Introduction
Cardiovascular disease (CVD) remains the leading cause of morbidity and mortality among patients undergoing hemodialysis. These patients are particularly susceptible to cardiac complications because they face a combination of traditional risk factors and dialysis-specific pathophysiological mechanisms, such as rapid fluid shifts, recurrent hypotension, and chronic disturbances in calcium-phosphate metabolism [1]. Therefore, identifying accessible and reliable markers for cardiovascular risk stratification in this high-risk population is a critical clinical priority.
Electrocardiography (ECG) is a noninvasive, cost-effective modality that reflects the electrical remodeling of the myocardium. Among various ECG parameters, the presence of a positive T wave in lead augmented vector right (TaVR)—although often overlooked—has been increasingly recognized as a marker of adverse cardiovascular outcomes in the general population. This finding may reflect subendocardial ischemia or abnormal ventricular repolarization [2–5]. However, its prognostic value has yet to be thoroughly investigated among patients undergoing hemodialysis, where cardiovascular risk is markedly elevated.
Echocardiography provides complementary insights into structural and functional cardiac status. In particular, the E/e′ ratio, derived from tissue Doppler imaging, serves as a surrogate for left ventricular (LV) filling pressure and is commonly used to assess diastolic dysfunction [6,7]. Although elevated E/e′ has been associated with poor cardiovascular outcomes [8,9], its incremental prognostic utility—when combined with ECG findings such as a positive TaVR—remains under-explored in hemodialysis.
Therefore, we aimed to evaluate the prognostic significance of positive TaVR for predicting major adverse cardiovascular events (MACE) in patients undergoing hemodialysis. In addition, we examined whether incorporating E/e′ measurements could enhance the predictive performance of positive TaVR to determine whether the combined model improves risk stratification.
Methods
Study population
This retrospective cohort study was conducted at the dialysis clinic of Kangdong Sacred Heart Hospital, a 600-bed academic teaching hospital in Seoul, South Korea, between October 2018 and April 4, 2024. Eligible participants were adult patients (≥18 years) receiving maintenance hemodialysis three times per week for at least 3 months, with available baseline ECG and transthoracic echocardiography (TTE) performed according to standardized protocols. The cohort comprised two groups: patients regularly undergoing hemodialysis at Kangdong Sacred Heart Hospital, and patients referred from external dialysis centers who were followed at the nephrology outpatient clinic as potential recipients for deceased donor kidney transplantation when brain-death donors became available at our institution. Patients were excluded if they were receiving hemodialysis due to chronic rejection following a prior kidney transplantation, had a planned transfer to another dialysis center during the follow-up period, lacked baseline ECG or TTE, presented with atrial fibrillation or atrial flutter on baseline ECG, or had a follow-up duration shorter than 3 months or markedly incomplete medical records.
The study protocol was approved by the Institutional Review Board (IRB) of Kangdong Sacred Heart Hospital (No. 2025-04-010) and conducted in accordance with the Declaration of Helsinki. Given the retrospective design using fully anonymized and de-identified data, the IRB waived the requirement for written informed consent.
Data collection
Baseline cardiovascular and clinical data were obtained from two sources. First, 106 patients were derived from a previously published prospective observational study titled “Vascular calcification and cardiac function according to residual renal function in patients on hemodialysis with urination” [10]. All participants in this prospective study had provided written informed consent and underwent protocol-driven ECG, TTE, and comprehensive clinical assessment. Second, additional patients from external dialysis centers were evaluated at the time of transplant candidacy registration at our hospital. For these patients, standardized cardiovascular assessment (ECG and TTE) was performed as part of the routine workup. Follow-up for clinical outcomes—including cardiovascular events and death—was conducted through serial electronic medical record review and ongoing nephrology clinic visits.
For patients referred from external dialysis centers, outcome data were supplemented by transplant coordinator nurses, who regularly contacted patients or their caregivers to update clinical status as part of the deceased donor waiting list management process. This ensured longitudinal data integrity even for patients not dialyzing within our hospital system.
Definition of variables and outcomes
MACE was defined as a composite outcome comprising cardiovascular death (including sudden cardiac death), nonfatal myocardial infarction, nonfatal stroke (ischemic or hemorrhagic), hospitalization for heart failure, or coronary revascularization procedures (percutaneous coronary intervention or coronary artery bypass grafting).
A positive TaVR was defined as a positive deflection in amplitude greater than 0 mV on baseline 12-lead ECG. Prior CVD was defined as a documented history of coronary artery disease, myocardial infarction, stroke, or peripheral artery disease. Heart failure was defined as heart failure with reduced LV ejection fraction (LVEF <40%) on baseline TTE.
