Abstract

Background:

Delayed cerebral ischemia (DCI) remains a major determinant of poor outcome after aneurysmal subarachnoid hemorrhage (aSAH). Growing evidence suggests that gut microbiota–derived metabolites, including short-chain fatty acids (SCFAs) and tryptophan-related indole compounds, modulate neuroinflammation and cerebrovascular vulnerability. However, their temporal dynamics and clinical relevance after aSAH are insufficiently characterized.

Methods:

In this prospective observational study, 80 consecutive patients with aSAH were enrolled at a tertiary neurocritical care center. Serum concentrations of SCFAs (propionic, butyric, isobutyric, valeric, isovaleric, caproic acids) and tryptophan-derived metabolites (tryptophan, indole-3-propionic acid [IPA], indole-3-acetic acid, indole-3-lactic acid) were quantified using LC–MS on Day 1 and Day 9 after hemorrhage. Functional outcome at 3 months was assessed using the modified Rankin Scale (mRS), and DCI was diagnosed according to consensus criteria. Associations were analyzed using non-parametric statistics, ROC analyses, and multivariable logistic regression adjusted for established clinical confounders.

Results:

Patients with unfavorable 3-month outcomes (mRS 4–6) showed significantly lower Day 1 levels of propionic, isobutyric, and isovaleric acids, persistently reduced tryptophan at both time points, and markedly lower IPA concentrations on Day 9. DCI was associated with reduced tryptophan and propionic acid levels on both days and a pronounced decrease in IPA on Day 9. Tryptophan and propionic acid demonstrated excellent discriminative performance for outcome and DCI (AUCs up to 0.99). In multivariable models, low Day 1 propionic acid and low Day 9 IPA independently predicted unfavorable outcome, while Day 9 tryptophan, IPA, and propionic acid independently predicted DCI.

Conclusion:

Distinct temporal alterations in gut microbiota–derived metabolites after aSAH are strongly associated with functional outcome and DCI. SCFAs and tryptophan-related metabolites—particularly propionic acid, tryptophan, and IPA—emerge as promising biomarkers and potential mechanistic mediators in secondary brain injury after aSAH.

1 Introduction

Aneurysmal subarachnoid hemorrhage (aSAH) is a devastating cerebrovascular event, accounting for approximately 5–10% of all strokes yet contributing disproportionately to morbidity and mortality (1). Epidemiological analyses indicate a 28-day mortality approaching 25–30%, and among survivors, nearly half never regain their premorbid level of functioning (1, 2). Beyond the initial hemorrhage, secondary complications play a critical role in patient outcomes. Among these, delayed cerebral ischemia (DCI) remains the most clinically significant late sequela, substantially worsening neurological recovery and long-term prognosis (3). In recent years, increasing attention has been directed toward the gut–brain axis as a key modulator of neuroinflammatory and neurovascular processes relevant to acute brain injury (4, 5). The gut microbiota and its metabolites—including short-chain fatty acids (SCFAs; such as propionic, butyric, isobutyric, valeric, isovaleric, and caproic acids) and tryptophan-derived indole compounds (e.g., indole-3-propionic acid (IPA), indole-3-acetic acid (IAA), indole-3-lactic acid (ILA), and tryptophan itself)—have emerged as potential contributors to cerebrovascular vulnerability and post-stroke recovery (4–7). SCFAs are known to reinforce the intestinal barrier, exert anti-inflammatory effects, and modulate systemic immune responses, thereby influencing the severity and trajectory of neurological injury (4–8). Clinical and experimental studies in ischemic stroke have shown that reduced abundance of SCFA-producing bacteria and lower fecal SCFA levels correlate with more severe neurological deficits and poorer 90-day outcomes (7, 8). Conversely, transplantation of SCFA-rich microbiota in animal models enhances post-stroke neurobehavioral recovery, supporting a causal role for microbiota-derived metabolites in neuroprotection (9). Disturbances in gut microbiota–mediated tryptophan metabolism play a significant role in ischemic stroke pathology through effects on the gut–brain axis, and targeting key tryptophan metabolic pathways and their metabolites may offer promising therapeutic avenues (10), as indole-3-propionic acid (IPA) emerges as a particularly promising neuroprotective tryptophan metabolite that acts through antioxidant, anti-inflammatory, receptor-mediated, and neurotrophic mechanisms to disrupt the gut–inflammation–brain cycle and potentially delay or mitigate neurodegenerative and ischemic brain disorders (11). Klepinowski et al. demonstrated that patients who developed cerebral vasospasm or delayed cerebral ischemia after aSAH displayed early alterations in gut microbiome diversity—particularly reduced butyrate-producing taxa—suggesting a microbial signature associated with these complications (12). A recent study found that patients with ruptured intracranial aneurysms exhibit significantly different gut microbial compositions compared with those with unruptured aneurysms, including shifts in SCFA-producing bacterial taxa and associated metabolic pathways (13). Taken together, these observations highlight a biologically plausible and clinically relevant interplay between gut microbiota, its metabolic products, and cerebrovascular disease processes. In this context, our study aimed to quantify serum levels of key SCFAs and tryptophan-derived metabolites (IPA, tryptophan, indole-acetic acid, indole-lactic acid, propionic acid, butyric acid, isobutyric acid, valeric acid, isovaleric acid, and caproic acid) at two time points following aSAH (Day 1 and Day 9), using liquid chromatography–mass spectrometry (LC–MS), an analytical platform combining chromatographic separation with mass spectrometric detection for sensitive and specific metabolite quantification. By characterizing their temporal profiles and associations with clinical outcomes, we sought to explore their potential roles as biomarkers or mechanistic mediators in the pathophysiology of aSAH and DCI. Based on prior evidence linking gut microbiota–derived metabolites to neuroinflammation, vascular dysfunction, and post-stroke recovery, we formulated the a priori hypothesis that reduced circulating levels of SCFAs and microbiota-derived tryptophan metabolites—particularly propionic acid, tryptophan, and IPA—would be associated with unfavorable 3-month functional outcome after aSAH. Furthermore, we hypothesized that early alterations in these metabolites would predict the subsequent development of DCI, reflecting a dysregulated gut–brain–immune axis contributing to secondary brain injury.

