Evaluation of carotid intima-media thickness and ultrafast pulse wave velocity for cardiovascular events and all-cause mortality in hemodialysis patients: impact of diabetes Authors: Xiaoye Zhu, Linna Zheng, Weibin Zhang, Xiaowan Liang, Yuan Ren, Qiaozhi Gu, Yuelin Zhang, Li You, Jun Xue, Chuanming Hao, Yuanhao Wu Published: 2026 ABSTRACT AbstractArterial remodeling is common in end-stage renal disease (ESRD). This prospective cohort study explored the potential prognostic value of carotid intima-media thickness (CIMT) and ultrafast pulse wave velocity (ufPWV) - including pulse wave velocity at the beginning (PWV-BS) and end of systole (PWV-ES) for cardiovascular (CV) events and all-cause mortality in 115 maintenance hemodialysis (MHD) patients followed for 39 months. Compared with non-ESRD controls, MHD patients exhibited significantly higher CIMT (1.60 vs 0.80 mm, p = 0.004), while ufPWV parameters appeared comparable. During follow-up, 21 CV events and 17 deaths were recorded. In univariable analysis, CIMT was associated with CV events and positively correlated with diabetes (r = 0.271, p = 0.003). However, after adjusting for diabetes, this association was no longer statistically significant (HR 1.350, 95% CI 0.865–2.106, p = 0.186). Neither PWV-BS nor PWV-ES was associated with outcomes in the overall cohort. Preliminary subgroup analysis of patients with diabetes (n = 36, 15 CV events) observed an inverse association between left PWV-ES and CV events. In conclusion, CIMT was no longer significantly associated with CV events after adjustment for diabetes in the limited model, whereas local ufPWV parameters were not significantly associated with CV events or all-cause mortality in the analyses performed within the overall MHD population. Given the issue of multiple comparisons and small sample size, the isolated finding between left PWV-ES and CV events in the diabetic subgroup must be treated strictly as hypothesis-generating, requiring independent validation in a larger cohort.Keywords: Carotid intima-media thickness, ultrafast pulse wave velocity, hemodialysis, cardiovascular events, all-cause mortality Introduction Cardiovascular disease (CVD) is the leading cause of morbidity and mortality in patients undergoing maintenance hemodialysis (MHD) [1]. Accelerated vascular remodeling, including both atherosclerosis and arteriosclerosis, contributes substantially to this excess risk. In dialysis patients, both atherosclerosis (mainly affecting the intima of the arteries) and arteriosclerosis (affecting predominantly the media of large and middle-sized arteries diffusely) are highly prominent [2]. This excess risk is attributed to accelerated vascular injury driven by uremia, inflammation, and mineral metabolism abnormalities [3]. Carotid intima-media thickness (CIMT) [4] and carotid-femoral pulse wave velocity (PWV) [5] are commonly used noninvasive markers reflecting structural and functional vascular changes. Although CIMT was markedly increased in MHD patients and had been reported to predict cardiovascular (CV) events in several studies [2,6], its independent prognostic value in this population remains inconsistent [7,8]. Similarly, elevated carotid-femoral pulse wave velocity (cfPWV) had been associated with increased risks of all-cause and CV mortality in MHD patients [9–11]. However, conflicting evidence exists, as some studies had reported no significant association between cfPWV and cardiac death or CV events [12], whereas other studies had identified cfPWV as an independent predictor of CV events [13]. Taken together, although CIMT and cfPWV are established markers of atherosclerosis and arterial stiffness, their independent prognostic value in the dialysis population remains controversial. Recently, ultrafast PWV (ufPWV) is a novel imaging technique that enables real-time assessment of local arterial stiffness [14,15]. However, its clinical relevance in dialysis populations has not been fully established. In addition, diabetes is highly prevalent in MHD patients and is strongly associated with both vascular remodeling and CV risk. Whether the predictive value of vascular markers is independent of diabetes remains unclear. Therefore, this study aimed to evaluate the prognostic value of CIMT and ufPWV for CV events and mortality in MHD patients, with particular attention to the influence of diabetes. Methods Study design and participants This single-center prospective cohort study was conducted at the Blood Purification Center of Huashan Hospital, Baoshan Branch, Fudan University. A total of 194 MHD patients were initially assessed between July 2022 and September 2022. The inclusion and exclusion criteria for MHD patient enrollment were summarized below: Inclusion Criteria: (1) Age ≥ 18 years; (2) Clinically diagnosed with end-stage renal disease (ESRD) and undergoing stable MHD for at least 6 consecutive months who had no residual urine output; (3) Hemodialysis prescription consisting of a standard regimen (typically 4 h per session, 3 times per week). Exclusion Criteria: (1) Acute myocardial infarction, severe or refractory heart failure defined as NYHA functional