P663Identifying patients at risk for atrial fibrillation in the community: a systematic review and meta-analysis of prediction models

Abstract

Background Atrial fibrillation (AF) presents a considerable burden on our health care systems. Early detection of AF may prevent AF-associated complications, such as stroke and heart failure. Given our aging populations, the number of new AF cases is expected to double over the next decades. As such, there is renewed interest to screen for AF in the community. To optimise screening efforts, risk prediction models may help us identify at-risk patients. Purpose To identify and evaluate the performance of prediction models for AF that may be applicable for screening in community settings. Methods We searched PubMed, Embase, and CINAHL databases for studies that derived and/or validated AF risk models from population-based cohorts. Three investigators independently assessed risk of bias (CHARMS checklist), and performed data extraction and evidence synthesis. The primary expression of associations in meta-analysis was the C-statistic for discrimination between AF and non-AF cases during follow-up, using a random effects model. We calculated 95% prediction intervals (PI) due to high heterogeneity (I2 >30%) in all analyses. Results We identified 23 studies that presented data on 8 risk models derived from 18 cohorts with a total of 1,4 million participants from across the globe. Average age in these cohorts ranged from 43–76 years and follow-up ranged from 3 to 20 years. Two of the 8 risk models had a sufficient number of validation studies to be included in the meta-analysis. The CHARGE-AF (Cohorts for Heart and Aging Research in Genomic Epidemiology) score had a summary C-statistic of 0.72 (95%-PI: 0.67–0.77; n=7 cohorts, n=53.040 patients). The FHS (Framingham Heart Study) score for AF had a summary C-statistic of 0.71 (95%-PI: 0.59–0.83; n=4 cohorts, n=19.300 patients). Both models include age, height and weight, blood pressure, prevalent heart failure, and antihypertensive medication use as variables. CHARGE-AF additionally includes race, current smoking, and history of diabetes and myocardial infarction. FHS additionally includes sex, PR interval, and significant murmur. Conclusions Currently two risk scores, CHARGE-AF and FHS, have been rigorously tested for predicting atrial fibrillation in general populations. The CHARGE-AF score may present the more promising, user-friendly score for future community screening efforts, as it solely relies on readily available clinical parameters. Acknowledgement/Funding This work was supported by the Netherlands Organisation for Health Research and Development (ZonMw) [80-83910-98-13046]