2 Materials and methods2.1 Participants and study design

Institutional ethical approval for this prospective observational study was obtained prior to patient enrollment (IV/8468-1/2021/EKU, 27.10.2021 and BM/4629-1/2024), and written informed consent was secured from all patients or their legal representatives before participation. Consecutive adults diagnosed with aSAH at our tertiary neurocritical care center were enrolled between September 2023 and July 2025. Inclusion criteria were: (i) age ≥18 years; (ii) confirmation of aSAH by non-contrast head CT and verification of an intracranial aneurysm by CTA or DSA; and (iii) diagnosis established within 24 h following ictus. Patients with traumatic SAH, pregnancy, delayed hospital admission (>24 h after ictus), untreated aneurysms, arteriovenous malformation–related bleeding, absence of informed consent, or underlying chronic systemic diseases—including malignancy, hepatic or renal insufficiency, chronic pulmonary disease, inflammatory bowel disease, or any known chronic gastrointestinal disorder—were excluded. Additional exclusion criteria included chronic or acute infection at admission, as well as rerupture or clinical deterioration after the initial bleeding event. For all eligible patients, comprehensive clinical data were systematically collected, including demographic characteristics, vascular risk factors, presenting symptoms, neurological severity scores (WFNS and modified Fisher scale), radiological findings, and all neurocritical care interventions performed during hospitalization such as mechanical ventilation, extraventricular or lumbar drainage, and decompressive surgery. Laboratory parameters at admission—including C-reactive protein, creatinine, white blood cell count, and neutrophil–lymphocyte ratio—were also recorded. Blood samples for metabolomic measurements were drawn at two prespecified time points: Day 1 (D1), within the acute phase of hemorrhage (arterial blood sampling was performed 24 h after the ictus), and Day 9 (D9), corresponding to the period of highest DCI risk (arterial blood sampling was performed within 216 ± 4 h after the ictus). Serum levels of short-chain fatty acids (SCFAs) and tryptophan-derived indole metabolites were quantified using LC–MS analysis. All patients were followed for 3 months after the index event using structured telephone or in-person evaluations, conducted by trained assessors who had completed formal mRS certification and were blinded to biomarker data. No formal a priori sample size calculation was performed. This study was designed as a prospective exploratory observational investigation, and patient enrollment was based on consecutive inclusion during the predefined recruitment period at our tertiary neurocritical care center. At the time of study planning, insufficient prior data were available on circulating SCFA and indole metabolite effect sizes in aSAH to permit reliable power estimation. The sample size therefore reflects the total number of eligible patients enrolled within the study period.