class III or IV, cerebral infarction, cerebral hemorrhage within the preceding 3 months; (2) Short life expectancy, non-renal organ failure, untreated malignancy diagnosed within 5 years; (3) Active gastrointestinal bleeding and active systemic infection within 1 month; (4) Prior history of carotid endarterectomy, carotid stenting, or severe local neck anatomical distortion; (5) Dementia, inability to cooperate or provide informed consent, or refusal to participate. We followed all MHD patients until December 2025. Seventy-nine patients were excluded. Consequently, a final cohort of 115 eligible MHD patients was enrolled in this study. Concurrently, non-ESRD controls (n = 82) were established by recruiting consecutive outpatients who underwent routine carotid ultrasonography in our hospital’s ultrasound department during the same period. The key inclusion criterion was the absence of stage 5 chronic kidney disease; patients receiving maintenance hemodialysis or peritoneal dialysis were strictly excluded. A random sampling and rough frequency-matching strategy based on age and sex were utilized to ensure that the overall demographic distributions remained statistically comparable between the MHD patients and non-ESRD controls (p > 0.05). The study was conducted in accordance with the Declaration of Helsinki. The study protocol has been approved by the Institutional Ethics Review Board of Huashan Hospital, Fudan University (approval # KY2016-394). All participants provided written informed consent before the study began. Measurement of CIMT and PWV CIMT and PWV were measured before the hemodialysis session by two expert radiologists who were blinded to the clinical and laboratory data of patients. The Aixplorer ultrafast commercial scanner (SuperSonic Imagine, Aix-en-Provence, France) was used to determine the CIMT and ufPWV with a linear array probe (SL 10-2, frequency: 2–10 MHz; center frequency: 8.0 MHz). The patient was placed in a supine position and the neck was fully exposed and relaxed, and the head was turned to one side. The common carotid artery was longitudinally scanned in B-mode to achieve a high-resolution image of the intima-media layer. Following image optimization, an automated tracking system was deployed, utilizing a sampling window with a standardized width of 1.0 cm. This window was targeted at the posterior wall of the distal common carotid artery, specifically located 1.0–1.5 cm proximal to the carotid bulb. A strict quality control threshold was implemented: the valid CIMT tracking segment length had to exceed 70% of the total sampling window width for the measurement to be accepted (Figure 1(A)). The presence of carotid plaque was defined as focal CIMT ≥ 1.5 mm or a focal thickening ≥ 50% of the surrounding CIMT value. The maximum thickness of CIMT was used for analysis. When there were focal protrusions or plaques, the maximum thickness of the plaques was taken as CIMT [2]. Protocols for assessing CIMT and ufPWV. (A) For CIMT measurement, a white ROI box was positioned on the posterior wall of the commoncarotid artery. The intimal and medial borders of the common carotid artery were automatically detected and traced as two dotted lines, yielding a mean CIMT within the ROI at a 78% fit accuracy. (B) PWV-BS and PWV-ES were evaluated using ufPWV imaging. A yellow ROI box was placed to encompass the common carotid artery, where a red line automatically tracked its anterior and posterior walls. The PWV-BS and PWV-ES values within the ROI were then calculated, maintaining a tolerance of Δ ± less than 20% of the values of both PWV-BS and PWV-ES. The participants were instructed to hold their breath and the ‘PWV’ key was pressed to activate ultrafast imaging acquisition, which was completed within 3 s. After obtaining a stable ufPWV image, a region of interest (ROI) was selected and adjusted to cover a section of the common carotid artery. The software automatically recorded the PWV at the beginning of systole (PWV-BS) and PWV at the end of systole (PWV-ES) with their respective standard deviations (SDs) (denoted by Δ ±; Figure 1(B)). For quality control, ufPWV acquisition was considered valid when Δ ± was less than 20% of the values of both PWV-BS and PWV-ES [16]. To evaluate measurement stability, an inter-observer and intra-observer reproducibility assessment were performed in a random subset of 20 participants. For CIMT, the inter-observer and intra-observer Intraclass Correlation Coefficient (ICC) were 0.80 and 0.85, and the coefficient of variation (CoV) was 6.0% and 4.0%, respectively. For PWV-BS, the inter- and intra-observer ICCs were 0.68 and 0.72, with corresponding CoVs of 17.0% and 14.5%, respectively. Similarly, PWV-ES demonstrated an inter-observer ICC of 0.72 (CoV = 16.5%) and an intra-observer ICC of 0.77 (CoV = 14.0%). Data collection We collected the dialysis and clinical feature data of the participants included in the study, including sex, age, dialysis vintage, body mass index, average pre-dialysis systolic blood pressure (SBP) and diastolic blood pressure (DBP), average net ultrafiltration volume (UFV), history of diabetes and CV disease. Kt/V