Characteristics of Hepatocellular Carcinoma Aggressiveness Factors in Turkish Patients

The classification of tumors and prognostication for patients with hepatocellular carcinoma (HCC) has generally been considered to be a reflection of 2 sets of separate factors since the first such report by Okuda et al. [1]. These are liver factors and tumor factors, which are separate yet likely related [2,3]. All modern classification schemata incorporate parameters from both sets of factors [4,5]. However, there are many characteristics that distinguish HCC patients in various parts of the world from each other. We report here, for the first time, on a large HCC database from several collaborating institutions in Turkey, which is a Mediterranean country that constitutes a land bridge between Europe and Asia, and we focus on 3 parameters of tumor behavior, namely maximum tumor diameter (MTD), portal vein thrombosis (PVT), and α-fetoprotein (AFP) levels, and the correlates of each parameter.

We analyzed a database of 1,332 prospectively accrued HCC patients who had full baseline tumor parameter data, including computed tomography scan information on HCC size, number of tumor nodules, presence or absence of PVT, and plasma AFP levels; complete blood count; routine blood liver function tests (total bilirubin, γ-glutamyl transpeptidase [GGTP], alkaline phosphatase [ALKP], albumin, and transaminases); and patient demographics. Diagnosis was made either via tumor biopsy or according to international guidelines [6,7]. Database management conformed to legislation on privacy, and this study conforms to the ethical guidelines of the Declaration of Helsinki; approval for this retrospective study on de-identified HCC patients was obtained from the institutional review board.

Means and standard deviations for continuous variables and relative frequencies for categorical variables were used as indices of centrality and dispersion of the distribution. For categorical variables, the χ² test and z test for proportions were used. The Pearson correlation was used to measure the association between 2 continuous variables, the Wilcoxon rank-sum (Mann-Whitney) test to test the difference between 2 categories, and the Kruskal-Wallis rank test to test the difference among categories.

A linear regression model was used to evaluate the associations between MTD and single variables examined. The final multiple linear regression was obtained with the backward stepwise method, and the variables that showed associations with p < 0.10 were left in the models.

When testing the null hypothesis of no association, the probability level of α error, 2 tailed, was 0.05. All the statistical computations were made using STATA 10.0 Statistical Software (StataCorp, 2007, Stata Statistical Software, release 10, College Station, TX, USA).

Table 1 shows the descriptive characteristics of the total patient cohort: 81.11% were male, and 81.42% had cirrhosis. The predominant etiological factor was hepatitis B virus (HBV) in 60.86%, followed by hepatitis C virus (HCV) in 20.72% of patients. The mean MTD was 5.89 cm, with approximately one-third of patients having tumors <3.5, 3.5-6.5, and >6.5 cm. The mean AFP level was 5,686.54 IU/mL; 42.13% of patients had normal AFP laboratory values, and 41.97% of patients had AFP values >100 IU/mL. The mean albumin was low at 3.09 g/dL, and the mean total bilirubin was elevated at 2.96 mg/dL.

The MTD was then further examined. Patients were divided into 3 size tertiles, and the means and distributions of the associated peripheral blood platelet counts were then calculated. Figure 1 shows that the mean platelet count increased with an increase in MTD tertile, as has been reported in other cohorts [8,9]. Patients with the largest MTD tumors had normal platelet counts and, thus, lesser degrees of fibrosis, as noted previously [10]. Characteristics of HCC patients with small and large MTD tumors were then further characterized after merging tertiles II and III, due to the small patient number in tertile III. Patient groups with either MTD <5 or ≥5 cm tumors were then each dichotomized according to blood platelet counts of <125 or ≥125 × 10³/μL (Table 2), as low blood platelets are a cirrhosis surrogate [11]. In the small-tumor group, there was little difference in tumor characteristics between the platelet subgroups, although albumin and bilirubin levels were significantly worse in the low-platelet subgroup, as expected. In contrast, in the larger-tumor group, AFP was significantly higher and PVT percent of patients was significantly higher in the high-platelet subgroup compared to the low-platelet subgroup. In the large-MTD subgroup with higher platelets and larger tumors, cirrhosis was significantly less, as was total bilirubin. Thus, platelet dichotomization selects for a patient phenotype with better liver function but more advanced HCC. A linear regression model of MTD showed several significant single variables (Table 3). However, in the final multiple linear regression model, only platelet count and presence of PVT were significant.

Patients were next examined according to presence or absence of PVT (Table 4). Patients who were PVT positive had significantly larger tumors (in terms of MTD) and higher platelet levels, significantly higher AFP and C-reactive protein levels, and higher total bilirubin, ALKP, and GGTP levels. Interestingly, levels of “good” HDL cholesterol were significantly lower in the PVT-positive group. To our knowledge, this has not been previously reported.

Patients were next divided into 3 groups according to blood AFP levels of <100, 100-1,000, and >1,000 IU/mL (Table 5), and their tumor and nontumor parameters were examined. As expected, MTD and PVT percent significantly increased with an increase in AFP. There was also a trend to tumor multifocality with an increase in AFP. There were also increases in ALKP, GGTP, and total bilirubin levels with an increase in AFP as well as a statistically significant increase in platelet counts, likely reflective of the increase in MTD.

We then examined the associations for MTD and platelets (Fig. 2a) and for MTD and PVT (Fig. 2b). We found a significant correlation for MTD and platelets (r = 0.3348; p < 0.0001). Box plots were then created for MTD and PVT categories. There were significant differences in the MTD between PVT-positive and -negative patients (p < 0.0001).

