Blood pressure variability and cerebrovascular risk in maintenance hemodialysis versus peritoneal dialysis: a comparative study Authors: Ting Hu, Ting Yu, Xiao Xu, Zhuanping Huang, Zhanxing Cao, Geping Yu Published: 2025 ABSTRACT AbstractThis study aimed to compare blood pressure variability (BPV) between maintenance hemodialysis (HD) and peritoneal dialysis (PD) patients and evaluate its association with cerebrovascular event risk. A prospective cohort of 374 end-stage renal disease (ESRD) patients, including 218 on HD and 156 on PD, was enrolled between January 2020 and July 2024. Clinical and laboratory data, along with 24-hour ambulatory blood pressure measurements, were collected. BPV was assessed using standard deviation (SD), coefficient of variation (CV), and average real variability (ARV). Patients were followed for 48 months to document cerebrovascular events, including stroke and transient ischemic attack (TIA). Compared to HD patients, those on PD had significantly lower 24-hour systolic blood pressure (SBP) and SBP variability (SD, ARV, and range) (all p < 0.01), while diastolic BPV showed no significant difference. The incidence of cerebrovascular events was higher in HD patients (14.68%) than in PD patients (7.69%) (p = 0.039). Elevated SBP-ARV was an independent predictor of cerebrovascular events in both groups. In HD patients, additional independent risk factors included diabetes, hypertension, HbA1c and elevated creatinine levels. Kaplan–Meier analysis demonstrated better cerebrovascular event-free survival in PD patients (p = 0.043). These findings suggest that HD patients experience greater BPV than PD patients, and increased BPV is independently associated with higher cerebrovascular risk. BPV monitoring and management may be beneficial in reducing cerebrovascular complications, particularly among patients undergoing HD. Introduction Cardio-cerebrovascular complications remain the leading causes of mortality and disability in patients with end-stage renal disease (ESRD), with cerebrovascular events, including ischemic stroke, hemorrhagic stroke, and transient ischemic attack (TIA), occurring at significantly higher rates compared to the general population [1]. Globally, the risk of stroke in ESRD patients is more than threefold greater than that in healthy individuals, with certain subgroups exhibiting risk elevations of 5- to 10-fold [2], particularly among females and younger adults [3]. Although renal replacement therapies [hemodialysis (HD) and peritoneal dialysis (PD)] have markedly improved patient survival, their hemodynamic impacts, notably exacerbating blood pressure (BP) fluctuations, may contribute to cerebrovascular instability [4]. In HD patients, cyclical volume shifts and solute clearance during dialysis induce pronounced BP oscillations. Intradialytic hypotension (occurring in 20-30% of sessions) and interdialytic hypertension collectively manifest as a “roller-coaster phenomenon of BP”, potentially triggering abnormal cerebral perfusion [5,6]. Data from the Chronic Renal Insufficiency Cohort (CRIC) study revealed that higher systolic BP levels outside dialysis units correlated with a significantly increased risk of cardiovascular events among patients undergoing maintenance HD, with the risk in the highest quartile being 2.76 times higher than that in the lowest quartile [7]. In contrast, PD, owing to its continuous ultrafiltration profile, theoretically offers more stable BP control. These disparities may arise from race-specific vascular reactivity, dialysis prescription variations (e.g. dialysate sodium concentration), or residual renal function preservation, underscoring the need for refined subgroup analyses and mechanistic investigations. Recent attention has focused on BP variability (BPV) as an independent cardiovascular risk factor distinct from absolute BP values [8–10]. Accumulating evidence links elevated short-term (24-h) and long-term (day-to-day) BPV to atherosclerosis progression, endothelial dysfunction, and target organ damage [11–13]. However, the differential impacts of HD versus PD on BPV and their associations with cerebrovascular events remain poorly characterized in dialysis populations, constituting the central scientific question of this study. In general hypertensive and cardiovascular cohorts, BPV reduction confers additional long-term benefits. ESRD patients on maintenance dialysis frequently exhibit heightened BPV due to prevalent hypertension and autonomic dysfunction [14]. Nevertheless, evidence regarding modality-specific BPV effects and their relationship to stroke risk remains sparse and inconsistent. Furthermore, cerebrovascular event heterogeneity (ischemic versus hemorrhagic) may be partly driven by distinct BP fluctuation patterns. BPV pathophysiology involves dysregulated autonomic modulation, vascular stiffening, and renin-angiotensin-aldosterone system (RAAS) activation [15,16]. In dialysis patients, uremic toxin accumulation, chronic inflammation [e.g. elevated C-reactive protein (CRP) and interleukin-6 (IL-6)], and electrolyte imbalances amplify BP instability [17,18]. Despite growing recognition of BPV’s clinical significance, standardization of its measurement remains challenging. No international consensus exists on optimal BPV metrics for dialysis populations. Parameters such as standard deviation (SD) of ambulatory BP, coefficient of variation (CV), and weighted SD are variably employed with differing sensitivities [19–21]. The American Society of Nephrology recommends combining ambulatory and home BP monitoring to capture interdialytic fluctuations [22]. Additionally, traditional survival analyses may underestimate BPV effects due to competing