was calculated using the Daugirdas formula. The biochemical parameters included N-terminal pro-B-type natriuretic peptide (NT-proBNP), C-reactive protein (CRP), intact parathyroid hormone (iPTH), calcium (Ca), phosphorus (P), calcium-phosphorus product (Ca × P), hemoglobin, albumin, serum creatinine, lipid profile. All clinical, biochemical, and imaging evaluations (including blood sampling, carotid ultrasonography, and echocardiography) for all participants were performed at the beginning of the study (baseline). For every enrolled MHD patient, to minimize the confounding effects of fluctuating volume status, uremic toxin levels, and blood pressure throughout the week, all assessments were strictly and uniformly scheduled immediately prior to the first hemodialysis session of the week. During the follow-up period, CV events were rigorously defined as a composite of fatal and non-fatal CV conditions, including ischemic heart disease (acute myocardial infarction and unstable angina requiring hospitalization), coronary revascularization procedures (percutaneous coronary intervention or coronary artery bypass grafting), cerebrovascular events (ischemic stroke, hemorrhagic stroke, or transient ischemic attack confirmed by computed tomography or magnetic resonance imaging), congestive heart failure requiring acute hospitalization, new-onset severe arrhythmias (such as atrial fibrillation, ventricular tachycardia/fibrillation, or high-degree atrioventricular block) requiring clinical intervention, and CV mortality (including sudden cardiac death, fatal myocardial infarction, fatal stroke, and death due to end-stage heart failure). Statistical analysis For categorical variables, frequency and percentage are used for descriptive statistics, while chi-square test is used for group comparisons. For continuous variables with normal distribution, mean ± standard deviation is used for presentation, and t-test is used for group comparisons. For continuous variables with skewed distribution, median and quartiles are used for presentation, and Wilcoxon rank sum test is used for group comparisons. Univariable analysis was conducted using Cox proportional hazards models, with all clinically relevant factors being included. The number of variables in each multivariable model was restricted based on the low event count, with each model adjusting for one selected covariate. To formally assess whether diabetes functions as an effect modifier on the prognostic value of carotid geometry, testing for statistical interaction was performed by introducing a multiplicative interaction term (CIMT × DM) into the Cox proportional hazards model. Additionally, subgroup analyses stratified by diabetes status were performed using univariable and multivariable Cox proportional hazards models to evaluate potential effect modification. Two-sided p < 0.05 was considered statistically significant. Statistical analyses were performed using IBM SPSS Statistics for windows, Version 27.0 (IBM Corp., Armonk, NY, USA). Results Comparison of CIMT and ufPWV between MHD and non-ESRD controls Baseline demographic and vascular characteristics are presented in Table 1. There were no significant differences in age (63 vs 59 years, p = 0.061) or sex distribution (57.4% vs 51.2% male, p = 0.391) between the two groups. CIMT was significantly higher in MHD patients compared with non-ESRD controls [1.60 (0.60–2.30) mm vs 0.80 (0.60–1.20) mm, p = 0.004)]. The prevalence of carotid plaque was also higher in the MHD group than non-ESRD controls (56.5% vs 35.4%, p = 0.003). In contrast, no significant differences were observed in ufPWV parameters between two groups. Table 1.Baseline demographic and carotid artery characteristics of MHD and non-ESRD controls. MHD(n = 115)Non-ESRD controls(n = 82)p-valueAge (years)63 (57, 68)59 (51, 67)0.061Male, n/N (%)66 /115 (57.4)42/82 (51.2)0.391CIMT (mm)1.60 (0.60, 2.30)0.80 (0.60, 1.20) 0.004 Carotid Plaque, n/N (%)65/115 (56.5)29/82 (35.4) 0.003 Left PWV-BS (m/s)6.21 (5.14, 7.22)6.23 (5.43, 7.12)0.834Left PWV-ES (m/s)8.93 ± 2.469.08 ± 2.580.664Right PWV-BS (m/s)5.63 (4.58, 6.47)5.41 (4.33, 6.53)0.402Right PWV-ES (m/s)8.04 (6.29, 10.44)7.52 (5.62, 10.48)0.237MHD: maintenance hemodialysis; ESRD: end-stage renal disease; CIMT: carotid intima-media thickness; PWV-BS: pulse wave velocity at the beginning of systole; PWV-ES: pulse wave velocity at the end of systole.Bold values indicate statistical significance with p ≤ 0.05. Baseline demographic and carotid artery characteristics of MHD and non-ESRD controls. MHD: maintenance hemodialysis; ESRD: end-stage renal disease; CIMT: carotid intima-media thickness; PWV-BS: pulse wave velocity at the beginning of systole; PWV-ES: pulse wave velocity at the end of systole. Bold values indicate statistical significance with p ≤ 0.05. Clinical and vascular factors associated with CV events The baseline characteristics of the MHD cohort (n = 115) are summarized in Table 2. The median dialysis vintage was 49 (20–79) months, and 31.3% of patients had diabetes. During a median follow-up of 39 months, 21 patients (18.3%) experienced CV events. Patients who developed