Misclassification of Atrial Fibrillation and Implications for Clinical Studies and Patient Management: The Elephant in the Room?

Introduction: Clinical atrial fibrillation (AF) classifications are important for patient management and design of clinical trials. However, the ability of the physician to accurately classify the intensity of AF is unknown. The aim of the present study was to investigate how accurately a physician can evaluate a patient's AF classification in patients continuously monitored with implantable devices.

Methods: Cardiac rhythm histories of 1,195 patients (age 73.0 ± 10.1 years, follow-up: 349 ± 40 days) with implantable devices were reconstructed and analyzed. Patients were classified as having paroxysmal or persistent AF by physicians at baseline in accordance with current guidelines. After the end of the study we measured the device registered AF burden as the proportion of time spent in AF. For the purposes of this study, 258 patients with more than 300 days of recorded data and more than 5% AF burden recurrence were included in the present study.

Results: A significant misclassification between clinical and device related data occurred (Cohen's kappa 0.06 95%C.I. [0.02–0.11]). The clinical AF classification agreed with the device AF classification only in 35.6% of patients. 7% of patients who experienced paroxysmal AF (device data AF burden 0.17[0.08–0.23] median[IQR]) were instead clinically classified as having persistent AF. On the contrary 80% of patients who experienced documented persistent AF (device data AF burden 0.49[0.21–0.89]) were instead classified as having clinical paroxysmal AF.

Conclusion: Our data show that physicians misclassify AF in the vast majority of cases. This misclassification occurs in low and high AF burden spectra; however, it seems to be more pronounced in mid to high AF burden spectra. 80% of patients who experienced persistent AF were instead classified by physicians as having paroxysmal AF. This has important implications for patient management and design of clinical studies.

No conflict of interest has been declared by the author(s).