Retinoic acid (RA) is an important regulator of normal cellular proliferation and differentiation and suppressor of tumor growth by cell cycle arrest and apoptosis. Furthermore, RA showed a chemo-preventive activity in preclinical and/or clinical models of lung, head and neck, breast, and hepatocellular carcinoma (HCC). In this study, we examined the effect of RA on the proliferation of human HCC cells, in order to analyze its mode of action and, finally, we attempted to find a surrogate biomarker of RA for HCC chemopreventive treatment. Our findings suggested that the growth inhibition of RA in HCC cells differed according to G(1) phase delay by CDK2 or 4, finally induction of apoptosis. No correlation was found between RA sensitivity and the expression of nuclear retinoid receptors, such as RARs or RXRs in HCC cells. RA treatment caused cell cycle arrest at G(1) and decreased the expressions and activities of CDK2 or CDK4 in RA-sensitive HepG2 and SNU354 cells. On the other hand, RA-resistant Hep3B and SNU449 cells progressed into the S/G(2)+M phase and showed increased CDK2 and CDK4 expression and activity. Since the inhibition of CDK2 or 4 activities resulted in sensitization of HCC cells to RA, the combination of RA and compounds of inhibiting CDKs such as UCN01 and flavopiridol might be a useful targeted therapy strategy for HCC.