Abstract

Background:

Delayed cerebral ischemia (DCI) remains a major determinant of poor outcomes after aneurysmal subarachnoid hemorrhage (aSAH), yet early risk stratification is challenging.

Methods:

A retrospective cohort of 44 aSAH patients was analyzed (DCI, n = 22; non-DCI, n = 22). Serum collected within 24 h of admission underwent untargeted liquid chromatography–tandem mass spectrometry metabolomics. Between-group separation was evaluated using supervised multivariate modeling with permutation testing. Differential metabolites were identified using a combined multivariate and univariate strategy (two-sided P < 0.05), followed by pathway enrichment analysis. Independent clinical predictors were assessed using multivariate logistic regression, and aneurysm morphology was quantified on admission CT angiography using radiomics-derived parameters.

Results:

The DCI group was older, had a higher proportion of females, worse admission neurological status, higher vasospasm incidence, and lower hemoglobin. Vasospasm and age were independent predictors. Aneurysm morphological parameters showed no between-group differences. Metabolic profiles showed clear separation, supported by permutation testing (R2 = 0.6485; Q2 intercept = −0.4173). A total of 110 differential metabolites were identified (39 upregulated and 71 downregulated in DCI). Representative changes included increased sphinganine, 2-octenoylcarnitine, guanidoacetic acid, and 5-aminopentanoic acid, with decreased lysophosphatidylcholine 16:0, lysophosphatidylethanolamine 16:0, xanthine, and dehydroepiandrosterone sulfate. Enrichment highlighted coordinated alterations in energy-related, amino-acid, and nucleotide-related pathways.

Conclusion:

Early serum metabolomics within 24 h after aSAH revealed a DCI-associated systemic metabolic signature, supporting the identification of exploratory serum metabolic features associated with DCI in a Chinese cohort, which may inform future biomarker development following targeted validation.

Introduction

Aneurysmal subarachnoid hemorrhage (aSAH) is a devastating subtype of stroke with high mortality and substantial long-term disability (Hoh et al., 2023). Compared with ischemic stroke, aSAH often affects patients at a younger age, resulting in a disproportionate loss of productive life-years and a considerable socioeconomic burden (Lv et al., 2024; Yang et al., 2025). Despite advances in aneurysm occlusion strategies and neurocritical care, many survivors experience secondary neurological deterioration (Hoh et al., 2023; Neifert et al., 2021). Delayed cerebral ischemia (DCI) remains one of the most clinically consequential complications, typically occurring several days after the initial hemorrhage and strongly associated with poor functional recovery (Abdulazim et al., 2023; Vergouwen et al., 2010).

The understanding of DCI has evolved beyond the classic vasospasm-centered model (Chou, 2021). Although angiographic vasospasm is common after aSAH, clinical trials have shown that reducing vasospasm alone does not consistently translate into improved outcomes (Macdonald et al., 2011, 2012). Current evidence supports a multifactorial process involving microcirculatory dysfunction, neuroinflammation, microthrombosis, impaired cerebrovascular autoregulation, and cortical spreading depolarizations (Rowland et al., 2012; Mehra et al., 2023). In routine practice, early identification of patients at high risk for DCI remains challenging, particularly when neurological examination is limited (Abdulazim et al., 2023). This gap underscores the need for accessible, non-invasive biomarkers that can support early risk stratification and guide timely monitoring and intervention.

Metabolomics offers a system-level approach to capture downstream biochemical changes that reflect real-time pathophysiology (Batista et al., 2023). Emerging studies suggest that aSAH is accompanied by widespread metabolic disturbances, including altered energy-related metabolism, shifts in lipid mediators, and changes in amino-acid and nucleotide metabolism (Orban et al., 2024; Lu et al., 2018; Chen et al., 2022b; Yang et al., 2025). These systemic changes may mirror brain–body responses to hemorrhage and may also relate to pathways implicated in secondary injury and inflammation (Gusdon et al., 2022). However, the early peripheral metabolic features that distinguish patients who develop DCI from those who do not remain incompletely defined (Gusdon et al., 2022; Abdulazim et al., 2023).