CV events had higher CIMT (p = 0.019), carotid plaque (p = 0.044) and SBP (p = 0.005), and lower ejection fraction (p = 0.013) and spKt/V (p = 0.023). They were also more likely to be male (p = 0.004), have diabetes (p < 0.001), prior CV disease (p = 0.043), and smoking (p = 0.004) (Table 3). In univariable Cox regression analyses, CIMT was significantly associated with CV events. However, after adjustment for diabetes in the single-variable model, this association was attenuated and no longer statistically significant (HR: 1.350, 95% CI: 0.865–2.106, p = 0.186) (Table 4). Based on the baseline distribution of the total study population, the median value of CIMT was identified as 1.6 mm. When stratified by CIMT (cut-off 1.6 mm), patients in the higher CIMT group were older males and had a higher prevalence of diabetes, CV disease, smoking and carotid plaque, as well as lower DBP, spKt/V and high-density lipoprotein (HDL) levels (Table 2). Diabetes patients had significantly higher CIMT (p = 0.004), SBP (p = 0.003), and BMI (p < 0.001), along with a higher prevalence of CV disease and carotid plaque, and lower spKt/V and HDL (Supplementary Table S1). CIMT was positively correlated with diabetes (r = 0.271, p = 0.003). Kaplan-Meier analysis showed that patients with higher CIMT had a lower CV event-free survival (log-rank p = 0.009) (Figure 2(A)). Similarly, patients with diabetes had significantly worse CV event-free survival (log-rank p < 0.001) (Figure 3(A)). To further dissect the methodological role of diabetes status, a formal interaction analysis was executed within the total MHD cohort. The multiplicative interaction term between CIMT and diabetes did not reach statistical significance (Pinteraction = 0.708; HR: 0.840, 95% CI: 0.338–2.087), whereas the effect of diabetes remained significant (p = 0.044; HR: 7.871, 95% CI: 1.057–58.601). No significant associations were observed between ufPWV parameters and CV events (Table 4). Table 2.Baseline characteristics of MHD patients stratified by CIMT.CharacteristicsTotal (n = 115)CIMT < 1.6 mm (n = 57)CIMT ≥ 1.6 mm (n = 58)p-valueAge (year)63 (57, 68)60 (54, 66)66 (61, 69) 0.004 Male, n/N (%)66/115 (57.4)26/57 (45.6)40/58 (69.0) 0.011 Dialysis vintage (months)49 (20, 79)42 (15, 71)54 (31, 90)0.088Hypertension, n/N (%)84/115 (73.0)42/57 (73.7)42/58 (72.4)0.878Diabetes, n/N (%)36/115 (31.3)9/57 (15.8)27/58 (46.6) <0.001 History of CVD*, n/N (%)28/113 (24.8)9/56 (16.1)19/57 (33.3) 0.034 Smoking, n/N (%)24/103 (23.3)6/52 (11.5)18/51 (35.3) 0.004 SBP (mmHg)135 ± 17134 ± 18136 ± 160.413DBP (mmHg)75 ± 1077 ± 1073 ± 9 0.020 spKt/V1.48 (1.31, 1.64)1.55 (1.37, 1.66)1.44 (1.25, 1.59) 0.049 NT-proBNP (pg/mL)3036 (1771, 5765)2827 (2047, 4935)3569 (1643, 6380)0.561Cholesterol (mmol/L)3.84 ± 0.963.91 ± 0.853.78 ± 1.060.483Triglycerides (mmol/L)1.60 (1.07, 2.79)1.48 (0.87, 2.22)1.66 (1.23, 2.90)0.147HDL (mmol/L)0.94 (0.80, 1.19)1.07 (0.83, 1.34)0.88 (0.77, 1.08) 0.005 LDL (mmol/L)2.21 ± 0.822.19 ± 0.712.23 ± 0.930.797iPTH (pg/mL)302 (162, 468)302 (187, 402)313 (148, 499)0.687Carotid Plaque, n/N (%)65/115 (56.5)7/57 (12.3)58/58 (100.0) <0.001 Left PWV-BS (m/s)6.21 (5.14, 7.22)6.41 (5.21, 7.36)6.10 (4.92, 7.02)0.305Left PWV-ES (m/s)8.93 ± 2.469.27 ± 2.398.59 ± 2.500.142Right PWV-BS (m/s)5.63 (4.58, 6.47)5.65 (4.56, 7.03)5.61 (4.68, 6.38)0.814Right PWV-ES (m/s)8.39 ± 2.578.46 ± 2.778.32 ± 2.380.766CV events, n/N (%)21/115 (18.3)5/57 (8.8)16/58 (27.8) 0.009 MHD: maintenance hemodialysis; CVD: cardiovascular disease; CIMT: carotid intima-media thickness; SBP: systolic blood pressure; DBP: diastolic blood pressure; spKt/V: single-pool Kt/V; NT-proBNP: N-terminal pro-B-type natriuretic peptide; HDL: high-density lipoprotein; LDL: low-density lipoprotein; iPTH: intact parathyroid hormone; PWV-BS: pulse wave velocity at the beginning of systole; PWV-ES: pulse wave velocity at the end of systole; CV: cardiovascular.*CVD was defined as the presence of either congestive heart failure or atherosclerotic CV disease, including: previous myocardial infarction, history of coronary artery angioplasty and/or stent placement, angina pectoris, evidence of coronary atherosclerotic disease, stroke, transient ischemic attacks, claudication, or aortic aneurysm.Bold values indicate statistical significance with p ≤ 0.05. Baseline characteristics of MHD patients stratified by CIMT. MHD: maintenance hemodialysis; CVD: cardiovascular disease; CIMT: carotid intima-media thickness; SBP: systolic blood pressure; DBP: diastolic blood pressure; spKt/V: single-pool Kt/V; NT-proBNP: N-terminal pro-B-type natriuretic peptide; HDL: high-density lipoprotein; LDL: low-density lipoprotein; iPTH: intact parathyroid hormone; PWV-BS: pulse wave velocity at the beginning of systole; PWV-ES: pulse wave velocity at the end of systole; CV: cardiovascular. *CVD was defined as the presence of either congestive heart failure or atherosclerotic CV disease, including: previous myocardial infarction, history of coronary artery angioplasty and/or stent placement, angina pectoris, evidence of coronary atherosclerotic disease, stroke, transient ischemic attacks, claudication, or aortic aneurysm. Bold