Predicting recurrent atrial fibrillation after catheter ablation: a systematic review of prognostic models

Research output: Contribution to journalArticlepeer-review

11 Citations (Scopus)
173 Downloads (Pure)


We assessed the performance of modelsf (risk scores) for predicting recurrence of atrial fibrillation (AF) in patients who have undergone catheter ablation.

Methods and results
Systematic searches of bibliographic databases were conducted (November 2018). Studies were eligible for inclusion if they reported the development, validation, or impact assessment of a model for predicting AF recurrence after ablation. Model performance (discrimination and calibration) measures were extracted. The Prediction Study Risk of Bias Assessment Tool (PROBAST) was used to assess risk of bias. Meta-analysis was not feasible due to clinical and methodological differences between studies, but c-statistics were presented in forest plots. Thirty-three studies developing or validating 13 models were included; eight studies compared two or more models. Common model variables were left atrial parameters, type of AF, and age. Model discriminatory ability was highly variable and no model had consistently poor or good performance. Most studies did not assess model calibration. The main risk of bias concern was the lack of internal validation which may have resulted in overly optimistic and/or biased model performance estimates. No model impact studies were identified.

Our systematic review suggests that clinical risk prediction of AF after ablation has potential, but there remains a need for robust evaluation of risk factors and development of risk scores.
Original languageEnglish
Pages (from-to)748–760
Issue number5
Early online date30 Mar 2020
Publication statusPublished - 1 May 2020


  • Atrial fibrillation
  • catheter ablation
  • model performance
  • prognostic model
  • recurrence
  • systematic review


Dive into the research topics of 'Predicting recurrent atrial fibrillation after catheter ablation: a systematic review of prognostic models'. Together they form a unique fingerprint.

Cite this