CATCH-AF Score
(2026)Objective
To develop and validate a simple, clinically applicable tool for stratifying early and long-term atrial fibrillation risk after embolic stroke of undetermined source
Study Summary
• High-risk patients (≥5 points) had a 19-fold higher hazard of AF detection versus low-risk, with an AUC of 0.85.
• The score remained stable over 4.5 years of follow-up and across 7 centres in internal-external cross-validation.
Intervention
Observational cohort study with systematic ICM monitoring to detect atrial fibrillation after ESUS
Inclusion Criteria
Consecutive ESUS patients who received implantable cardiac monitors at 7 participating centres
Study Design
Arms: Array
Patients per Arm: 543 total patients
Outcome
Bottom Line
The CATCH-AF score (range 0-9) reliably stratified AF risk in ESUS patients monitored with ICMs. High-risk patients (≥5 points) had a 19-fold higher hazard of AF detection versus low-risk, with an AUC of 0.85. The score remained stable over 4.5 years of follow-up and across 7 centres in internal-external cross-validation.
Major Points
- First externally validated prediction score specifically designed for AF detection in ESUS patients with long-term ICM monitoring. Addresses a key clinical gap: which ESUS patients benefit most from prolonged cardiac monitoring.
- 543 consecutive ESUS patients across 7 UK and Italian centres (2021-2024), with 1558.5 patient-years of ICM follow-up (median 1104 days). 118 patients (22%) developed AF.
- CATCH-AF components derived via LASSO-penalized Cox regression: Coronary artery disease (2 points), Age ≥65 (1pt) / ≥70 (2pts) / ≥75 (3pts), previous TIA or Stroke (2 points), CHF (2 points). Total range 0-9.
- Discrimination: AUC 0.85 (95% CI 0.82-0.89) overall, stable at 1-year (0.86), 2-year (0.85), 3-year (0.80), and 4.5-year (0.85) timepoints.
- Risk stratification: Low risk (0-2 pts) — 5% AF rate; Medium risk (3-4 pts) — 23% AF rate, HR 4.7; High risk (≥5 pts) — 65% AF rate, HR 19.2 (95% CI 9.4-39.4, p<0.001) vs low risk.
- High-risk patients experienced 918 fewer AF-free survival days compared to low-risk patients, highlighting a dramatically accelerated time to AF detection.
- Internal-external cross-validation (leave-one-centre-out) demonstrated robust generalizability with pooled AUC 0.84 and well-calibrated predictions across geographically diverse centres.
Study Design
- Study Type
- Retrospective, multicentre, observational cohort
- Randomization
- No
- Blinding
- Not applicable (observational)
- Sample Size
- 543
- Follow-up
- Median 1104 days (1558.5 patient-years)
- Centers
- 7
- Countries
- United Kingdom, Italy
Primary Outcome
Definition: New AF detected during ICM follow-up
| Control | Intervention | HR/OR | P-value |
|---|---|---|---|
| - | - | - (9.4-39.4) | <0.001 |
Limitations & Criticisms
- Retrospective design — inherent risk of selection and information bias
- Moderate sample size (N=543) limits power for rare subgroup analyses
- Geographically limited to UK and Italy — generalizability to non-European populations unknown
- No comparison with existing AF prediction scores (e.g., CHA₂DS₂-VASc, HAVOC) in the same cohort
- ICM monitoring protocols may vary across centres, potentially affecting AF detection rates
- No prospective validation — score needs testing in a real-world implementation study before clinical adoption
Citation
D'Anna L, et al. Int J Stroke. 2026. doi:10.1177/17474930261428118.