values indicate statistical significance with p ≤ 0.05. Table 3.Baseline clinical characteristics of MHD patients with and without CV events.VariablesTotal (n = 115)Patients without CV events (n = 94)Patient with CV events (n = 21)p-valueAge (year)63 (57, 68)63 (57, 69)64 (55, 66)0.314Male, n/N (%)66/115 (57.4)48/94 (51.1)18/21 (85.7) 0.004 Dialysis vintage (months)49 (20, 79)53 (20, 87)48 (33, 75)0.797Diabetes, n/N (%)36/115 (31.3)21/94 (22.3)15/21 (71.4) <0.001 History of CVD, n/N (%)28/113 (24.8)19/93 (20.4)9/20 (45.0) 0.043 Smoking, n/N (%)24/103 (23.3)14/83 (16.9)10/20 (50.0) 0.004 SBP (mmHg)135 ± 17133 ± 17144 ± 16 0.005 DBP (mmHg)75 ± 1074 ± 1078 ± 70.134spKt/V1.48 (1.31, 1.64)1.48 (1.36, 1.65)1.31 (1.19, 1.55) 0.023 NT-proBNP (pg/mL)3036 (1771, 5765)2728 (1605, 5572)4025 (2215, 10512)0.078iPTH (pg/mL)302 (162, 468)319 (196, 466)221 (92, 482)0.076CIMT (mm)1.60 (0.60, 2.30)1.15 (0.60, 2.23)2.20 (1.15, 2.75) 0.019 Carotid Plaque, n/N (%)65/115 (56.5)49/94 (52.1)16/21 (76.2) 0.044 Left PWV-BS (m/s)6.20 (5.18, 7.17)6.36 (5.25, 7.29)5.63 (4.74, 6.74)0.080Left PWV-ES (m/s)8.93 ± 2.469.09 ± 2.608.18 ± 1.49 0.036 Right PWV-BS (m/s)5.63 (4.58, 6.47)5.60 (4.60, 6.64)5.66 (4.28, 6.23)0.556Right PWV-ES (m/s)8.39 ± 2.578.43 ± 2.688.21 ± 2.000.731EF (%)68 (65, 70)68 (66, 70)66 (60, 68) 0.013 LVM (g)153.0 (128.0, 207.1)153.0 (127.8, 201.6)164.0 (137.7, 209.6)0.207LVM/BSA (g/m2)95.7 (78.7, 120.0)96.4 (77.5, 120.0)94.7 (82.0, 125.1)0.472MHD: maintenance hemodialysis; CV: cardiovascular; SBP: systolic blood pressure; DBP: diastolic blood pressure; spKt/V: single pool Kt/V; NT-proBNP: N-terminal pro-B-type natriuretic peptide; iPTH: intact parathyroid hormone; CIMT: carotid intima-media thickness; PWV-BS: pulse wave velocity at the beginning of systole; PWV-ES: pulse wave velocity at the end of systole; EF: ejection fraction; LVM: left ventricular mass; LVM/BSA: left ventricular mass/body surface area.Bold values indicate statistical significance with p ≤ 0.05. Baseline clinical characteristics of MHD patients with and without CV events. MHD: maintenance hemodialysis; CV: cardiovascular; SBP: systolic blood pressure; DBP: diastolic blood pressure; spKt/V: single pool Kt/V; NT-proBNP: N-terminal pro-B-type natriuretic peptide; iPTH: intact parathyroid hormone; CIMT: carotid intima-media thickness; PWV-BS: pulse wave velocity at the beginning of systole; PWV-ES: pulse wave velocity at the end of systole; EF: ejection fraction; LVM: left ventricular mass; LVM/BSA: left ventricular mass/body surface area. Bold values indicate statistical significance with p ≤ 0.05. Table 4.Cox proportional hazards analysis for CV events in the overall MHD patients.VariablesUnadjustedModel 1Model 2Model 3HR (95% CI)p-valueHR (95% CI)p-valueHR (95% CI)p-valueHR (95% CI)p-valueAge (per year)0.987 (0.947–1.029)0.5480.971 (0.928–1.016)0.202 Male (yes vs. no)5.118(1.507–17.385) 0.009 4.333 (1.248–15.042) 0.021 CIMT (per 1 mm)1.578 (1.054–2.364) 0.027 1.688 (1.122–2.539) 0.012 1.374 (0.910–2.075)0.1311.350 (0.865–2.106)0.186Carotid plaque (yes vs. no)2.665 (0.976–7.280)0.056 Diabetes (yes vs. no)6.497 (2.517–16.772) < 0.001 5.648 (2.147–14.860) < 0.001 Left PWV-BS (per 1 m/s)0.801 (0.612–1.050)0.109 Left PWV-ES (per 1 m/s)0.884 (0.738–1.060)0.183 Right PWV-BS (per 1 m/s)0.911 (0.692–1.199)0.504 Right PWV-ES (per 1 m/s)0.973 (0.820–1.153)0.750 Due to the low event count (n = 21), separate parsimonious Cox proportional hazards models were constructed, each adjusting for a single selected covariate (age, sex, or diabetes separately). Model 1 of CIMT adjusted for age. Model 2 of CIMT adjusted for sex. Model 3 of CIMT adjusted for diabetes.CV: cardiovascular; MHD: maintenance hemodialysis; CIMT: carotid intima-media thickness; PWV-BS: pulse wave velocity at the beginning of systole; PWV-ES: pulse wave velocity at the end of systole.Bold values indicate statistical significance with p ≤ 0.05. Cox proportional hazards analysis for CV events in the overall MHD patients. Due to the low event count (n = 21), separate parsimonious Cox proportional hazards models were constructed, each adjusting for a single selected covariate (age, sex, or diabetes separately). Model 1 of CIMT adjusted for age. Model 2 of CIMT adjusted for sex. Model 3 of CIMT adjusted for diabetes. CV: cardiovascular; MHD: maintenance hemodialysis; CIMT: carotid intima-media thickness; PWV-BS: pulse wave velocity at the beginning of systole; PWV-ES: pulse wave velocity at the end of systole. Bold values indicate statistical significance with p ≤ 0.05. Kaplan–Meier curves for (A) CV event-free survival and (B) all-cause mortality according to CIMT categories in the overall MHD patients. Patients were stratified into two groups based on CIMT (< 1.6 mm vs ≥ 1.6 mm). Kaplan–Meier curves for (A) CV event-free survival and (B) all-cause mortality in hemodialysis patients stratified by diabetes status. Furthermore, to address the distinct impact of diabetes, subgroup analyses stratified by diabetes status were performed (Supplementary Table S1). Within the diabetic subgroup (n = 36), univariable Cox regression revealed that the left PWV-ES was significantly associated