Abstract
BACKGROUND: MicroRNAs (miRNAs) are non-coding RNAs consisting of approximately 20 base pairs. Several studies have reported the usefulness of serum miRNAs in the diagnosis of gliomas. However, the results differ among studies, and reproducibility remains an issue. In this study, we investigated the usefulness of serum miRNAs in the diagnosis of gliomas using prospectively collected samples with strictly defined procedures. Furthermore, we integrated the data set with retrospectively collected samples stored at the same temperature and constructed a diagnostic model. MATERIALS AND METHODS: Fifty-eight gliomas and 273 healthy controls from prospective cohort, and 17 gliomas from retrospective cohort were included in the study. First, 58 gliomas from prospective cohort and 273 healthy subjects were divided into two groups: training set and validation set. Serum miRNAs were comprehensively analyzed, and a diagnostic model for gliomas was constructed. Second, 100 miRNAs useful for diagnosis of gliomas were extracted from both cohorts, and the diagnostic model was constructed based on the 45 miRNAs commonly extracted from both cohorts. RESULTS: A diagnostic model for gliomas was generated using 24 miRNAs in the prospective cohort. The AUC was 0.818 and 0.925 for the prospective and retrospective cohorts, respectively. However, the miRNAs used in the model did not include miRNAs that have been reported to be useful in the diagnosis of gliomas. The expression profiles of miRNAs differed between the prospective and retrospective cohorts. The integrated analysis using both prospective and retrospective datasets constructed diagnostic model using seven miRNAs. The AUC was 0.810 for both datasets. CONCLUSION: The serum miRNA profiles are greatly affected by conditions from sample collation to storage. Although interpretation of the results should be done with caution, the diagnostic model using miRNAs that are stably expressed under different conditions is expected to be useful for screening gliomas.