Performances Of The GALAD And GAAP Models For The Detection Of Hepatocellular Carcinoma

BackgroundHepatocellular carcinoma(HCC)is a common primary hepatic malignancy and a major global health problem.When the symptoms are obvious,patients are usually at the advanced stage when effective treatment options are limited.Therefore,early detection and diagnosis are of great significance to improve the prognosis of HCC patients.Although CT and MRI scanning have higher sensitivity and specificity,they are not used as routine screening methods due to their high price and complex operation.Ultrasound(US)is currently the most widely used imaging method for HCC surveillance.However,US as a surveillance method for HCC has some major limitations.AFP、AFP-L3%and DCP have been acknowledged to be specific for HCC and are currently used in clinical practice In 2014,Johoson and colleagues constructed GALAD model based on gender,age and three tumor markers(AFP,AFP-L3%and DCP).The performance of GALAD model in detection of HCC was significantly higher than that of AFP,AFP-L3%and DCP alone.Then,the value of GALAD model has been validated in Europe,the United States and Japan.Recently,some domestic scholars constructed GAAP model based on gender,age,AFP and DCP/PIVKA-Ⅱ,showing excellent performance in distinguishing HCC from other diseases in the Chinese population where most HCC are associated with chronic hepatitis B.However,up to now,values of these two models have not been fully validated in the Chinese HCC population.ObjectiveThe study aimed to validate the use of the GALAD and GAAP models in detection of HCC in Chinese population,and compare the ability of the two models to distinguish HCC from other diseases.Research methodsIn this study,113 newly diagnosed HCC patients and 639 non-HCC patients were enrolled from October 2018 to June 2020 at the Second Hospital of Shandong University.AFP,AFP-L3%and DCP were measured in the same serum sample,and the GALAD and GAAP scores were calculated according to equations.We compared the levels of three tumor markers and two model scores in different groups.We analyzed the performances of AFP,AFP-L3%,DCP alone,the three biomarkers in combination and the GALAD and GAAP scores to identify patients with HCC using ROC curves in the entire cohort and different subgroups.We also analyzed the performances of AFP、AFP-L3%、DCP alone,the three biomarkers in various combinations and the GALAD and GAAP models for the detection of HCC at specific cut-off values.Results1.A total of 752 patients were enrolled in this study,113 were HCC patients and 639 were non-HCC patients.The median age in the HCC group was higher than that in the non-HCC group(61 years vs 50 years,P<0.001).The proportion of males in HCC group and non-HCC group was significantly higher than that of females(89/24 vs 449/190),but the difference was not statistically significant.The majority of HCC patients(92.92%)suffered from cirrhosis whereas only 48.83%in thenon-HCC group were cirrhotic(P<0.001).The predominant liver disease in the HCC group and non-HCC group was HBV infection(74.34%vs 73.71%,P=0.889).Median serum marker levels of AFP,AFP-L3%,DCP,the GALAD scores and GAAP scores were significantly higher in HCC patients compared to non-HCC patients(p<0.001).45.13%of HCC patients met the Milan criteria,and 34.5%had tumors<3cm in diameter.2.In the entire cohort,the GAAP model achieved a higher AUC value compared with the GALAD model(0.908 vs 0.900,P=0.0948),but the difference was not statistically significant.The AUC value of the GAAP model was significantly higher than that of the three biomarkers alone or in combination(AUC=0.908,0.784,0.722,0.847,and 0.869,respectively,P<0.05).The AUC value of GALAD model was significantly higher than that of three biomarkers alone(AUC=0.900,0.784,0.722,0.847,and 0.869,respectively,P<0.05),but there was no statistical difference between the GALAD model and the three biomarkers in combination(0.900 vs 0.869,P=0.0665).3.In the Milan criteria subgroup,the GAAP model achieved a higher AUC value compared with the GALAD model(0.852 vs 0.849,P=0.6778),but the difference was not statistically significant.Both of the AUC values of GAAP and GALAD models were higher than that of AFP,AFP-L3%and DCP alone(AUC=0.852,0.849,0.758,0.716,and 0.766,respectively,P<0.05).Both of the AUC values of the GAAP and GALAD models were higher than that of AFP,AFP-L3%and DCP in combination(AUC=0.852,0.849,and 0.819,respectively,P>0.05).In the maximal size<3cm subgroup,the GAAP model achieved a higher AUC value compared with the GALAD model(0.832 vs 0.829,P=0.7536),but the difference is not statistically significant.The AUC value of the GAAP model was higher than those of three biomarkers alone or in combination(AUC=0.832,0.743,0.704,0.735,and 0.580,respectively,P<0.05).The AUC value of the GALAD model was higher than those of the three biomarkers alone(AUC=0.829,0.743,0.704,and 0.580,respectively,P<0.05),while there was no statistical difference between the AUC values of the GALAD model and that of the DCP alone(P=0.0659).4.In the HBV subgroup,the GAAP model achieved a higher AUC value compared with GALAD model(0.927 vs 0.915,P=0.0106),the difference was statistically significant.The AUC values of the two models were significantly higher than those of AFP,AFP-L3%and DCP alone or in combination(AUC=0.804,0.739,0.865 and 0.895,respectively,P<0.05).In the non-HBV subgroup,the GAAP model achieved a lower AUC value compared with GALAD model(0.837 vs 0.842,P=0.7181),but there was no statistical difference.The AUC values of two models were significantly higher than those of AFP and AFP-L3%alone(0.679 and 0.642 respectively,P<0.05),but there was no statistical difference between the two models and DCP alone or the three biomarkers in combination(0.788 and 0.776 respectively,P>0.05).5.No significant statistical difference was shown between the GAAP model and the GALAD model in AFP-negative subgroup and AFP-positive subgroup.The AUC values of the two models were higher than those of AFP,AFP-L3%alone or the three biomarkers in combination,but there was no significant statistical difference between the two models and DCP alone.6.With the specific cut-off values(AFP,20ng/mL;AFP-L3%,10%;DCP,40ng/mL;GALAD model,0.1763;GAAP model,-1.7443),the sensitivity of the GALAD and GAAP models were higher than that of AFP,AFP-L3%and DCP alone or in different combinations while maintaining high specificity.Conclusions1.In the entire cohort in which most HCC are associated with chronic hepatitis B,the GAAP model achieved similar performance to the GALAD model in detection of HCC and from other diseases.The performances of these two models were significantly better than that of AFP,AFP-L3%,and DCP alone.2.In the Milan criteria subgroup,the GAAP model achieved a similar performance to the GALAD model in detection of HCC from other diseases.The performances of these two models were significantly better than that of AFP,AFP-L3%,and DCP alone.3..In HBV subgroup,the GAAP model achieved significantly better performance than the GALAD model.These two models showed significantly better performances than AFP,AFP-L3%,and DCP alone,and the three biomarkers in combination.The excellent performance of the GAAP model for HCC detection in Chinese population needs to be further validated.4.With the specific cut-off values(AFP,20ng/mL;AFP-L3%,10%;DCP,40ng/mL; GALAD model,0.1763;GAAP model,-1.7443),the GALAD and GAAP models achieved higher sensitivities than AFP,AFP-L3%,DCP alone and the three biomarkers in different combinations while maintaining high specificities.