CASTLE-AF: Complete Statistical Analysis of Catheter Ablation in Atrial Fibrillation and Heart Failure

1. Trial at a Glance

CASTLE-AF was a randomized, parallel-design phase 4 treatment study evaluating radiofrequency ablation compared with conventional treatment in patients with left ventricular dysfunction and atrial fibrillation.

2. Clinical Question

3. Trial Design

4. Endpoints

5. Planned Analysis

The registry identifies the primary endpoint as a time-to-event clinical outcome combining mortality and worsening heart failure requiring unplanned hospitalization. Time-to-event endpoints of this type are commonly evaluated using methods such as Kaplan-Meier estimation, log-rank comparisons, and Cox proportional-hazards models when prespecified by the statistical analysis plan.

The ClinicalTrials.gov record does not report posted statistical analyses or numerical results for the primary endpoint. No formal statistical analyses were posted to ClinicalTrials.gov.

6. Statistical Methodology

Composite time-to-event endpoints

A composite endpoint combines multiple clinically meaningful events into one outcome. For CASTLE-AF, the registry-defined primary endpoint combines death from any cause with worsening heart failure requiring unplanned hospitalization.

Kaplan-Meier estimation

Kaplan-Meier methods estimate the probability of remaining free from the endpoint over time while accounting for patients whose follow-up ends before an event occurs.

Cox proportional-hazards model

A Cox model can estimate a hazard ratio comparing event rates between randomized groups. The hazard ratio is a relative measure and does not directly represent the percentage of patients experiencing an event.

7. Statistical Methods Explained

Why use a composite endpoint?

A composite endpoint can capture more total clinical events and may improve statistical efficiency when all components are clinically relevant.

Why is randomization important?

Random allocation helps balance known and unknown factors between treatment groups, supporting a fair comparison.

What does a hazard ratio describe?

A hazard ratio compares estimated event rates over time between groups. It does not describe individual patient outcomes.

Why are confidence intervals important?

Confidence intervals describe uncertainty around an estimated effect and provide information about precision.

Why does censoring matter?

Time-to-event analyses must account for patients who have incomplete follow-up before experiencing the endpoint.

8. Limitations

• No posted statistical analyses: The ClinicalTrials.gov record does not report numerical endpoint analyses.
• Composite interpretation: A combined endpoint may reflect contributions from different clinical components.
• Time-to-event assumptions: Model-based approaches such as Cox regression depend on assumptions that should be evaluated.
• Generalizability: Applicability depends on how closely patients in practice resemble the enrolled population.

9. Why This Trial Matters Statistically

10. Sources

• ClinicalTrials.gov record: NCT00643188
• PubMed reference
• PubMed reference