This was a retrospective observational cohort study. We consecutively collected 467 patients with AIS who have been tested for ceramide levels from the affiliated hospital of Xuzhou Medical University from December 2021 to May 2023. The inclusion criteria were as follows: (1) confirmed diagnosis of AIS based on typical symptoms of neurological deficits and CT scans; (2) hospitalization within 72 hours of symptoms onset. Patients with active infection or any other infection within the past two weeks (n = 24), as well as patients with severe heart diseases (such as New York Heart Association class III-IV heart failure, recent myocardial infarction), lung diseases (such as chronic obstructive pulmonary disease requiring oxygen therapy), liver diseases (such as cirrhosis or liver failure), kidney diseases (such as end-stage renal disease or nephrotic syndrome), autoimmune diseases or malignant tumors (n = 25). In addition, patients with ventilator-associated pneumonia (n=30) or incomplete baseline records (n=122) were eliminated. Finally, 266 patients with completed laboratory tests within 48 hours of admission and baseline records were included and divided into the SAP group (n=96) and the non-SAP group (n=170). The process of selecting patients is provided in Figure 1. In addition, 93 age- and sex-matched healthy controls were included from individuals undergoing routine health examinations at our hospital during the study period. All controls had no history of stroke, infection, autoimmune disease, malignant tumor, or chronic inflammatory conditions, and none had a history of smoking or alcohol consumption. A post hoc sample size estimation was conducted based on plasma C16:0-ceramide levels. With a significance level (α) of 0.05 and a statistical power (1−β) of 90%, a minimum total of 78 patients was required assuming a SAP-to–non-SAP ratio of 1:2, indicating that the final sample size was adequate for statistical analysis.

The diagnosis of definite SAP was determined according to the criteria established by the Centers for Disease Control and Prevention (3). Diagnosis required the simultaneous fulfillment of systemic, respiratory, and radiological criteria:

(1) Systemic evidence of infection was defined by the presence of at least one of the following: fever (>38 °C) without another explanation; leukocytosis (>12×10⁹/L) or leukopenia (<4×10⁹/L); or, for patients aged ≥70 years, an acute alteration in mental status without another cause. (2) Respiratory manifestations required the presence of at least two of the following: new or exacerbated cough, dyspnea, or tachypnea; increased volume or purulence of respiratory secretions, or a change in sputum character; abnormal findings on lung auscultation; or impaired gas exchange, evidenced by oxygen desaturation or increased oxygen demand. (3) Radiological confirmation was established by chest imaging demonstrating new, progressive, and persistent pulmonary infiltrates, consolidation, or cavitation on at least two serial examinations.

Meanwhile, to avoid confusion with community-acquired pneumonia, all patients underwent routine clinical evaluation at admission to exclude pre-existing pulmonary infection and all the patients with SAP included in this study developed SAP more than 48 hours after admission.

Demographic data, including age, gender and premorbid risk factors containing hypertension, diabetes mellitus, coronary heart disease, atrial fibrillation and history of smoking and stroke, was gathered from medical records. Peripheral venous blood was collected from all patients within 48 hours of hospital admission. In those who later developed SAP, samples were taken before clinical onset to minimize confounding by infection-related inflammation and to evaluate ceramide as early biomarker of SAP. The laboratory results included WBC, neutrophil, lymphocyte, platelet, and albumin levels were collected. Dysphagia was diagnosed at the time of admission with a swallow test. Then, the inflammatory and immune-related indices were selected to reflect different aspects of immune status after stroke. High-sensitivity C-reactive protein (hs-CRP) was used as a marker of systemic inflammatory activation, whereas the neutrophil-to-lymphocyte ratio (NLR) reflects the imbalance between innate immune activation and adaptive immune suppression (19, 20). The platelet-to-lymphocyte ratio (PLR) and the systemic immune-inflammation index (SII), which integrate platelet, neutrophil, and lymphocyte counts, have been proposed as composite indicators of immune–inflammatory status and infection susceptibility in patients with acute ischemic stroke and stroke-associated pneumonia (20, 21). The severity of AIS was assessed by the National Institutes of Health Stroke Score (NIHSS) and modified Rankin scale (mRS). A2DS2 score was calculated (age ≥75 years=1, male=1, AF = 1, dysphagia=2, NIHSS scores: 0-4 = 0, 5-1 = 3, ≥16 = 5) at admission (10). All measurements of laboratory examination were performed by professional laboratory technicians in our hospital.

