Abstract

Objective:

To explore the influencing factors of left ventricular diastolic dysfunction (LVDD) in maintenance hemodialysis (MHD) patients and construct a nomogram prediction model.

Methods:

Data was collected from 357 patients who received MHD treatment in our hospital from April 2022 to December 2024. According to a 7:3 ratio, the patients were grouped into a modeling group of 250 cases and a validation group of 107 cases. The modeling group was grouped into LVDD group of 61 cases and non LVDD group of 189 cases based on whether LVDD occurred. Multivariate logistic regression was used to analyze the risk predictive factors of LVDD in MHD patients in the modeling group. R software was used to draw nomograms. The calibration curve and Hosmer-Lemeshow goodness of fit test were used to evaluate the calibration of the nomogram. The receiver operating characteristic (ROC) curve was used to evaluate the discriminative power of the nomogram. Clinical decision curve analysis (DCA) was used to evaluate the clinical utility of nomograms.

Results:

Age, left ventricular hypertrophy, hypertension, diabetes, LVMI and hemoglobin were risk predictors of LVDD in MHD patients (P < 0.05). The consistency between the predicted values in the calibration curves of the modeling and validation groups and actual observed values was good, with Hosmer-Lemeshaw test P = 0.317 and 0.320, and the AUC of the ROC curve was 0.922 and 0.896, indicating good calibration and discrimination ability of the model. When the high-risk threshold probabilities of the modeling group and validation group were within the range of 0.02–0.98 and 0.04–0.86, the net benefit of using a nomogram prediction model to make treatment decisions by clinical physicians was high.

Conclusion:

The nomogram prediction model constructed in this study can help clinicians identify LVDD high-risk patients in MHD and improve management strategies.

1 Introduction

Cardiovascular disease is a common comorbidity in patients with chronic kidney disease and the leading cause of death in patients receiving maintenance hemodialysis (MHD) (1–3). Left ventricular diastolic dysfunction (LVDD) is highly prevalent among patients with chronic kidney disease, occurring early in the course of heart disease, and progressive diastolic dysfunction is independently associated with a higher risk of mortality (4–6). Early detection and intervention of LVDD can reduce the incidence of cardiovascular events and improve the prognosis of MHD patients. In this study, we aimed to investigate the relevant risk factors for LVDD by integrating baseline data and blood laboratory results of MHD patients, and to evaluate its impact on cardiac function. Furthermore, this study utilized the identified risk factors to construct a nomogram prediction model, aiming for early prediction of LVDD and providing guidance for the clinical diagnosis and functional management of cardiovascular disease in MHD patients.

2 Participants and study design2.1 Study subjects

Data were collected from 357 patients receiving MHD treatment at our hospital between April 2022 and December 2024. Patients were divided into a modeling group (n = 250) and a validation group (n = 107) at a 7:3 ratio, with the modeling group used to construct the nomogram prediction model. This study received approval from the hospital's Medical Ethics Committee.

Inclusion criteria: (1) Age ≥ 18 years; (2) Received regular MHD treatment for more than 3 months, with 3 sessions per week, each lasting 4 h. Exclusion criteria: (1) Incomplete medical records; (2) Presence of malignant tumors or severe liver dysfunction; (3) History of cardio-cerebrovascular events such as arrhythmia, cerebrovascular accident, or heart failure within 1 month prior to admission; (4) History of trauma, acute inflammation, or other surgical treatments within 1 month prior to admission. The case collection flowchart is shown in Figure 1.

Case collection process diagram.

