Objective: To establish a diagnostic model for alpha-fetoprotein-negative hepatocellular carcinoma (AFP-NHCC) by combining multiple laboratory hematological indicators and explore its clinical diagnostic efficiency. Methods: A total of 124 inpatients, including 110 males and 14 females, aged 57 (51, 66) years, who were first diagnosed with AFP-NHCC in the PLA General Hospital were included from December 2011 to June 2017. Meanwhile, 331 cases of non-HCC were enrolled as the control group, including 279 males and 52 females, aged 58 (51, 63) years old, with 47 cases of hepatitis B virus (HBV) infection, 40 cases of liver cirrhosis, 64 cases of hepatic hemangioma or cysts, 7 cases of liver nodules, 8 cases of fatty liver, 146 cases of non-liver disease and 19 health controls. Subjects in the AFP-NHCC group and the control group were divided into a training group and a validation group. A total of 196 subjects were involved in the training group, including 103 AFP-NHCC patients and 93 non-HCC patients (19 healthy controls, 25 patients with HBV infection, 22 patients with liver cirrhosis, 23 patients with hepatic hemangioma or cyst, and 4 patients with liver nodules). The differences in laboratory parameters were analyzed, and a diagnostic model of AFP-NHCC under different AFP levels was established. Likewise, 259 subjects, including 113 patients with liver disease, were involved in the validation group to verify the diagnostic efficiency of the model for AFP-NHCC. The receiver operating characteristic (ROC) curve was used to analyze the sensitivity and specificity of different models, and the area under the curve (AUC) was calculated to evaluate the diagnostic performance of different models. Results: In the training group, the indicators of AFP-NHCC diagnostic model included platelet (PLT), prothrombin activity (PTA), serum albumin (ALB), prothrombin time (PT) and carbohydrate antigen 19-9 (CA19-9), and the AUC of the model was 0.848 (95%CI: 0.786-0.911) when AFP≤5 μg/L. Similarly, the indicators of AFP-NHCC diagnostic model included PLT, PTA, ALB, PT and hematocrit (HCT), and the AUC of the model was 0.839 (95%CI: 0.780-0.897) when AFP≤10 μg/L. When AFP≤20 μg/L, the indicators of AFP-NHCC diagnostic model contained PLT, PTA, ALB, PT, HCT and AFP, and the AUC of the model was 0.866 (95%CI: 0.815-0.917). The AUC values of these three models were higher than those of AFP and CA19-9 alone for the diagnosis of AFP-NHCC [0.634 (95%CI: 0.560-0.709), 0.691 (95%CI:0.620-0.761), all P<0.05]. The indicators screened by these three models were combined to establish the final diagnostic model, and the AUC of the model was 0.873 (95%CI: 0.824-0.923), with the sensitivity of 78.6% (81/103) and the specificity of 81.7% (76/93). In the validation group, the predictive AUC of the final model in liver disease patients was 0.892 (95%CI: 0.832-0.951), with the sensitivity of 100% (21/21) and the specificity of 71.7% (66/92), while in the total validation population, the predictive AUC was 0.931 (95%CI: 0.890-0.972), with the sensitivity of 100.0% (21/21) and the specificity of 75.6% (180/238). Conclusion: The final diagnostic model includes PLT, PTA, ALB, PT, HCT, CA19-9 and AFP, which has higher sensitivity and specificity, and has good diagnostic efficiency for the clinical diagnosis of AFP-NHCC.
Read full abstract