A model for predicting colorectal cancer (CRC) based on determining the expression level of miRNA-21 in saliva and plasma

Abstract

Introduction. The development of new approaches to screening colorectal cancer (CRC), including those based on such molecular genetic methods as the assessment of microRNA(miRNA) expression, is an urgent clinical task. Purpose of the study. The study was aimed at development of a model for CRC prediction based on the miRNA-21 expression level in saliva and plasma.

Materials and Methods. The miRNA-21 expression levels in plasma (P-miRNA-21) and saliva(S-miRNA-21) samples obtained from 65 patients with diagnosed colon cancer (CRC group) and 66 healthy volunteers were measured by reverse transcription polymerase chain reaction and expressed in conventional expression units (CU). Correlation analysis of the levels of miRNA-21 expression in plasma and saliva with gender, age, localization and stage of the process was performed. Univariate analysis was used to identify CRC predictors. Based on a combination of leading factors, risk classes were formed with the subsequent construction of a decision tree. ROC analysis was used to estimate the quality of the simulated decision tree and the CRR predicting model. The level of statistical significance was fixed at 0.05.

Results. The P-miRNA-21 (3.71±7.38 CU) and S-miRNA-21 (9.67±18.52 CU) in CRC patients was higher than in healthy control (p=0.0089 and p<0.0001, respectively). The sensitivity and specificity of P-miRNA-21>1.6 CU and S-miRNA-21>2 CU as CRC markers were 52% and 89% and 61% and 83%, respectively. No correlation of P-miRNA-21 or S-miRNA-21 with gender, age, process localization and disease stage was found. In univariate analysis the predictors of CRC were: S-miRNA-21>2 CU (Relative risk (CI), RR [CI): 2.51 (1.76; 3.58)], P-miRNA-21>1.6 CU [RR (CI): 2.41 (1.75; 3.31)], age >61 years [RR (CI): 2.65 (1.67; 4.21)], high level of cancer embryonic antigen (CEA) [RR (CI): 2.4 (1.93; 2.99)], high level of СA-19.9 [RR (CI): 2.27 (1.85; 2.78)]. An algorithm for CRC diagnostics was proposed in the form of a decision tree model based on S-miRNA-21, P-miRNA-21 and age, which has high sensitivity (88.7%) and specificity (76.6%) of the model, AuROC 0.86.

Conclusion. Evaluation of the expression miRNA-21 both in saliva and in blood plasma can be suitable non-invasive biomarkers for screening colon cancer, especially as part of the proposed diagnostic model.