Background/Aims: Several noninvasive scoring systems have been developed to determine the risk of advanced fibrosis in nonalcoholic fatty liver disease (NAFLD). We examined the diagnostic accuracy of the fibrosis-4 (FIB-4) score and NAFLD fibrosis score (NFS) in patients with biopsy-proven NAFLD. Methods: For this meta-analysis, various databases including PubMed (MEDLINE), EMBASE, OVID Medline and the Cochrane Library were systematically searched. After the acquired abstracts were reviewed by two investigators, manuscripts were chosen for a full-text examination. Results: Thirty-six studies evaluating biopsy-proven NAFLD were selected for meta-analysis. A total of 14,992 patients were analyzed. The lower cutoff sensitivity of the FIB-4 score predicting histological fibrosis stage 3 or more (≥F3) was 69%, with specificity of 64%, positive likelihood ratio (LR+) of 1.96, and negative likelihood ratio (LR–) of 0.47. The low baseline sensitivity of the NFS score predicting ≥F3 was 70%, with a specificity of 61%, LR+ of 1.83, and LR– of 0.48. The area under the receiver operating characteristic curve (AUC) values of the FIB-4 score predicting ≥F3 and ≥F2 were 76% and 68%, respectively. The AUC values of the NFS score predicting ≥F3 and ≥F2 were 74% and 60%, respectively. Conclusions The FIB-4 or NFS test can be used to predict the degree of liver fibrosis in NAFLD, and the diagnostic accuracy resulted as relatively high in fibrosis stages of F3 or higher.