Utility of the CHA2DS2-VASc score for predicting ischaemic stroke in patients with or without atrial fibrillation: a systematic review and meta-analysis
Tariq Jamal Siddiqi, Muhammad Usman, Izza Shahid, Jawad Ahmed, Safi U. Khan, Lina Ya’qoub, Charanjit S. Rihal, Mohamad Alkhouli
Abstract
AIMS: Anticoagulants are the mainstay treatment for stroke prevention in patients with non-valvular atrial fibrillation (NVAF), and the CHA2DS2-VASc score is widely used to guide anticoagulation therapy in this cohort. However, utility of CHA2DS2-VASc in NVAF patients is debated, primarily because it is a vascular scoring system, which does not incorporate atrial fibrillation related parameters. Therefore, we conducted a meta-analysis to estimate the discrimination ability of CHA2DS2-VASc in predicting ischaemic stroke overall, and in subgroups of patients with or without NVAF. METHODS AND RESULTS: PubMed and Embase databases were searched till June 2020 for published articles that assessed the discrimination ability of CHA2DS2-VASc, as measured by C-statistics, during mid-term (2-5 years) and long-term (>5 years) follow-up. Summary estimates were reported as random effects C-statistics with 95% confidence intervals (CIs). Seventeen articles were included in the analysis. Nine studies (n = 453 747 patients) reported the discrimination ability of CHA2DS2-VASc in NVAF patients, and 10 studies (n = 138 262 patients) in patients without NVAF. During mid-term follow-up, CHA2DS2-VASc predicted stroke with modest discrimination in the overall cohort [0.67 (0.65-0.69)], with similar discrimination ability in patients with NVAF [0.65 (0.63-0.68)] and in those without NVAF [0.69 (0.68-0.71)] (P-interaction = 0.08). Similarly, at long-term follow-up, CHA2DS2-VASc had modest discrimination [0.66 (0.63-0.69)], which was consistent among patients with NVAF [0.63 (0.54-0.71)] and those without NVAF [0.67 (0.64-0.70)] (P-interaction = 0.39). CONCLUSION: This meta-analysis suggests that the discrimination power of the CHA2DS2-VASc score in predicting ischaemic stroke is modest, and is similar in the presence or absence of NVAF. More accurate stroke prediction models are thus needed for the NVAF population.