Abstract 3336: A novel clinical and biomarker model accurately predicts hepatocellular carcinoma (HCC) across racial groups

Abstract

Abstract Purpose: HCC incidence, which is highest in Asians and Pacific Islanders (API), followed by Black/African Americans (AA), Hispanics (H), and non-Hispanic Whites (W), is rising at an alarming rate in the United States. Most HCC is diagnosed at an advanced incurable stage, emphasizing the need for accurate biomarkers for early diagnosis. The predictive accuracy in available biomarkers such as alpha-fetoprotein (AFP) is sub-optimal, and these have not been systematically assessed across ethnic and racial subgroups. Here, we report on the performance characteristics of a model derived from an ethnically diverse population including API, distinct from the mainland USA. This model includes biologically relevant proteins, as well as clinical characteristics and seeks to capture the molecular changes underlying hepatocarcinogenesis. Experimental Procedures: We combined human genomic studies using The Cancer Genome Atlas (TCGA) and mechanistic insight from animal models to identify biologically relevant proteins. A proteomic analysis of 7,000 proteins (SomaScan) utilizing serum samples (55 uL/each) from 520 individuals with cirrhosis, 158 of whom had HCC by imaging criteria. Patients were recruited from three participating centers with significant cohorts of ethnic minorities (27% AA, 3% H, 2.3% API, 63%W patients, 3.8% other). We combined serum biomarkers with clinical and demographic risk factors to develop a risk prediction model. Results: Using multivariate logistic regression analysis, we developed a model that incorporated race and ethnicity in predicting the likelihood of HCC in cirrhotic patients. This functional model included 4 clinical variables (age, serum AFP and Bilirubin levels, HBV infection or not) and 3 serum protein biomarkers (AKR1B10, COL15A1, and INHBB), resulting in an area under the receiver operating characteristics curve (AUC) of 0.86 and sensitivity of 57% at 91% specificity. Focusing on the TGF-β superfamily signaling, clustering analysis identified several proteomic signatures that performed well in distinguishing HCC patients from cirrhosis. Our model outperformed both AFP alone and the Doylestown model by clinically relevant margins: at 90% specificity, the sensitivity of our model was 20% higher than AFP alone and 33% higher than the Doylestown model in the derivation samples. Conclusions and Next Steps: This study successfully demonstrates the performance of our biologically functional model combining 4 serum biomarkers with 3 clinical parameters to predict HCC across cirrhotic patients from a racially and ethnically diverse population, which outperformed the Doylestown model or AFP alone. The next steps will be large-scale validations of this biomarker panel in cirrhotic patients across all racial and ethnic subgroups in Phase II/III studies. Citation Format: Xiyan Xiang, Anil K. Vegesna, Richard L. Amdur, Kirti Shetty, Herbert Yu, Linda L. Wong, Sharon S. Fox, Michael G. Ryan, Nyasha Chambwe, Sanjaya K. Satapathy, James M. Crawford, Gregory M. Grimaldi, Vikas Kundra, Lopa Mishra. A novel clinical and biomarker model accurately predicts hepatocellular carcinoma (HCC) across racial groups [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 1 (Regular and Invited Abstracts); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(7_Suppl):Abstract nr 3336.