liated Hospital, School of Medicine, Zhejiang University, between January 1, 2013 and December 31, 2018 were retrospectively reviewed. All patients were admitted to the hospital within 24 hours after onset. AN-SAH was de ned as non-traumatic SAH with negative ndings in the rst cerebral digital subtraction angiography (DSA) examination, which was performed according to strict standards within 72 hours of admission [4]. Some patients underwent a repeated DSA examination 10-14 days later. In addition, the following patients were excluded: (1) history of head injury or suspicious head injury; (2) history of previous diabetes; (3) missing/lost radiological data; (4) missing/lost serum glucose data on admission or during hospitalization.
All aspects of this study were approved by the local Institutional Review Board. Due to the approval of the Institutional Review Board, patient consent was not required in this study.
Statistical analysis was performed using IBM-SPSS V24.0 (SPSS Inc, Armonk, NY) and Prism 8 (GraphPad Software, Inc, LA Jolla, CA). All p-values were two-tailed with statistical signi cance set at pvalue < 0.05. Demographic data, clinical data, radiological data, laboratory data, and SAH-related complications were compared between the DCI and the non-DCI groups. The Shapiro-Wilk test was used to evaluate normality of the data. Normally and abnormally distributed variables were respectively expressed as means ± standard deviations (SD), as well as median and interquartile range (IQR). Categorical variables were expressed as the number of patients (percentage). Student's t-test was used to compare the normally distributed variables. Mann-Whitney U-test was used to compare the non-normally distributed variables. Chi-square or Fisher's exact test was used to compare the categorical variables. Subsequently, the variables of PAN-SAH and NPAN-SAH patients were compared. After dividing AN-SAH patients into PAN-SAH patients and NPAN-SAH patients, the variables between the DCI group and non-DCI group were compared again. In addition, the characteristics of the high glucose and the low glucose groups were compared. Finally, univariate and multivariate binary logistic regression analysis were performed. The odds ratio (OR) and 95% con dence interval (Cl) were calculated. Multivariate logistic regression analysis, using a backward selection method, was performed for the variables with a p-value < 0.15 in univariate analysis to identify the independent predictors of DCI.

Patient characteristics
A total of 296 patients were diagnosed with AN-SAH. Thirteen patients had a history of head injury or suspicion of head injury. Twenty-two patients suffered from diabetes. Ten patients were missing radiological data and seven patients were missing serum glucose data on admission or during hospitalization. Therefore, 244 patients were included in the nal cohort, with 164 (67.2%) PAN-SAH patients and 80 (32.8%) NPAN-SAH patients. Males comprised 136 (55.7%) of the patients. The average age of the patients was 55.7 years.

Comparison of variables between PAN-SAH and NPAN-SAH patients
The evaluation of associations regarding bleeding patterns with patient characteristics is shown in Table   2. Hospital stay, WFNS, HH grade, mFS, SEBES, IVH, and admission serum glucose presented signi cant differences between PAN-SAH patients and NPAN-SAH patients (all p < 0.001). Additionally, there were signi cant differences in some complications, including DCI (p < 0.001), rebleeding (p = 0.001), and acute hydrocephalus (p < 0.001) between the two populations.
Comparison of variables between DCI and non-DCI groups in two bleeding patterns Due to the signi cant differences in the characteristics between two bleeding patterns, we analyzed the data of PAN-SAH and NPAN-SAH patients ( Table 3). The DCI group was still signi cantly associated with high admission serum glucose in both populations (both p < 0.001; Fig. 1A, B). There was a signi cant association between high SD of serum glucose in the rst three days and DCI group in both PAN-SAH (p = 0.038; Fig. 1C) and NPAN-SAH (p = 0.001; Fig. 1D) patients. In addition, there were signi cant correlations between the DCI group and an extended stay in the hospital and ICU stay, as well as high WFNS, high HH grade, high mFS, IVH, and high serum glucose in the rst ve days after admission in both PAN-SAH patients and NPAN-SAH patients (all p < 0.05).
Dynamic changes of serum glucose after admission Serum glucose levels of DCI and non-DCI groups showed similar trends after admission in both PAN-SAH ( Fig. 2A) and NPAN-SAH ( Fig. 2B) patients. Shortly after admission, the serum glucose levels gradually decreased and reached the lowest value on the third day. However, in the four or ve days after admission, glucose levels increased slightly. In both bleeding patterns, the serum glucose of the DCI group was consistently higher than that of the non-DCI group in the rst ve days after admission.

