This retrospective observational study continuously enrolled AIS patients admitted to the Affiliated Hospital of Shaanxi University of Chinese Medicine between February 2022 and December 2023. AIS was defined as sudden neurological dysfunction due to focal cerebral ischemia, with symptoms lasting >24 h or neuroimaging evidence of acute infarction (22). Inclusion criteria: (1) Age ≥18 years; (2) Imaging-confirmed acute infarction; (3) Complete clinical data, including demographics, cardiovascular risk factors, medical and medication history, echocardiography, and stroke causative mechanisms. Exclusion criteria: (1) Intracerebral hemorrhage or venous infarction; (2) Persistent atrial fibrillation during hospitalization (a condition precluding reliable P-wave detection); (3) History of carotid stenting; (4) Brain tumor, malignancy, head trauma, severe systemic disease (e.g., heart failure, renal failure/dialysis, respiratory failure), liver failure, substance/alcohol abuse, psychiatric disorders, or other neurological diseases (e.g., multiple sclerosis, encephalopathy, chorea, Parkinson’s); (5) Non-vascular white matter lesions (e.g., autoimmune, metabolic, toxic, infectious, sarcoidosis, radiation-induced); (6) MRI contraindications.

The study was approved by the Ethics Committee of Shaanxi University of Chinese Medicine Affiliated Hospital [Approval No. 2023(77)] and adhered to the Declaration of Helsinki. Written informed consent was obtained from all participants.

Comprehensive baseline data were collected: Demographics (sex, age, educational years, height, weight); Cardiovascular risk factors (systolic/diastolic blood pressure (SBP/DBP), smoking, alcohol use, heart rate, fasting plasma glucose (FPG), HbA1c%, total cholesterol (TC), low density lipoprotein cholesterol (LDL), triglycerides (TG)); Medical history (hypertension, diabetes, atrial fibrillation); Medication history (antihypertensives, antiplatelets, anticoagulants) within 3 months; Inflammatory markers [fibrinogen (Fbg), homocysteine (HCY)]. The presumed stroke causative mechanism was assessed based on the Trial of ORG 10172 in Acute Stroke Treatment (TOAST) classification. Data were extracted from medical records via dual verification.

Brain MRI scans (Siemens 1.5 T/3.0 T MAGNETOM SKYRA) were retrieved from the hospital’s imaging system. Sequences included DWI, T1WI, T2WI, FLAIR, and SWI. Measurements followed Standards for Reporting Vascular Changes on Neuroimaging (STRIVE) (23): (1) Lacunes: Defined as round or ovoid, subcortical CSF-signal lesions (3–15 mm diameter) on T1/T2/FLAIR. Number counted visually. (2) White Matter Hyperintensities (WMH): Abnormal hyperintense signals in white matter on T2/FLAIR. Severity was quantified using the Fazekas scale (24) for periventricular WMH (PV-WMH) and deep WMH (D-WMH), summed for a total WMH score (0–6). WMH volume (mm³) was measured using 3D Slicer (v5.0.3) by semi-automated thresholding and manual correction. WMH severity was expressed as a percentage of intracranial volume (ICV), measured similarly. (3) Enlarged Perivascular Spaces (EPVS): Defined as linear (parallel to vessels) or round/ovoid (diameter <3 mm, CSF-signal) hyperintensities on T2WI. Centrum semiovale (CSO) and basal ganglia (BG) EPVS counts were visually graded (25): Grade 0 (none), 1 (≤10), 2 (11–20), 3 (21–40), 4 (>40). (4) Cerebral Microbleeds (CMBs): Defined as small, round/ovoid, homogeneous, signal-void lesions (2–5 mm, max 10 mm) on SWI. Burden was graded (26): 0 (none), 1 (1–5), 2 (6–15), 3 (>15). (5) Brain Atrophy: Assessed by reduced brain volume, ventricular enlargement, and sulcal widening on MRI. Graded using the Global Cortical Atrophy (GCA) scale (27): 0 (none), 1 (mild), 2 (moderate), 3 (severe). Two trained raters, blinded to clinical data, assessed all CSVD markers. Disagreements were resolved by a senior physician. (6) Total CSVD Burden: Wardlaw score (0–4) (25): 1 point each for: D-WMH (Fazekas scale 2 or 3) or irregular PV-WMH (Fazekas score 3), presence of≥1 lacunes, presence of≥1 CMB, moderate/severe BG-EPVS (N > 10). The markers on MRI images were independently assessed by Q. H. and M. Y. This assessment was conducted while the raters were blinded to the clinical data of the participants. Any discrepancies between their assessments were resolved by Y. J. The kappa coefficients and corresponding p-values of CSVD markers on brain MRI between the two raters are listed as follows: Lacunes (0.724, p < 0.01), PV-WMH (0.868, p < 0.01), D-WMH (0.863, p < 0.01), tolat WMH (0.826, p < 0.01), EPVS grade (0.859, p < 0.01), CSO-EPVS (0.559, p < 0.01), BG-EPVS (0.630, p < 0.01), CMBs-grade (0.888, p < 0.01), CBM numbers (0.689, p < 0.01), and Atrophy grade (0.839, p < 0.01).

Transthoracic echocardiography (Philips EPIQ5) assessed cardiac structure/function during hospitalization. Parameters recorded: Left atrial diameter index (LADI); Measurements followed American Society of Echocardiography guidelines by an experienced sonographer blinded to other data. LADI were derived by LAD/BSA. Body surface area (BSA, m²) was calculated (28): Male S = 0.0057*height (cm) + 0.0121*weight (kg) + 0.0882; Female S = 0.0073*height (cm) + 0.0127*weight (kg) - 0.2106.

