Biological ageing-related plasma proteomics improves chronic liver disease prediction: a prospective UK Biobank study

Biological ageing may predict chronic liver disease better than chronological age, but its value alongside genetic susceptibility and circulating proteomics is unclear. We assessed ageing acceleration and developed an integrated prediction framework. This prospective UK Biobank study included 325,580 participants without baseline liver-related disease. BioAgeAccel and PhenoAgeAccel were calculated, polygenic risk scores (PRSs) were built for metabolic dysfunction-associated steatotic liver disease (MASLD) and cirrhosis, and Olink proteomics derived protein scores. Cox models, spline analyses, PRS-stratified analyses, and time-dependent receiver operating characteristic analyses were performed. During follow-up, 9,876 incident liver disease events occurred, including 4,513 MASLD, 1,521 cirrhosis, 258 liver cancers, and 651 liver-specific deaths. Both ageing metrics were independently associated with all outcomes, with stronger effects for PhenoAgeAccel on severe endpoints. In the highest versus lowest group, hazard ratios for PhenoAgeAccel were 2.01 for overall liver disease, 4.19 for cirrhosis, 4.62 for liver cancer, and 3.98 for liver-specific mortality; corresponding BioAgeAccel estimates were 1.54, 2.34, 1.85, and 1.89. Proteomic signatures for MASLD and cirrhosis showed limited overlap and immune-inflammatory enrichment. Protein scores improved discrimination more than PRS alone; for 3-year MASLD, the area under the curve increased to 0.858 with protein-score addition versus 0.827 with PRS addition, and the full model reached 0.870. Accelerated biological ageing independently predicts chronic liver disease. The main innovation is a multi-omics framework integrating ageing acceleration, PRS, and proteomic scores, demonstrating that protein-level information adds more predictive value than genetic risk alone for early risk stratification, especially for MASLD and cirrhosis. Biological age acceleration independently predicted chronic liver disease. PhenoAgeAccel showed stronger links to cirrhosis, liver cancer, and liver death. MASLD and cirrhosis displayed largely distinct proteomic signatures. Protein scores improved risk prediction more than polygenic risk scores. A multi-omics model achieved the best discrimination for MASLD and cirrhosis.