In this study, we performed untargeted Liquid chromatography–tandem mass spectrometry (LC–MS/MS) metabolomics on serum samples collected within 24 h of admission to characterize early systemic metabolic differences associated with DCI after aSAH. We aimed to identify discriminating metabolites and enriched pathways and to place these findings in the context of clinical risk factors and aneurysm morphology. We hypothesized that early alterations in energy-related metabolism, lipid signaling, and nucleotide turnover would be associated with subsequent DCI. By delineating a serum metabolic signature linked to DCI, this work seeks to support improved prognostication and provide mechanistic clues for future therapeutic exploration.

MethodsStudy population

This study was approved by the Ethics Committee of Shaoxing People's Hospital and was conducted in accordance with the Declaration of Helsinki. This retrospective study was designed as a case–control analysis. Patients with aSAH admitted to the Department of Neurosurgery between January 2025 and October 2025 were reviewed. Patients who developed DCI were first identified from the institutional database, and an equal number of non-DCI patients were subsequently selected from the same admission period for balanced comparison.

The diagnosis of aSAH was confirmed by non-contrast computed tomography (CT), and the presence of an intracranial aneurysm was verified using CT angiography (CTA) or digital subtraction angiography (DSA). A total of 44 patients were included and divided into the DCI group (n = 22) and the non-DCI group (n = 22).

DCI was defined according to the multidisciplinary consensus recommendations as: (1) the occurrence of focal neurological impairment (e.g., hemiparesis, aphasia, apraxia, hemianopia, or neglect), or (2) a decrease of at least 2 points on the Glasgow Coma Scale, lasting for at least 1 h, which was not apparent immediately after aneurysm occlusion and could not be attributed to other causes (such as hydrocephalus, rebleeding, infection, or metabolic disturbances) by means of clinical assessment, CT/MRI scanning, and appropriate laboratory studies. Exclusion criteria were as follows: (1) traumatic or mycotic aneurysms, or SAH caused by arteriovenous malformations; (2) history of previous neurological diseases; (3) presence of severe systemic diseases (e.g., malignancy, severe liver or kidney dysfunction, autoimmune diseases) that might significantly affect the serum metabolome; and (4) admission > 24 h after symptom onset.

CTA-based aneurysm morphological assessment

Admission head and neck CT angiography images were obtained using a 64-detector CT system (Philips Healthcare) with dual-phase bolus injection of 50 mL ioversol followed by 40 mL saline at 4.5–5.0 mL/s. Acquisition parameters were 120 kVp with automatic tube current, a 512 × 512 matrix, 1.0-mm slice thickness, and pitch 0.8. Scanning was triggered at the common carotid bifurcation when attenuation reached 150 HU, and images were subsequently used for post-processing and aneurysm morphological assessment. To minimize measurement bias, an experienced neuroradiologist blinded to the clinical grouping performed the region of interest segmentation and feature extraction.

Based on the reconstructed images, quantitative morphological and density features were extracted to assess their correlation with DCI. Specifically, the maximum diameter was measured as the largest cross-sectional diameter of the aneurysm sac in three-dimensional space, while the neck width was defined as the maximum width of the aneurysm orifice at the level of the parent artery. The height-width ratio was calculated as the ratio of the maximum aneurysm height to the neck width to reflect the geometric aspect ratio. In terms of density features, the maximum Hounsfield unit was recorded as the highest density value within the aneurysm sac region of interest, serving as a surrogate for contrast filling efficiency or potential intra-aneurysmal thrombus. Finally, regarding shape regularity, aneurysms were qualitatively classified as either regular, defined by a smooth surface, or irregular, characterized by the presence of blebs, multiple lobes, or daughter sacs.

Sample collection and preparation

Peripheral venous blood samples were obtained within 24 h of admission as part of routine clinical care and subsequently retrieved from the institutional biobank for metabolomic profiling. Blood samples were drawn into serum separator tubes. Within 30 min of collection, the samples were centrifuged at 3,000 rpm for 10 min at 4 °C to separate the serum. The supernatant was immediately aliquoted and stored at −80 °C until metabolomic analysis to prevent freeze-thaw cycles.