2.2 Clinical definitions, data collection methods and outcome evaluation

Clinical data were collected prospectively from all enrolled patients at admission and throughout the acute hospitalization period. Demographic variables, vascular risk factors (hypertension, diabetes mellitus, ischemic heart disease, smoking history), and clinical presentation features—including loss of consciousness at ictus—were systematically recorded. Initial neurological severity was assessed using the World Federation of Neurosurgical Societies (WFNS) scale, while the extent of subarachnoid blood was graded using the modified Fisher (mFisher) scale. Aneurysm location, radiological findings, and the need for neurosurgical or neurocritical care interventions—such as lumbar drainage, extraventricular drainage (EVD), mechanical ventilation, and decompressive craniotomy—were extracted from clinical charts and imaging reports. Laboratory parameters obtained at admission included C-reactive protein, creatinine, white blood cell count, and neutrophil–lymphocyte ratio. The primary clinical outcome was functional status at 3 months, assessed using the modified Rankin Scale (mRS), with favorable outcome defined as mRS 0–3 and unfavorable outcome as mRS 4–6. Delayed cerebral ischemia (DCI) was defined according to the 2010 consensus criteria proposed by Vergouwen et al. (3) as either: (i) the occurrence of a new focal neurological deficit or a decrease of at least 2 points on the Glasgow Coma Scale lasting for ≥1 h, not attributable to rebleeding, hydrocephalus, seizures, metabolic disturbances, infection, or other identifiable causes after clinical and radiological evaluation; and/or (ii) the presence of a new cerebral infarction on follow-up CT or MRI not present on the immediate post-treatment scan and not attributable to procedural complications or other causes. Symptomatic angiographic vasospasm in the absence of clinical deterioration or radiologically confirmed infarction was not considered sufficient for DCI diagnosis. The diagnosis was established by agreement of at least two experienced neurointensivists blinded to biomarker data. These outcome categories were used for all subsequent comparative analyses of metabolite concentrations. The occurrence of infectious complications during hospitalization was also documented. Blood samples for metabolomic analyses were obtained on Day 1 (D1) and Day 9 (D9) following aSAH. Serum was processed and stored according to standardized protocols prior to LC–MS analysis. Quantified metabolites included short-chain fatty acids (propionic, butyric, isobutyric, valeric, isovaleric, and caproic acids) and indole-derived tryptophan metabolites (indole-3-propionic acid, tryptophan, indole-3-acetic acid, and indole-3-lactic acid).

2.3 LC–MS-based metabolite determination

Liquid chromatography–mass spectrometry (LC–MS) is an analytical technique that couples chromatographic separation of compounds with mass-based detection, allowing precise quantification of small-molecule metabolites in complex biological samples. Short-chain fatty acids (SCFAs) and indole derivatives were quantified following an established method (14) with several modifications. Human serum samples were thawed at 4 °C under constant shaking using a Genie 2 Vortex (Scientific Industries, Bohemia, NY, United States) for 30 min. Sixty microliters of serum were mixed with 420 μL ice-cold methanol (Supelco 1.06035) in a 1.5 mL microcentrifuge tube for protein precipitation. Samples were vortexed at 1000 rpm for 10 min, then centrifuged at 15,000 rpm for 10 min at 4 °C to obtain the serum extract. Sixty microliters of the resulting supernatant were combined on ice with 30 μL of 0.2 M 3-NPH and 30 μL of 0.12 M EDC. The reaction was incubated at 40 °C for 25 min without shaking, followed by centrifugation at 15,000 rpm for 10 min at 4 °C. The clarified supernatant was transferred into HPLC vials (VP91, Zhejiang Aijiren Technology Inc., Qujiang, Quzhou, Zhejiang, China) and sealed with PTFE/silicone septa (SC9291, Zhejiang Aijiren Technology Inc.). SCFAs were separated on a Waters HSS T3 column (1.7 μm, 2.1 × 100 mm) using a Waters ACQUITY Premier UPLC coupled to a Waters TQS-Micro tandem mass spectrometer. Mobile phase A consisted of LC–MS–grade water (VWR 83645.320) with 0.1% formic acid (Sigma 5.33002), and mobile phase B consisted of 0.1% formic acid in 2-propanol (VWR 84881.320):acetonitrile (VWR 83640.320). The injection volume was 3 μL and the flow rate was 0.3 mL/min. The LC gradient was programmed as follows: 0–0.2 min, 0% B; 0.2–0.7 min, 0–15% B; 0.7–2 min, 15% B; 2–6 min, 15–30% B; 6–12 min, 30–72% B; 12–12.4 min, 72–100% B; 12.4–14 min, 100% B; 14–14.1 min, 100–0% B; 14.1–15.6 min, 0% B. The autosampler was maintained at 10 °C and the column at 55 °C. Mass spectrometric detection was performed in negative electrospray ionization mode. Source settings were: capillary voltage 0.5 kV, desolvation temperature 600 °C, desolvation gas flow 1,000 L/h, and cone gas flow 1 L/h. Data were acquired in multiple-reaction monitoring mode using previously described transitions (14). The cone voltage was set to 10 V for all target analytes. Collision energy was set to 10 V for propionic acid and 20 V for all other SCFAs. Calibration was performed using an external dilution series of an SCFA mix as described (14). Samples were analyzed in four experimental batches. Each batch included three pooled QC samples and one blank to monitor and correct for batch effects. Quantitative data were obtained using an eight-point external calibration curve fitted either with linear regression or a second-order polynomial model. All R2 values were above 0.99. For LOQ estimation, we evaluated the signal-to-noise (S/N) ratios of analyte peaks across calibration levels. In parallel, we assessed S/N values in blank samples. LOQ was defined as the lowest calibration level with a signal-to-noise ratio ≥10 and at least twofold higher than the blank signal. Quantitative data were normalized using EigenMS batch normalization in MetaboAnalyst 6.0 (15). Acetic acid was excluded from downstream analysis due to high blank signal. Analytical performance of the transferred LC–MS method was evaluated by assessing calibration linearity, signal-to-noise ratios, and technical reproducibility using pooled QC samples across analytical batches. Calibration curves demonstrated excellent linearity (R2 > 0.99), and limits of quantification were defined based on signal-to-noise criteria. Technical reproducibility assessed from repeated QC injections showed low analytical variability. Statistical analysis and data visualization were performed using IBM SPSS Statistics v25 (IBM, Armonk, New York, United States) and GraphPad Prism 10 (GraphPad Software, LLC, San Diego, California, United States).

