Abstract
BACKGROUND: Type 2 diabetes mellitus (T2DM) is closely associated with chronic inflammation. However, the value of low-cost, readily accessible complete blood count (CBC)-derived inflammatory indicators in predicting all-cause mortality among T2DM patients has not been fully evaluated. This study evaluated their predictive potential. METHODS: This single-center retrospective cohort included 619 T2DM patients. Associations between nine inflammatory indicators and all-cause mortality were assessed using multivariable-adjusted Cox regression models. Time-dependent ROC curves, Kaplan-Meier (K-M) analysis, restricted cubic spline (RCS) models, and subgroup analyses evaluated predictive performance and stability. RESULTS: Over a median follow-up of 1058 days, 137 deaths (22.1%) occurred. Patients in the highest quartile (Q4) of the systemic inflammation response index (SIRI, HR=2.486, 95% CI:1.291,4.788), neutrophil-to-lymphocyte ratio (NLR, HR=2.275, 95% CI:1.217,4.252), and neutrophil-monocyte to lymphocyte ratio (nMLR, HR=2.212, 95% CI:1.200,4.077) had significantly elevated mortality risk versus the lowest quartile (Q1) after adjustment (P for trend<0.01 for all). RCS confirmed significant positive linear associations (P for nonlinearity>0.05). Predictive ability attenuated over time per time-dependent ROC. Subgroup analyses demonstrated consistent associations across clinical subgroups (P for interaction>0.05). CONCLUSION: SIRI, NLR, and nMLR can independently predict all-cause mortality in T2DM patients and demonstrate stable performance across clinical subgroups. These low-cost, routinely tested biomarkers possess the potential to optimize risk stratification in resource-limited settings.