Objective: To investigate the clinical value of serum glypican-3 (GPC3) detection in predicting recurrence of primary hepatocellular carcinoma (HCC). Methods: Through univariate and multivariate logistic regression analysis, the patients pathologically diagnosed with HCC in our hospital from March 2019 to January 2021 were enrolled as the experimental group (n=113), and patients with follow-up time longer than 6 months were included in the prognosis group(n=64). At the same time,20 healthy individuals and 20 individuals with benign liver disease from the physical examination center were enrolled by simple random sampling as control group (n=40). The serum GPC3 and alpha-fetoprotein (AFP) levels were respectively detected by ELISA and chemiluminescence. Then, the study explored the influential factors of the recurrence in HCC patients and constructed the HCC-GPC3 recurrence predicting model by logistic regression. Results: In the research, the sensitivity of GPC3 for the diagnosis of HCC was 61.95% (70/113) and AFP was 52.21% (59/113), meanwhile, the specificity of GPC3 could reach 87.50% (35/40) and AFP was 90.00% (36/40),respectively; The serum GPC3 levels of HCC patients with progressive stage, tumor size≥3 cm, vascular cancer thrombosis and portal venous thromboembolism were significantly higher than that of HCC patients with early stage, tumor size<3 cm, vascular cancer thrombosis and portal venous thromboembolism (Z=2.677, 2.848, 2.995, 2.252, P<0.05), independent of different ages, presence or absence of ascites, peritoneal metastasis, cirrhosis, intrahepatic metastasis (Z=-1.535, 1.011, 0.963, 0.394, 1.510, P>0.05), respectively. Univariate analysis showed that there were no statistically significant differences between the recurrence group and the non-recurrence group in terms of different age, tumor size, presence or absence of vascular cancer thrombosis, ascites, peritoneal metastasis, cirrhosis and AFP levels (χ2=2.012, 0.119, 2.363, 1.041, 0.318, 0.360, Z=0.748, P>0.05); The ratio of those with the progressive stage, portal venous thromboembolism and intrahepatic metastasis and GPC3 levels were all higher in the recurrence group than in the non-recurrence group (χ2=4.338, 11.90, 4.338, Z=2.805, P<0.05).Including the above risk factors in the logistic regression model, the logistic regression analysis showed that the stage, the presence of portal venous thromboembolism,intrahepatic metastasis and GPC3 levels were correlated with the prognosis recurrence of HCC patients (Wald χ2=4.421, 5.681, 4.995, 4.319, P<0.05), and the HCC-GPC3 recurrence model was obtained as: OcScore=-2.858+1.563×[stage]+1.664×[intrahepatic metastasis]+2.942×[ portal venous thromboembolism]+0.776×[GPC3]. According to the receiver operating characteristic curve(ROC), the area under the curve(AUC)of the HCC-GPC3 prognostic model was 0.862, which was better than that of GPC3 alone (AUC=0.704). The cut-off value of model SCORE was 0.699 (the cut-off value of GPC3 was 0.257 mg/L), furthermore, the total sensitivity and specificity of model were 83.3% and 82.4%, which were better than those of GPC3(60.0% and 79.4%).Kaplan-Meier showed that the median PFS was significantly shorter in HCC patients with high GPC3 levels (≥0.257 mg/L) and high values of the model SCORE (≥0.700) (χ2=12.73, 28.16, P<0.05). Conclusion: Besides diagnosing of HCC, GPC3 can may be an independent risk indicator for the recurrence of HCC and can more efficiently predicting the recurrence of HCC patients when combined with the stage, the presence or absence of intrahepatic metastasis and portal venous thromboembolism.