2.4 Statistical analysis

All statistical analyses were conducted using IBM SPSS Statistics version 25 (IBM Corp., Armonk, NY, United States) and GraphPad Prism version 10 (GraphPad Software, LLC, San Diego, CA, United States). Continuous variables were inspected for distributional characteristics using histograms, Q–Q plots, and the Shapiro–Wilk test. Because most biochemical and clinical parameters exhibited non-normal distributions, non-parametric tests were applied throughout the analyses unless otherwise indicated. Categorical variables were analyzed using contingency-table statistics. A two-sided p-value <0.05 was considered statistically significant. Baseline demographic and clinical characteristics were summarized as mean ± standard deviation (SD) for normally distributed variables, median with interquartile range (IQR) for non-normally distributed data, and frequencies with percentages for categorical variables. Between-group comparisons (favorable vs. unfavorable 3-month functional outcome) were performed using: Independent-samples t-tests for continuous variables with approximate normality, Mann–Whitney U tests for non-normally distributed continuous data (e.g., WFNS, mFisher score, inflammatory markers), Chi-square tests or Fisher’s exact tests for categorical variables (e.g., hypertension, diabetes, aneurysm location, DCI), as appropriate. Serum concentrations of SCFAs and indole-derived metabolites measured on Day 1 and Day 9 were compared between clinical subgroups. For each metabolite, differences between groups (favorable vs. unfavorable outcome; DCI vs. No-DCI) were analyzed using Mann–Whitney U tests due to non-parametric distributions and unequal variances. Results are illustrated as box-and-whisker plots with median, IQR, and full data range. Statistical significance is indicated using the standard notation (p < 0.05; p < 0.01; p < 0.001; p < 0.0001). Given the number of biomarker comparisons, false discovery rate (FDR) correction using the Benjamini–Hochberg procedure was applied as a sensitivity analysis. To assess the discriminative performance of individual metabolites for predicting (i) unfavorable 3-month outcome and (ii) the development of delayed cerebral ischemia (DCI), ROC curve analyses were conducted for each metabolite at both time points (Day 1 and Day 9). For each ROC model, the following were computed: Area under the curve (AUC) with 95% confidence intervals (DeLong method), Sensitivity and specificity at optimal cut-off values defined by Youden’s index, Corresponding p-values for AUC significance testing. ROC analyses were performed in a univariable manner for each metabolite separately and were not adjusted for clinical covariates. Adjustment for established clinical predictors was conducted exclusively in the multivariable logistic regression models. To evaluate the independent association between selected metabolites and clinical outcomes, multivariable logistic regression models were constructed for: 3-month functional outcome (favorable vs. unfavorable), occurrence of DCI. Each model included clinical covariates chosen a priori based on known prognostic relevance and to control for confounding: Outcome prediction models (Table 1): age, WFNS grade, and infection status. DCI prediction models (Table 2): age, hypertension, and modified Fisher score. For each metabolite (analyzed in separate models): Odds ratios (ORs) and 95% confidence intervals (CIs) were calculated per unit increase in metabolite concentration, Wald χ2 statistics assessed predictor significance. Model performance was summarized by Nagelkerke’s R2, reflecting explained variance. To evaluate the interrelations among SCFAs and indole-derived metabolites at both time points, Spearman’s rank correlation coefficients were calculated. Correlation matrices were visualized as heatmaps, using a color gradient from −1 to +1. Autocorrelations between Day 1 and Day 9 levels were examined to assess temporal stability of metabolite patterns.

