Salivary amylase: a potential non-invasive diagnostic indicator for identifying diabetes in men with overweight/obesity

Saliva has emerged as a promising source of non-invasive biomarkers for systemic diseases. However, reliable salivary indicators for diabetes screening in individuals with overweight/obesity remain limited. This study aimed to evaluate the diagnostic value of salivary α-amylase (sAA), fructosamine, and cortisol, individually and in combination with body mass index (BMI), for identifying diabetes in men with overweight/obesity. Ninety-three men with overweight/obesity were recruited and divided into a discovery cohort ( n = 43) and an independent validation cohort ( n = 50). sAA activity, fructosamine, and cortisol were measured before and after citric acid stimulation. Their diagnostic performance, individually and in combination with BMI, was evaluated using receiver operating characteristic (ROC) analysis. The optimal diagnostic model was further assessed using random forest analysis, externally validated in the validation cohort, and re-evaluated following age adjustment. Among the salivary biomarkers evaluated, the combination of sAA before acid stimulation (B-sAA) and BMI achieved the highest diagnostic accuracy in the discovery cohort (AUC = 0.8152), approaching the performance of fasting blood glucose (AUC = 0.8484). Random forest analysis showed an overall classification accuracy of 79.1%, with accuracies of 85.0% for the group of people without diabetes and 73.9% for the group of people with diabetes. External validation confirmed the diagnostic performance of the BMI plus B-sAA model (AUC = 0.8736). Following age adjustment, the AUC increased to 0.8207 in the discovery cohort and 0.9100 in the validation cohort, indicating that the model retained robust predictive performance. Basal sAA activity combined with BMI represents a promising non-invasive approach for identifying diabetes in men with overweight/obesity in this research. These findings support the potential application of saliva-based biomarkers as convenient tools for diabetes screening and monitoring, although further validation in larger and more diverse populations is warranted.