Evaluation of electrocardiogram findings
Standard 12-lead ECGs were obtained at baseline using the Philips PageWriter TC30 cardiograph (Philips Medical Systems, Inc.). ECGs were recorded at a paper speed of 25 mm/sec and a calibration of 10 mm/mV. All ECGs were independently reviewed and interpreted by two experienced cardiologists blinded to clinical data. Discrepancies in interpretation were resolved by consensus.
Parameters assessed included heart rhythm, QRS duration, corrected QT interval (QTc, using Bazett’s formula), and T wave morphology in lead augmented vector right (aVR).
Echocardiography
Baseline TTE was performed using Vivid7 Ultrasound Systems (GE Vingmed Ultrasound AS), following the American Society of Echocardiography guidelines [11]. Standard two-dimensional, M-mode, Doppler, and tissue Doppler imaging techniques were used. Studies were conducted on the morning following a dialysis session to minimize volume-related confounding.
LV systolic function was quantified as the ejection fraction (EF) derived from the modified Simpson biplane method on apical four- and two-chamber views. Diastolic function was evaluated using the transmitral E/A ratio, mitral annular Ea and Aa velocities (averaged from septal and lateral walls), deceleration time, and the E/e′ ratio (early transmitral E divided by septal e′) [11]. LV mass was calculated with the Devereux ‘cube’ formula and indexed to body-surface area to give the LV mass index (LVMI) [12].
Statistical analyses
All statistical analyses were performed using R version 4.4.3 (R Foundation for Statistical Computing); a full package list with versions is provided in Supplementary Methods (available online). Categorical variables were expressed as counts and percentages, and continuous variables as mean ± standard deviation or median with interquartile range, depending on distribution. Group comparisons were conducted using the chi-square test or Fisher exact test for categorical variables and the Student t test or Mann-Whitney U test for continuous variables.
The primary outcome was the time from baseline to the first MACE event. For patients with multiple events, only the first was counted. Patients who underwent kidney transplantation, were transferred to another facility, or died from non-cardiovascular causes were censored at the time of last follow-up.
The Kaplan-Meier curves were generated to compare MACE-free survival across groups, and log-rank tests were used for significance testing. To identify independent predictors of MACE, Cox proportional hazards regression was performed. Variables with p < 0.10 in univariate analysis or clinically relevant factors—including age, sex, dialysis vintage, diabetes, prior CVD, calcium, phosphate, heart failure, and LVMI—were included in multivariate models. Multicollinearity was also assessed using variance inflation factors (VIF), and the event-per-variable (EPV) ratio was confirmed to be sufficient, ensuring the stability of multivariate estimates (Supplementary Methods, available online). The proportional hazards assumption was evaluated using Schoenfeld residuals. Visual inspection of scaled Schoenfeld residual plots for each covariate and the global model showed no systematic trends over time, with smoothed lines remaining within the confidence bands around zero (Supplementary Fig. 1, available online). In addition, all individual covariates had nonsignificant p-values (all p > 0.05), and the global test yielded p = 0.31, indicating that the proportional hazards assumption was satisfied (Supplementary Table 1, available online).
To evaluate the incremental prognostic value of positive TaVR and E/e′ ratio, four nested Cox models were constructed as follows: 1) a base model including standard clinical and biochemical variables; 2) Model 1: base model plus E/e′ ratio; 3) Model 2: base model plus positive TaVR; and 4) Model 3: base model including both E/e′ ratio and positive TaVR.
For categorical modeling, the E/e′ ratio was dichotomized at the cohort-specific median value of 19.0 to ensure balanced group sizes. To assess the robustness of this dichotomization, a sensitivity analysis was performed using the guideline-recommended E/e′ threshold of 15.0 [13]. Model discrimination was assessed using Harrell’s C-index, with 95% confidence intervals (CIs) derived from 1,000 bootstrap replicates. Reclassification performance was assessed via the integrated discrimination improvement (IDI) and continuous net reclassification improvement (NRI). Likelihood-ratio tests were used for comparing nested models. Statistical significance was set at two-sided p < 0.05. For reclassification metrics, both IDI and continuous NRI were calculated at the 5-year time point using 500 perturbation resamples. Model diagnostics also confirmed that multicollinearity was minimal (all VIFs <2.0; Supplementary Table 2, available online).
