Background In patients with end stage renal disease and atrial fibrillation (AF), undergoing chronic dialysis, direct oral agents are contraindicated and warfarin does not fully prevent embolic events while increasing the bleeding risk. The high hemorrhagic risk represents the main problem in this population. Aim of the study was to estimate the safety and efficacy for thromboembolic prevention of left atrial appendage (LAA) occlusion in a cohort of dialysis patients with AF and high hemorrhagic risk. Methods Ninety-two dialysis patients with AF who underwent LAA occlusion were recruited. For comparative purposes, two cohorts of dialysis patients with AF, one taking warfarin (oral anticoagulant therapy, OAT cohort, n = 114) and the other not taking any OAT (no-therapy cohort, n = 148) were included in the study. Primary endpoints were (1) incidence of peri-procedural complications, (2) incidence of 2-year thromboembolic and hemorrhagic events, (3) mortality at 2 years. In order to evaluate the effect of the LAA occlusion on the endpoints with respect to the OAT and No-therapy cohorts, a multivariable Cox regression model was applied adjusted for possible confounding factors. Results The device was successfully implanted in 100% of cases. Two major peri-procedural complications were reported. No thromboembolic events occurred at 2-year follow-up. The adjusted multivariable Cox regression model showed no difference in bleeding risk in the OAT compared to the LAA occlusion cohort in the first 3 months of follow-up [HR 1.65 (95% CI 0.43-6.33)], when most of patients were taking two antiplatelet drugs. In the following 21 months the bleeding incidence became higher in OAT patients [HR 6.48 (95% CI 1.32-31.72)]. Overall mortality was greater in both the OAT [HR 2.76 (95% CI 1.31-5.86)] and No-Therapy [HR 3.09 (95% CI 1.59-5.98)] cohorts compared to LAA occlusion patients. Conclusions The study could open the way to a non-pharmacological option for thromboembolic protection in dialysis patients with AF and high bleeding risk.