Reducing the heterogeneity in hepatocellular carcinoma. A cluster analysis based on clinical variables in patients treated at a quaternary care hospital.

Abstract Introduction and aims Even though the term hepatocellular carcinoma designates the most common type of primary liver cancer, the disease has a high level of heterogeneity due to its etiology, geographic variation, behavior, and association with specific genetic alterations. The aim of the present study was to establish, through a cluster analysis, the clinical characteristics that enable homogeneous conglomerates to be defined. Materials and methods An exploratory cluster analysis was developed utilizing the K-means method for sub-classifying 119 cases of patients with hepatocellular carcinoma. Sixty-two of those patients met the inclusion criteria, as well as none of the exclusion criteria. For the cluster analysis, an n-dimensional space was defined, in which n was equal to the number of variables included in the study (n = 17). The spatial coordinates corresponded to any possible magnitude between the minimum and maximum values of the variables analyzed (age, sex, tumor volume, AFP, AST, DB, Alb, Na, INR, Cr, HBV, HCV, OH, NASH, cirrhosis, multiple tumors, and neotumor). Results Four patterns with homogeneous clinical characteristics were identified, in which age at presentation, history of hepatitis B virus infection, altered liver profile with cholestatic dominance, and low albumin levels were associated with an apparently worse outcome. Conclusions How heterogeneity in hepatocellular carcinoma could be reduced was shown through utilizing an unsupervised learning method to define specific subgroups, in whom known pathophysiologic mechanisms could better explain tumor behavior and define the determining prognostic factors related to the subgroups.