Incremental prognostic value of the T-helper 1/T-helper 2 immune phenotype and 72-h change in sequential organ failure assessment for in-hospital mortality in sepsis with and without type 2 diabetes: a retrospective two-cohort study

BackgroundType 2 diabetes mellitus (T2DM) is accompanied by chronic immunometabolic dysregulation and may alter the host response to sepsis. We compared immunometabolic profiles and early organ dysfunction trajectories in patients with sepsis with and without T2DM and evaluated 72-h mortality risk reassessment models.MethodsOf 1,200 consecutive adult intensive care unit (ICU) patients screened between January 2020 and December 2025, 1,023 met Sepsis-3 criteria and constituted the 72-h landmark cohort, defined by an ICU stay of at least 72 h and paired day-1/day-3 Sequential Organ Failure Assessment (SOFA) measurements; all model comparisons were therefore conditional on survival and evaluability at 72 h. Patients were classified into a T2DM-sepsis cohort (n = 513) or non-T2DM-sepsis cohort (n = 510). Model 1 included day-1 SOFA; Model 2 added the T-helper 1/T-helper 2 (Th1/Th2) ratio and interleukin-6 (IL-6); and Model 3 further added the 72-h change in SOFA (ΔSOFA; day-3 minus day-1 SOFA). Post hoc secondary analyses evaluated age-only and age-adjusted benchmarks and a pooled model containing diabetes status, age, and interaction terms. Internal validation was performed using 1,000 bootstrap resamples.ResultsCompared with patients without T2DM, those with T2DM had higher glycemic, inflammatory, renal, and physiological severity measures. In the T2DM cohort, the apparent area under the receiver operating characteristic curve (AUC) was 0.781 for Model 1, 0.838 for Model 2, and 0.861 for Model 3; the optimism-corrected Model 3 AUC was 0.858. Age alone had an AUC of 0.634, whereas adding age to Model 3 yielded an AUC of 0.870 (DeLong P = 0.099 vs. Model 3). The age-adjusted pooled model had an apparent AUC of 0.862, and the global predictor-by-diabetes interaction test was not statistically significant (P = 0.191).ConclusionA 72-h reassessment model combining baseline organ dysfunction, early SOFA trajectory, the Th1/Th2 ratio, and IL-6 may improve mortality risk stratification beyond SOFA alone. The model is intended to prioritize structured clinical reassessment, not replace SOFA or prescribe treatment. Temporal and multicenter, external validation is required before clinical translation.