Results
Study population
A total of 328 patients undergoing hemodialysis were screened. Thirty-two were excluded: missing or uninterpretable baseline ECG (n = 1); absence of echocardiographic E/e′ ratio measurements on baseline echocardiography (n = 2); atrial fibrillation or atrial flutter on baseline ECG (n = 12); and a follow-up duration of less than 3 months or markedly incomplete medical records (n = 17). The final analysis therefore included 296 patients with complete and interpretable ECG and echocardiographic data (Supplementary Fig. 2, available online).
Baseline characteristics
The baseline demographic, clinical, and biochemical characteristics of the 296 patients included in the study are summarized in Table 1. The median age was 64.0 years (range, 30.0–88.0 years), and 147 patients (49.7%) were male. Based on ECG criteria, 111 patients (37.5%) were the positive TaVR.
Among clinical characteristics, patients with positive TaVR had a significantly longer dialysis vintage (medians of 46.8 months vs. 36.2 months, p = 0.01). Regarding biochemical parameters, serum calcium was lower in positive TaVR (p = 0.04). Echocardiographic assessment revealed pronounced diastolic dysfunction in patients with positive TaVR, as reflected by a significantly higher E/e′ (medians of 26.0 vs. 15.0, p < 0.001). These patients also showed a trend toward greater LVMI (140.3 g/m² vs. 127.7 g/m², p = 0.08), although EF and heart failure prevalence (LVEF <40%) were comparable between groups.
Major adverse cardiovascular events
During the median follow-up duration of 56.5 months (approximately 4.7 years), a total of 118 patients experienced MACE. The overall MACE incidence in our cohort was 8.9 events per 100 person-years. The most frequent component of MACE was cardiovascular death, which occurred in 34 patients (28.7%), followed by nonfatal myocardial infarction in 27 patients (23.0%), hospitalization for heart failure in 24 patients (20.7%), nonfatal stroke in 19 patients (16.1%), and coronary revascularization procedures in 14 patients (11.5%).
Cardiovascular outcomes according to the presence of a positive augmented vector right
Among those with positive TaVR, MACE occurred in 82 out of 111 patients (73.9%), compared to 30 out of 185 patients (16.2%) in the negative aVR group. This corresponds to incidence rates of 16.0 events vs. 3.7 events per 100 person-years, respectively (p < 0.001). The Kaplan-Meier analysis confirmed that MACE-free survival was significantly lower in the positive aVR group (p < 0.001) (Fig. 1).
Predictors of cardiovascular events
As summarized in Table 2, the presence of a positive TaVR was a strong and independent predictor of MACE. In univariate Cox regression analysis, positive TaVR was significantly associated with an increased risk of MACE (hazard ratio [HR], 4.67; 95% CI, 3.05–7.15; p < 0.001). This association remained significant in the multivariate model after adjustment for clinical and echocardiographic covariates (adjusted HR, 3.19; 95% CI, 2.00–5.08; p < 0.001).
Similarly, an elevated E/e′ (≥19.0)—a marker of diastolic dysfunction—was also independently associated with adverse outcomes (adjusted HR, 2.92; 95% CI, 1.71–4.96; p < 0.001). Although both positive TaVR and elevated E/e′ provided significant prognostic value, positive TaVR demonstrated greater predictive strength in this cohort.
To explore potential effect modification, we added an interaction term between positive TaVR and a high E/e′ (≥19) to the fully adjusted Cox model (Supplementary Table 3 lists the full model coefficients and Supplementary Table 4 reports the stratified HRs for each TaVR × E/e′ subgroup; available online). The interaction coefficient was a significant negative term (β = –1.539 ± 0.559; HR, 0.21; 95 % CI, 0.07–0.64; p = 0.01), indicating that the prognostic impact of positive TaVR was attenuated in the presence of a high E/e′—and, conversely, that the effect of a high E/e′ was attenuated when TaVR was positive.
Enhanced predictive performance with E/e′ ratio integration
Adding the mechanical marker high E/e′ (≥19) to the clinical base model raised the C-index from 0.65 to 0.71 (Δ +0.06, p < 0.001) and improved reclassification (IDI, 0.08; continuous NRI, 0.08). Introducing the electrical marker positive TaVR produced an even larger gain (C-index, 0.73; Δ +0.08; p < 0.001) and a higher NRI (0.16). The combination of positive TaVR and elevated E/e′ (≥19) yielded the highest improvement in model performance, with a C-index of 0.75 (Δ +0.10), IDI of 0.10, and continuous NRI of 0.16 (all p < 0.01), indicating additive prognostic value beyond clinical covariates alone (Table 3). Although positive TaVR alone provided a slightly larger gain in C-index, both markers together yielded the greatest improvement (C index, 0.75; Δ 0.10; IDI, 0.10; NRI, 0.16; all p < 0.01). Sensitivity analysis with a guideline-recommended E/e′ cut off of 15 produced similar improvements (C index, 0.74; Δ 0.09; IDI, 0.09; NRI, 0.17), confirming robustness (Supplementary Table 5, available online). Subgroup analyses by diabetes status and prior CVD showed generally consistent patterns, though the magnitude of improvement varied (Supplementary Table 6, available online).
