**Background:** Non-alcoholic fatty liver disease (NAFLD) affects an estimated quarter of the world's adult population, with prevalence between 31.6% and 40.8%. Liver fibrosis is the most relevant prognostic factor for clinical outcomes, yet it often remains undiagnosed until cirrhosis develops. Liver biopsy is the gold standard for fibrosis staging but is invasive, costly, subject to sampling error, and not scalable. Over the past 20 years, noninvasive serum biomarkers have been developed and validated against biopsy to identify ≥F3 fibrosis. This narrative review examines current biomarker utility, potential for identifying F2 fibrosis, prediction of patient outcomes, and use in monitoring disease progression.
**Methods:** This is a narrative review summarizing evidence on widely available serum biomarkers including the enhanced liver fibrosis (ELF) test, fibrosis-4 (FIB-4) index, NAFLD fibrosis score (NFS), aspartate aminotransferase to platelet ratio index (APRI), and FibroTest (FibroSURE in the USA). The authors also discuss PRO-C3, a less widely available biomarker, and machine learning algorithms (LiverAID models). They reviewed 21 NASH drug trials from a recent systematic review and meta-analysis by Ampuero et al., tabulating 13 studies with available biomarker data to assess whether biomarkers can track biopsy-observed changes in fibrosis.
**Key Results:** For identifying ≥F2 fibrosis, systematic reviews show APRI, FIB-4, FibroTest, and NFS demonstrate 'fair' performance, while ELF shows 'good' performance (though with specificity of only 12%, indicating high false-positive rates). PRO-C3 has sensitivity 68% (95% CI 0.50–0.82), specificity 79% (95% CI 0.71–0.86), and AUC 0.81 (95% CI 0.77–0.84), matching ELF and outperforming other biomarkers, but is not clinically available. The SAFE score (sex, BMI, diabetes, AST, ALT, platelet, globulin) outperformed NFS for ≥F2 detection. Machine learning models (LiverAID) achieved AUCs from 0.86 to 0.94 for ≥F2 detection, significantly outperforming FIB-4 (AUC 0.70) and APRI (AUC 0.74) (P≤0.001). For predicting outcomes, a large US retrospective study (n=5,123) found FIB-4 >2.67 associated with markedly increased risk of progression to cirrhosis (HR 56.26; 95% CI 25.77–122.83; P<0.001) and death (HR 3.26; P<0.001). A one-point increase in ELF score was associated with a twofold increase in risk of liver-related clinical outcomes. In NASH drug trials, only 1 of 13 studies provided sufficient data to compare biomarker changes against histological changes, and most trials included patients with F1–F2 fibrosis using biomarkers validated only for ≥F3.
**Clinical Implications:** Current biomarkers are useful for excluding advanced fibrosis (≥F3) in high-prevalence settings but require confirmation with vibration-controlled transient elastography. F2 fibrosis, which affects approximately 20% of patients who progress to F3/F4 within 5 years, cannot be reliably identified with existing biomarkers. NFS and FIB-4 are reliable prognostic markers for all-cause mortality, and NFS may predict cardiovascular death. ELF has FDA marketing authorization as a prognostic tool for fibrosis progression risk. Serial biomarker measurement for monitoring disease progression lacks sufficient validation, though repeating FIB-4 within 5 years may help identify low-risk individuals who develop severe liver disease. The authors conclude that developing biomarkers for accurate F2 detection and validating biomarkers for monitoring therapeutic response are crucial priorities.