ABSTRACT: OBJECTIVE:Midkine (MDK) has been proposed as one of the most promising markers for hepatocellular carcinoma (HCC). This meta-analysis was conducted to compare the diagnostic accuracy of MDK and ?-fetoprotein (AFP) for HCC. METHODS:We systematically searched PubMed/MEDLINE, Ovid/EMBASE, and the Cochrane Library for all relevant studies up to 18 May 2019. The Revised Quality Assessment for Studies of Diagnostic Accuracy tool (QUADAS-2) was used to assess the methodological quality of the included studies. The sensitivity, specificity, and the area under the curve (AUC) of MDK and AFP for detecting HCC were pooled using random-effects model. RESULTS:Seventeen studies from five articles with a total of 1122 HCC patients and 2483 controls were included. The summary estimates using MDK and AFP for detecting HCC were as follows: sensitivity, 85 vs 52%, specificity, 82 vs 94%, and AUC, 0.90 vs 0.83. The summary estimates using MDK and AFP for detecting hepatitis virus-related HCC as follows: sensitivity, 93 vs 74%, specificity, 85 vs 97%, and AUC, 0.95 vs 0.97. The summary estimates using MDK and AFP for detecting early-stage HCC were as follows: sensitivity, 83.5 vs 44.4%, specificity, 81.7 vs 84.8%, and AUC, 0.87 vs 0.52. The summary estimates using MDK for detecting AFP-negative HCC as follows: sensitivity, 88.5%, specificity, 83.9%, and AUC, 0.91. CONCLUSION:MDK is more accurate than AFP in diagnosing HCC, especially for early-stage HCC and AFP-negative HCC. Both MDK and AFP had excellent diagnostic performance for hepatitis virus-related HCC.