Untargeted metabolomics analysis

Serum samples (100 μL) were thawed on ice and mixed with 400 μL of ice-cold methanol to precipitate proteins. The mixture was vortexed and incubated at −20 °C for 30 min. Subsequently, the samples were centrifuged at 20,000 g for 10 min at 4 °C. The supernatant was collected and dried under a gentle stream of nitrogen. Dried samples were reconstituted in acetonitrile:water (1:1, v/v) for LC-MS/MS analysis. A quality control (QC) sample was prepared by pooling equal volumes of all serum samples to monitor system stability. QC samples were generated by mixing equal aliquots of the supernatant from each study sample and were used throughout the analytical workflow to assess signal stability and data quality.

Chromatographic separation was performed on an ACQUITY UPLC HSS T3 column (100 mm × 2.1 mm, 1.8 μm). Mobile phase A consisted of 5 mmol/L ammonium acetate plus 5 mmol/L acetic acid in water, and mobile phase B was acetonitrile. The column temperature was maintained at 40 °C and the injection volume was 4 μL. The flow rate was 0.35 mL/min. The gradient elution program was as follows: 0–0.8 min, 2%−70% B; 0.8–2.8 min, 70%−90% B; 2.8–5.3 min, 90%−99% B; 5.3–5.9 min, 99% B; 5.9–7.5 min, 99%−2% B; 7.5–7.6 min, 2% B; and 7.6–10.0 min, 2% B for re-equilibration.

Mass spectrometric detection was carried out using a Thermo Q Exactive Plus high-resolution tandem mass spectrometer equipped with a heated electrospray ionization source operating in both positive and negative ion modes. Key parameters were: spray voltage, +3.8 kV (positive) / −3.4 kV (negative); capillary temperature, 320 °C. Data were acquired in full scan and data-dependent acquisition modes with a scan range of m/z 70–1,050 and a resolution of 70,000, AGC target of 3E6, and maximum ion injection time of 100 ms. The top 5 precursor ions with intensities greater than 100,000 were selected for MS/MS acquisition, with a DDA resolution of 17,500 and a maximum injection time of 50 ms. Dynamic exclusion was set to 6 s.

Raw LC–MS data were converted to mzXML format and processed in the R environment using XCMS, CAMERA, and the metaX toolbox for peak picking, peak grouping, retention time correction, secondary grouping, and annotation of isotopes and adducts. Each ion feature was defined by retention time and m/z, and a three-dimensional matrix consisting of feature indices, sample identifiers, and ion intensities was generated. Metabolites were annotated by matching exact mass data against the Kyoto Encyclopedia of Genes and Genomes (KEGG) and the Human Metabolome Database (HMDB) databases, with a mass tolerance of 10 ppm; molecular formulas were further supported by isotopic distribution measurements. An in-house MS/MS fragment library was also used to validate metabolite identification. Data preprocessing followed the vendor's standardized workflow in R, including data filtering, missing-value imputation, and normalization. Features were filtered if missing values exceeded 80% across study samples or 50% across QC samples. Remaining missing values were imputed using the k-nearest neighbors method, and data were normalized using probabilistic quotient normalization.

Metabolomics data were processed and analyzed using a standardized untargeted workflow. Group separation was assessed by partial least squares discriminant analysis (PLS-DA), and model validity was evaluated using permutation testing with R2 and Q2 metrics. Differential metabolites between the DCI and non-DCI groups were identified using a combined multivariate and univariate strategy, and were defined by a variable importance in projection (VIP) value > 1 together with a two-sided P value < 0.05. The numbers of upregulated and downregulated metabolites were summarized, and results were visualized using a volcano plot and hierarchical clustering heatmap of the top discriminating metabolites. For the eight representative metabolites, group comparisons were performed using log2-transformed normalized intensities. KEGG pathway enrichment analysis was conducted based on the set of differential metabolites, and enrichment results were visualized as a bubble plot. More specifically, PCA and differential metabolite analyses were performed using the R package metaX, hierarchical clustering was visualized using the R package pheatmap, and PLS-DA was conducted using the R package ropls, from which VIP values were derived. KEGG pathway enrichment analysis was performed using a hypergeometric test, and pathways with P < 0.05 were considered significantly enriched.

