A liver-aware cardio-kidney-liver-metabolic risk architecture in adults with diabetes: a population-based and hospital-based cohort study

Cardiovascular-kidney-metabolic (CKM) health frames diabetes as a multiorgan condition, yet liver involvement is not explicitly represented. We examined whether liver involvement contributes distinct risk information within a cardio-kidney-liver-metabolic (CKLM) architecture. Adults with diabetes in NHANES 1999–2018 were linked to mortality records; patients with type 2 diabetes in a hospital database entered a day-180 landmark cohort. CKM comprised cardiovascular, kidney, and metabolic domains. Liver involvement was represented by steatosis and fibrosis-risk axes in NHANES and a pre-landmark clinical phenotype in the hospital. Survey-weighted and hospital Cox models assessed domain overlap, associations, internally validated discrimination, calibration, and net benefit. NHANES included 7648 participants, 2122 all-cause deaths, and 722 cardiovascular deaths. CKM domains explained 2.4-29.8% of liver-measure variance (all variance inflation factors ≤ 1.42). Adding both liver axes changed the cross-validated C-index by 0.0036 (95% CI 0.0019–0.0052) for all-cause mortality and 0.0029 (0.0013–0.0048) for cardiovascular mortality. The hospital cohort included 18,418 patients, with 348 confirmed eGFR declines ≥ 40%, 1,199 cardiorenal events, and 866 subsequent strict ASCVD events in the corresponding risk sets. Clinical liver involvement was associated with subsequent ASCVD in the architecture model (HR 1.54, 95% CI 1.28–1.85) and after broader adjustment (HR 1.55, 1.31–1.83); the association persisted after excluding prevalent ASCVD. Changes in C-index were − 0.0017 (-0.0043 to 0.0008) for confirmed eGFR decline, 0.0048 (0.0013–0.0082) for the cardiorenal composite, and 0.0081 (0.0019–0.0148) for subsequent ASCVD. FLI, HSI, FIB-4, and recorded ultrasound steatosis did not reproduce the hospital ASCVD association. Liver involvement provided information incompletely represented by CKM domains, supporting explicit liver assessment in multiorgan diabetes research. Its contribution varied by measurement and outcome, with the clearest hospital signal for a care-recorded liver phenotype and ASCVD. Prospective standardization and external validation are required before clinical staging or treatment use. Across a population-based NHANES cohort and a hospital-based cohort of adults with diabetes, explicit assessment of liver involvement provided distinct, outcome-dependent risk information beyond the cardiovascular, kidney, and metabolic domains represented by CKM. The clearest association was observed for subsequent strict atherosclerotic cardiovascular disease in the hospital cohort.