Venous blood was collected from on empty stomach at early morning and plasma was separated by centrifugation at 3000 rpm for 10 minutes at 4 °C.

For sample preparation, 100 µL of plasma was combined with 100 µL of internal standard (IS) working solution and 400 µL of ethyl acetate:isopropanol (1:4, v/v). Calibration standards were prepared similarly by replacing plasma with standard solution. After vortexing for 5 minutes, the mixture was centrifuged at 11,000 rpm for 10 minutes at 4 °C. A 300 µL aliquot of the supernatant was recentrifuged under the same conditions, and 100 µL of the resulting supernatant was used for analysis.

High performance liquid chromatography-tandem mass spectrometry (HPLC-MS/MS) analysis was performed on a triple quadrupole mass spectrometer (HPLC-MS/MS, Waters TQD Mass Detector LC/MS/MS, United States) with an electrospray ionization ion source in positive mode. Chromatographic separation was performed on an ACQUITY UPLC BEH C18 column (2.1 mm × 50 mm, 1.7 µm) using a Waters TQD triple quadrupole mass spectrometer equipped with an electrospray ionization source in positive mode. The mobile phases consisted of (A) water containing 1% formic acid and 5 mmol/L ammonium acetate, and (B) isopropanol:acetonitrile (4:3, v/v) with 5 mmol/L ammonium acetate. The flow rate was 0.3 mL/min, and the injection volume was 15 µL. Ceramides were quantified via multiple reaction monitoring (MRM).

Statistical Package for the Social Sciences 26.0 (SPSS 26.0), GraphPad Prism software 9, and R statistical software (version 4.5.0) were utilized for statistical analysis and making figures. Results were presented as percentages for categorical variables and median (interquartile ranges, IQR) for continuous variables. Mann–Whitney U test (abnormal distribution) was utilized to compare two independent samples. Categorical variables were compared by the Chi-squared test or Fisher’s exact test. Propensity score matching (PSM) was applied to minimize potential confounding between groups. Propensity scores were estimated using a multivariable logistic regression model incorporating age, sex, and medical history (hypertension, diabetes, coronary heart disease, stroke, smoking, and atrial fibrillation). Patients were matched using 1:1 nearest-neighbor matching without replacement and covariate balance before and after matching was assessed using standardized mean differences (SMDs). The least absolute shrinkage and selection operator (LASSO) regression analysis was operated with the “glmnet” package. Multivariable logistic regression analysis was utilized to identify the risk factors of SAP. ROC curves were conducted to evaluate the predictive value. The statistical significance was considered as P < 0.05 (two-tailed).

To identify the abnormal expression of ceramide in AIS, plasma ceramide levels were compared between AIS patients and healthy controls. A total of 266 eligible patients with AIS were included into analysis, with a median age of 66.0 (57.0, 72.0) years, a median NIHSS score of 4.5 (2.0, 10.0), and a median mRS score 3.0 (2.0, 4.0). Meanwhile, 93 age-and-sex matched healthy controls were included. There was no statistical difference in age and sex between two groups (all P>0.05, Table 1). Significantly, the levels of C16:0-Cer, C18:0-Cer, C24:1-Cer, C24:0-Cer, and the CERT1 score were higher in patients with AIS comparing with healthy controls (all P<0.05, Table 1).