VariableOR (95% CI)Wald χ2P-valueModel fit (Nagelkerke R2)Propionic acid D10.49 (0.29–0.81)7.60.0060.68IPA D90.09 (0.02–0.38)10.70.0010.70Tryptophan D10.53 (0.28–1.00)3.80.0510.64Isobutyric acid D10.95 (0.62–1.48)0.040.8420.64Isobutyric acid D90.67 (0.49–0.90)6.60.010.67Isovaleric acid D10.28 (0.1–0.85)4.90.0260.66

SCFA levels, IPA and 3-month outcomes in aSAH patients: ORs at D1 and D9.

ORs (odds ratios, 95% CI: confidence intervals), Wald χ2, and p-values for SCFAs (short-chain fatty acids: propionic acid, IPA: indole-3-propionic acid, isobutyric acid, isovalerianic acid) measured at D1 (day 1) and D9 (day 9) in aSAH patients; model fit indicated by Nagelkerke R2. IPA, indol-3-propionic acid; aSAH, aneurysmal subarachnoid hemorrhage; OR, odds-ratio.

VariableOR (95% CI)Wald χ2P-valueModel fit (Nagelkerke R2)IPA D10.55 (0.17–1.82)0.90.3290.34IPA D90.13 (0.02–0.86)4.50.0340.5Trytophane D10.87 (0.38–1.9)0.10.7310.32Tryptophan D90.02 (0.01–0.05)6.7<0.0010.43Propionic acid D10.66 (0.41–1.09)2.60.1080.37Propionic acid D90.51 (0.32–0.81)8.20.0040.52Isovaleric acid D10.31 (0.1–0.8)3.90.0480.41Caproic acid D10.24 (0.08–0.77)5.70.0170.46

Binary logistic regression analysis of metabolite levels associated with the development of DCI in aSAH patients (adjusted for age, hypertension, and mFisher score).

ORs (odds ratios, 95% CI: confidence intervals), Wald χ2, and p-values for SCFAs (short-chain fatty acids: propionic acid, IPA: indole-3-propionic acid, isobutyric acid, isovalerianic acid) measured at D1 (day 1) and D9 (day 9) in aSAH patients; model fit indicated by Nagelkerke R2. IPA, indol-3-propionic acid; aSAH, aneurysmal subarachnoid hemorrhage; OR, odds-ratio.

3 Results3.1 Patients characteristics

A total of 131 patients were screened for eligibility. Of these, patients were excluded for the following reasons: traumatic subarachnoid hemorrhage (n = 4), delayed hospital admission (n = 6), arteriovenous malformation–related hemorrhage (n = 1), chronic systemic disease (n = 12), chronic obstructive pulmonary disease (n = 3), chronic gastrointestinal diseases (including inflammatory bowel disease, celiac-related disorders, and gastritis; n = 4), incomplete biomarker sampling (n = 11), loss to follow-up (n = 6), aneurysm rerupture prior to treatment (n = 1), and chronic infection or undefined immunological disease (including pyelonephritis; n = 3). After exclusions, 80 patients were included in the final analysis. All included patients underwent endovascular treatment; no patients were treated with open surgical clipping. The mean age of the total cohort was 58 ± 11 years, with patients in the favorable outcome group being slightly younger (56 ± 9 years) than those in the unfavorable group (60 ± 13 years). Females represented 74% of the study population, accounting for 64% of the favorable group and 84% of the unfavorable group. Hypertension was present in 56% of all patients, while diabetes occurred in 11%, the latter being more frequent among those with unfavorable outcomes (18% vs. 5%). Smoking was reported in 49% of patients, with comparable proportions across outcome groups. Ischemic heart disease was documented in 50% of the cohort, occurring in 43% of favorable and 57% of unfavorable cases. Loss of consciousness during ictus was observed in 35% of all patients. Baseline clinical severity differed markedly between groups: the median WFNS score was 1 (IQR 1–2) in the favorable group and 4 (IQR 2–5) in the unfavorable group, while the median modified Fisher score was 2 (IQR 2–3) and 3 (IQR 3–4), respectively. Inflammatory markers showed notable differences, with median C-reactive protein levels of 7 mg/L (IQR 2–13) in the favorable group and 29 mg/L (IQR 9–73) in the unfavorable group, and neutrophil–lymphocyte ratio values of 4.2 (IQR 3–7) versus 9.1 (IQR 5–12). Among infected patients treated with antibiotics (n = 31), antibiotic therapy was initiated a median of 7 days after D1 sampling (IQR 7–8; range 4–10). The need for extraventricular drainage (10% vs. 71%) and mechanical ventilation (7% vs. 82%) was substantially more common in the unfavorable outcome group, and delayed cerebral ischemia and infections were also more prevalent in this group (53 and 68%, respectively) (Table 1).