2.2 Data collection

Clinical characteristic data of MHD patients were collected, including baseline information, dialysis-related data, and blood laboratory examinations. Baseline information included: age, sex, body mass index (BMI), systolic blood pressure, diastolic blood pressure, smoking status, alcohol consumption, left ventricular hypertrophy, hypertension, coronary artery disease, diabetes, and use of angiotensin-converting enzyme inhibitors/angiotensin II receptor blockers (ACEI/ARB). Dialysis-related data included: dialysis vintage, vascular access type. Blood laboratory examinations included(All measurements were obtained at a single time point): brain natriuretic peptide (BNP), cardiac troponin T (cTnT), high-density lipoprotein cholesterol (HDL-C), low-density lipoprotein cholesterol (LDL-C), triglycerides (TG), total cholesterol (TC), creatinine, urea clearance index (Kt/V), calcium, phosphorus, hemoglobin, albumin, uric acid, 25-hydroxyvitamin D3 [25-(OH)D3], intact parathyroid hormone (iPTH), and arteriovenous fistula laterality. Cardiac ultrasound measurements were performed by experienced sonographers on non-dialysis days. Patients were placed in the left lateral decubitus position, and measurements were obtained at rest. Before the examination, patients were instructed to rest in the supine position for at least 10 min and to avoid strenuous activity and caffeine intake that could affect hemodynamics. All parameters were obtained during the same examination session to minimize the potential influence of pre- and post-dialysis volume status changes on diastolic function indices. Measurements included left ventricular end-diastolic diameter (LVDd), interventricular septal thickness (IVST), and posterior wall thickness (PWT). Endocardial borders at end-diastole and end-systole were traced in apical two- and four-chamber views. The tricuspid regurgitation peak velocity (TRV) was measured using continuous-wave Doppler, and the maximum regurgitant velocity was recorded. In this study, TRV was measurable in all patients, and the regurgitation signals were clearly visualized. Left ventricular ejection fraction (LVEF) was calculated using the Simpson method. Left ventricular mass was calculated as: 0.8 × 1.04 × [(LVDd + IVST + PWT)³—LVDd³] + 0.6. Left ventricular mass index (LVMI) was calculated as left ventricular mass/body surface area. All echocardiographic images were independently analyzed by two experienced cardiologists, both of whom were blinded to the patients' clinical data (including laboratory parameters and group allocation). In cases of discrepant measurements, a third senior cardiologist reviewed the images and the average value was recorded. The final diagnosis was based on agreement between the two primary reviewers.

LVDD Diagnosis: According to the 2016 recommendations by the American Society of Echocardiography and the European Association of Cardiovascular Imaging (7), the following four criteria were used: (1) Septal e′ velocity < 7 cm/s or lateral e′ velocity < 10 cm/s; (2) Average E/e′ ratio > 14; (3) Left atrial volume index (LAVI) > 34 mL/m2; (4) TRV > 2.8 m/s. Diastolic dysfunction was indicated if two or more criteria exceeded the threshold values. The modeling group was divided into an LVDD group (n = 61) and a non-LVDD group (n = 189) based on the presence or absence of LVDD. In the LVDD group, septal e′ was (6.2 ± 0.8) cm/s, lateral e′ was (8.1 ± 1.0) cm/s, mean E/e′ was (15.8 ± 2.5), LAVI was (38.5 ± 4.2) mL/m2, and TRV was (3.1 ± 0.4) m/s.In the non-LVDD group, septal e′ was (8.5 ± 1.2) cm/s, lateral e′ was (11.2 ± 1.5) cm/s, mean E/e′ was (9.2 ± 1.8), LAVI was (28.6 ± 3.1) mL/m2, and TRV was (2.4 ± 0.3) m/s.

2.3 Statistical analysis

Statistical analysis was performed using SPSS software (version 25.0). Normally distributed continuous data were expressed as mean ± standard deviation (mean ± SD) and compared using independent samples t-tests; Categorical data were expressed as counts (percentages) [n (%)] and compared using the chi-square (χ²) test; Multivariable logistic regression analysis was used to identify risk predictors for LVDD in the modeling group of MHD patients, with results expressed as odds ratios (OR); A nomogram was plotted based on the logistic regression results using the RMS package in R software; Internal validation was conducted using 1,000 bootstrap resamples;The calibration of the nomogram was assessed using calibration curves and the Hosmer-Lemeshow goodness-of-fit test; The discrimination of the nomogram was evaluated using the receiver operating characteristic (ROC) curve; The clinical utility of the nomogram was assessed using decision curve analysis (DCA). A two-sided P < 0.05 was considered statistically significant.

3 Results3.1 Comparison of clinical characteristics between modeling and validation groups

There were no statistically significant differences (P > 0.05) between the modeling and validation groups in terms of age, sex, BMI, dialysis vintage, systolic blood pressure, diastolic blood pressure, smoking status, alcohol consumption, vascular access type, ACEI/ARB medication use, left ventricular hypertrophy, hypertension, coronary artery disease, diabetes, BNP, cTnT, LVEF, LVMI, HDL-C, LDL-C, TG, TC, creatinine, Kt/V, calcium, phosphorus, hemoglobin, albumin, uric acid, 25-(OH)D3, iPTH, and arteriovenous fistula laterality. See Table 1.