risks (e.g. infection-related deaths or sudden cardiac arrest) leading to premature follow-up termination [23]. Application of competing risk models (e.g. Fine-Gray subdistribution hazards) could enhance accuracy [24], though such approaches remain underutilized. Against this backdrop, this study systematically compares HD and PD patients across multiple BPV indices and evaluates the predictive utility of these parameters for cerebrovascular events. By elucidating modality-specific BPV-cerebrovascular outcome relationships, we aim to provide evidence-based guidance for individualized BP management and stroke risk stratification in ESRD patients. Materials and methods Study design This single-center prospective cohort study enrolled adult ESRD patients undergoing maintenance dialysis for > 3 months at our hospital between January 2020 and July 2024. Participants were stratified into two groups based on dialysis modality: maintenance HD and continuous ambulatory PD. All study participants were of Han Chinese ethnicity. The study protocol was approved by the Ethics Committee of our hospital. Among 374 screened patients meeting eligibility criteria, 218 were allocated to the HD group and 156 to the PD group. Baseline demographic data, dialysis vintage, primary renal etiology, and comorbidities were systematically recorded. Laboratory parameters, including hemoglobin, albumin, CRP, lipid profiles, electrolytes, blood urea nitrogen, serum creatinine, uric acid, residual urine volume, and intact parathyroid hormone (iPTH), were measured at enrollment using standardized assays. Patients were prospectively followed for up to 48 months, with study endpoints defined as incident cerebrovascular events (ischemic/hemorrhagic stroke or TIA) or all-cause mortality. Cerebrovascular events were identified from hospital records and verified by treating neurologists. Endpoint adjudication was performed independently by two neurologists blinded to dialysis modalities, with discrepancies resolved through consensus review. All events were confirmed by neuroimaging—non-contrast cranial CT or brain MRI—performed at the time of presentation, and ischemic versus hemorrhagic subtypes were classified based on imaging findings and discharge diagnoses. To assist in organizing literature and summarizing preliminary findings, we used ChatGPT (4o model, [25]); this tool was not used for the study design, statistical analysis, or interpretation of results. Inclusion criteria were as follows: (1) adults aged 18–75 years receiving maintenance HD or continuous ambulatory PD in our Department of Nephrology; (2) stable dialysis therapy > 3 months without active infections, liver dysfunction, recent transfusions, trauma, surgery, malignancies, or autoimmune disease flares; (3) provision of written informed consent. Exclusion criteria comprised: (1) age < 18 years at dialysis initiation; (2) malignancy or hematologic disorders diagnosed within the preceding 3 years; (3) pregnancy/lactation or planned pregnancy within 2 years; (4) life expectancy < 1 year as clinically assessed; or (5) transitions between dialysis modalities (e.g. HD to PD, or PD to HD) or failed renal transplant recipients returning to dialysis. Written informed consent for publication of the detailed clinical information and any accompanying images was obtained from all individuals included in the study (or their legal guardians, if applicable). All necessary steps were taken to ensure the anonymity of the subjects. This publication consent is distinct from the consent to participate in the study. HD protocol Maintenance HD was performed using Fresenius 5008 dialysis machines (Fresenius Medical Care, Jiangsu, China; NMPA registration No. 20183101861) with bicarbonate-based dialysate [Wego Terumo (Weihai) Medical Products Co., Ltd., NMPA registration No. H20183286; 2,500 mL/bag]. Vascular access was established via central venous catheters or arteriovenous fistulae. Blood flow rates were maintained at 200–300 mL/min, with anticoagulation achieved using low-molecular-weight heparin calcium (Shenzhen Sciprogen Bio-Pharmaceutical Co., Ltd., NMPA registration No. H20060190; 1.0 mL: 5,000 AXa IU). Each dialysis session lasted 4 h, administered three times weekly. Ultrafiltration volumes were adjusted according to patients’ dry weight. Dialysis adequacy was ensured by achieving single-pool Kt/V ≥ 1.2 and urea reduction ratio (URR) ≥ 65%. The dialysis prescription for the HD group included a standard dialysate temperature of 36.5 °C, sodium concentration of 138 mmol/L, potassium concentration of 1.5 mmol/L, bicarbonate concentration of 31.1 mmol/L, and dialysate conductivity of 14.3 mS/cm. These parameters were kept constant for all patients unless otherwise clinically indicated. PD protocol Continuous ambulatory PD was performed using Baxter 5C4469C peritoneal dialysis systems (Baxter Healthcare Corporation, NMPA registration No. 2017210708). Glucose-based PD solutions (Huaren Pharmaceutical Co., Ltd., NMPA registration No. H20113088; 2,000 mL/bag, 1.5% glucose concentration) were administered at a frequency of 3–4 exchanges daily (2,000 mL per exchange). Dialysate prescription, including glucose concentration, dwell time, and exchange frequency, was individualized based on peritoneal equilibration test (PET) results and ultrafiltration volumes. PET assessments, conducted every 6 months, guided therapeutic adjustments for high-transport patients, such as increasing dialysate glucose concentration or optimizing exchange schedules. Exit-site care involved daily disinfection with povidone-iodine solution, with peritonitis incidence maintained below 0.5 episodes per patient-year. Dialysis adequacy targets included a