Journal of Clinical Hypertension · Nov 7, 2026

Continuous Associations and Predictive Performance of Electrocardiographic Voltage‐ and Product‐Based Indices for Incident Hypertension: A Prospective Cohort Study

By Tanaka F, Noda S, Shimoda H et al.

Background: Electrocardiographic (ECG) indices—Sokolow–Lyon voltage (SLV), Cornell voltage (CV), and Cornell product (CP)—have traditionally been used to detect left ventricular hypertrophy (LVH), including that associated with established hypertension. However, it remains unclear whether these ECG indices, including values below conventional LVH thresholds, provide predictive information for incident hypertension before overt blood pressure (BP) elevation. Methods: We included 2860 adults from a prospective cohort with a baseline BP < 130/80 mmHg, no antihypertensive therapy, and no prior cardiovascular disease. Associations of SLV, CV, and CP with incident hypertension—defined as BP ≥ 140/90 mmHg or initiation of antihypertensive medication—were evaluated using multivariable Cox regression, and their predictive performance was assessed. Results: During follow‐up (mean 6.2 ± 3.0 years), 943 participants (33.0%) developed hypertension. Each 1‐SD increase in sex‐specific standardized ECG indices was independently associated with a higher risk of incident hypertension, with multivariable‐adjusted hazard ratios (95% confidence intervals) of 1.30 (1.22–1.38) for SLV, 1.15 (1.08–1.22) for CV, and 1.15 (1.08–1.22) for CP. Restricted cubic spline analyses showed predominantly linear associations, although modest non‐linearity was observed for SLV ( p = 0.047). The addition of SLV, CV, or CP to a conventional risk model improved discrimination and reclassification at 5 and 7 years (e.g., 5‐year area under the curve: base model, 0.774 vs. base model + SLV, 0.787; p < 0.001). Conclusions: SLV, CV, and CP provide complementary predictive information for incident hypertension beyond conventional risk factors, suggesting that the predictive utility of these ECG indices extends across their full range of values.

doi.org/10.1111/jch.70382 Open access PubMed 42998883 Biomarkers