Comparison of diagnostic accuracy and utility of non-invasive tests for clinically significant liver disease in a general population with metabolic dysfunction

Background: Screening for liver disease in the general population requires accurate non-invasive tests (NITs). A head-to-head comparison of NITs for early detection of clinically relevant liver disease among the target population for screening is lacking. Methods: Among meta-cohort (Rotterdam Study and NHANES) with metabolic dysfunction aged 18-80 years, 10 NITs were investigated. The diagnostic accuracy for clinically relevant conditions (increased liver stiffness measurement [LSM], at-risk MASH, advanced fibrosis or cirrhosis) was assessed. Subgroup analysis included stratification by age group and diabetes/obesity status. Results: We analysed 11,404 participants. MAF-5 obtained the highest AUC for increased LSM (≥8kPa:0.80; ≥12kPa:0.87) and advanced fibrosis (AUC:0.90). FNI and MAF-5 performed best for detecting MASH (AUC:0.93 and AUC:0.92, P =ns) and SAFE for cirrhosis (AUC:0.92). To obtain 80% sensitivity for LSM ≥8kPa, the corresponding MAF-5 cut-off resulted in fewer referrals (42%) compared to FIB-4 (77%) and higher specificity (62% vs. 24%); MAF-5 was also superior for detection of LSM ≥12kPa and advanced fibrosis. Age-dependent scores yielded lower sensitivity amongst younger individuals e.g., by referring 20% of the population with highest NIT-scores, the FIB-4, SAFE, NFS, FORNS and HFS yielded <10% sensitivity for LSM ≥8kPa amongst individuals aged 18-35y while FNI and MAF-5 obtained 40% and 71%. Conclusions: Of the 10 investigated NITs, MAF-5 discriminated best between all conditions except cirrhosis, for which SAFE yielded the highest accuracy. The performance of FIB-4 was poor, implying that referral pathways for significant liver disease in low-prevalence populations can be improved when more accurate NITs such as MAF-5 are employed.