ABSTRACT Aims Cirrhosis and hepatocellular carcinoma are severe major adverse liver outcomes (MALOs) of metabolic dysfunction‐associated steatotic liver disease (MASLD), yet the relationship between proteomics and MALOs remains unclear. This study aimed to identify plasma proteomic features associated with MALOs and evaluate the potential of proteomics to enhance the prediction of MALOs risks in individuals with MASLD. Materials and Methods This prospective cohort study included 14 166 participants with MASLD from the UK Biobank. Associations between 2920 plasma proteins and incident MALOs were assessed using multivariable Cox proportional hazards regression models. Proteomic predictors were selected using elastic net‐penalised Cox regression with five‐fold cross‐validation and a protein risk score (ProRS) was derived. Predictive performance was evaluated in validation sets using Harrell's C ‐index, category‐free net reclassification improvement (NRI) and integrated discrimination improvement (IDI). Associations between the ProRS and liver disease activity were further examined using multivariable linear regression with MRI‐derived iron‐corrected T1 (cT1) values. Results Over a median follow‐up of 13.4 years, 151 participants (1.1%) developed MALOs. We identified 834 proteins (803 positively and 31 inversely) associated with MALOs in multivariable analysis (false discovery rate‐adjusted p ‐value < 0.01), primarily enriched in pathways related to cell adhesion, extracellular region, cytokine activity and cytokine–cytokine receptor interactions. The top five proteins positively associated with increased risks of MALOs were CDHR2, CEACAM1, FCAMR, LTBP2 and GGT1, while TTR, APOM, KITLG, C1orf56 and PROC showed inverse associations. Incorporation of the ProRS derived from 13 elastic net‐selected proteins significantly improved the prediction of MALOs, achieving a C‐index of 0.880 (95% CI: 0.847, 0.910) with an increment of 0.072 (95% CI: 0.041, 0.102). Furthermore, each standard deviation increase in ProRS was associated with higher cT1 values ( β = 7.50; 95% CI: 1.91, 13.10). Conclusions This study identified plasma proteins associated with MALOs in individuals with MASLD. Incorporation of a proteomics‐derived risk score into the clinical model significantly improved predictive performance and was associated with greater liver disease activity, suggesting the potential to enable more precise risk stratification and personalised management.
Yu et al. (Sun,) studied this question.