standardized weekly total Kt/V ≥ 1.7. For patients with residual urine output > 200 mL/day, residual renal creatinine clearance was additionally quantified. BP monitoring during exchanges documented episodes of systolic BP (SBP) reduction ≥ 20 mmHg, triggering protocolized volume status reassessment. Ambulatory BP monitoring and variability indices Ambulatory BP monitoring was conducted using the Mobil-O-Graph portable oscillometric device (Japan). All enrolled patients underwent thrice-weekly HD sessions, with intradialytic BP measurements performed in accordance with the National Kidney Foundation Kidney Disease Outcomes Quality Initiative (NKF-K/DOQI) guidelines. During HD sessions, certified dialysis nurse - each holding valid national nursing licensure and a minimum of three years of nephrology clinical experience - measured BP of patients in the supine position using a calibrated cuff placed 2 fingerbreadths proximal to the antecubital crease on the non-autogenous arteriovenous fistula (AVF)-free arm, aligned at heart level. Pre- and post-dialysis BP readings were obtained after 5 min of rest, with additional measurements recorded every 30 min during dialysis. Both SBP and diastolic BP (DBP) were measured. Protocolized supplementary BP assessments were triggered by intradialytic symptoms (e.g. muscle cramps, diaphoresis). The following indices were ultimately selected for analysis: SD, CV, average real variability (ARV), and BP range. SD and CV reflect the dispersion of BP readings, with the former representing the absolute fluctuation amplitude and the latter quantifying the normalized fluctuation proportion relative to the mean, thereby enabling comparisons of variability across different BP levels. In contrast, ARV emphasizes fluctuations between consecutive measurements, capturing the frequency of dynamic variability, while BP range directly illustrates the maximum amplitude of BP fluctuations during the monitoring period. Specifically, SD serves as a metric for assessing the dispersion of BP measurements, reflecting the absolute magnitude of BP variability. These indices were calculated as: SD=∑i=1nBPi−BP¯2n−1, CV=SDBP×100%, ARV=1n−1∑i=1n−1BPi+1−BPi, RANGE = BPmax−BPmin, Statistical analysis Data were analyzed using SPSS version 26.0 (IBM Corp., Armonk, NY, USA). Continuous variables were assessed for normality and expressed as mean ± SD for parametric distributions or median (interquartile range, IQR) for non-parametric distributions. Between-group comparisons employed independent samples t-tests or Mann-Whitney U-tests. Categorical variables were reported as n (%) and compared using chi-square (χ2) or Fisher’s exact tests, as appropriate. Cerebrovascular event-free survival was evaluated via Kaplan-Meier curves, with between-group (HD vs. PD) differences assessed using log-rank tests. Cox proportional hazards regression models were applied to examine associations between BPV metrics and cerebrovascular outcomes, stratified by dialysis modality. Univariate analyses identified candidate variables (p < 0.05), which were subsequently entered into multivariate models after excluding collinear variables [variance inflation factor (VIF) > 5] through stepwise enter method selection. All statistical tests were two-tailed, with p < 0.05 considered statistically significant. All BPV indices showed VIF values <5, suggesting low collinearity. However, because of sample-size limitations, additional stepwise or bootstrapped stability analyses were not performed, and we note that future multicenter studies should reassess model stability or consider consolidating related BPV metrics. Results Baseline characteristics of HD and PD patients The study enrolled 374 patients, comprising 218 in the HD group and 156 in the PD group. There were no significant differences between groups in age, gender distribution, body mass index (BMI), smoking history, or the prevalence of diabetes mellitus and hypertension (all p > 0.05). Dialysis vintage was significantly longer in the HD group than in the PD group (median 2.45 [1.40, 4.30] vs. 2.15 [0.78, 3.99] years, p = 0.025). Laboratory indices, including hemoglobin, albumin, C-reactive protein (CRP), total cholesterol, LDL, HDL, calcium, phosphorus, potassium, blood urea nitrogen, uric acid, and parathyroid hormone, were generally comparable between groups, except that serum creatinine was significantly higher in the HD group (900.17 ± 26.58 vs. 597.48 ± 124.91 µmol/L, p < 0.001) and urine output was lower in the HD group (median 199.65 [137.48, 286.33] vs. 404.75 [284.10, 504.50] mL/24h, p < 0.001). The prevalence of carotid plaque, statin use, antiplatelet use, mean HbA1c, and insulin therapy was similar between the HD and PD cohorts (all p > 0.05). As seen in Table 1. Table 1.Baseline characteristics of HD and PD patients [ x¯±s , n (%), M (P25, P75)].VariablePD group (n = 156)HD group (n = 218)t/χ²/z P Age (years)60.32 ± 9.2660.00 ± 9.420.3320.740SexMale150 (68.81)95 (60.90)2.5180.113Female68 (31.19)61 (39.10)BMI (kg/m2)22.55 ± 2.8822.10 ± 2.901.5020.134Dialysis vintage (years)2.15 (0.78, 3.99)2.45 (1.40, 4.30)−2.2440.025Smoking historyYes105 (48.17)87 (55.77)2.1050.147No113 (51.83)69 (44.23)Diabetes mellitusYes68 (31.19)63 (40.38)0.0400.842No150 (68.81)93 (59.62)HypertensionYes85 (38.64)70 (44.87)1.4650.226No133 (61.36)86 (55.13)Hemoglobin (g/L)105.54 ± 15.06104.92 ± 14.290.4020.688Albumin (g/L)36.84 ± 4.2037.21 ± 3.72−0.8960.371CRP (mg/L)5.20 (3.24, 8.46)5.21 (2.79, 8.27)−0.70850.479Total cholesterol (mmol/L)5.20 ± 0.845.21 ± 0.78−0.1760.860LDL (mmol/L)3.30 ± 0.693.18 ± 0.681.5810.115HDL (mmol/L)1.19 ± 0.281.20 ± 0.30−0.2660.791Calcium (mmol/L)2.28 (2.15, 2.44)2.33 (2.19, 2.46)−1.3410.180Phosphorus (mmol/L)1.61 ± 0.291.58 ± 0.280.9450.345Potassium (mmol/L)4.84 ± 0.704.82 ± 0.750.2710.786Creatinine (µmol/L)597.48 ± 124.91900.17 ± 26.58−34.715<0.001Blood urea nitrogen (mmol/L)21.80 ± 5.6822.28 ± 5.65−0.8020.423Uric acid (µmol/L)451.05 ± 66.40447.89 ± 66.260.4540.650Urine output (mL/24 h)404.75 (284.10, 504.50)199.65 (137.48, 286.33)−11.538<0.001Parathyroid hormone (pg/mL)238.65 (147.91, 410.02)247.46 (181.08, 331.87)−0.2850.775Carotid plaque91 (58.33)136 (62.39)0.6260.429Statin use83 (53.21)122 (55.96)0.2790.597Antiplatelet use32 (19.23)48 (22.02)0.4280.513HbA1c (%)7.40 ± 1.247.29 ± 1.040.9830.326Insulin therapy62 (36.54)87 (38.07)0.0910.762Note: PD = Peritoneal dialysis; HD = Hemodialysis; BMI = Body mass index; CRP = C-reactive protein; LDL = low-density lipoprotein; HDL = high-density lipoprotein. Baseline characteristics of HD and PD patients [ x¯±s , n (%), M (P25, P75)]. Note: PD = Peritoneal dialysis; HD = Hemodialysis; BMI = Body mass index; CRP = C-reactive protein; LDL = low-density lipoprotein; HDL = high-density lipoprotein. Comparison of BPV indices between dialysis modalities Patients in the PD group exhibited significantly lower 24-h SBP compared to the HD group (128.4 mmHg vs. 135.9 mmHg, p < 0.001). PD patients also demonstrated reduced SBP variability across multiple indices (SBP-SD: 7.08 ± 1.95 vs. 7.90 ± 2.40 mmHg; SBP-ARV: 8.02 vs. 8.85 mmHg; SBP-RANGE (19.8 vs. 22.0 mmHg; all p < 0.01). However, the SBP-CV showed no significant intergroup difference. DBP and its associated variability indices, including DBP-SD, DBP-ARV, DBP-RANGE, and DBP-CV, demonstrated no significant intergroup differences (Table 2). Table 2.Comparison of BPV indices between HD and PD groups [ x¯±s , M(P25, P75)].VariablePD group (n = 156)HD group (n = 218)t/z p SBP-SD7.08 ± 1.957.90 ± 2.40−3.522<0.001SBP-CV5.53 ± 1.535.80 ± 1.74−1.5960.111SBP-ARV8.02 (6.13, 10.45)8.85 (7.03, 11.28)−2.6680.008SBP-RANGE19.80 (16.30, 24.45)22.00 (17.70, 26.73)−2.9210.003SBP-MEAN128.41 (126.21, 129.79)135.88 (133.73, 138.02)−15.388<0.001DBP-SD4.92 (3.65, 6.74)5.29 (3.92, 7.08)−1.0690.285DBP-CV6.01 (4.23, 8.05)6.18 (4.58, 8.45)−0.9480.343DBP-ARV6.48 (4.70, 8.90)6.36 (4.60, 9.34)−0.0140.989DBP-RANGE15.45 (10.96, 20.25)26.19 (11.65, 21.83)−1.1100.267DBP-MEAN85.19 ± 8.0985.20 ± 7.96−0.0120.991Note: PD = Peritoneal dialysis; HD = Hemodialysis; SBP = Systolic blood pressure; SD = Standard deviation; CV = Coefficient of variation; ARV = Average real variability; DBP = Diastolic blood pressure. Comparison of BPV indices between HD and PD groups [ x¯±s , M(P25, P75)]. Note: PD = Peritoneal dialysis; HD = Hemodialysis; SBP = Systolic blood pressure; SD = Standard deviation; CV = Coefficient of variation; ARV = Average real variability; DBP = Diastolic blood pressure. Cerebrovascular event incidence between dialysis modalities During the 48-month follow-up period, 44 cerebrovascular events occurred (12 in the PD group vs. 32 in the HD group). The cumulative incidence of cerebrovascular events was significantly lower in PD patients compared to HD patients (7.7 vs. 14.7%, p = 0.039). Kaplan-Meier survival curves demonstrated consistently higher event-free survival rates in the PD cohort (p = 0.033 by log-rank test). While the distribution of stroke subtypes (ischemic vs. hemorrhagic) did not differ significantly between groups (p = 0.21), PD patients exhibited a numerically higher proportion of hemorrhagic strokes. Overall, HD patients showed elevated rates of both ischemic and hemorrhagic stroke compared to PD counterparts, though statistical significance was confined to the composite endpoint (Table 3). Table 3.Incidence of cerebrovascular events between HD and PD groups [n (%)].GroupOverall incidenceIschemic strokeHemorrhagic strokeTIAχ² p PD group (n = 156)12 (7.69)3 (1.92)7 (4.49)2 (1.28)4.2760.039HD group (n = 218)32 (14.68)12 (5.53)15 (6.91)4 (1.84)Note: PD = Peritoneal dialysis; HD = Hemodialysis; TIA = Transient ischemic attack. Incidence of cerebrovascular events between HD and PD groups [n (%)]. Note: PD = Peritoneal dialysis; HD = Hemodialysis; TIA = Transient ischemic attack. Univariate analysis of cerebrovascular events in HD patients Compared to patients without cerebrovascular events, those who experienced such events were significantly more likely to have diabetes mellitus (53.1 vs. 27.4%, p = 0.004) and hypertension (62.5 vs. 34.9%, p = 0.003). The cerebrovascular event group also had higher serum creatinine levels (916.90 ± 26.90 vs. 897.29 ± 25.51 µmol/L, p < 0.001), however, the absolute difference was minimal, suggesting limited clinical relevance and likely reflecting normal biological or measurement variability. With respect to blood pressure variability parameters, patients with cerebrovascular events exhibited significantly greater systolic BP variability, as reflected by higher SBP-SD (9.36 ± 2.75 vs. 7.65 ± 2.25, p < 0.001), SBP-CV (6.67 ± 1.89 vs. 5.65 ± 1.68, p = 0.002), SBP-ARV (median 10.24 [7.77, 13.82] vs. 8.59 [6.58, 11.01], p = 0.004), SBP-RANGE (27.13 ± 9.15 vs. 21.59 ± 6.52, p < 0.001), and SBP-MEAN (139.99 ± 3.59 vs. 135.51 ± 2.93, p < 0.001). Carotid plaque was more prevalent in the event group (78.1 vs. 59.7%, p = 0.047). Statin use (71.9 vs. 53.2%, p = 0.050) and mean HbA1c (7.65 ± 1.19 vs. 7.24 ± 1.00, p = 0.032) were also higher among patients with cerebrovascular events. No significant differences were observed between groups in age, gender, BMI, smoking status, duration of dialysis, albumin, CRP, lipid profile, calcium, phosphorus, potassium, uric acid, urine output, parathyroid hormone, diastolic BP variability indices, antiplatelet use, or the proportion receiving insulin therapy (all p > 0.05). (Table 4). Table 4.Univariate