Bidirectional associations and shared plasma proteomic signatures of CKD and MASLD: evidence from a population-based prospective cohort and a real-world hospital cohort.

Background

AND

Aim

Chronic kidney disease (CKD) and metabolic dysfunction-associated steatotic liver disease (MASLD) coexist, but their bidirectional associations, clinical burden and shared proteomic features remain unclear.

Methods

AND

Results

UK Biobank cohorts were analysed for CKD-to-MASLD (n = 462,516) and MASLD-to-CKD (n = 438,126) associations using time-dependent Cox models with 1- and 2-year lags. Baseline plasma proteomics from 50,941 participants informed exploratory cross-sectional comparisons and disease-specific LASSO-Cox prediction. A hospital cohort included 32,898 Chinese adults with coding-defined MASLD first admitted during 2018-2023; operational CKD combined diagnostic codes, eGFR and albuminuria. In fully adjusted primary analyses, CKD was associated with incident MASLD (HR = 1.36, 95% CI = 1.26-1.48), and MASLD with incident CKD (HR = 1.64, 95% CI = 1.45-1.85; both P < 0.001). Associations persisted with a 2-year lag. Operational CKD prevalence exceeded discharge-coded CKD (9.23% versus 2.34%), although acute kidney dysfunction could contribute. Each additional cardiovascular-kidney-metabolic comorbidity was associated with higher operational CKD odds (OR = 1.64, 95% CI = 1.56-1.72; P < 0.001). Exploratory profiling identified 41 shared proteins. The 37-protein CKD and 19-protein MASLD scores improved discrimination beyond clinical factors in internal testing and geographical validation within UK Biobank. In internal testing, C-indices rose from 0.780 to 0.844 for CKD and from 0.775 to 0.817 for MASLD. Scores were associated with their corresponding outcomes in internal testing (HR = 2.37 and 1.98) and geographical validation (HR = 2.78 and 2.34; all P < 0.001). FSTL3 and GDF15 were retained in both models, and the two-protein score added predictive information.

Conclusions

Findings support integrated kidney-liver risk assessment, and proteomic features and scores require validation in independent, ethnically diverse non-UK cohorts and across proteomic platforms.