Artificial intelligence-augmented risk assessment frameworks using inflammatory biomarkers for detection and progression of type 2 diabetes mellitus: a systematic review and meta-analysis

Background/objectives Chronic low-grade inflammation causes complex metabolic disease type 2 diabetes (T2DM). Early biomarkers of glycaemic worsening are hard to detect since glucose-based measurements miss this inflammatory component. Blood count–derived indices may predict inflammation according to new research. From fundamental statistics to ensemble artificial intelligence (AI) models, modern risk-prediction methods analyse multidimensional data to explain inflammation-T2DM. T2DM pathway progression is predicted and monitored using inflammatory biomarkers and AI-augmented models for risk stratification and linked treatment decision-making. Design and Data sources The study followed Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines to search major scholarly databases, including PubMed, Scopus, Cochrane Library and Google Scholar. AI-based and conventional T2DM risk-prediction frameworks were found during screening. RevMan Web was used to retrieve data, measure prediction accuracy, sensitivity, specificity, area under the curve and ORs. The ROBINS-I technique assessed bias risk within non-randomised studies of interventions. Results Fourteen studies were found to meet the inclusion criteria. Though studies varied, AI models using inflammatory biomarkers improved T2DM pathway detection and progression prediction in the meta-analysis. Some research backed traditional approaches; however, AI models were overwhelmingly supported (Z=4.88, p<0.00001). However, the considerable heterogeneity (χ²=311.86, df=13, p<0.0001; I²=96%) emphasises the need to improve AI frameworks for more consistent predictive outcomes. Conclusions AI frameworks combining inflammatory biomarkers demonstrate a better potential in improving prognosis for T2DM pathway detection by enhancing both prediction accuracy and sensitivity. The outlines of AI model specificity and the extended variability across research present fundamental requirements and linked latent opportunities to enhance AI model standardisation and refinement procedures. PROSPERO registration number CRD420251076209.