Statistical analysis

All tests were two-sided. Continuous variables were first assessed for normality using the Shapiro–Wilk test and were then compared using either the independent-samples Student's t-test or the Mann–Whitney U-test, as appropriate. Categorical variables were compared using the chi-square test or Fisher's exact test when applicable. Variables significant in univariate analyses were entered into a multivariate logistic regression model, and odds ratios (OR) with 95% confidence intervals (CI) were reported. A P value < 0.05 was considered statistically significant. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

ResultsBaseline clinical characteristics and independent predictors of DCI

A total of 44 aSAH patients were included in the final analysis (22 in the DCI group and 22 in the non-DCI group). As detailed in Table 1, patients who developed DCI were significantly older (63.09 ± 13.82 vs. 51.82 ± 15.27 years, P = 0.014) and had a higher proportion of females (P = 0.005) compared to the non-DCI group. Clinically, the DCI group presented with a more severe initial neurological status, evidenced by lower admission Glasgow Coma Scale (GCS) scores (P = 0.021) and higher Hunt-Hess grades (P = 0.003). Additionally, the incidence of cerebral vasospasm was markedly higher in the DCI group (54.5% vs. 13.6%, P = 0.004). Regarding laboratory findings, admission hemoglobin levels were significantly lower in the DCI group compared to the non-DCI group (114.36 ± 18.19 vs. 127.18 ± 20.24 g/L, P = 0.033). This suggests a potential reduction in oxygen-carrying capacity in patients prone to DCI.

VariablesTotal (n = 44)Non-DCI (n = 22)DCI (n = 22)StatisticPAge, Mean ± SD57.45 ± 15.4851.82 ± 15.2763.09 ± 13.82t = −2.570.014Max diameter, Mean ± SD4.84 ± 2.764.72 ± 2.094.96 ± 3.35t = −0.290.773Neck width, Mean ± SD3.00 ± 2.532.87 ± 2.073.14 ± 2.96t = −0.350.726HWR, Mean ± SD1.07 ± 0.321.04 ± 0.261.10 ± 0.38t = −0.620.539Maximum CT value, Mean ± SD58.68 ± 11.0456.42 ± 9.2960.94 ± 12.35t = −1.370.178HGB, Mean ± SD120.77 ± 20.09127.18 ± 20.24114.36 ± 18.19t = 2.210.033DBIL, Mean ± SD5.91 ± 2.206.22 ± 2.545.59 ± 1.80t = 0.950.347IBIL, Mean ± SD9.92 ± 4.069.01 ± 3.2610.82 ± 4.63t = −1.500.141STB, Mean ± SD15.76 ± 5.5315.10 ± 5.0616.41 ± 6.01t = −0.780.437CRP, Mean ± SD27.83 ± 50.8927.02 ± 51.1628.64 ± 51.82t = −0.100.918Neutrophil, Mean ± SD10.03 ± 10.9311.94 ± 14.878.13 ± 3.95t = 1.160.253Lymphocyte, Mean ± SD1.36 ± 0.881.29 ± 0.581.44 ± 1.11t = −0.560.578Monocyte, Mean ± SD0.79 ± 1.061.03 ± 1.450.55 ± 0.27t = 1.500.140PLT, Mean ± SD210.82 ± 74.84228.09 ± 81.06193.55 ± 65.36t = 1.560.127NLR, Mean ± SD10.34 ± 12.0511.89 ± 15.838.80 ± 6.48t = 0.850.401SII, Mean ± SD2,222.91 ± 3,199.772,797.89 ± 4,272.241,647.93 ± 1,421.21t = 1.200.238SIRI, Mean ± SD8.62 ± 15.3612.28 ± 20.664.96 ± 5.29t = 1.610.115GCS, M (Q1, Q3)15.00 (9.00, 15.00)15.00 (15.00, 15.00)13.00 (8.25, 15.00)Z=-2.300.021Sex, n(%)χ2 = 7.760.005Female27 (61.36)9 (40.91)18 (81.82)Male17 (38.64)13 (59.09)4 (18.18)Hunt and Hess, n(%)-0.003115 (34.09)9 (40.91)6 (27.27)213 (29.55)10 (45.45)3 (13.64)312 (27.27)1 (4.55)11 (50.00)44 (9.09)2 (9.09)2 (9.09)Fisher, n(%)-0.96218 (18.18)4 (18.18)4 (18.18)224 (54.55)13 (59.09)11 (50.00)34 (9.09)2 (9.09)2 (9.09)48 (18.18)3 (13.64)5 (22.73)Shape regular, n(%)χ2 = 0.070.795Irregular11 (25.58)5 (23.81)6 (27.27)Regular32 (74.42)16 (76.19)16 (72.73)Vasospasm, n(%)χ2 = 8.190.004No29 (65.91)19 (86.36)10 (45.45)Yes15 (34.09)3 (13.64)12 (54.55)