Comparison of variables between high and low glucose groups in two bleeding patterns
We then divided the PAN-SAH patients and the NPAN-SAH patients into either high or low glucose groups according to the median serum glucose on admission of each group (6.575 mmol/L in PAN-SAH patients and 7.235 mmol/L in NPAN-SAH patients). The results are shown in Additional le 1: Table S1. In the PAN-SAH population, patients in the high glucose group had higher age (p = 0.036), WFNS (p < 0.001), HH grade (p = 0.004), and DCI incidence (p = 0.003). In the NPAN-SAH population, patients in the high glucose group had an extended hospital stay (p = 0.009), as well as high WFNS (p = 0.004) and increased incidence of DCI (p < 0.001) and acute hydrocephalus (p = 0.039). The distributions of WFNS, HH grade, mFS, SEBES, and DCI in the two groups are shown in Additional le 2: Figure S1. The percentages of patients with high WFNS, HH grade, mFS, SEBES, and DCI incidence were often higher in the high glucose group than in the low glucose group.

Multivariate analysis of predictors for DCI in PAN-SAH and NPAN-SAH patients
The results of univariate and multivariate analyses concerning variables that in uence the occurrence of DCI in PAN-SAH and NPAN-SAH patients are provided in Table 4. Multivariate logistic regression analysis revealed admission serum glucose (p = 0.001, OR 1.705, 95% CI 1.232-2.360) and WFNS (p = 0.008, OR 2.889, 95% CI 1.322-6.311) as signi cant and independent predictors for the occurrence of DCI in PAN-SAH patients. Admission serum glucose (p = 0.016, OR 2.307, 95% CI 1.167-4.562) and SD of the serum glucose in the rst three days after admission (p = 0.049, OR 5.684, 95% CI 1.006-32.114) and mFS (p = 0.033, OR 1.859, 95% CI 1.051-3.288) were signi cant and independent predictors for the occurrence of DCI in NPAN-SAH patients.

