Development and Clinical Validation of a Cross-Sex Translator of ECG Drug Responses

Sex differences in cardiac electrophysiology critically affect arrhythmia risk and therapeutic responses. Female subjects are at a higher risk of drug-induced torsade de pointes and sudden cardiac death, largely due to longer QTc intervals, compared with male subjects. However, the underrepresentation of female subjects in both basic and clinical research creates biases that limit our understanding of sex-specific arrhythmia mechanisms, risk metrics, and treatment outcomes.

The authors aimed to develop a quantitative tool that predicts ECG features in female subjects based on data from male subjects (and vice versa) by combining biophysical models of human ventricular electrophysiology and statistical regression models.

Male and female ventricular tissue models were constructed incorporating transmural heterogeneity and sex-specific parameterizations, and pseudo-ECGs were derived from these models. Multivariable lasso (least absolute shrinkage and selection operator) regression was used to generate sets of regression coefficients (a cross-sex translator) that map male ECG features to female ECG features.

The translator successfully predicted drug-induced effects on simulated female ECG features using simulated male ECG data as input, with an average discrepancy <5%. In addition, a proof-of-concept clinical application using ECG data from age-matched male and female subjects showed that the translator predicted relative drug-induced changes in female ECG features from corresponding male data under various drug regimens with an average error <6%.

We propose our cross-sex ECG translator as a novel digital health tool that can facilitate sex-specific cardiac safety assessments, ensuring that pharmacotherapy is safe and effective across sexes, which is a major step forward in addressing disparities in cardiac treatment for female subjects.