The aim of this study was to develop and validate a molecular index for the diagnosis of hepatocellular carcinoma (HCC) based on genes whose specificity and level of expression are the most discriminating for the diagnosis of HCC. The level of expression of 219 genes was assessed with a real-time reverse transcription-polymerase chain reaction approach in a training set of samples including normal livers (15), cirrhosis (12), and HCC (16). The most informative genes were selected for the molecular index. This index was prospectively validated in a new set of 40 samples (testing set) and in a set of 45 cirrhotic macronodules. 44 out of the 219 genes were differentially expressed in HCC. 13 out of these 44 genes were finally selected for the molecular index according to their diagnostic performance and after exclusion of most redundant genes. Using this index, 42 out of 43 samples of the training set and 39 out of the 40 samples of the testing set were correctly ranked as HCC or not HCC (normal liver or cirrhosis). The index also enabled correct ranking of 44 out of 45 cirrhotic macronodules into 2 groups: benign (including macroregenerative and dysplastic macronodules) and malignant macronodules. This molecular diagnostic index is an efficient tool both for identification of overt HCC as well as minute lesions (cirrhotic macronodules). It might be useful to correctly diagnose borderline lesion or small well-differentiated hepatocellular carcinomas whose diagnosis is often difficult on a histopathological basis.