Letters to the Editor Staging for hepatocellular carcinoma: look for a perfect classification system To the Editor: We have read with interest the paper entitled “The critical issue of hepatocellular carcinoma prognostic classification: which is the best tool available?” by Cillo et al. published in a recent issue of the Journal of Hepatology [1]. By utilizing a sophisticated analysis, this study looked at the discriminatory ability of the currently used five (Okuda, BCLC, CLIP, CUPI and French) staging systems for hepatocellular carcinoma (HCC) in terms of prognostic prediction for HCC patients. The authors demonstrated that the BCLC classification system from the Spanish group is the best one in differentiating the clinical outcomes of these patients. The main cause why authors have reached this conclusion could be closely related to the distinct characteristics of the study population as most (85%) patients had undergone radical treatment (resection or loco-regional therapy), suggesting the majority of them belonged to early or intermediate clinical stage, and only 15% of patients were treated with supportive measures. It would be interesting to know if the same findings still hold true when the BCLC staging system is fitted to other study groups and their patient populations. The other two validated systems, the Okuda and CLIP systems, were originally derived from a large unselected patient population and the majority of them had been treated conservatively. Therefore, although the risk factors for the currently used staging systems are not mutually exclusive, the derived predictive models from these predictors may have otherwise variable differentiating power. Certain important risk factors (tumor size . 3 or , 3 cm) could only be significant in highly selected patient populations that undergo surgical resection, liver transplantation or loco- regional therapies [2–6], and such strict criteria may not be able to differentiate the outcome in some circumstances. Therefore, the choice of different cut-offs or the extent of disease in a staging system may have a substantial impact on the clinical applicability. In these instances, the predictive ability of a given staging model constructing from the selected risk factors could be significantly impaired if the majority of patients do not have early stage HCCs. Such an effect may explain why CLIP or Okuda staging systems did not have a better discriminatory power in this study. Consistent with this notion is that the investigators from Canada have demonstrated that CLIP was a good predictive model for HCC patients in their patient population that more than half (52%) had been only treated conservatively due to a relatively advanced tumor or cirrhotic stage [7]. Alter- natively, we have recently observed that the CLIP system did not efficiently predict the survival among patients predominantly undergoing surgical resection or transarterial chemoembolization in a multicenter, nation-wide survey in Taiwan [8], probably also because these patients had been pre-selected for analysis. Another important factor that may alter the predictive ability of a staging system is the distribution of patients in each category. In the study by Cillo et al., the proportion of study patients was considered fairly balanced from stage A to D in the BCLC system (39, 23, 25 and 13%, respectively) given the patient number was relatively small ðn ¼ 187Þ; whereas the distributions in other staging systems were quite uneven and sometimes to the extreme. It is apparent that there will be substantial magnitude of overlapping (or swing) in the survival curves if the denominator was too small or the predictive variables are not weighted in a balanced target population. We concur with Cillo et al. that their results well support BCLC is an ideal system for a surgically oriented unit or centers characterized by periodi- cal surveillance for HCC and early referral. However, since the clinical presentation of HCC is tremendously hetero- geneous, it is necessary to consider all known predictive factors from early to advanced stage in building a ‘perfect’ staging system to fit all patient populations. Teh-Ia Huo, Shou-Dong Lee, Jaw-Ching Wu Department of Medicine, Taipei Veterans General Hospital and National Yang-Ming University School of Medicine, Taipei, Taiwan, ROC E-mail address: tihuo@vghtpe.gov.tw References [1] Cillo U, Bassanello M, Vitale A, Grigoletto FA, Burra P, Fagiuoli S, et al. The critical issue of hepatocellular carcinoma prognostic classification: which is the best tool available? J Hepatol 2004;40: 124–131. [2] Mc Peake JR, O’Grady JG, Zaman S, Portmann B, Wight DG, Tan KC, 0168-8278/$30.00 q 2004 European Association for the Study of the Liver. Published by Elsevier B.V. All rights reserved. Journal of Hepatology 40 (2004) 1041–1049 www.elsevier.com/locate/jhep