analysis of cerebrovascular events in HD patients [ x¯±s , n (%), M (P25, P75)].VariableNon-cerebrovascular event group (n = 186)Cerebrovascular event group (n = 32)t/χ²/z P Age (years)59.86 ± 9.4660.79 ± 9.28−0.5170.605SexMale131 (70.43)19 (59.38)1.5550.212Female55 (29.57)13 (40.63)BMI (kg/m2)22.16 ± 2.8321.75 ± 3.260.7310.466Dialysis vintage (years)2.45 (1.40, 4.08)2.60 (1.23, 6.10)−0.2850.775Smoking historyYes87 (46.77)18 (56.25)0.9820.322No99 (53.23)14 (43.75)Diabetes mellitusYes51 (27.42)17 (53.13)8.4060.004No135 (72.58)15 (46.88)HypertensionYes65 (34.95)20 (62.50)8.7140.003No121 (65.05)12 (37.50)Hemoglobin (g/L)105.35 (92.42, 115.26)108.13 (98.74, 115.83)−0.7660.444Albumin (g/L)37.24 ± 3.6637.06 ± 4.110.2480.805CRP (mg/L)5.17 (2.73, 8.27)5.41 (2.88, 8.72)−0.2460.806Total cholesterol (mmol/L)5.20 ± 0.795.26 ± 0.70−0.4010.689LDL (mmol/L)3.16 ± 0.683.34 ± 0.69−1.3790.169HDL (mmol/L)1.21 ± 0.291.17 ± 0.300.7460.457Calcium (mmol/L)2.34 (2.19, 2.46)2.29 (2.20, 2.49)−0.0490.961Phosphorus (mmol/L)1.59 ± 0.291.54 ± 0.250.9750.330Potassium (mmol/L)4.82 ± 0.754.82 ± 0.760.0180.986Creatinine (µmol/L)897.29 ± 25.51916.90 ± 26.90−3.985<0.001Blood urea nitrogen (mmol/L)22.38 ± 5.5921.66 ± 6.000.6730.502Uric acid (µmol/L)450.79 ± 66.53431.09 ± 63.011.5590.121Urine output (mL/24 h)213.82 ± 104.93192.82 ± 88.041.0690.286Parathyroid hormone (pg/mL)241.93 (179.80, 335.18)257.96 (189.90, 330.02)−0.4580.647SBP-SD7.65 ± 2.259.36 ± 2.75−3.833<0.001SBP-CV5.65 ± 1.686.67 ± 1.89−3.1080.002SBP-ARV8.59 (6.58, 11.01)10.24 (7.77, 13.82)−2.8520.004SBP-RANGE21.59 ± 6.5227.13 ± 9.15−4.155<0.001SBP-MEAN135.51 ± 2.93139.99 ± 3.59−7.735<0.001DBP-SD5.16 (3.91, 6.93)6.21 (4.23, 7.53)−1.2050.228DBP-CV6.06 (4.51, 8.30)7.27 (4.51, 9.19)−1.2590.208DBP-ARV6.31 (4.58, 9.21)8.35 (4.84, 9.87)−0.9800.327DBP-RANGE15.73 (11.55, 21.75)18.31 (13.61, 24.81)−1.2110.226DBP-MEAN85.42 ± 7.9283.94 ± 8.180.9720.332Carotid plaque111 (59.68)25 (78.13)3.9600.047Statin use99 (53.23)23 (71.88)3.9530.050Antiplatelet use39 (20.97)9 (28.13)0.8150.367HbA1c (%)7.24 ± 1.007.65 ± 1.19−2.1530.032Insulin therapy69 (37.10)14 (43.75)0.5130.474Note: BMI = Body mass index; CRP = C-reactive protein; LDL = low-density lipoprotein; HDL = high-density lipoprotein; SBP = Systolic blood pressure; SD = Standard deviation; CV = Coefficient of variation; ARV = Average real variability; DBP = Diastolic blood pressure. Univariate analysis of cerebrovascular events in HD patients [ x¯±s , n (%), M (P25, P75)]. Note: BMI = Body mass index; CRP = C-reactive protein; LDL = low-density lipoprotein; HDL = high-density lipoprotein; SBP = Systolic blood pressure; SD = Standard deviation; CV = Coefficient of variation; ARV = Average real variability; DBP = Diastolic blood pressure. Univariate analysis of cerebrovascular events in PD patients Among 156 PD patients, 12 experienced cerebrovascular events, while 144 remained event-free. The prevalence of hypertension was significantly higher in the event group compared to the nonevent group (75.0 vs. 42.4%, p = 0.029). Paradoxically, baseline serum albumin levels were elevated in the event cohort (38.52 ± 4.18 vs. 35.75 ± 4.20 g/L, p = 0.030). SBP variability indices were markedly increased in the event group (SBP-SD: 8.62 ± 1.31 vs. 6.95 ± 1.95 mmHg, p = 0.004; SBP-CV: 6.52 ± 0.99 vs. 5.44 ± 1.53%, p = 0.018; median SBP-ARV: 10.61 vs. 7.88 mmHg, p = 0.003; and SBP-RANGE (24.73 ± 4.30 vs. 19.81 ± 5.80 mmHg, p = 0.005). Mean SBP was also significantly higher in the event group (132.16 ± 3.15 vs. 127.86 ± 2.65 mmHg, p < 0.001). In contrast, DBP variability indices showed no significant intergroup differences. Also, there were no statistically significant differences between patients with and without cerebrovascular events in the prevalence of carotid plaque (75.0 vs. 56.9%, p = 0.223), statin use (58.3 vs. 52.8%, p = 0.711), antiplatelet use (25.0 vs. 18.8%, p = 0.702), mean HbA1c (7.60 ± 1.16 vs. 7.38 ± 1.25, p = 0.572), or the proportion of patients receiving insulin therapy (41.7 vs. 36.1%, p = 0.759) (Table 5). Table 5.Univariate analysis of cerebrovascular events in PD patients [ x¯±s , n (%), M (P25, P75)].VariableNon-cerebrovascular event group (n = 144)Cerebrovascular event group (n = 12) t/χ²/z/F P Age (years)60.57 ± 9.2657.33 ± 9.071.1650.246SexMale87 (60.24)8 (66.67)0.1820.670Female57 (39.85)4 (33.33)BMI (kg/m2)22.66 ± 2.8721.21 ± 2.781.6840.094Dialysis vintage (years)2.25 (0.78, 4.22)19.4 (0.71, 3.17)−0.9150.360Smoking historyYes78 (54.17)9 (75.00)1.9490.163No66 (45.83)3 (25.00)Diabetes mellitusYes55 (38.19)8 (66.67)3.7300.530No89 (61.81)4 (33.33)HypertensionYes61 (42.36)9 (75.00)4.7700.029No83 (57.64)3 (25.00)Hemoglobin (g/L)105.71 ± 15.12103.47 ± 14.840.4950.622Albumin (g/L)35.75 ± 4.2038.52 ± 4.18−2.190.030CRP (mg/L)5.22 (3.39, 9.02)4.53 (1.85, 6.75)−1.4270.154Total cholesterol (mmol/L)5.18 ± 0.845.44 ± 0.78−1.0540.293LDL (mmol/L)3.31 ± 0.703.21 ± 0.690.4550.650HDL (mmol/L)1.19 ± 0.281.25 ± 0.31−0.6690.505Calcium (mmol/L)2.28 (2.15, 2.45)2.24 (2.15, 2.36)−0.6790.497Phosphorus (mmol/L)1.61 ± 0.301.64 ± 0.23−0.3210.748Potassium (mmol/L)4.82 ± 0.715.09 ± 0.44−1.3040.194Creatinine (µmol/L)597.84 ± 123.81593.12 ± 143.330.1260.900Blood urea nitrogen (mmol/L)21.77 ± 5.5822.14 ± 7.01−0.2130.832Uric acid (µmol/L)448.85 ± 66.93477.49 ± 55.30−1.4400.152Urine output (mL/24 h)399.04 ± 154.96460.64 ± 125.62−1.3400.182Parathyroid hormone (pg/mL)238.65 (146.48, 410.02)241.98 (171.20, 499.72)−0.4990.618SBP-SD6.95 ± 1.958.62 ± 1.31−2.9090.004SBP-CV5.44 ± 1.536.52 ± 0.99−2.3950.018SBP-ARV7.87 (6.00, 9.54)10.61 (8.58, 12.95)−3.0160.003SBP-RANGE19.81 ± 5.8024.73 ± 4.30−2.8650.005SBP-MEAN127.86 ± 2.65132.16 ± 3.15−5.323<0.001DBP-SD4.95 (3.71, 6.66)4.47 (3.05, 6.93)−0.6050.545DBP-CV6.04 (4.26, 8.00)5.45 (3.72, 8.34)−0.3860.700DBP-ARV6.48 (4.70, 8.91)6.36 (4.44, 8.75)−0.1730.863DBP-RANGE15.52 (11.01, 20.32)14.53 (9.38, 20.13)−0.4860.627DBP-MEAN85.27 ± 7.9784.29 ± 9.590.7090.479Carotid plaque82 (56.94)9 (75.00)1.4860.223Statin use76 (52.78)7 (58.33)0.1370.711Antiplatelet use27 (18.75)3 (25.00)–0.702HbA1c (%)7.38 ± 1.257.60 ± 1.16−0.5670.572Insulin therapy52 (36.11)5 (41.67)–0.759Note: BMI = Body mass index; CRP = C-reactive protein; LDL = low-density lipoprotein; HDL = high-density lipoprotein; SBP = Systolic blood pressure; SD = Standard deviation; CV = Coefficient of variation; ARV = Average real variability; DBP = Diastolic blood pressure. Univariate analysis of cerebrovascular events in PD patients [ x¯±s , n (%), M (P25, P75)]. Note: BMI = Body mass index; CRP = C-reactive protein; LDL = low-density lipoprotein; HDL = high-density lipoprotein; SBP = Systolic blood pressure; SD = Standard deviation; CV = Coefficient of variation; ARV = Average real variability; DBP = Diastolic blood pressure. Cox regression analysis of cerebrovascular events in HD patients Following collinearity analysis of significant variables identified in univariate analyses, five parameters with VIF ≤ 5, including diabetes mellitus, hypertension, serum creatinine, SBP-ARV, DBP-ARV, carotid plaque, statin use, and HbA1c were entered into the Cox proportional hazards regression model (LR, forwards). Multivariate Cox regression analysis in the HD group revealed that diabetes mellitus (HR = 2.249, 95% CI 1.095–4.620, p = 0.027) and hypertension (HR = 3.914, 95% CI 1.849–8.284, p < 0.001) were both independent risk factors for cerebrovascular events. Higher serum creatinine was also significantly associated with increased risk (HR = 1.026 per µmol/L, 95% CI 1.010–1.042, p = 0.001). In addition, greater systolic blood pressure average real variability (SBP-ARV) independently predicted cerebrovascular events (HR = 1.087 per mmHg, 95% CI 1.001–1.179, p = 0.046). Notably, elevated HbA1c was associated with a 45.6% higher risk per 1% increase (HR = 1.456, 95% CI 1.015–2.089, p = 0.041), highlighting the role of poor glycemic control in the development of cerebrovascular complications among HD patients, as seen in Table 6. Table 6.Cox regression analysis of cerebrovascular events in HD patients. BSEWald p HR95%CIDiabetes mellitus0.8110.3674.8760.0272.2491.095–4.620Hypertension1.3650.38312.723<0.0013.9141.849–8.284Creatinine0.0250.00810.7250.0011.0261.010–1.042SBP-ARV0.0830.0423.9810.0461.0871.001–1.179HbA1c0.3760.1844.1630.0411.4561.015–2.089Note: SBP = Systolic blood pressure; ARV = Average real variability; DBP = Diastolic blood pressure; SE = Standard error; HR = Hazard Ratio; 95%CI = 95% Confidence Interval. Cox regression analysis of cerebrovascular events in HD patients. Note: SBP = Systolic blood pressure; ARV = Average real variability; DBP = Diastolic blood pressure; SE = Standard error; HR = Hazard Ratio; 95%CI = 95% Confidence Interval. Cox regression analysis of cerebrovascular events in PD patients Following collinearity screening (VIF > 5 exclusion), hypertension, albumin levels, and SBP-ARV were retained for Cox proportional hazards regression analysis. Multivariate modeling identified SBP-ARV as an independent risk factor for cerebrovascular events in PD patients (HR = 1.26, 95% CI: 1.07–1.50, p = 0.007), indicating a 26% elevation in cerebrovascular event risk per unit increment in SBP-ARV. Conversely, neither hypertension (HR = 0.48, 95% CI: 0.14–1.60, p = 0.231) nor albumin levels (HR = 1.12, 95% CI: 0.96–1.31, p = 0.151) demonstrated significant associations with event risk, and the unadjusted difference should be interpreted with caution given the small number of PD events. (Table 7). Table 7.Cox regression analysis of cerebrovascular events in PD patients. BSEWald p HR95%CIHypertension−0.7380.6161.4340.2310.4780.143-1.600Albumin0.1160.082.0660.1511.1220.959-1.314SBP-ARV0.2340.0877.2180.0071.2641.065-1.499Note: SBP = Systolic blood pressure; ARV = Average real variability; SE = Standard error; HR = Hazard Ratio; 95%CI= 95% Confidence Interval. Cox regression analysis of cerebrovascular events in PD patients. Note: SBP = Systolic blood pressure; ARV = Average real variability; SE = Standard error; HR = Hazard Ratio; 95%CI= 95% Confidence Interval. Kaplan-Meier survival curves between dialysis modalities Kaplan-Meier survival analysis revealed significant divergence in cerebrovascular event-free survival rates between HD and PD groups during the 48-month follow-up period (Figure 1). PD patients consistently exhibited a higher probability of event-free survival at most time points compared to HD counterparts. The log-rank test confirmed a statistically significant intergroup difference (p = 0.043). Further HR analysis demonstrated that HD patients faced a 1.955-fold increased risk of cerebrovascular events relative to PD patients (95% CI: 1.075–3.556). Kaplan-Meier survival curves for cerebrovascular event-free survival in PD and HD groups. Discussion This study systematically compared BPV indices during maintenance HD and PD and explored their associations with cerebrovascular event risk. The findings revealed significantly lower SBP variability in PD patients compared to HD counterparts, alongside a reduced incidence of cerebrovascular events in the PD cohort. Multivariate analyses further identified elevated SBP variability during dialysis as an independent predictor of cerebrovascular events in both groups, while mean BP levels demonstrated limited predictive value. These results underscore the critical importance of ambulatory BP management during dialysis and provide novel insights for optimizing stroke prevention strategies in