Discussion
In this cohort of 296 hemodialysis patients, positive TaVR emerged as the single most powerful independent predictor of MACE. A positive TaVR increased the adjusted risk of MACE more than threefold, exceeding the prognostic strength of the echocardiographic E/e′ ratio and every conventional clinical covariate we tested. Although a high E/e′ (≥19.0) was itself an independent predictor, its greatest value appeared when combined with positive TaVR: the dual-marker model improved the C-index by 0.10 and produced significant gains in both IDI and continuous NRI. Interestingly, traditional cardiovascular risk factors—such as diabetes, prior CVD, dialysis vintage, and LVMI—did not retain independent prognostic significance in the multivariable model. Several explanations merit consideration. First, the number of events per covariate was sufficient (EPV ≈ 10.2), minimizing the risk of overfitting. Second, multicollinearity among predictors was assessed using the VIF, with all values below 2.0, confirming model stability. These findings suggest that the lack of statistical significance was not due to statistical artifact or model instability, but rather reflects overlapping prognostic information. It is plausible that TaVR and E/e′ serve as surrogate or composite indicators that integrate multiple pathophysiologic processes—such as myocardial ischemia, concentric remodeling, and elevated LV filling pressure—thereby capturing cardiovascular risk more effectively than individual clinical or structural parameters. To our knowledge, this is the first study to show that integrating a simple ECG repolarization marker with a Doppler-derived index of diastolic load meaningfully enhances cardiovascular risk stratification in the hemodialysis setting.
The overall MACE incidence in our cohort was 8.9 events per 100 person-years, closely mirroring the rate reported in Japan (7.5 per 100 person-years) but markedly lower than contemporary observations in North America (19.4 per 100 person-years) and Europe (17.4 per 100 person-years) [14]. Differences in patient age, dialysis vintage, and case-mix likely explain some of this geographic variation, yet our findings confirm that cardiovascular risk among Asian hemodialysis populations remains substantial.
Our results extend prior work in non-dialysis cohorts linking positive TaVR to cardiac death after myocardial infarction, ventricular arrhythmias in cardiomyopathy, and broad cardiovascular mortality in the general population [2,3,15–20]. In advanced chronic kidney disease, Kurisu et al. [5] associated positive TaVR with larger LV volumes, but outcome data were sparse. By demonstrating robust associations with hard cardiovascular endpoints, we show that repolarization abnormalities captured by lead aVR remain clinically relevant even amid the profound ventricular remodeling.
While prior studies, including the work by Jaroszyński et al. [21], have demonstrated the prognostic relevance of a positive T wave in lead aVR among hemodialysis patients, our study advances this knowledge by introducing a practical two-step strategy that combines ECG-based repolarization abnormality (TaVR) with echocardiographic diastolic burden (E/e′). This approach captures both electrical and mechanical stress in a dialysis-specific context and is feasible for routine use given the widespread availability of both tests. Notably, despite a modest increase in the C-index (Δ 0.10), the combined model significantly improved net reclassification (IDI and NRI), highlighting meaningful gains in patient-level risk discrimination that are not fully reflected in global discrimination metrics alone. To facilitate balanced group sizes for statistical modeling, we dichotomized the E/e′ ratio at the cohort-specific median value of 19.0. However, using the guideline-recommended threshold of ≥15 yielded nearly identical results (C-index, +0.09; similar IDI/NRI), confirming robustness across clinically relevant cutoffs (Supplementary Table 5, available online).
Mechanistically, the observed negative interaction between TaVR and E/e′ supports the idea that these markers reflect overlapping but distinct pathways of dialysis-induced cardiomyopathy. TaVR likely reflects subendocardial ischemia, concentric hypertrophy, and myocardial fibrosis, whereas E/e′ captures elevated LV filling pressures and impaired diastolic relaxation [5,8,9,22]. In the hemodialysis setting—characterized by chronic volume overload, arterial stiffness, calcium-phosphate imbalance [23], and frequent intradialytic myocardial stunning—these processes often co-occur, converging on shared fibrotic or ischemic remodeling. The lack of additive prognostic value may thus reflect biological redundancy rather than statistical artifact, reinforcing the interpretation of TaVR and E/e′ as complementary surrogate markers rather than entirely independent predictors.
