Serum MicroRNA-Based Risk Prediction for Stroke.

Abstract

Background and Purpose- Numerous studies have shown that circulating microRNAs (miRNAs) can be used as noninvasive biomarkers of various diseases. This study aimed to identify serum miRNAs that predict the risk of stroke. Methods- The cases were individuals who had been diagnosed with cerebrovascular disorder by brain imaging. The controls were individuals with no history of stroke who had undergone a medical checkup. Serum miRNA profiling was performed for all participants using microarray analysis. Samples were divided into discovery, training, and validation sets. In the discovery set, which consisted of control samples only, serum miRNAs that correlated with the predicted risk of stroke, as calculated using 7 clinical risk factors, were identified by Pearson correlation analysis. In the training set, a discriminant model between cases and controls was constructed using the identified miRNAs, Fisher linear discrimination model with leave-one-out cross-validation and DeLong test. In the validation set, the predictive accuracy of the constructed model was calculated. Results- First, in 1523 control samples (discovery set), we identified 10 miRNAs that correlated with a predicted risk of stroke. Second, in 45 controls and 87 cases (training set), we identified 7 of 10 miRNAs that significantly associated with cerebrovascular disorder (miR-1228-5p, miR-1268a, miR-1268b, miR-4433b-3p, miR-6090, miR-6752-5p, and miR-6803-5p). Third, a 3-miRNA combination model (miR-1268b, miR-4433b-3p, and miR-6803-5p) was constructed in the training set with a sensitivity of 84%, a specificity of 98%, and an area under the receiver operating characteristic curve of 0.95 (95% CI, 0.92-0.98). Finally, in 45 controls and 86 cases (validation set), the 3-miRNA model achieved a sensitivity of 80%, a specificity of 82%, and an area under the receiver operating characteristic of 0.89 (95% CI, 0.83-0.95) for cerebrovascular disorder. Conclusions- We identified 7 serum miRNAs that could predict the risk of cerebrovascular disorder before the onset of stroke.