Discussion
The main ndings of our study were that high early serum glucose levels signi cantly correlated with high incidence of DCI in both PAN-SAH and NPAN-SAH patients, despite similar time-dependent changes in DCI and non-DCI groups. In addition, high serum glucose levels on admission were associated with high WFNS. Multivariate logistic regression analysis showed that admission serum glucose was a signi cant and independent predictor for DCI occurrence in both PAN-SAH and NPAN-SAH patients. To our knowledge, this is the rst article to illustrate the association of serum glucose and occurrence of DCI after AN-SAH.
AN-SAH occurs in approximately 15% of SAH patients, and is characterized as either PAN-SAH or NPAN-SAH according to their bleeding patterns [3]. Although the overall prognosis is good, a certain percentage of AN-SAH patients still have DCI and poor outcome, especially NPAN-SAH [4,5]. It is generally believed that patients with NPAN-SAH have a worse neurological outcome and a higher incidence of complications than PAN-SAH patients, which is consistent with our ndings [3,25]. The prognosis of NPAN-SAH is considered similar to that of aSAH [5]. Therefore, it is necessary to identify early risk factors of DCI and unfavorable outcome in AN-SAH. However, few studies have focused on factors related to the occurrence of DCI in AN-SAH patients thus far.
Previous studies have veri ed the relationship between serum glucose and DCI or vasospasm after aSAH.
One study found that post-aSAH symptomatic vasospasm signi cantly correlated with admission serum glucose levels (p = 0.003), which was congruent with our ndings. However, in multivariate analysis, admission glucose levels were not a signi cant predictor of symptomatic vasospasm (OR, 0.99 [95% CI, 0.99-1.01]) [26]. In another study, maximum serum glucose levels shortly after aSAH were associated with an increased risk of DCI (p = 0.002). Multivariate analysis showed that glucose was an independent predictor of DCI (OR, 1.17 [95% CI, 1.05-1.30]) [6]. This is consistent with the results we found in patients with AN-SAH. However, this study differs in that it de ned DCI as a new hypodensity on CT not otherwise explained by cerebral infarction due to DCI after admission. To more accurately understand the relationship between serum glucose and DCI after AN-SAH, we also analyzed the mean, maximum, minimum, range, SD, and CV of serum glucose shortly after admission. In addition, another study found that the serum glucose/potassium ratio was an independent predictor of cerebral vasospasm after aSAH [27].
Although the mechanisms underlying the relation between serum glucose and risk of DCI are still unclear, the SAH-induced stress response may explain this association. Catecholamines, glucagon, and corticosteroids are the main hormones involved in causing hyperglycemia [27]. Neurogenic stress can cause the release of these hormones, which may induce in ammation and cause systemic damage [28]. Elevation of serum catecholamine concentrations, which induce sympathetic activation, has been con rmed after aSAH, and was found to be associated with a poor outcome [29]. One study found that early sympathetic activation after bleeding re ected the severity of aSAH, and was related to the development of DCI and poor outcomes [30]. A cohort study showed that using a beta-blocker to inhibit sympathetic activity was associated with a lower incidence of cerebral vasospasm in patients with aSAH [31]. An experimental study also con rmed the correlation between sympathetic suppression and decreased cerebral vasospasm after SAH [32]. Therefore, elevated serum glucose after SAH may re ect the stress response and severity of the neurological insult. In the present study, patients with higher serum glucose on admission showed had increased severity when assessed using WFNS, as well as increased risk of DCI. NPAN-SAH patients have higher serum glucose levels, as well as higher clinical and radiological severity than PAN-SAH patients. This may re ect that NPAN-SAH patients have a stronger stress response and more severe neurological damage when compared with PAN-SAH patients, and may explain the higher risk of DCI in patients with NPAN-SAH. However, after controlling for some risk factors, including WFNS and mFS, the admission serum glucose was still associated with the occurrence of DCI, suggesting that there might be other mechanisms mediating the link between serum glucose and risk of DCI.
It remains unclear whether hyperglycemia plays a role in the occurrence of DCI or if it is just a stress response to SAH. Previous studies have shown that hyperglycemia may exacerbate secondary brain injury after stroke [33]. Moreover, hyperglycemia provides an abundant substrate for anaerobic glycolysis in ischemic brain tissue, which leads to excessive lactate accumulation, acidosis, and cell death [34]. In a rat SAH model, hyperglycemia exacerbated cerebral vasospasm by dysregulating endothelial nitric oxide synthase (eNOS) and inducing nitric oxide synthase (iNOS) [35]. Another experimental study indicated that hyperglycemia may activate the extrinsic caspase cascade through the extracellular regulated kinase (ERK) signal pathway to contribute to neuronal apoptosis after SAH [36]. Aggressive glucose management may help improve outcome in aSAH [33]. However, some studies held an opposite view. A previous prospective study found that hyperglycemia preceded aSAH onset, as evidenced by the elevated glycated hemoglobin (HbA1c) levels, but did not lead to poor outcome [37]. Low cerebral glucose, which is related to severe metabolic distress, may exert deleterious effects in patients with aSAH [38]. Intensive glycemic control with insulin after aSAH may reduce cerebral glucose, leading to worse outcome [39, 40]. However, some studies have shown that increasing serum glucose through enteral nutrition could increase cerebral glucose levels without causing abnormal cerebral glucose metabolism, which may improve the prognosis of patients with aSAH [41, 42]. Therefore, post-SAH hyperglycemia may be a protective factor in the brain that compensates for insu cient cerebral glucose after brain injury.
Our study is the rst to examine the relationship between early serum glucose levels and risk of DCI in patients with AN-SAH. In this study, high serum glucose at admission correlated with high WFNS, indicating that admission serum glucose may re ect the severity of AN-SAH. Serum glucose levels at admission and early after admission, which may re ect the degree of SAH-induced stress, were signi cantly associated with the occurrence of DCI in both PAN-SAH patients and NPAN-SAH patients. Multivariate analysis showed that admission serum glucose was an independent predictor of DCI risk after PAN-SAH or NPAN-SAH. Since serum glucose is routinely collected in SAH patients, it seems to be a convenient method to identify patients at high risk of DCI after AN-SAH to guide intensive care. It remains unclear whether the elevated serum glucose after AN-SAH has a harmful effect. Therefore, tight glycemic control is not recommended.
Several limitations of this study should be considered. First, we did not study stress-related indicators after AN-SAH. Other parameters must be collected to further establish the relationship between serum glucose and DCI risk. Second, due to the mild condition and short hospital stay of AN-SAH patients, we only collected the serum glucose data in the rst ve days after admission. In addition, glucose infusion, diet, drug use, and number of glucose measurements during hospitalization were not taken into account as a potential confounder, which may have introduced bias. Third, potential bias may exist in how DCI is de ned. To address this problem, we de ned the DCI according to the criteria of the previous study [24]. Two senior neurologists who were blind to the clinical information independently evaluated the DCI. Additionally, we did not record the speci c time when the patients developed DCI. Fourth, because patients with missing radiological and serum glucose data often have better neurological status at admission, our results likely apply to patients in a moderately worse condition at admission. Finally, our study was retrospective and conducted at a single center. A multicenter collaborative prospective validation with an increased cohort size is recommended in future studies.

Conclusions
Among patients with PAN-SAH or NPAN-SAH, those with higher early serum glucose tend to have more severe conditions and a higher incidence of DCI. Admission serum glucose is an independent predictor of the occurrence of DCI after PAN-SAH or NPAN-SAH, which can serve as an early biomarker to predict DCI risk, and may guide intensive care in AN-SAH patients. Further studies with larger cohorts are needed to verify our ndings.   Figure 1 Comparisons of admission glucose and SD of serum glucose in the rst three days between DCI and non-DCI groups. DCI group was signi cantly related to higher admission serum glucose than non-DCI group in both PAN-SAH (A, p < 0.001) and NPAN-SAH (B, p < 0.001) patients. There was a signi cant association between high SD of serum glucose in the rst three days and DCI group in both PAN-SAH (C, p = 0.038) and NPAN-SAH (D, p = 0.001) patients. SD, standard deviation; DCI, delayed cerebral ischemia; PAN-SAH, perimesencephalic angiogram-negative subarachnoid hemorrhage; NPAN-SAH, non-perimesencephalic angiogram-negative subarachnoid hemorrhage