To investigate the association between ceramide and the severity of stroke, we divided 266 AIS patients into the minor ischemic stroke (MIS) group (n=149) and the non-MIS group (n=117). MIS was defined as an NIHSS score ≤5-points. The levels of plasma C16:0-Cer, C18:0-Cer, and the CERT1 score were significantly lower in the MIS group compared with those in patients with moderate and severe ischemic stroke (all P<0.001, Table 2). The incidence of SAP was higher in the non-MIS group (P<0.001, Table 2). Besides, there were significant differences in hypertension, dysphagia, atrial fibrillation history, mRS score, A2DS2 score, WBC count, neutrophil count, lymphocyte count, hs-CRP, albumin, NLR, PLR, and SII between the two groups (all P < 0.05, Table 2). To adjust for confounding factors between the two groups, we matched age, gender, and medical history variables (hypertension, diabetes, coronary heart disease, stroke, smoking, and atrial fibrillation) using PSM. A total of 94 patients with MIS were matched 1:1 with 94 non-MIS patients. After matching, most categorical covariates achieved acceptable balance, with SMDs below 0.1, only age remained a mild imbalance, with a SMD of 0.177. The significant results obtained from this matching process were consistent with the findings before matching (Table 2).

Furthermore, Spearman’s correlation analysis revealed statistically significant yet modest positive associations of C16:0-Cer, C18:0-Cer, and CERT1 score with NIHSS score (r=0.299, P<0.001; r=0.305, P<0.001; r=0.304, P<0.001) and mRS score (r=0.256, P<0.001; r=0.230, P<0.001; r=0.244, P<0.001) in AIS patients. However, no meaningful correlation was detected in C24:1-Cer, and C24:0-Cer (all P>0.05).

The demographic data, clinical data, and ceramide levels before and after PSM were compared between the SAP group (n=96) and the non-SAP group (n=170). Before matching, SAP patients were older than non-SAP patients and had a higher incidence of atrial fibrillation and dysphagia (all P<0.001, Table 3). There were significant differences in NIHSS score, mRS score, A2DS2 score, WBC count, neutrophil count, lymphocyte count, hs-CRP, albumin, NLR, PLR and SII between the two groups (all P<0.001, Table 3). The levels of C16:0-Cer, C18:0-Cer, and the CERT1 score were significantly higher in patients with SAP compared with those in the non-SAP group (all P<0.001, Table 3). However, no difference was detected in C24:1-Cer and C24:0-Cer. After matching for age, gender, and medical history variables, a total of 76 non-SAP patients were matched 1:1 with 76 SAP patients. After matching, most categorical covariates achieved acceptable balance, with SMDs below 0.1, only coronary heart disease history remained a mild imbalance, with a SMD of 0.156. After matching, there were significant differences in the NIHSS score, mRS score, A2DS2 score, WBC count, neutrophil count, lymphocyte count, hs-CRP, albumin, NLR, PLR, and SII between the two groups (all P<0.001, Table 3). The levels of C16:0-Cer, C18:0-Cer, and the CERT1 score were still significantly higher in SAP patients (all P<0.05, Table 3).

The LASSO regression was performed on all 26 variables, and eight variables (dysphagia, A2DS2 score, neutrophil count, hs-CRP, albumin, NLR, C16:0-Cer, and CERT1 score) that had non-zero coefficients were chosen (Figures 2A, B). The eight variables were included in the multivariate regression analysis, and the results were displayed in a forest plot. The A2DS2 score, hs-CRP, C16:0-Cer, and NLR were identified as being independently associated with SAP after stroke (Figure 3). ROC curve analysis revealed that C16:0−Cer exhibited a moderate ability to discriminate SAP (AUC: 0.725; 95% CI: 0.664–0.787; P<0.001). However, its predictive performance did not exceed that of established markers: the A2DS2 score (AUC: 0.873; 95% CI: 0.827–0.919; P<0.001), hs−CRP (AUC: 0.807; 95% CI: 0.752–0.862; P<0.001), and NLR (AUC: 0.866; 95% CI: 0.813–0.920; P<0.001) (Figure 4). This suggests C16:0-Cer may serve as a complementary biomarker rather than a replacement. The optimal cut-off values, determined by maximizing the Youden index, were 4.5 points for the A2DS2 score, 0.465 μmol/L for C16:0−Cer, 1.25 mg/L for hs−CRP, and 4.3 for NLR, yielding specificities of 0.876, 0.612, 0.624, and 0.888, and sensitivities of 0.781, 0.740, 0.854, and 0.781, respectively.