Salivary microRNA Expression and Co-Expression Patterns in Periodontal Disease: A Cross-Sectional Analysis of...

MicroRNAs (miRNAs) have been proposed as promising biomarkers for periodontal disease because of their role in regulating inflammation and tissue remodeling. This study evaluated the salivary expressions of miR-34a, miR-155, and miR-146a and their relationships with periodontal disease and its clinical severity. Saliva samples were collected from 35 patients with periodontal disease and 35 healthy controls, and miRNA expression was quantified by quantitative real-time PCR (qRT-PCR). Fold-change values were log2-transformed, and analyses included differential expression (Mann-Whitney U with Benjamini-Hochberg correction), Spearman correlation with clinical attachment loss (CAL), logistic regression/ROC modeling with internal cross-validation, and an exploratory differential co-expression analysis. None of the three miRNAs was significantly differentially expressed between groups (all FDR-adjusted p > 0.40; the result was unchanged after adjustment for age), and none correlated with CAL after accounting for the healthy-versus-diseased separation. miR-34a showed poor stand-alone diagnostic performance (AUC = 0.58; five-fold cross-validation 0.56). Patients were significantly older than controls (41.1 vs. 37.1 years; p < 0.001), and age was the strongest single discriminator (AUC = 0.82), indicating that age acts as an important confounder. In an exploratory analysis, the positive co-expression observed among the three miRNAs in healthy controls was attenuated in disease (e.g., miR-155/miR-146a Spearman rho = 0.68 vs. 0.27; permutation p approximately 0.06). These findings indicate that salivary miR-34a, miR-155 and miR-146a abundance do not discriminate periodontitis in this cohort, but underline age as a critical confounder and suggest that altered miRNA co-expression, rather than differential abundance, may warrant further investigation in larger, age-matched studies. In addition, miR-34a and miR-146a showed reduced expression dispersion in disease; because the reduced dispersion and the reduced co-expression cannot be fully separated in a sample of this size, these network-level observations are regarded as jointly hypothesis-generating.