Arrhythmias & electrophysiology
QTNet: Predicting Drug-Induced QT Prolongation With Artificial Intelligence-Enabled Electrocardiograms
Authors' conclusion
An ECG-based CNN can accurately predict diLQTS in the outpatient setting while maintaining its predictive performance over time.
Study context
- Population
- Adult outpatients newly prescribed a QT-prolonging medication with a 12-lead sinus ECG in the preceding 6 months, 2003–2022held-out test dataset
- Participants
- 44,386 outpatients
- Comparison
- QTNet vs models using risk factors, ECG signal or baseline QTc alone
- Outcome
- Prediction of drug-induced long QT syndrome (diLQTS)
Methodological detailsheld-out test dataset, external validation cohort
- Held-out test dataset
- External validation cohort
Performance
As reported in the abstract
diLQTS prediction
| Test or model | Mean AUC |
|---|---|
| QTNet | 0.802 (95% CI: 0.786-0.818) |
QTNet
Mean AUC0.802 (95% CI: 0.786-0.818)
- QTNet achieved the best performance vs 3 other models
- P < 0.001
Further findings
| Outcome | QTNet | Risk factors | ECG signal | Baseline QTc | |
|---|---|---|---|---|---|
| IPCW AUC (survival analysis) · day 2 | |||||
| IPCW AUC | 0.875; 95% CI: 0.848-0.904 | Not reported in the abstract | Not reported in the abstract | Not reported in the abstract | Risk factors, ECG signal, Baseline QTc: not reported in the abstract |
IPCW AUC (survival analysis) · day 2
- QTNet had the highest IPCW AUC at day 2 and up to 6 months
Subgroups
- QTNet performed best among males and patients ≤50 years or with baseline QTc <450 ms
External validation (suburban outpatient practices)
- QTNet maintained the highest predictive performance
Original publication
QTNet: Predicting Drug-Induced QT Prolongation With Artificial Intelligence-Enabled Electrocardiograms. JACC Clin Electrophysiol. Published 1 May 2024. doi:10.
Summarised from the published abstract; every number is checked against it