with a decreased risk of CV events (HR: 0.736, 95% CI: 0.570–0.950, p = 0.018). This association remained significant after separate adjustment for age (HR: 0.736, 95% CI: 0.568–0.954, p = 0.020) or sex (HR: 0.735, 95% CI: 0.574–0.942, p = 0.015) (Supplementary Table S2). Neither CIMT nor other ufPWV parameters showed any significant association with CV events in either subgroup (Supplementary Table S2–S3). Factors associated with all-cause mortality During a median follow-up of 39 months, 17 patients (14.8%) died, and 4 (3.5%) were CV deaths. Comparisons of baseline clinical characteristics between survivors and non-survivors revealed several significant differences. Non-survivors had a significantly higher prevalence of diabetes (52.9% vs. 27.6%, p = 0.037) and preexisting CVD (52.9% vs. 19.8%, p = 0.009) compared to survivors. Non-survivors exhibited markedly elevated plasma NT-proBNP levels [6027 (2272, 16,960) vs. 2728 (1665, 5164) pg/mL, p = 0.009] and significantly higher left ventricular mass [219.1 (142.5, 241.2) vs. 152.6 (127.8, 199.3) g, p = 0.045]. No statistically significant differences were observed between the CIMT and ufPWV parameters. In univariable Cox regression analysis, neither CIMT nor ufPWV parameters were associated with all-cause mortality (Table 5). Figure 2(B) shows the Kaplan–Meier curves for all‑cause mortality according to CIMT categories. There was no significant difference in overall survival between the CIMT < 1.6 mm and ≥ 1.6 mm groups (log‑rank p = 0.198). Diabetes was significantly associated with all-cause mortality in univariable Cox analysis (p = 0.046) (Table 5). Kaplan-Meier analysis demonstrated that patients with diabetes had significantly lower overall survival compared with those without diabetes (log-rank p = 0.038) (Figure 3(B)). Table 5.Cox proportional hazards analysis for all-cause mortality in the overall MHD patients.VariablesUnivariable analysis for all-cause mortalityHR (95% CI)p-valueAge (per year)1.014 (0.963–1.067)0.607Male (yes vs. no)2.645 (0.862–8.116)0.089CIMT (per 1 mm)1.387 (0.884–2.178)0.155Carotid plaque (yes vs. no)1.448 (0.535–3.916)0.466Diabetes (yes vs. no)2.643 (1.019–6.850) 0.046 Left PWV-BS (per 1 m/s)0.909 (0.681–1.212)0.516Left PWV-ES (per 1 m/s)0.968 (0.796–1.178)0.747Right PWV-BS (per 1 m/s)0.865 (0.632–1.185)0.366Right PWV-ES (per 1 m/s)1.044 (0.872–1.250)0.642History of CVD (yes vs. no)3.992 (1.538–10.366) 0.004 Albumin (per 1 g/L)0.860 (0.749–0.986) 0.031 CRP (per 1 mg/L)1.020 (1.003–1.037) 0.021 NT-proBNP (per 100 pg/mL)1.000 (1.000–1.000) 0.002 LVM (per 1 g)1.010 (1.002–1.018) 0.011 MHD: maintenance hemodialysis; CIMT: carotid intima-media thickness; PWV-BS: pulse wave velocity at the beginning of systole; PWV-ES: pulse wave velocity at the end of systole; CVD: cardiovascular disease; CRP: C-reactive protein; NT-proBNP: N-terminal pro-B-type natriuretic peptide; LVM: left ventricular mass.Bold values indicate statistical significance with p ≤ 0.05. Cox proportional hazards analysis for all-cause mortality in the overall MHD patients. MHD: maintenance hemodialysis; CIMT: carotid intima-media thickness; PWV-BS: pulse wave velocity at the beginning of systole; PWV-ES: pulse wave velocity at the end of systole; CVD: cardiovascular disease; CRP: C-reactive protein; NT-proBNP: N-terminal pro-B-type natriuretic peptide; LVM: left ventricular mass. Bold values indicate statistical significance with p ≤ 0.05. Furthermore, exploratory subgroup analyses stratified by diabetes status were performed to evaluate the potential prognostic impact of vascular parameters on all-cause mortality. Within both the diabetic (n = 36, 9 patients died) and non-diabetic (n = 79, 8 patients died) sub-cohorts, univariable Cox proportional hazards analysis demonstrated that neither CIMT nor any of the bilateral ufPWV parameters were significantly associated with all-cause mortality (Supplementary Table S4). However, these mortality baseline predictors must be interpreted with extreme caution. Due to the low number of recorded endpoint events, these specific survival analyses are inherently underpowered. Discussion Our findings confirmed that CIMT is significantly elevated in MHD patients compared to non-ESRD controls, reflecting the advanced, uremia-induced atherosclerotic remodeling widely reported in the literature [4,6, 17,18]. However, its independent prognostic value remains contentious [2,6–8]. While some cross-sectional data linked maximum CIMT to CV disorders [19], our prospective analysis demonstrates that the association between CIMT and CV events was attenuated and lost statistical significance after adjustment for diabetes in the limited model. Indeed, diabetes mellitus represented a major driver of CV risk in our cohort, characteristically clustering with other traditional risk factors such as elevated SBP, BMI and reduced high-density lipoprotein. Mechanistically, diabetes inflicts diffuse vascular injury via medial calcification, advanced glycation end-product accumulation, and chronic inflammation – pathological features that transcend mere intimal thickening and may explain the overriding clinical impact of