Journal of Electrocardiology, cilt.97, 2026 (SCI-Expanded, Scopus)
Anthracycline-induced cardiotoxicity remains a significant clinical challenge. We evaluated longitudinal electrocardiographic (ECG) repolarization changes in 36 lymphoma patients receiving doxorubicin-based chemotherapy and explored their association with subclinical cancer therapy-related cardiac dysfunction (CTRCD) using an exploratory machine learning-assisted approach. Standard 12‑lead ECGs, echocardiography including left ventricular ejection fraction (LVEF) and global longitudinal strain (GLS), and cardiac biomarkers were assessed across sequential chemotherapy cycles. Significant prolongation of corrected QT, corrected JT interval (JTc), and T-peak to T-end interval was detected as early as the second cycle (all p < 0.001), whereas QT dispersion and corrected QT dispersion increased significantly from the fourth cycle onward (p < 0.001). Upon treatment completion, GLS declined significantly (−19.66 ± 2.95% to −18.36 ± 2.51%, p < 0.001), while LVEF remained preserved (p = 0.286). Exploratory Random Forest feature importance analysis was used for feature prioritization rather than predictive modeling, and it prioritized changes in JTc and corrected QT dispersion as the ECG parameters most strongly associated with GLS reduction, with JTc accounting for approximately 28% of the model-derived importance. These findings suggest that doxorubicin-induced ventricular repolarization abnormalities precede overt systolic dysfunction and accompany GLS-defined subclinical CTRCD. Focused ECG monitoring of repolarization parameters may serve as a cost-effective adjunct to echocardiographic surveillance for early cardiotoxicity detection. Structured abstract: Background: Anthracycline-induced cardiotoxicity remains a major clinical challenge in cancer survivors. Although global longitudinal strain (GLS) is a sensitive marker of early myocardial dysfunction, the temporal behavior of electrocardiographic (ECG) repolarization parameters across sequential chemotherapy cycles remains incompletely characterized.Purpose: To investigate longitudinal ECG repolarization changes during doxorubicin therapy and to explore, using an exploratory machine learning-assisted approach, which ECG markers demonstrate the greatest relative importance in association with subclinical cancer therapy-related cardiac dysfunction (CTRCD).Methods: This retrospective study included 36 patients with lymphoma treated with doxorubicin-based chemotherapy. Standard 12‑lead ECGs were analyzed at baseline and after the 2nd, 4th, and 6th chemotherapy cycles. Echocardiographic parameters, including left ventricular ejection fraction (LVEF) and GLS, and cardiac biomarkers were assessed at baseline and after treatment completion. An exploratory Random Forest regressor was used solely for feature prioritization to evaluate the relative importance of delta (Δ) ECG repolarization parameters in modeling absolute GLS change (ΔGLS).Results: QTcB, corrected JT (JTc), and T-peak to T-end (Tp–e) intervals were significantly prolonged as early as the second cycle (p < 0.001), whereas QT dispersion (QTd) and corrected QT dispersion (QTcd) increased significantly from the fourth cycle onward (p < 0.001). At chemotherapy completion, GLS declined significantly (−19.66 ± 2.95% to −18.36 ± 2.51%, p < 0.001), while LVEF remained preserved (p = 0.286). Feature importance analysis prioritized ΔJTc and ΔQTcd as the ECG markers most strongly associated with GLS reduction, with JTc accounting for approximately 28% of the model-derived importance.Conclusions: Doxorubicin-induced ventricular repolarization abnormalities precede overt systolic dysfunction and emerge as early as the second chemotherapy cycle. Exploratory machine learning-assisted analysis highlights JTc and QTcd as prioritized ECG markers associated with subclinical CTRCD. Focused ECG monitoring may serve as a cost-effective and widely accessible adjunct to echocardiographic cardiotoxicity surveillance.