IndexModeling group(n = 250)Validation group(n = 107)χ2/tPAge (years old)1.4920.222 ≤60146 (58.40)55 (51.40) >60104 (41.60)52 (48.60)Gender [n(%)]1.2860.257 Male138 (55.20)66 (61.68) Female112 (44.80)41 (38.32)BMI(kg/m2)22.66 ± 2.7522.89 ± 2.930.7100.478Dialysis age (years)3.61 ± 1.483.90 ± 1.421.7170.087Systolic pressure(mmHg)146.37 ± 21.79149.82 ± 22.561.3560.176Diastolic pressure(mmHg)83.11 ± 10.8585.37 ± 11.041.7940.074Smoke [n(%)]121 (48.40)62 (57.94)2.7320.098Drink [n(%)]34 (13.60)18 (16.82)0.6250.429Vascular access [n(%)]1.7370.187 Venous catheterization88 (35.20)30 (28.04) Arteriovenous fistula162 (64.80)77 (71.96)Left ventricular hypertrophy [n(%)]113 (45.20)53 (49.53)0.5650.452Hypertension [n(%)]154 (61.60)75 (70.09)2.3500.125Diabetes [n(%)]43 (17.20)11 (10.28)2.7950.095BNP(pg/mL)487.59 ± 120.74499.76 ± 128.930.8550.393cTnT(ng/L)46.52 ± 12.9844.53 ± 13.821.3010.194LVEF (%)60.48 ± 6.9462.01 ± 6.791.9210.056LVMI(g/m2)116.64 ± 22.90120.85 ± 24.461.5590.120HDL-C(mmol/L)1.03 ± 0.221.05 ± 0.190.8190.414LDL-C(mmol/L)2.24 ± 0.522.13 ± 0.471.8830.060TG(mmol/L)1.26 ± 0.361.32 ± 0.341.4670.143TC(mmol/L)4.19 ± 1.094.24 ± 1.130.3930.695Creatinine(μmol/L)862.73 ± 235.18882.54 ± 237.690.7270.468Kt/V1.61 ± 0.281.55 ± 0.271.8750.062Ca(mmol/L)2.25 ± 0.192.29 ± 0.221.7360.083P(mmol/L)1.77 ± 0.451.74 ± 0.480.5660.572Hemoglobin(g/L)96.90 ± 18.75100.42 ± 22.371.5310.127Albumin(g/L)40.11 ± 3.9740.75 ± 4.621.3270.185Uric acid(μmol/L)450.30 ± 90.57464.92 ± 89.491.4020.16225-(OH)D321.27 ± 5.9422.08 ± 5.731.1930.234iPTH274.66 ± 62.03287.35 ± 65.761.7390.083Arteriovenous fistula side [n(%)]0.6950.405 Left233 (93.20)97 (90.65) Right17(6.80)10(9.35)

Comparison of clinical characteristics between modeling group and validation group[n(%)/()].

MHD patients were divided into a modeling group of 250 cases and a validation group of 107 cases based on a 7:3 ratio; BNP, B-type natriuretic peptide; cTnT, cardiac troponin T; LVEF, left ventricular ejection fraction; LVMI, left ventricular mass index; HDL-C, high-density lipoprotein cholesterol; LDL-C, low-density lipoprotein cholesterol; TG, triglycerides; TC, total cholesterol;25-(OH)D3, 25-hydroxyvitamin D3; iPTH, intact parathyroid hormone; ACEI/ARB, angiotensin converting enzyme inhibitor/angiotensin II receptor blocker.

3.2 Comparison of clinical characteristics between non-LVDD and LVDD groups in the modeling group

Within the modeling group, there were no statistically significant differences (P > 0.05) between the two groups in terms of sex, BMI, dialysis vintage, systolic blood pressure, diastolic blood pressure, smoking status, alcohol consumption, coronary artery disease, vascular access type, ACEI/ARB medication use, BNP, cTnT, LVEF, HDL-C, LDL-C, TG, TC, creatinine, Kt/V, calcium, phosphorus, albumin, uric acid, uric acid, 25-(OH)D3, iPTH, and arteriovenous fistula laterality. In the modeling group, patients in the LVDD group had higher rates of age >60, left ventricular hypertrophy, hypertension, diabetes, and higher LVMI compared to the non-LVDD group, while hemoglobin levels were lower than in the non-LVDD group (P < 0.05). See Table 2.