diabetes over CIMT alone [20–22]. A relevant methodological consideration is how diabetes influences the association between CIMT and CV outcomes. Diabetes is strongly associated with both increased CIMT and elevated CV risk [20,23]. In our analysis, adjustment for diabetes attenuated the association between CIMT and CV events, and no statistically significant interaction between CIMT and diabetes was detected. However, because our interaction analysis was limited by the small sample size and low number of events, these findings do not definitively rule out effect modification. Furthermore, as formal mediation analysis was not performed, we cannot draw firm conclusions regarding whether diabetes acts primarily as a confounder or a mediator. In contrast to CIMT, our study observed that ufPWV parameters did not differ significantly between MHD patients and non-ESRD controls, nor did they demonstrate significant associations with clinical outcomes in the overall cohort. Traditionally, conventional metrics such as cfPWV [24] have served as the gold standard for assessing systemic arterial stiffness, backed by robust evidence predicting CV events [25]. However, data evaluating localized ufPWV within the MHD population remain exceptionally scarce. While a prior study indicated higher ufPWV in advanced CKD, that cohort explicitly excluded patients undergoing hemodialysis [16]. This discrepancy between the prognostic performance of systemic cfPWV and our overall local ufPWV findings likely reflects the unique, confounding pathophysiological milieu of the dialyzed state. MHD patients experience dramatic interdialytic fluid shifts and acute blood pressure variability. Because local carotid ufPWV captures instantaneous, real-time vascular mechanics within a highly restricted 1–2 cm anatomical segment, it is inherently sensitive to these transient fluctuations in intravascular volume and acute transmural pressure. In the overall cohort, this pronounced ‘hemodynamic noise’ may temporarily distort or mask the true, chronic baseline arterial stiffness – a limitation that regional cfPWV might partially circumvent by integrating vascular properties over a much longer anatomical distance. An isolated inverse association between left PWV-ES and CV events was observed exclusively within the small diabetic sub-cohort (n = 36, 15 events, Supplementary Table S2). This finding should be interpreted with extreme caution and must not be viewed as evidence of superior prognostic performance or a clinically meaningful protective effect. Given the small subgroup size, low event count, and the evaluation of multiple bilateral vascular metrics, this unadjusted association carries a high risk of Type I error (false-positive finding) and may simply reflect statistical artifact or sample variability. Therefore, this observation remains strictly exploratory and hypothesis-generating, requiring rigorous validation in larger, adequately powered cohorts before any meaningful pathophysiological or clinical conclusions can be drawn. Beyond the statistical adjustment for diabetes, several unique pathophysiological and epidemiological characteristics of the MHD patients may inherently decouple macrovascular structural parameters (such as CIMT and ufPWV) from long-term all-cause and CV mortality. First, MHD patients are fundamentally impacted by the malnutrition-inflammation syndrome, which drives the well-documented phenomenon of reverse epidemiology. Traditional CV risk factors are often overshadowed by chronic low-grade systemic inflammation and wasting, weakening the linear prognostic correlation between established arterial remodeling indices and clinical survival. Most importantly, mortality trajectories in long-term hemodialysis patients are frequently dictated by uremic cardiomyopathy, characterized by profound myocardial fibrosis, left ventricular hypertrophy, and structural cardiac remodeling. Driven by rapid fluid-electrolyte shifts during interdialytic intervals and systemic microinflammation, these structural heart alterations predispose patients to electrophysiological instability and sudden cardiac death (SCD) due to fatal ventricular arrhythmias, rather than isolated macrovascular atherothrombotic occlusions. Because CIMT and ufPWV primarily capture the structural and functional compliance of large conduit arteries, they inherently lack the capacity to reflect the microvascular, metabolic, and cardiomyocyte-level triggers responsible for cardiac remodeling and arrhythmic mortality in this specialized population. We also observed a higher prevalence of carotid plaques in MHD patients compared with non-ESRD controls. Savage et al. also found plaques in 71% of dialysis (nondiabetic) patients compared to 21% of healthy controls [26]. Several studies showed that the presence of carotid plaques is an independent marker of increased risk and the likelihood of CV events or mortality in MHD patients [27,28] and is a better prognostic marker than CIMT [8,27,29]. In our univariable analysis, the presence of carotid plaque indeed displayed a strong clinical trend toward predicting future CV events (HR = 2.665, 95% CI: 