3.2 Associations of SCFAs, IPA, and tryptophan metabolites with 3-month outcome after aSAH

Figure 1 presents the serum concentrations of short-chain fatty acids (SCFAs), IPA and tryptophan on D1 and D9 according to 3-month functional outcome. For IPA (Figure 1A), no significant difference appeared between Fav-D1 and Unfav-D1 (ns), whereas Fav-D9 showed substantially higher concentrations than Unfav-D9 (p < 0.0001). Tryptophan (Figure 1B) levels were markedly higher in the Fav-D1 group compared with Unfav-D1 (p < 0.0001), and a similarly strong difference was observed between Fav-D9 and Unfav-D9 (p < 0.0001). Propionic acid (Figure 1C) levels were significantly lower in the unfavorable outcome group on Day 1 (p < 0.001), whereas no difference was observed on D9. Isobutyric acid (Figure 1D) also differed between groups, with lower concentrations in the unfavorable group on D1 (p < 0.05) and on D9 (p < 0.0001). In contrast, butyric acid (Figure 1E) showed no significant difference at either time point. Isovaleric acid (Figure 1F) was significantly reduced in the unfavorable outcome group on D1 (p < 0.05), while Day 9 levels did not differ between groups. Valeric acid (Figure 1G) and caproic acid (Figure 1H) showed no significant differences on either Day 1 or Day 9 (all p > 0.05). Indole acetic acid (Figure 1I) levels did not differ significantly between groups at either time point (both comparisons marked as ns). Likewise, indole lactic acid (Figure 1J) displayed no significant differences between Fav and Unfav groups on day 1 or day 9 (ns for both comparisons).

IPA, tryptophan, and short-chain fatty acid concentrations on Day 1 and Day 9 according to 3-month functional outcome in patients with aSAH. Boxplots illustrating serum concentrations of IPA, tryptophan, and short-chain fatty acids (SCFAs) in aSAH patients with favorable (Fav) and unfavorable (Unfav) 3-month outcomes, measured on Day 1 (D1) and Day 9 (D9). Lines above the boxplots indicate between-group comparisons for each time point. Statistical significance is denoted as follows: **** p < 0.0001; *** p < 0.001; ** p < 0.01; * p < 0.05; ns, not significant; IPA, indole-3-propionic acid. (A) Indole-3-propionic acid (IPA); (B) Tryptophan; (C) Propionic acid; (D) Isobutyric acid; (E) Butyric acid; (F) Isovalerianic acid; (G) Valeric acid; (H) Caproic acid; (I) Indole-3-acetic acid; (J) Indole-3-lactic acid. In the favorable outcome group (Fav), the number of patients was n = 42, while in the unfavorable outcome group (Unfav), n = 38.

