Introduction. The natural course of atrial fibrillation (AF) is not well defined. We aimed to investigate the change in AF burden over time and its associated risk factors among AF patients. Methods. Fifty-four participants with recently documented paroxysmal or persistent AF were enrolled. Main exclusion criteria were permanent AF or previous catheter ablation for AF. AF burden was calculated as time in AF divided by total recording time using yearly continuous 7-day Holter-ECG recordings. A relative change ≥10% or an absolute change >0.5% in AF burden between two yearly Holter-ECG recordings was considered significant. Results. Mean age was 67 years, 72% were men. The proportion of patients with no recorded AF increased from 53.7% at baseline to 78.6% (p=0.1) after 4 years of follow-up. In 7-day Holter-ECG recordings performed after baseline, 23.7% of participants had a decrease and 23.7% an increase in AF burden. In separate mixed effect models, AF burden over time was associated with prior stroke (β 42.59, 95% CI (23.40; 61.77); p<0.0001), BNP (β 0.05, CI (0.02; 0.09); p=0.005) end-diastolic (β 0.49, CI (0.23; 0.74); p=0.0003) as well as end-systolic (β 0.25, CI (0.05; 0.46); p=0.02) left atrial volume, left atrial ejection fraction (β −0.43, CI (−0.76;−0.10); p=0.01), E-wave (β 36.67, CI (12.96; 60.38); p=0.003), and deceleration time (β −0.1, CI (−0.16; −0.05); p=0.002). In a multivariable model, a history of prior stroke (β 29.87, CI (2.61; 57.13); p=0.03) and BNP levels (β 0.05, CI (0.01; 0.08); p=0.007) remained significantly associated with AF burden. Conclusions. Few patients with paroxysmal or persistent AF have AF episodes on yearly 7-day Holter-ECG recordings, and AF progression is rare. AF burden was independently associated with a history of prior stroke and BNP levels.