本帖最后由 kennyu 于 2011-3-23 17:52 编辑
回复 StephenW 的帖子
http://www1.easl.eu/easl2011/program/Posters/Abstract182.htm
Poster Presentations Session Title: Category 07b: Viral Hepatitis B & D: Clinical (except therapy)
Presentation Date: 31 MAR, 2011
GENERATION OF THE PREDICTION MODELS FOR THE DEVELOPMENT OF HEPATOCELLULAR CARCINOMA IN CHRONIC HEPATITIS B AND EXTERNAL VALIDATION BY INDEPENDENT COHORTS M. Kurosaki1*, M.-F. Yuen2, W.-K. Seto2, T. Kumada3, K. Tanaka1, Y. Suzuki1, Y. Hoshioka1, N. Tamaki1, T. Kato1, Y. Yasui1, T. Hosokawa1, K. Ueda1, K. Tsuchiya1, T. Kuzuya1, H. Nakanishi1, J. Itakura1, Y. Takahashi1, Y. Asahina1, N. Izumi1
1Musashino Red Cross Hospital, Tokyo, Japan, 2University of Hong Kong, Hong Kong, Hong Kong S.A.R., 3Ogaki Municipal Hospital, Gifu, Japan. *[email protected]
Aims: Data mining analysis was used to generate predictive models for the development of HCC and to evaluate the efficacy of nucleos(t)ide analog (NUC) therapy to reduce HCC in patients at high risk.
Methods: A cohort of 588 chronic hepatitis B patients was screened for HCC (average period 7.5 years). Data mining analysis (IBM-SPSS Modeler 13) was used to identify risk factors for HCC and to generate predictive models. Independent cohort in Hong Kong (cohort-H, n=525) and in Ogaki, Japan (cohort-O, n=576) was used for external validations. Sensitivity to identify patients at risk for HCC was compared between models and the guideline criteria for antiviral treatment. Data from 151 patients on long-term NUC therapy was applied on these models to evaluate the efficacy of NUC to prevent HCC.
Results: The 5 year prevalence of HCC in untreated cohort was 5.3%. Age (≥40), platelet (< 150 x 109/L), HBV DNA (>5.8 log copies/ml), and mutations in core promoter 1762/1764 were identified as risk factors and were used to build prediction model of HCC (model-1). The 5 year prevalence of HCC in patients having 0, 1, 2, 3, and 4 risk factors was 0%, 0-3.4%, 3.2-4.3%, 17.6% and 30%, respectively. In the model without incorporating core promoter mutations (model-2), the 5 year prevalence of HCC in patients having 0, 1, 2, and 3 risk factors was 0%, 1.1-1.8%, 5.0-9.1%, and 28.1%, respectively. The external validations confirmed the reproducibility of these models (cohort-H: r2=0.87-0.95, cohort-O: r2=0.93). Sensitivity for identifying HCC development was as follows: AASLD guideline 19%, EASL guideline 67%, model-1 83%, and model-2 98%. In patients with high risk of HCC, long-term NUC therapy reduced the 5 year incidence of HCC by 11-23% (p < 0.05).
Conclusions: Prediction models that include age, platelet counts, HBV DNA, and core promoter mutations had high reproducibility and sensitivity to identify patients with risk for the development of HCC. These models may be used to extract patients at risk for HCC out of those excluded from therapy by treatment guidelines. The NUC therapy significantly reduced the incidence of HCC among patients with high risk for HCC. |