0.976–7.280, p = 0.056). However, our limited sample size and the low number of CV events strictly precluded us from performing an independent, fully-adjusted multivariable analysis dedicated to separate plaque burden parameters. For all-cause mortality, a number of positive associations were observed. Non-survivors had significantly higher NT-proBNP levels and higher left ventricular mass. In univariable Cox regression, higher NT-proBNP and higher LVM were all significantly associated with increased mortality risk. NT-proBNP is a well-established marker of volume overload and left ventricular stress in dialysis patients, and its elevation often reflects inadequate ultrafiltration or underlying cardiomyopathy [30]. Left ventricular hypertrophy (LVH) is highly prevalent in MHD patients and is driven by pressure overload (hypertension, arterial stiffness) and volume overload [31]. Our finding that both markers predict mortality reinforces the importance of regular cardiac assessment and optimal volume management in this population. To the best of our knowledge, this is the first study to apply ufPWV within MHD population. As an innovative and exploratory pilot study, it addresses a critical gap in conventional regional stiffness assessments by introducing a highly sensitive, real-time method to capture localized carotid artery hemodynamics in this high-risk population. Diabetic hemodialysis patients represent a heavily burdened subgroup requiring heightened clinical vigilance. Taken together, these exploratory insights lay a crucial foundation for utilizing localized vascular phenotyping to refine CV risk stratification in patients with end-stage renal disease. Several limitations of the present study warrant consideration. First, the single-center design and relatively small sample size with a low number of outcome events limited the statistical power and precluded comprehensive multivariable adjustment and formal mediation analysis. Second, potential selection bias and restricted generalizability must be acknowledged. Our stringent exclusion criteria intentionally eliminated several high-risk patient groups, specifically individuals with recent CV events within the preceding three months or those presenting with severe, refractory heart failure (NYHA class III–IV). While this selection protocol was methodologically vital to ensure hemodynamic and cardiac rhythm stability during specialized ufPWV imaging and to minimize acute confounding parameters, it inevitably limits the extrapolation of our conclusions. Third, all clinical, biochemical, and imaging parameters were measured only once at baseline. Consequently, we were unable to account for potential longitudinal changes or dynamic trajectories of these metrics during the follow-up period. Fourth, carotid plaque burden was not evaluated independently in dedicated multivariable models due to event constraints. Fifth, the lack of established clinical consensus or thresholds (such as the standard 10 m/s utilized for cfPWV) for newer local metrics like ultrafast PWV, combined with our sample size constraints, restricted our ability to fully characterize non-linear or threshold-specific prognostic relationships. Sixth, the measurement reproducibility for local ufPWV parameters was only moderate in this hemodialysis cohort, which may have increased random variability and attenuated potential prognostic associations. Finally, residual confounding from unmeasured factors, such as the duration of diabetes, long-term glycemic control (HbA1c), or specific dietary patterns, cannot be entirely excluded. Conclusion In conclusion, within the statistical limitations of this single-center cohort with a low event count, CIMT was no longer significantly associated with CV events after adjustment for diabetes in the limited model, whereas local ufPWV parameters were not significantly associated with CV events or all-cause mortality in the analyses performed within the overall MHD population. Although an inverse association between left PWV-ES and CV events was observed in the small diabetic subgroup, this isolated finding may have arisen by chance and should be regarded strictly as preliminary and hypothesis-generating. Supplementary Material Supplementary Appendix.docx Acknowledgments XZ and YW conceived and designed the study. XZ, XL, YR and YZ were responsible for data collection. LZ and WZ conducted ultrasonography for patients. XZ, QG and LY conducted formal analysis. XZ and LZ wrote the original draft. YW, JX and CH reviewed and edited the article. XZ and LZ contributed equally to this work. All authors reviewed and agreed to be accountable for all aspects of the work. Funding Statement The authors declare that no funds, grants, or other support were received during the preparation of this manuscript. Disclosure statement No potential conflict of interest was reported by the author(s). Data availability statement The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request. References Associated Data Supplementary Materials Supplementary Appendix.docx Data Availability Statement The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.