The ROC analyses show that tryptophan (Figures 2A,B) provides excellent discriminatory performance at both time points. On day 1, tryptophan reached an AUC of 0.990 (95% CI: 0.97–1.00, p < 0.0001) with a sensitivity of 97.5% and specificity of 97%. Similarly, on day 9, its AUC remained high at 0.980 (95% CI: 0.94–1.00, p < 0.0001), again with 97.5% sensitivity and 97% specificity. Indole-3-propionic acid (Figure 2C) measured on day 9 showed an AUC of 0.873 (95% CI: 0.79–0.95, p < 0.0001), accompanied by 90.5% sensitivity and 75.8% specificity. Propionic acid (Figure 2D) on day 1 exhibited a moderate AUC of 0.734 (95% CI: 0.62–0.85, p < 0.001) with sensitivity and specificity of 82.5 and 59.5%, respectively. Isobutyric acid (Figures 2E,F) on day 1 showed an AUC of 0.688 (95% CI: 0.57–0.81, p < 0.005) with sensitivity of 67.5% and specificity of 59.5%, while on day 9 its AUC increased to 0.780 (95% CI: 0.68–0.88, p < 0.001) with sensitivity of 73.2% and specificity of 65.6% (Figure 2).

ROC curve analysis of tryptophan, indole- and short-chain fatty acid–related metabolites on Day 1 and Day 9 after aSAH in relation to the 3-month outcome. Each ROC panel includes the corresponding 95% confidence interval (CI), p-value, sensitivity, and specificity. (A) Tryptophan D1; (B) Tryptophan D9; (C) Indole propionic acid D9; (D) Propionic acid D1; (E) Isobutyric acid D1; (F) Isobutyric acid D9. AUC, area under the curve; CI, confidence interval; D1, Day 1; D9, Day 9.

The multivariable logistic regression model—including age, WFNS grade, and the presence of infection as covariates—evaluated the association between selected metabolites and 3-month outcomes in patients with aSAH. Propionic acid on day 1 was significantly associated with 3-month outcome (OR = 0.49, 95% CI 0.29–0.81; Wald χ2 = 7.6; p = 0.006; R2 = 0.68). IPA measured on day 9 showed a strong association with the outcome (OR = 0.09, 95% CI 0.02–0.38; Wald χ2 = 10.7; p = 0.001; R2 = 0.70). Tryptophan on day 1 demonstrated a near-significant relationship (OR = 0.53, 95% CI 0.28–1.00; Wald χ2 = 3.8; p = 0.051; R2 = 0.64). Additional metabolites—such as isobutyric acid on day 1 (OR = 0.95, p = 0.842) and day 9 (OR = 0.67, p = 0.01), as well as isovaleric acid on day 1 (OR = 0.28, p = 0.026) showed variable associations with 3-month outcome, with Nagelkerke R2 values ranging from 0.64 to 0.67 (Table 2).

In incremental value analyses for 3-month outcome, adding Day 1 propionic acid to the baseline clinical model (including WFNS, age, and covariates) significantly improved model fit (Δχ2 = 13.45, df = 1, p < 0.001). Discrimination increased from an AUC of 0.862 (95% CI 0.770–0.954) for the clinical model alone to 0.916 (95% CI 0.845–0.986) after inclusion of propionic acid. Model calibration remained acceptable (Hosmer–Lemeshow p = 0.142). These findings indicate that Day 1 propionic acid provides incremental prognostic value beyond established clinical severity measures.

After false discovery rate (FDR) correction using the Benjamini–Hochberg procedure, the principal associations (Day 1 propionic acid, Day 1 and Day 9 tryptophan, and Day 9 IPA) remained statistically significant.

3.3 Associations of SCFAs, IPA, and tryptophan metabolites with DCI after aSAH

The metabolite analyses revealed distinct concentration patterns between patients who developed DCI and those who did not (Figure 3). IPA levels (Figure 3A) showed no significant difference on D1 (ns), whereas a marked decrease was observed in DCI patients on D9 (p < 0.0001); correspondingly, IPA exhibited a moderate predictive performance (Figures 3F,G) on D1 (AUC = 0.672) and a substantially higher accuracy on D9 (AUC = 0.812). Tryptophan concentrations (Figure 3B) were significantly lower in DCI patients on both D1 (p < 0.0001) and D9 (p < 0.001), and its ROC curves (Figures 3J,K) demonstrated excellent discriminative ability at both time points (AUC = 0.823 on D1 and AUC = 0.824 on D9). Propionic acid (Figure 3C) similarly showed strong group differences, with elevated levels in patients without DCI on D1 (p < 0.001) and D9 (p < 0.0001); its predictive performance (Figures 3H,I) was consistently high across time points (AUC = 0.762 on D1 and AUC = 0.811 on D9). In contrast, isovaleric acid (Figure 3D) displayed a significant difference only on D1 (p < 0.05) but not on D9 (ns), while its predictive capacity (Figure 3L) remained moderate (AUC = 0.664 on D1). Caproic acid (Figure 3E) followed a similar pattern, showing a significant decrease in DCI patients on D1 (p < 0.01) but no difference on D9, with a moderate AUC of 0.737 on D1 (Figure 3M). Tryptophan and propionic acid emerged as the strongest discriminators between DCI and No-DCI patients, supported by both their consistent group differences and their high AUC values, while IPA exhibited improved predictive value at the later time point.

