Patients with atrial fibrillation (AF) are associated with increased thrombotic events. Our previous case-control study showed low-density lipoprotein cholesterol (LDL-C) was an independent predictor of ischemic stroke in AF patients. To investigate the risks of thrombosis in relation to LDL-C among AF patients at different stroke risks by long-time follow-up. Atrial fibrillation patients without history of thrombosis enrolled from five hospitals were classified into low-risk (LR) and high-risk (HR) group according to CHA(2)DS(2)VASc score and followed up with a median period of 26 months. Univariate and multivariate logistic regression analysis were performed in each group. The best cut-off value calculated by receiver operating characteristic (ROC) analysis was used to divide patients into low LDL-C (L-LDL) and high LDL-C (H-LDL) subgroups. Propensity score matching (PSM) and inverse probability of treatment weighted (IPTW) were utilized in both subgroups, after which Kaplan-Meier curves for thrombosis were performed. Univariate and multivariate analysis showed LDL-C was significantly related to thrombosis in LR, but less significantly in HR group. The best cut-off value was 2.155 mmol/L in LR and 2.795 mmol/L in HR group. Lower LDL-C was associated with decreased thrombosis in both groups by PSM and IPTW. Kaplan-Meier curves displayed that H-LDL subgroup was at higher thrombosis risk with significant difference at 24th month in LR patients. LDL-C independently predicts thrombosis with different cut-off values in AF patients at different risks. A stricter control of LDL-C level is necessary for thrombosis reduction in patients with lower score.