Baseline demographic, clinical, and radiological characteristics of aSAH patients stratified by DCI status.

Data are presented as mean ± standard deviation (SD) for normally distributed variables, median (interquartile range, IQR) for non-normally distributed variables, or number (percentage) for categorical variables. P-values were calculated using Student's t-test, Mann-Whitney U test, or Chi-square test, as appropriate.

aSAH, aneurysmal subarachnoid hemorrhage; DCI, delayed cerebral ischemia; HWR, height-width ratio; CT, computed tomography; HGB, hemoglobin; DBIL, direct bilirubin; IBIL, indirect bilirubin; STB, serum total bilirubin; CRP, C-reactive protein; PLT, platelets; NLR, neutrophil-to-lymphocyte ratio; SII, systemic immune-inflammation index; SIRI, systemic inflammation response index; GCS, Glasgow Coma Scale; SD, standard deviation; IQR, interquartile range.

To identify independent risk factors for DCI, a multivariate logistic regression analysis was performed including variables that showed statistical significance in the univariate analysis (age, sex, Hunt-Hess grade, hemoglobin, and cerebral vasospasm). As shown in Table 2, cerebral vasospasm was identified as the strongest independent predictor of DCI (OR 7.16, 95% CI 1.52–33.67, P = 0.013). Age was also independently associated with DCI outcomes (OR 1.06, 95% CI 1.01–1.11, P = 0.044). Other factors, such as sex and Hunt-Hess grade, were not statistically significant in the multivariate model.

VariablesUnivariate analysisMultivariate analysisβS.EZPOR (95%CI)βS.EZPOR (95%CI)SexFemale1.00 (Reference)Male−1.870.70−2.660.0080.15 (0.04–0.61)Hunt and Hess11.00 (Reference)2−0.800.84−0.950.3440.45 (0.09–2.35)32.801.172.400.01716.50 (1.67–163.42)15.6-8,-14498pt40.411.130.360.7201.50 (0.16–13.75)Fisher11.00 (Reference)2−0.170.82−0.200.8380.85 (0.17–4.20)3−0.001.22−0.001.0001.00 (0.09–11.03)15.6-8,-14498pt40.511.020.500.6151.67 (0.23–12.22)Shape regularIrregular1.00 (Reference)Regular−0.180.70−0.260.7950.83 (0.21–3.29)VasospasmNo1.00 (Reference)1.00 (Reference)Yes2.030.752.690.0077.60 (1.73–33.35)1.970.792.490.0137.16 (1.52–33.67)Age0.060.022.290.0221.06 (1.01–1.11)0.050.032.020.0441.06 (1.01–1.11)GCS−0.150.09−1.760.0790.86 (0.72–1.02)Longest diameter0.030.110.300.7671.03 (0.83–1.28)N0.040.120.360.7201.05 (0.82–1.33)HWR0.600.950.630.5301.82 (0.28–11.80)Maximum CT value0.040.031.350.1771.04 (0.98–1.10)HGB−0.040.02−2.040.0410.96 (0.93–0.99)DBIL−0.140.14−0.950.3430.87 (0.66–1.16)IBIL0.120.081.450.1461.13 (0.96–1.32)STB0.040.060.790.4291.05 (0.94–1.17)CRP0.000.010.110.9151.00 (0.99–1.01)Neutrophil−0.060.06−0.910.3610.94 (0.83–1.07)Lymphocyte0.200.350.570.5701.22 (0.61–2.45)Monocyte−2.141.18−1.820.0690.12 (0.01–1.18)PLT−0.010.00−1.480.1380.99 (0.98–1.00)