Hepatic biomarker trajectories and mortality in septic adults with type 2 diabetes and recorded dipeptidyl peptidase-4 inhibitor exposure versus metformin-based therapy: A propensity-score-matched cohort study.

Aims: To determine whether recorded dipeptidyl peptidase-4 (DPP-4) inhibitor exposure, compared with metformin, is associated with worse hepatic biomarker trajectories or mortality in adults with type 2 diabetes and sepsis. Methods: In three critical care databases (MIMIC-IV, eICU-CRD, NWICU), adults with type 2 diabetes and sepsis with a DPP-4 inhibitor record were matched 1:3 within each database to metformin comparators on age, sex, body mass index, 24-h SOFA and day-1 estimated glomerular filtration rate. The day 1-7 total bilirubin trajectory was analysed by linear mixed-effects regression under nine specifications including log transformation, flexible day terms, doubly robust adjustment and inverse-probability-of-observation weighting. Results: Matching yielded 105 DPP-4 inhibitor and 297 metformin patients; a pre-admission exposure record was present in 60.0%. The between-group difference in bilirubin slope was +0.048 mg/dL/day (95% CI -0.029 to 0.126; p = 0.221), with p between 0.109 and 0.436 across specifications. No other hepatic outcome survived multiplicity correction. Alkaline phosphatase rose more steeply on the primary specification (+6.97 U/L/day; q = 0.005) but was fragile: 12 of 28 sensitivity analyses retained p < 0.05. In-hospital mortality was 21.0% versus 13.5% (risk ratio 1.56, 95% CI 0.97-2.49), attenuating to 1.23 (0.67-2.29) with documented pre-admission exposure. Conclusions: There was no statistically detectable difference in bilirubin trajectory between recorded DPP-4 inhibitor and metformin exposure, although differences below the cohort's detectable-effect threshold cannot be excluded. An ALP signal was fragile and hypothesis-generating. Mortality findings are exploratory and non-causal. These associations with recorded exposure cannot inform inpatient management.