The worldwide incidence of hepatocellular carcinoma (HCC) is approximately one million cases a year. This makes HCC one of the most frequent human malignancies, especially in Asia and Africa, although the incidence is increasing also in the western world. HCC is a complication of chronic liver disease, with cirrhosis as the most important risk factor. Viral co-pathogenesis makes cirrhosis due to hepatitis B (HBV) and hepatitis C virus (HCV) infection a very important factor in the development of HCC. As curative therapy is often ruled out due to the late detection of HCC, it would be attractive to find parameters which predict malignant transformation in HBV- and HCV-infected livers. This study has used comparative genomic hybridization (CGH) to analyse 26 HCCs (11 non-viral, nine HBV, six HCV) and 12 concurrent dysplasias (five non-viral, five HBV, two HCV). Frequent gain (> or =25% of all tumours) was detected, in decreasing order of frequency, on 8q (69%), 1q (46%), 17q (46%), 12q (42%), 20q (31%), 5p (27%), 6q (27%), and Xq (27%). Frequent loss (> or =25% of all tumours) was found, in decreasing order of frequency, on 8p (58%), 16q (54%), 4q (42%), 13q (39%), 1p (35%), 4p (35%), 16p (35%), 18q (35%), 14q (31%), 17p (31%), 9p (27%), and 9q (27%). Minimal overlapping regions could be determined at multiple locations (candidate genes in parentheses). Minimal regions of overlap for deletions were assigned to 4p14-15 (PCDH7), 8p21-22 (FEZ1), 9p12-13, 13q14-31 (RB1), 14q31 (TSHR), 16p12-13.1 (GSPT1), 16q21-23 (CDH1), 17p12-13 (TP53), and 18q21-22 (DPC4, DCC). Minimal overlapping amplified sites could be seen at 8q24 (MYC), 12q15-21 (MDM2), 17q22-25 (SSTR2, GH1), and 20q12-13.2 (MYBL2, PTPN1). A single high level amplification was seen on 5q21 in an HBV-related tumour. Aberrations appeared more frequent in HBV-related HCCs than in HCV-associated tumours (p=0.008). This was most prominent with respect to losses (p=0.004), specifically loss on 4p (p=0.007), 16q (p=0.04), 17p (p=0.04), and 18q (p=0.03). In addition, loss on 17p was significantly lower in non-viral cancers than in HBV-related HCC (p<0.001). Furthermore, loss on 13q was more prevalent in HCCs in non-cirrhotic livers (p=0.02), thus suggesting a different, potentially more aggressive, pathway in neoplastic progression. A tendency (p=0.07) was observed for loss on 9q in high-stage tumours; no specific changes were found in relation to tumour grade. A subset of the HCC-associated genetic changes was disclosed in the preneoplastic stage, i.e. liver cell dysplasia. This group of dysplasias showed frequent gain on 17q (25%) and frequent loss on 16q (33%), 4q (25%), and 17p (25%). The majority of the dysplasias with alterations revealed genetic changes that were also present in the primary tumour. In conclusion, firstly, this study has provided a detailed map of genomic changes occurring in HCC of viral and non-viral origin, and has suggested candidate genes. Loss on 17p, including the TP53 region, appeared significantly more prevalent in HBV-associated liver cancers, whereas loss on 13q, with possible involvement of RB1, was distinguished as a possible genetic biomarker. Secondly, CGH analysis of liver cell dysplasia, both viral and non-viral, has revealed HCC-specific early genetic changes, thereby confirming its preneoplastic nature. Finally, genes residing in these early altered regions, such as CDH1 or TP53, might be associated with hepatocellular carcinogenesis.