Beyond Voltage: Independent Validation of the SV3  +  SV4 Criterion for LVH Detection

We read with great interest the study by Yoosefi et al. proposing the SV3 + SV4 voltage sum—defined as the algebraic addition of the S-wave amplitudes in leads V3 and V4—as a new ECG criterion for left ventricular hypertrophy (LVH) and exploring whether age, sex, and hypertension improve diagnostic performance (Yoosefi et al. 2025). Using their thresholds, the authors reported a sensitivity of 0.609 and specificity of 0.669 overall; with sex-specific cutoffs, performance reached 0.500/0.809 in women and 0.556/0.910 in men (Yoosefi et al. 2025). These values represent a modest improvement over traditional voltage criteria but still illustrate limited sensitivity, especially in women. To assess generalizability, we externally validated SV3 + SV4 in a Mexican cohort (n = 664) using both the global and sex-specific thresholds. The non–sex-specific cutoff achieved an AUC of 0.685 (95% CI 0.644–0.726), accuracy 0.654, and sensitivity 0.606. Applying sex-specific thresholds increased specificity to 0.879 but reduced sensitivity to 0.412, indicating that the anticipated benefit of separate cutoffs did not translate into improved diagnostic balance. We compared these findings with our recently published Marcos VCG-ECG model, a clinically interpretable, signal-only algorithm that integrates electrocardiographic and vectorcardiographic features through a rule-based C5.0 classifier (De la Garza Salazar and Egenriether 2025). Developed and validated against echocardiographic LVH, this model achieved an AUC of 0.779 (95% CI 0.715–0.844), accuracy 0.755, sensitivity 0.731, and specificity 0.775—outperforming SV3 + SV4 while maintaining full interpretability through explicit rule sets grounded in physiologic P-QRS-T wave and vector-loop relationships. Consistent with Yoosefi et al., adding demographic variables did not improve discrimination. The authors also present an SVM analysis that achieved an F1-score of 0.714 in men (Yoosefi et al. 2025). While encouraging, this estimate derives from a small dataset (n = 159; LVH 14.5%), which increases the risk of optimistic performance. Some feature combinations showed near-zero sensitivity but perfect specificity—formally “high-performing,” yet clinically uninformative. These results highlight the importance of class-balance awareness and transparent reporting of class-wise metrics in small ML datasets. Taken together, three practical messages arise. (1) External validation remains essential, as SV3 + SV4 performance decreased outside the derivation cohort. (2) Voltage-only summation may have reached its diagnostic ceiling, given persistent sensitivity–specificity trade-offs (Faggiano et al. 2024). (3) Interpretable, multiparametric ECG/VCG models may offer a better diagnostic equilibrium while preserving transparency and physiologic traceability (De la Garza Salazar and Egenriether 2025; Huang et al. 2025). Yoosefi et al. have provided a valuable stimulus to revisit ECG-based LVH detection (Yoosefi et al. 2025). Our external validation indicates that SV3 + SV4 may not generalize with the same performance profile, particularly under sex-specific cutoffs. Interpretable, signal-only models such as Marcos VCG-ECG can deliver higher and more balanced accuracy in practice. We support further multicenter validation of SV3 + SV4 and encourage that future criteria be benchmarked against interpretable, morphology-rich ECG/VCG approaches, with rigorous attention to class balance and subgroup performance (De la Garza Salazar and Egenriether 2025; Faggiano et al. 2024; Huang et al. 2025). Fernando De la Garza Salazar: conceptualization, methodology, formal analysis, software, data curation, visualization, investigation, resources, supervision, validation, writing original draft, reviewing and editing, project administration. During the preparation of this work, the author used ChatGPT-5 (OpenAI) in order to assist with editing, style refinement, grammar verification, and English–Spanish translation. After using this tool, the author carefully reviewed and edited the content as needed and takes full responsibility for the integrity and accuracy of the final manuscript. The author has nothing to report. The author declares no conflicts of interest. The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.