ESRD patients. Previous studies have established BPV as a key predictor of cardiovascular outcomes in hypertensive and chronic kidney disease populations, and this study extends such evidence to HD and PD cohorts [26]. While prior research on dialysis populations remains limited, analogous findings exist. A prospective study in HD patients linked interdialytic SBP variability elevation to heightened stroke risk [23]. Our study, employing multiple BPV indices (SD, ARV, etc.), consistently demonstrated robust associations between BPV and cerebrovascular events. Emerging evidence suggests that dynamic BP fluctuations reflect chronic vascular wall stress, impair cerebral autoregulation, and destabilize microcirculatory perfusion, thereby highlighting that dynamic assessment of BPV can compensate for the limitations of single-time BP measurements. This perspective aligns with findings by Jhee et al. [27], who emphasized the prognostic superiority of BPV over isolated BP readings. Although our study collected intradialytic BPV using 24-h ambulatory blood pressure monitoring during dialysis, ambulatory or home BP monitoring on non-dialysis days would better reflect interdialytic BP variability, which is increasingly recognized to correlate more strongly with cerebrovascular and cardiovascular outcomes in CKD and dialysis populations. Incorporating such home or interdialytic ABPM is therefore expected to improve the generalizability and reliability of BPV estimates across dialysis modalities. Indeed, Liu et al. demonstrated a strong correlation between extended 44-h ABPM and a fixed 24-h ABPM on nondialysis days—suggesting that shorter recording periods may serve as clinically feasible surrogates for longer interdialytic BP profiles [28]. However, extended ABPM monitoring (e.g. ≥44–48 h) poses practical challenges: in hemodialysis patients, approximately 25% refuse such extended recordings, and significant discomfort, sleep disturbance, or skin irritation is frequently reported [29]. Furthermore, in-center ABPM does not capture patients’ home BP behavior. For these reasons, we focused on in-unit BPV, which is logistically feasible and reproducible, but may underestimate the full extent of BP fluctuation across interdialytic intervals. Future prospective studies are warranted to integrate outpatient ABPM (including fixed 24‑hour or extended 44–48 h monitoring) and structured home blood pressure checks, to fully characterize real-world BPV and its incremental value in cerebrovascular risk prediction among CKD and dialysis patients. The differential impacts of dialysis modality on BPV and stroke risk have been documented at home and abroad, yet mechanistic consensus remains elusive. HD patients experience pronounced BP oscillations than PD patients due to rapid fluid shifts and solute concentration changes during sessions, termed the “BP roller-coaster phenomenon”, which may chronically exacerbate cerebrovascular injury [20]. Our findings corroborated that PD patients exhibited lower intradialytic BPV and reduced stroke risk, consistent with multiple observational cohort studies and meta-analyses. For instance, a meta-analysis by Zhan et al. [4] reported lower hemorrhagic stroke risk in PD versus HD patients, though ischemic stroke risk varied by ethnicity, with a higher risk in non-Asian populations while no significant difference observed in Asian populations. Similarly, Boonpheng et al. [30] conducted a systematic review and meta-analysis synthesizing 15 cohort studies (1,289,572 ESRD patients) and identified a 16% reduction in hemorrhagic stroke risk for PD patients compared with HD patients. A recent large-scale multinational study further reported a 19% lower overall stroke risk in PD patients, particularly for hemorrhagic subtypes [31]. Clinically, our findings emphasize the necessity of integrating dynamic BP monitoring and hemodynamic management into routine dialysis care. Traditional practices focusing on pre- and post-dialysis static BP measurements overlook the detrimental effects of intradialytic BP instability on organ perfusion and vascular integrity. The strong association between ambulatory BPV (such as SBP-ARV) and cerebrovascular outcomes observed in our study suggests that BPV indices should be standardized in clinical protocols. In this study, we evaluated BPV using four widely reported metrics: standard deviation (SD), coefficient of variation (CV), average real variability (ARV), and range. These were chosen based on their prevalence in the literature and guideline recommendations. However, no consensus exists on the optimal BPV measure in dialysis populations. Interestingly, only SBP-ARV was independently associated with cerebrovascular events in the PD group, indicating that different BPV metrics may vary in predictive sensitivity. Recent studies suggest that ARV, which reflects dynamic and sequential BP changes, may be more clinically relevant for risk stratification than traditional measures like SD or CV. Further research is needed to directly compare these metrics and establish standardized definitions for clinical use. For HD patients, strategies such as reducing ultrafiltration rates, prolonging session durations, increasing dialysis frequency, or utilizing “cold dialysate” may stabilize BP, mitigate BPV, and reduce stroke risks. It has been shown that significant BP fluctuations during dialysis are associated with an elevated risk of cardiovascular events and all-cause mortality. Notably, Flythe et al. [32] demonstrated that intradialytic BPV independently predicts all-cause and cardiovascular mortality. The cerebrovascular consequences of BPV likely involve impaired cerebral autoregulation, arterial stiffening, and