PTFV1 was calculated as the product of depth (μV) and duration (ms) of the terminal portion of the P-wave in V1. The 12-lead routine ECG was used, and the ECGs of all patients were collected in the resting state on the first day after admission. Paper ECGs were scanned, magnified 10 in Digimizer software, and three non-consecutive inverted P-waves in V1 were measured and averaged. Measurements were blinded. All ECGs were calibrated to standard settings (10 mm/mV, 25 mm/s). A P-wave terminal force in lead V1 (PTFV1) > 5,000 μV·ms was used to define atrial cardiomegaly (ACM), and a PTFV1 ≤ 5,000 μV·ms was adopted to define non-atrial cardiomegaly (NACM). Patients with unstable baselines or absent P-waves due to atrial fibrillation were excluded. Two measurers performed assessments; discrepancies were resolved by a senior cardiologist.

Continuous variables are presented as mean±SD or median (IQR); categorical variables as frequency (%). Group comparisons used Student’s t-test, Mann–Whitney U test, Chi-square test, or Fisher’s exact test. Baseline and echocardiographic characteristics were compared between ACM (PTFV1 > 5,000 μV·ms) and non-ACM (NACM, PTFV1 ≤ 5,000 μV·ms) groups. Ordinal logistic regression assessed associations between echocardiographic parameters and total Wardlaw scores. Univariate and multivariate ordinal logistic regression identified independent risk factors for higher CSVD burden (total Wardlaw score as dependent variable; clinical/laboratory parameters as independent variables). Variables with p value ≤ 0.05 in univariate analysis were entered multivariate analysis. The standardized variance inflation factor (SVIF) < 1.5 was adopted as the critical threshold for determining the absence of multicollinearity among variables. It is calculated as GVIF^(1/(2*df)), where GVIF denotes the generalized variance inflation factor and df denotes degrees of freedom. Multivariate ordinal logistic regression was constructed based on the total Wardlaw score, followed by overall significance testing and goodness-of-fit evaluation: (1) The Brant test was employed to validate the proportional odds assumption (POA), assessing whether the effects of independent variables on each category of the dependent variable were consistent; (2) The likelihood ratio test (LRT) was used to compare the full model (incorporating 8 independent variables including ACM and hypertension) with the null model (intercept-only), to verify the joint explanatory power of the independent variables for the ordered categories of the total Wardlaw score; (3) Pseudo R² and the Hosmer-Lemeshow (HL) test were utilized to evaluate the model’s goodness-of-fit to the data. The Cochran-Armitage trend test was used to assess the linear trend in the distribution of total Wardlaw scores across ACM and NACM groups. Statistical significance was set at p value < 0.05. Analyses used R software (v4.0.1).

Of 426 screened AIS patients, 103 were excluded (32 incomplete echo data, 15 unreadable ECG, 11 > 7 days post-stroke, 45 MRI artifacts/missing sequences), leaving 323 patients (240 NACM, 83 ACM). Baseline characteristics are shown in Table 1. No significant differences existed in age, sex, weight, height, BMI, education, hypertension/diabetes/atrial fibrillation history, medication use (antiplatelets, antihypertensives), or lab values (FPG, HbA1c, TG, TC, HDL, LDL, HCY, Fbg). Statistically significant differences were found only in heart rate, SBP, DBP, and LADI (higher in ACM group).

Univariate ordinal logistic regression (Table 2) identified age, hypertension, NIHSS scores, TOAST Subtypes, ACM, triglycerides, HDL, and Fbg as significant risk factors for higher total Wardlaw scores.

Eight variables screened by univariate ordinal regression analysis were entered into multivariate ordinal logistic regression analysis (Table 3). Multicollinearity analysis was conducted for the eight independent variables included in the ordinal logistic regression model. The results revealed that the SVIF were all less than 1.5, indicating no multicollinearity among all variables. The results of model validation and testing are as follows: (1) Validation of the proportional odds assumption: The Brant test yielded an omnibus chi-square value of 15.6 (p = 0.84), indicating that the model satisfied the proportional odds assumption. (2) Overall significance of the regression equation: The LRT results showed that, compared with the null model, the full model produced a chi-square value of 66.93 (P < 0.001). This demonstrates that the independent variables, when considered jointly, have significant explanatory power for the ordered categories of the total Wardlaw score, and the regression equation is statistically significant overall. (3) Model goodness-of-fit: The Nagelkerke pseudo-R² of the model was 0.218, suggesting a moderate level of explanatory power. The HL test yielded a chi-square value of 9.28 (p = 0.319), confirming that the model exhibits a good fit.

Multivariate analysis (adjusting for TOAST Subtypes, TG, HDL, and Fbg) confirmed ACM as an independent predictor of higher total Wardlaw scores (OR = 1.79, 95% CI = 1.07–3.02, p = 0.026). This indicates that, after adjusting for confounding factors, the cumulative odds of having a more severe small vessel disease burden (i.e., being in a higher Wardlaw score category) are 1.79 times higher for patients with ACM compared to those with NACM. In addition, age (OR = 1.05, 95% CI = 1.02–1.07, p < 0.001), hypertension (OR = 2.09, 95% CI = 1.29–3.42, p = 0.003), and NIHSS score (OR = 1.13, 95% CI = 1.03–1.24, p = 0.009) were also independent risk factors. This indicates that for each 1-year increase in age, the adjusted cumulative odds of being in a higher Wardlaw score category increase by a factor of 1.05. The cumulative odds of a higher Wardlaw score are 2.09 times greater in patients with hypertension than in those without. Each one-point increase in NIHSS score is associated with a 1.13-fold increase in the cumulative odds of a higher Wardlaw score.

The ACM group had significantly higher proportions of patients with scores of 3 and 4 compared to NACM (p value = 0.023 for ordinal difference). Figure 1 illustrates the distribution of total Wardlaw score.