Abstract: Objective To analyze the independent risk factors of long-term ischemic stroke and establish a nomogram for predicting the long-term risks in elderly patients with obstructive sleep apnea (OSA). Methods This multicenter prospective cohort study was conducted from January, 2015 to October, 2017 among consecutive elderly patients (≥60 years) with newly diagnosed OSA without a history of cardio-cerebrovascular diseases and loss of important clinical indicators. The follow-up outcome was the occurrence of ischemic stroke. The baseline demographic and clinical data, sleep parameters, laboratory and ultrasound results were collected from all the patients, who were randomized into the modeling group (n=856) and validation group (n=258) at a 3∶1 ratio. LASSO regression was used for variable reduction and dimension screening, and the risk score prediction model of ischemic stroke was established based on Cox proportional hazard regression. Results In the total of 1141 patients enrolled in this study, 58 (5.08% ) patients experienced ischemic stroke during the median follow-up of 42 months (range 41-54 months). The cumulative incidence of ischemic stroke was 5.14% in the model group and 4.91% in the verification group (P<0.05). Age (HR=3.44, 95% CI: 2.38- 7.77), fasting blood glucose (FPG) (HR=2.13, 95% CI: 1.22-3.72), internal diameter of the ascending aorta (HR=2.60, 95% CI: 1.0- 4.47), left atrial anteroposterior diameter (HR=1.98, 95% CI: 1.75-2.25) and minimum oxygen saturation (LSpO2) (HR=1.57, 95% CI: 1.20-1.93) were identified as independent risk factors for ischemic stroke (P<0.05 or 0.01). A long-term ischemic stroke risk score model was constructed based the regression coefficient ratios of these 5 risk variables. Before and after the application of the Bootstrap method, the AUC of the cohort risk score model was 0.84 (95% CI: 0.78- 0.90) and 0.85 (95% CI: 0.78- 0.89) in the model group and was 0.83 (95% CI: 0.73-0.93) and 0.82 (95%CI: 0.72-0.90) in the verification group, respectively, suggesting a good prediction efficiency and high robustness of the model. At the best clinical cutoff point, the cumulative incidence of ischemic stroke was significantly higher in the high-risk group than in the low-risk group (P=0.021). Conclusion This model can help to identify high-risk OSA patients for early interventions of the risks of ischemic stroke associated with OSA.

Key words: obstructive sleep apnea; elderly; ischemic stroke; ModeL; risk prediction