inflammatory cascades. Elevated BPV may reduce cerebral small vessel compliance, destabilize perfusion, and induce microcirculatory injury, predisposing to ischemic or hemorrhagic stroke. Mechanistically, BPV exacerbates white matter lesions, increases blood-brain barrier (BBB) permeability, and amplifies neuroinflammation. While our study did not directly assess these pathways, extant evidence supports their relevance. For example, BPV elevation correlates with cerebral small vessel disease (CSVD) severity, particularly white matter hyperintensity (WMH) burden. A meta-analysis estimated a 26% increase in WMH risk per SD rise in SBP variability [13]. Additionally, elevated BPV may increase the risk of stroke by impairing cerebral autoregulation, leading to unstable cerebral perfusion. Furthermore, BP fluctuations could disrupt BBB integrity, thereby triggering neuroinflammatory responses. A study has shown that elevated BPV is correlated with increased BBB permeability, which may contribute to neuroinflammation and white matter lesions [33]. Recent mechanistic evidence further reinforces the autonomic pathway: a study in chronic kidney disease patients demonstrated that CKD increases beat-to-beat BP variability and reduces baroreflex sensitivity [34], underscoring impaired autonomic regulation as a physiological link between BPV and cerebrovascular vulnerability. This finding complements our observations on SBP-ARV as a predictor of cerebrovascular risk and strengthens the mechanistic framework of our study. Although our current analyses focus exclusively on intradialytic BPV measured via 24‑hour ambulatory blood pressure monitoring during dialysis sessions, accumulating evidence suggests that interdialytic BPV—the variation in predialytic or home-measured BP between dialysis treatments—may have even greater prognostic value for cerebrovascular risk. In a prospective study of maintenance hemodialysis patients, higher interdialytic BPV was independently associated with incident stroke, after adjusting for age, sex, antihypertensive therapy, and mean BP levels [23]. Conversely, recent data indicate that intradialytic BPV may more strongly predict broader cardiovascular outcomes and overall mortality, likely reflecting acute hemodynamic stress during dialysis itself [20]. Thus, while intradialytic BPV remains a valuable marker of hemodynamic instability, it may capture a different clinical phenotype than interdialytic BPV. Nevertheless, our study has several limitations. First, the single-center cohort design and the limited sample size—particularly the low incidence of cerebrovascular events in the PD group—restricted our ability to perform robust subtype-specific stroke analyses and may affect the generalizability of our findings. Second, BPV measurements were exclusively obtained during dialysis sessions; this approach may underestimate true blood pressure variability by omitting fluctuations during interdialytic periods. Although 24-h ambulatory blood pressure monitoring (ABPM) is widely used, a single session may not fully capture long-term BP changes across the entire dialysis cycle, especially in hemodialysis patients with less frequent treatments. Future studies should incorporate structured home BP monitoring and extended or fixed-duration interdialytic ABPM to enhance the generalizability and reliability of BPV estimation and to validate our findings across diverse dialysis settings. Our study population consisted exclusively of Han Chinese patients, and well-documented racial and ethnic differences in blood pressure regulation and cardiovascular risk may limit generalizability. Despite rigorous enrollment and data collection, residual confounding remains possible, including unmeasured factors such as antihypertensive adjustments, inflammatory status, and nutritional profiles. Antihypertensive regimens were not standardized, and drug classes or dosing schedules were not uniformly recorded, which could influence BPV differences between HD and PD groups despite adjustment for hypertension status. Some baseline variables achieved statistical significance despite minimal absolute differences (e.g. serum creatinine), likely reflecting statistical sensitivity or measurement variability rather than clinical relevance. Because of the limited number of cerebrovascular events-particularly in the PD group-Fine-Gray competing-risks modeling and additional stability checks for multicollinearity were not feasible without risking overfitting. The counterintuitive higher albumin observed in PD event cases on univariate analysis did not persist after adjustment and should be regarded as unstable and non-causal. Finally, we did not evaluate specific pathophysiological pathways linking BPV to cerebrovascular risk, such as autonomic dysfunction or arterial stiffness. Larger, multicenter studies with diverse populations, extended follow-up, and granular phenotyping—including ischemic versus hemorrhagic subtype analyses—are needed to validate our findings, clarify mechanisms, and refine BPV-based risk stratification in ESRD. Conclusion Elevated intradialytic BPV was linked to higher cerebrovascular risk, with PD patients showing lower BPV and stroke incidence than HD patients. Although ambulatory SBP variability appears prognostically useful, the single-center, homogeneous cohort, limited PD events, potential confounders, and lack of competing-risks modeling warrant caution. These results are hypothesis-generating and need confirmation in larger, multi-ethnic, multicenter studies before routine clinical adoption. Tailored dialysis prescriptions and antihypertensive regimens for high-BPV patients may help improve cerebrovascular outcomes in ESRD.