Development and internal-external validation of a prediction model for premature ventricular contraction unresponsive to the medical treatment

Abstract Background Ventricular premature contractions (PVCs) are the most common ventricular arrhythmias. However, to date, no risk stratification tool exists to assess response to medical therapy in patients with frequent PVCs (>5% per 24 hours). We aimed to develop and validate a clinical prediction model evaluating response to medical therapy in frequent PVCs. Methods We conducted a retrospective cohort study of patients with frequent PVCs who were considered for medical treatment. The study outcome was unresponsiveness to the medical treatment (24-hour PVC count not diminishing by at least 80%) in patients with frequent PVCs at 3-6 months follow-up. Potential predictors were, age (years), PVC burden (%), LVEF % (<55, ³55), PVC QRS width (msec), mean heart rate (beat/min), sinus beat QTc (msec), PVC coupling interval, gender, presence of multifocal PVC and Non-sustained VT. Binary logistic regression analyses were performed to develop and internally validate the model. Finally, we used internal-external cross validation using leave-one-out center. Results 1644 patients were included in the study (mean age 52.2±13.5 years and 56.3% male). The frequency of unresponsiveness to the medical treatment in patients with frequent PVC was 31.2% (n=513). In the model, PVC burden (%), LVEF% (<55, ³55) and PVC QRS width were found to be the three strongest predictors. The apparent and internal validation discriminations of the model (C-statistics in internal validation 0.910) were quite satisfactory. Model discriminations and calibration metrics with internal-external cross validation were similar to the apparent model and were deemed acceptable (https://demonoreflow.shinyapps.io/dynnomapp/). Conclusions We developed and validated a model that accurately predicts unresponsiveness to medical treatment in patients with frequent PVC. Our model, once externally validated, has the potential to facilitate management decisions by providing individualized risk estimates in patients with frequent PVC for whom ablation may be suggested as a first line treatment modality.Figure-1Figure-2