O-013 - A global risk gene score predicts early and late tumor recurrence after resection of hepatocellular carcinoma

Date 04 July 2015
Event WorldGI 2015
Session Oral and LBA abstracts
Topics Hepatobiliary Cancers
Translational Research
Presenter J. Dekervel
Citation Annals of Oncology (2015) 26 (suppl_4): 108-116. 10.1093/annonc/mdv235
Authors J. Dekervel1, C. Verslype2, E. Van Cutsem3, D. Popovic4, H. van Malenstein1, P. Windmolders1, L. Libbrecht1, A. Bulle1, B. de Moor4, F. Nevens1, J. van Pelt1
  • 1University Hospitals Leuven & Department of Clinical and Experimental Medicine, Leuven/BE
  • 2University Hospitals Leuven and Department of Oncology, Leuven/BE
  • 3University Hospitals Leuven, Leuven/BE
  • 4Signal Processing and Data Analytics / iMinds Medical IT, Leuven/BE



Patients undergoing treatment with curative intent for hepatocellular carcinoma (HCC) remain at high risk for tumor recurrence by the emergence of intrahepatic metastasis (“early recurrence”) or the formation of new lesions (“late recurrence”). We aimed to develop and validate a gene expression signature that could predict this risk.


HepG2 liver cancer cells made resistant to sorafenib during several months were used as a model for hepatocyte dedifferentiation and aggressive tumor biology. The differential expressed genes between this cell line and its non-resistant parental lineage were determined using an Affymetrix microarray platform. To form a score, the number of genes was subsequently downsized by assessing the performance of each one in published microarray data sets of HCC samples (GSE25097, GSE9843) and samples of non-tumoral surrounding liver (GSE40873). This resulted in a 7 gene score (Global Risk Score, GRS) that was validated in two independent data sets (LEC cohort, GSE1898/GSE4024, n = 67 and NCI cohort, GSE14520, n = 247) using a tumor- and liver-specific cut-off for high and low GRS.


The GRS, when assessed in tumor tissue of patients treated with resection, identified patients at low and high recurrence risk at 3 years in the LEC (68 +/- 10% vs 35 +/- 7%, p = 0.03) and NCI (62 +/- 5% vs 37 +/- 4%, p < 0.001, figure 1A) cohort. Moreover, assessment of the GRS in non-cancerous surrounding liver tissue of the NCI patients revealed that, starting at about 19 months after resection, the score was correlated with the formation of new lesions in the cirrhotic liver (figure 1B). Taken together, based on the tumor and liver score value, the GRS identified 4 patient groups with differences in recurrence rates and time patterns of recurrence (figure 1C). Multivariable analysis by Cox regression showed that GRS group was a significant predictor of recurrence, independent from BCLC stage (p = 0.007).


We present the first gene score validated in both HCC tissue and surrounding liver based on an in vitro model. By identifying aggressive tumor biology as well as tumorigenic potential of the underlying liver disease, the Global Risk Score allows calculating a prognostic estimate for an individual patient facilitating therapeutic decisions.

Figure: O-013. Performance of the Global Risk Score in the NCI cohort (n = 247)