Metabolite profiles and their predictive performance in aSAH according to DCI status. Indol-3-propionic acid (A) serum concentrations on D1 and D9 comparing DCI and no-DCI groups, showing no difference on D1 and a highly significant increase in DCI patients on D9. Tryptophan (B) levels at D1 and D9, demonstrating strong and persistent group differences with higher values in the DCI group. Propionic acid (C) concentrations on D1 and D9, both time points showing marked elevations in DCI patients. Isovaleric acid (D) levels, with a significant group difference on D1 but no separation on D9. Caproic acid (E) concentrations, significantly differing between groups on D1 but not on D9. ROC curves for the examined metabolites: indole-propionic acid (F,G) on D1 (AUC = 0.672) and D9 (AUC = 0.812), propionic acid (H,I) on D1 (AUC = 0.762) and D9 (AUC = 0.811), tryptophan (J,K) on D1 (AUC = 0.823) and D9 (AUC = 0.824), isovaleric acid (L) on D1 (AUC = 0.664), and capric acid (M) on D1 (AUC = 0.737), demonstrating varying degrees of predictive performance. DCI, delayed cerebral ischemia; IPA, indole-3-propionic acid; D1/D9, Day 1/Day 9; AUC, area under the curve. In the DCI group, the number of patients was n = 22, while in the no-DCI group, n = 68.

Similarly, for DCI, the key associations (Day 1 and Day 9 tryptophan, Day 9 propionic acid, and Day 9 IPA) remained significant after FDR correction.

An other multivariable logistic regression models—including age, hypertension, and mFisher score as covariates—evaluated the associations between several metabolites and the development of delayed cerebral ischemia (DCI) in patients with aSAH. IPA measured on day 1 showed no significant relationship with DCI (OR = 0.55, 95% CI 0.17–1.82; Wald χ2 = 0.9; p = 0.329; R2 = 0.34), whereas IPA on day 9 was significantly associated with DCI (OR = 0.13, 95% CI 0.02–0.86; Wald χ2 = 4.5; p = 0.034; R2 = 0.50). Tryptophan levels on day 1 were not related to DCI (OR = 0.87, p = 0.731), while day 9 tryptophan levels showed a strong association (OR = 0.02, 95% CI 0.01–0.05; Wald χ2 = 6.7; p < 0.001; R2 = 0.43). Propionic acid on day 1 did not reach statistical significance (OR = 0.66, p = 0.108), in contrast to its day 9 level, which was significantly associated with DCI (OR = 0.51, 95% CI 0.32–0.81; Wald χ2 = 8.2; p = 0.004; R2 = 0.52). In addition, isovaleric acid on day 1 (OR = 0.31, p = 0.048; R2 = 0.41) and caproic acid on day 1 (OR = 0.24, p = 0.017) were also significantly associated with DCI occurrence (Table 3). In incremental value analyses for DCI prediction, addition of tryptophan (Day 9) to the baseline clinical model significantly improved model fit (Δχ2 = 18.68, df = 1, p < 0.001). Discrimination increased from an AUC of 0.783 (95% CI 0.645–0.921) for the clinical model alone to 0.909 (95% CI 0.830–0.989) after inclusion of the biomarker. Model calibration remained acceptable (Hosmer–Lemeshow p = 0.645). These findings indicate substantial incremental prognostic value beyond established clinical predictors. Among the 22 patients who developed DCI, the median time from ictus to DCI was 9 days (IQR 8–10; range 5–13). Seven patients developed DCI before Day 9 sampling, five on Day 9, and ten after Day 9.

VariableTotal (n = 80)Favorable (n = 42)Unfavorable (n = 38)P-valueAge (mean±SD)58 ± 1156 ± 960 ± 130.097Female, N (%)59 (74)27 (64)32 (84)0.074Hypertension, N (%)45 (56)21 (50)24 (63)0.266Diabetes, N (%)9 (11)2 (5)7 (18)0.078Smoking, N (%)39 (49)22 (52)17 (45)0.512IHD, N (%)40 (50)18 (43)22 (57)0.263Loss of consc. During ictus, N (%)28 (35)12 (29)16 (42)0.245WFNS, median (IQR)2 (1