QTNet: Predicting Drug-Induced QT Prolongation With Artificial Intelligence-Enabled Electrocardiograms

Area of Science:

• Cardiology
• Artificial Intelligence in Medicine
• Pharmacovigilance

Background:

• Drug-induced long QT syndrome (diLQTS) poses a risk for torsades de pointes.
• Current methods for outpatient diLQTS prediction are unreliable.

Purpose of the Study:

• To evaluate a convolutional neural network (CNN) for predicting diLQTS in outpatients using ECG data.
• To assess the performance of the CNN model, named QTNet, in a real-world outpatient setting.

Main Methods:

• Adult outpatients prescribed QT-prolonging medications were identified.
• QTNet, a CNN, was developed using risk factor data and ECG signals.
• The model was trained and validated on a large dataset of 44,386 patients.

Main Results:

• QTNet demonstrated superior predictive performance (AUC=0.802) compared to other models.
• The model showed strong predictive accuracy in survival analysis up to 6 months.
• External validation confirmed QTNet's consistent high predictive performance.

Conclusions:

• An ECG-based CNN (QTNet) can accurately predict diLQTS in the outpatient setting.
• The model maintains predictive performance over time and identifies high-risk patients.
• QTNet facilitates closer monitoring for individuals susceptible to diLQTS.