IndexNon LVDD group(n = 189)LVDD group(n = 61)χ2/tPAge (years old)14.2240.000 ≤60123 (65.08)23 (37.70) >6066 (34.92)38 (62.30)Gender [n(%)]1.1820.277 Male108 (57.14)30 (49.18) Female81 (42.86)31 (50.82)BMI(kg/m2)22.57 ± 2.8422.92 ± 2.560.8570.393Dialysis age (years)3.52 ± 1.483.90 ± 1.501.7380.083Systolic pressure(mmHg)145.07 ± 20.12150.41 ± 22.631.7470.082Diastolic pressure(mmHg)82.51 ± 10.9484.95 ± 10.711.5220.129Smoke [n(%)]95 (50.26)26 (42.62)1.0780.299Drink [n(%)]24 (12.70)10 (16.39)0.5360.464Vascular access [n(%)]0.2060.650 Venous catheterization68 (35.98)20 (32.79) Arteriovenous fistula121 (64.02)41 (67.21)Left ventricular hypertrophy [n(%)]71 (37.57)42 (68.85)18.2240.000Hypertension [n(%)]102 (53.97)52 (85.25)19.0730.000Coronary artery disease61 (32.28)27 (44.26)2.9050.088Diabetes [n(%)]22 (11.64)21 (34.43)16.8120.000Application of ACEI/ARB drugs3.0820.079 No117 (61.90)30 (49.18) Yes72 (38.10)31 (50.82)BNP(pg/mL)485.19 ± 126.50494.98 ± 117.230.5350.593cTnT(ng/L)45.74 ± 12.0648.95 ± 13.661.7490.082LVEF (%)60.89 ± 5.1759.22 ± 8.151.8810.061LVMI(g/m2)110.52 ± 23.75135.62 ± 22.877.2410.000HDL-C(mmol/L)1.04 ± 0.221.01 ± 0.240.9050.366LDL-C(mmol/L)2.23 ± 0.562.29 ± 0.480.7520.453TG(mmol/L)1.25 ± 0.371.31 ± 0.351.1160.266TC(mmol/L)4.18 ± 1.144.22 ± 1.060.2420.809Creatinine(μmol/L)846.48 ± 225.72913.06 ± 252.941.9440.053Kt/V1.60 ± 0.221.63 ± 0.450.6960.487Ca(mmol/L)2.25 ± 0.192.27 ± 0.200.7060.481P(mmol/L)1.78 ± 0.421.75 ± 0.500.4620.644Hemoglobin(g/L)102.43 ± 20.9479.76 ± 16.747.6960.000Albumin(g/L)40.39 ± 4.1539.26 ± 3.541.9130.057Uric acid(μmol/L)446.75 ± 91.32461.28 ± 89.241.0860.27825-(OH)D3(ng/mL)21.64 ± 6.3520.13 ± 5.151.6860.093iPTH(pg/mL)27 (1.45 ± 60.29)284.62 ± 67.331.4410.151Arteriovenous fistula side [n(%)]0.6250.429 Left178 (94.18)55 (90.16) Right11(5.82)6(9.84)

Comparison of clinical characteristics between non LVDD group and LVDD group in the modeling group [n(%)/()].

The modeling group was divided into LVDD group (61 cases) and Non LVDD group (189 cases) based on whether LVDD occurred; BMI, body mass index; BNP, B-type natriuretic peptide; cTnT, cardiac troponin T; LVEF, left ventricular ejection fraction; LVMI, left ventricular mass index; HDL-C, high-density lipoprotein cholesterol; LDL-C, low-density lipoprotein cholesterol; TG, triglycerides; TC, total cholesterol; 25-(OH)D3, 25-hydroxyvitamin D3; iPTH, intact parathyroid hormone; ACEI/ARB, angiotensin converting enzyme inhibitor/angiotensin II receptor blocker.

3.3 Multivariable logistic regression analysis of risk predictors for LVDD in MHD patients in the modeling group

Variables with P < 0.05 in the univariate analysis from Section 3.2 were included in the multivariable analysis. Variable assignments are shown in Table 3. Multivariable logistic regression analysis identified risk predictors for LVDD in MHD patients. The results showed that age (OR = 3.195), left ventricular hypertrophy (OR = 5.610), hypertension (OR = 6.088), diabetes (OR = 3.436), LVMI (OR = 1.047), and hemoglobin (OR = 0.928) were risk predictors for LVDD in MHD patients (P < 0.05). See Table 4.

VariableAssignmentAge1 = >60 years old, 0 = ≤60 years oldLeft ventricular hypertrophy1 = Yes, 0 = NoHypertension1 = Yes, 0 = NoDiabetes1 = Yes, 0 = NoLVMIContinuous variableHemoglobinContinuous variableDependent variable1 = LVDD, 0 = non LVDD

Variable assignment table.

LVMI, left ventricular mass index.

Influence factorβSEWaldχ2OR95% CIPAge1.1620.4466.7913.1951.334–7.6530.009Left ventricular hypertrophy1.7250.45814.1975.6102.288–13.7600.000Hypertension1.8060.51612.2746.0882.216–16.7230.000Diabetes1.2340.5215.6193.4361.238–9.5330.018LVMI0.0450.01022.3891.0471.027–1.0660.000Hemoglobin−0.0750.01430.5790.9280.903–0.9530.000Constant−2.8311.5003.5600.000–0.000

Multivariate logistic regression analysis of risk predictive factors for LVDD in MHD patients.

LVMI, left ventricular mass index.

3.4 Construction of a nomogram model for predicting LVDD in MHD patients

A nomogram model was constructed based on the risk predictors for LVDD in MHD patients: age, left ventricular hypertrophy, hypertension, diabetes, LVMI, and hemoglobin. See