From a clinical standpoint, lead aVR is recorded on every 12-lead ECG yet is rarely examined in routine interpretation. Because ECG is inexpensive, operator-independent, and universally available in dialysis units, incorporating TaVR status into routine surveillance could provide a pragmatic first-line screen to identify high-risk patients. A positive TaVR may prompt further echocardiographic evaluation to assess the diastolic burden. E/e′—readily obtained via standard TTE—adds granularity to risk stratification without requiring advanced imaging or expensive biomarkers. As a surrogate of LV filling pressure and diastolic dysfunction, E/e′ is particularly informative in hemodialysis settings, where chronic volume overload, arterial stiffness, and LV hypertrophy drive cardiovascular risk [13]. Prior studies have shown that elevated E/e′ outperforms conventional indices, such as EF, in predicting adverse events [8,9,24]. In clinical practice, this two-step ECG → E/e′ strategy may guide practical interventions:
• For patients with positive TaVR: consider focused echocardiographic evaluation.
• For patients with elevated E/e′: initiate or intensify volume management, refer to cardiology, increase arrhythmia monitoring, or consider ischemia workup.
This approach is feasible even in resource-limited dialysis units and may enable earlier identification of patients who would benefit from intensified cardiovascular care.
Key strengths of this study are: (i) a relatively large hemodialysis cohort (296 patients, 1,325 patient-years) with robust event capture; (ii) blinded, protocol-driven interpretation of ECG and TTE; and (iii) rigorous assessment of incremental prognostic value using bootstrap-validated C-indices, IDI, and continuous NRI.
Several limitations should be acknowledged. First, the retrospective, single-center design introduces potential for selection and survivor bias, limiting the generalizability of findings beyond Korean dialysis practice patterns. Second, a subset of patients was recruited from external dialysis centers during deceased donor kidney transplant evaluation; these individuals may differ systematically from the broader dialysis population, introducing selection bias. Third, heart failure was defined solely by LVEF <40%, which likely underestimated the prevalence of heart failure with preserved EF —a common but often subclinical phenotype in this population. Fourth, echocardiographic data were acquired at a single baseline time point, and adjudication of sudden cardiac death relied on chart review without continuous rhythm monitoring, introducing potential information bias. Fifth, data on dialysis adequacy (e.g., Kt/V) and medication burden (e.g., number of antihypertensive agents) were not consistently available across the cohort, precluding their inclusion in multivariable modeling and descriptive tables. Although both are known contributors to cardiovascular risk, missingness may have introduced residual confounding. Sixth, we used complete-case analyses; although this avoids the assumptions implicit in multiple imputation, it may have attenuated associations for variables with missing values. Seventh, subgroup analyses stratified by diabetes status and prior CVD showed consistent directional effects for the dual-marker model, but improvements in IDI and NRI did not reach statistical significance in these high-risk subgroups. Given that 59.5% of the cohort had diabetes and 35.8% had prior CVD, this finding suggests that the incremental prognostic gain of the combined strategy may be attenuated when baseline cardiovascular risk is already high.
Although our findings were internally robust, external validation using an independent hemodialysis cohort is necessary to assess the generalizability of this two-step strategy across different populations and clinical settings. Prospective, multicenter studies should determine whether targeted interventions—such as intensified volume control, dialysate calcium adjustment, or proactive arrhythmia surveillance—improve outcomes among TaVR-positive patients. Additionally, longitudinal monitoring of ECG and echocardiographic parameters could clarify whether temporal changes in E/e′ or TaVR status are modifiable and whether such changes correlate with reduced cardiovascular risk. Moreover, future research should explore the integration or comparison of dialysis-specific predictors like TaVR and E/e′ with traditional cardiovascular risk scores once comprehensive datasets, including lipid profiles, become available.
In conclusion, a positive TaVR is a powerful, easily obtained marker of cardiovascular risk in maintenance hemodialysis. When combined with a high E/e, risk discrimination improves even further, underscoring the value of integrating electrical and mechanical indices of dialysis-related cardiomyopathy. Routine assessment of TaVR and E/e′ may enable earlier identification of patients who would benefit most from intensive cardioprotective strategies.
PDF Links
PubReader
ePub Link
XML Download
Full text via DOI
Download Citation
Supplement 1
Print
