Abstract Background Screening for atrial fibrillation (AF) has been proposed as a method to prevent strokes in high-risk individuals. Recent trials were unable to prove efficacy for stroke prevention but suggested that a more refined approach may be beneficial. Low-normal thyroid-stimulating hormone (TSH) is a marker for AF and stroke risk and may be used to select individuals for AF screening. Purpose To explore TSH as a tool to select individuals for AF screening to prevent stroke. Methods Post hoc analysis of a clinical trial enrolling 6004 participants with stroke risk factors but no known AF randomized 1:3 to implantable loop recorder (ILR) screening for AF and initiation of oral anticoagulation (OAC) upon AF detection versus usual care (control group). The primary outcome was stroke or systemic embolism (SE). Secondary outcomes included the combination of stroke, SE, or cardiovascular (CV) death, all-cause death, AF diagnosis, and OAC initiation. All outcomes were adjudicated. Results Baseline TSH was measured in 6003 participants (99.98%); 1500 randomized to ILR vs 4503 to control. The mean age was 75±4 years, 2836 (47%) were women. The tertiles of TSH were at ≤1.2, 1.2-1.9, and >1.9 mIU/L. Participants in the first tertile had slightly higher levels of the thyroid hormones (T3, T4, and free T4), were slightly younger, had more smoking pack-years, and higher pulse rate compared to other tertiles. The median follow-up duration was 5.4 (4.9-5.8) years, and no participants were lost to follow-up. AF was diagnosed in 476 (31.7%) participants in the ILR group vs 550 (12.2%) in the control group and this difference was similar across tertiles (p-interaction=0.44). Similar patters were observed for OAC initiation. In the first TSH tertile, participants in the ILR group were significantly less likely to experience the primary outcome (hazard ratio 0.52 [0.30–0.90]; p=0.019) or the combination of stroke, SE, or CV death (hazard ratio 0.54 [0.35–0.84]; p=0.006, respectively) compared to control, while there was no effect in participants with higher TSH (p-interaction 0.029 and 0.014, respectively) and there was no effect on all-cause death in any tertile. When TSH was modelled as a continuous variable, the effect of ILR screening gradually increased with decreasing TSH, with the upper bound of the confidence interval reaching HR<1 at TSH<1.2-1.5 mIU/L. Sensitivity analyses excluding participants with baseline TSH outside reference range (0.3-4.0 mIU/L) and/or treated with thyroid hormone replacement or antithyroid agents yielded similar results. Conclusion In persons with stroke risk factors and TSH <1.2-1.5 mIU/L, ILR screening for AF and anticoagulation upon detection led to significantly decreased stroke risk. There was no effect in persons with higher TSH, though the yield of screening was similar across levels. This suggests that low-normal TSH may indicate benefit from AF screening while higher levels may lead to overdiagnosis.TableFigure