A risk model for CPVT that works better in its own cohort than in yours

Design

Prediction model development with internal and independent external validation, Cox regression

Population

743 patients with RYR2-mediated CPVT on beta-blocker monotherapy for derivation; 129 for external validation

Primary outcome

Arrhythmic events (arrhythmic syncope, appropriate ICD shock, sudden cardiac arrest, sudden cardiac death) and near-fatal or fatal events

Effect

Events in 13.7% (derivation, median 5.1 years) and 18.6% (validation, 2.4 years). C-index for arrhythmic events 0.67 (95% CI 0.62-0.72) falling to 0.59 (0.48-0.71) externally; for near-fatal or fatal events 0.74 (0.68-0.80) falling to 0.60 (0.47-0.72). Calibration slopes 1.00 in derivation

Patients with catecholaminergic polymorphic ventricular tachycardia (CPVT) still suffer arrhythmic events on beta-blockers, and deciding who needs flecainide added, sympathetic denervation or a defibrillator has been a matter of judgement. This study built prediction models from 743 patients with RYR2-mediated CPVT on beta-blocker monotherapy and validated them in an independent cohort of 129.

Arrhythmic events occurred in 13.7% of the derivation cohort over a median 5.1 years and 18.6% of the validation cohort over 2.4 years. Predictors were arrhythmic syncope or sudden cardiac arrest before diagnosis, and age at beta-blocker initiation; for near-fatal or fatal events, the severity of ventricular arrhythmia before beta-blockade entered as a fourth.

The honest reading is in the validation numbers. For near-fatal or fatal events the model achieved a C-index of 0.74 in derivation but 0.60 externally; for all arrhythmic events, 0.67 falling to 0.59. A C-index of 0.60 is weak discrimination — better than chance, not much better. Calibration in the derivation cohort was good. So the predictors are real and worth eliciting carefully at every CPVT consultation, but the model should inform a conversation rather than settle a decision about a defibrillator.

• Document arrhythmic syncope or cardiac arrest before diagnosis explicitly — it is the strongest predictor
• Record age at beta-blocker initiation and arrhythmia severity before treatment
• Do not let a low predicted risk override clinical judgement; external discrimination was weak
• Continue beta-blockade and the usual escalation pathways unchanged
• Re-elicit the pre-diagnosis history; it is often recorded vaguely years later

Why it matters

It names which pieces of history actually carry prognostic weight in CPVT, even if the composite score does not yet earn a decision.

Don't overread it

External discrimination of 0.60 with an interval reaching 0.47 means this model has not been shown to work reliably outside its derivation cohort.

The statistics, in plain English

A C-index falls between 0.5 (chance) and 1.0 (perfect). The drop from 0.74 in derivation to 0.60 in external validation is the expected pattern when a model is partly fitted to its own cohort's noise, and the external figure is the one that describes how it will behave in your clinic. The validation cohort had 24 events in 129 patients, so its confidence intervals are wide (0.47-0.72 for near-fatal events) and include values no better than chance.

Read the rest in the app

You have read your two free briefings this month. The app carries all 27 specialties, every morning, free — and this finding is waiting in it.

Get Daily Doctor — free

Scan to keep reading on your phone. No account needed to start.