Yield of two week holter monitoring and diagnostic accuracy of intermittent single-lead electrocardiogram recordings in screening for paroxysmal atrial fibrillation in primary care

Abstract

Background Timely detection of atrial fibrillation (AF) is important because of its increased risk of thrombo-embolic events. Single time point screening interventions for AF are hindered because of AF's often paroxysmal nature. Community based prevalence of paroxysmal AF (pAF), however, is scarcely investigated. To detect pAF, often a Holter for prolonged electrocardiographic monitoring is used. Holter monitoring, however, is less appropriate for screening purposes because of its high burden for asymptomatic patients. Intermittent screening using a single-lead electrocardiogram (1L-ECG) device can offer a less burdensome alternative. Purpose To determine A) the diagnostic yield of two week Holter monitoring in screening for pAF after a negative 12-lead (12L-)ECG, B) the diagnostic accuracy of an intermittently used 1L-ECG device, and C) how often an expert panel is able to accurately diagnose AF by visually assessing the 1L-ECG recordings of patients with ≥1 algorithm-positive 1L-ECG recording(s). We used the Holter monitor as reference standard. Methods Patients of 65 years and older participated in an opportunistic screening study for AF. They received three screening tests (pulse palpation, electronic sphygmomanometer and a 1L-ECG device both with built-in AF detection algorithm). A 12L-ECG was recorded whenever at least one screening test was positive, supplemented with 10% of patients with three negative tests. We invited patients with a negative 12L-ECG to use a 1L-ECG device thrice daily and perform Holter monitoring for two weeks. We report the prevalence of pAF found by Holter monitoring and calculate the diagnostic accuracy of the 1L-ECG device's built-in AF detection algorithm. All 1L-ECG recordings of patients with ≥1 positive algorithm result were subsequently visually assessed by an expert panel of three cardiologists. Results We included 270 patients, of whom four had pAF on a median of 8.5 days of Holter monitoring: prevalence 1.5% (95%-CI: 0.4–3.8%). 205 patients were simultaneously screened with 1L-ECG recordings and Holter monitoring. Sensitivity was 67% (95%-CI: 9–99%), specificity 69% (95%-CI: 62–75%), positive predictive value 3% (95%-CI: 1–7%) and negative predictive value 99% (95%-CI: 97–100%) for diagnosing pAF based on the 1L-ECG device's built-in AF detection algorithm. Out of 65 patients with ≥1 algorithm-positive 1L-ECG recording(s), seven were classified as pAF after visual assessment by the expert panel. Six of these were false-positive. In 24 cases the expert panel could not interpret the 1L-ECGs. Conclusion We found a low prevalence of pAF using Holter monitoring in elderly primary care patients, with a negative 12L-ECG. The diagnostic accuracy of the built-in AF detection algorithm of an intermittently used 1L-ECG device is limited with a high false-positive rate. Cardiologists' assessment of algorithm-positive intermittent 1L-ECGs did not sufficiently improve accuracy to make a reliable diagnosis. FUNDunding Acknowledgement Type of funding sources: Public grant(s) – National budget only. Main funding source(s): This work was supported by ZonMw (The Netherlands Organisation for Health Research and Development)