As this database was composed of patients from several geographic sites throughout Turkey, we were curious as to whether there might be any regional differences in the patterns of HCC presentation or the patients. Table 6 shows patient groups from the locations that contributed most patients, and several marked differences were observed. MTD differed by locale, with the highest mean MTD of 7.0 cm being found in Mersin and the lowest mean MTDs of 5.03 and 5.33 cm being found in Ankara and Hatay, respectively. There were also large differences in PVT percent, with 40.74% in Mersin (where there was also the highest MTD) and the lowest PVT percent in Hatay with 20.69% (where there was also the smallest MTD). Multifocality with >1 tumor nodule was highest in Diyarbakır (53.49% of patients) and lowest in Mersin (17.02% of patients). Mean AFP levels were highest in Mersin (10,109 ng/mL) and lowest in Mardin and Ankara (2,885 and 3,254 ng/mL, respectively). There were also large regional differences in underlying liver disease. Cirrhosis was present in 88.29% of patients in Diyarbakır but only in 62.92% of patients in Hatay. Furthermore, there were big regional differences in HBV, HCV, and hepatitis D virus (HDV). HBV was 73.95% in Diyarbakır but only 40.24% in Hatay. Conversely, HCV incidence in these HCC patients was highest in Hatay (30.49%) and lowest in Diyarbakır and Mardin (8.37 and 7.81%, respectively). In contrast, HDV was 17.13% in Diyarbakır but <10% elsewhere. Thus, Mersin patients had the largest tumors, highest incidence of PVT, and highest AFP levels. HBV was highest in Diyarbakır (Mardin a close second), and HCV was highest in Hatay. Mersin patients not only had the largest MTD, but also the highest total bilirubin levels (4.63 mg/dL), and the lowest bilirubin levels were found in Adana (2.20 mg/dL).

Tumor characteristics that are generally considered for HCC patients are predominantly maximum tumor size or MTD and PVT and, to a lesser extent, AFP (due to its variability). Larger tumors have a worse prognosis in cancer in general as well as in HCC [12]. However, 40% of this cohort of HCC patients had AFP levels <20 IU/mL. This is similar to findings elsewhere [13] and doubtless contributes to the uncertainty concerning the use of blood AFP levels as a screening tool [14]. Platelets have been previously found to be both a harbinger of HCC in patients that are predisposed as well as a cirrhosis surrogate and to be associated with tumor size [9,11,15]. In this cohort, 80.4% of patients had cirrhosis and 30.9% had thrombocytopenia (Table 1). Furthermore, there was a significant difference in platelet levels between patients having smaller versus larger MTD tumors (p = 0.0001; Fig. 1), and this was also confirmed by the significant Pearson correlation (Fig. 2). In a regression analysis, both platelets and presence of PVT were significant for MTD (Table 3).

PVT positivity was present in 28.55% of the total cohort (Table 1). When PVT-positive and -negative patients were compared (Table 4), the PVT-positive patients were found to have significantly larger tumors (in terms of MTD), as also seen in Figure 2, as well as tumor multifocality, higher AFP levels, and higher total bilirubin levels. Whether this is due to more aggressive tumors causing parenchymal destruction, or due to an increased PVT in patients with worse cirrhosis, is not addressed here, except that there was less cirrhosis in the PVT-positive group. Remarkably, however, there were significantly higher cardioprotective HDL levels in the PVT-negative group. We think this is the first report of this association, as we are not aware of this having been reported elsewhere.

Patients were trichotomized into 3 groups based on their AFP levels of <100, 100-1,000, and >1,000 IU/mL (Table 5). The majority of patients had levels <100 IU/mL (Table 1, 5). Patients in the high AFP group also had significantly larger tumors (in terms of MTD), tumor multifocality, and high PVT percent as well as platelet numbers. Patients with higher AFP also had worse liver function, as judged by total bilirubin, GGT, and ALKP levels, and lower albumin and higher cirrhosis percent, but the differences were not significant between the 2 elevated AFP groups.

An advantage of this multi-institutional study was the unusual possibility to compare HCC presentation in differing parts of a large country. Table 6 only includes the groups that contributed the largest patient numbers, predominantly in the center and east of Turkey. However, there were some remarkable differences. Firstly, the incidence of HDV was mainly found in patients from Diyarbakır and Mardin. Secondly, there was a considerable range of presence of cirrhosis, from 85.4% in Mersin to 62.9% in Hatay. Thirdly, tumor characteristics showed considerable heterogeneity. Thus, mean MTD was 7 cm in Mersin but 5 cm in Ankara; PVT was over 40% in Mersin and Ankara, yet only 20% in Hatay; tumor multifocality was 53% in Diyarbakır, yet only 17% in Mersin.

In Turkey, the prevalence of inactive HBV carrier patients is higher than in many countries, and these patients are usually ignored in terms of treatment. However, some of them may have severe fibrosis or cirrhosis. These patients may develop HCC over time. This could explain why HBV-related HCC patients more commonly have cirrhosis than other etiology-associated HCC patients in our cohort. In the Diyarbakır region, HDV infection is extremely prevalent (Table 6). Dual infection may cause cirrhosis more commonly.

In the Hatay region, HCV infection is more common than in other regions of Turkey. The high prevalence may be related to inappropriate dental treatments. It seems that this problem is being solved by dental awareness and added hygiene. However, it is difficult to explain why HCV-related HCC patients in Hatay had less cirrhosis. The average MTD of Hatay patients was the second lowest (above Ankara), possibly because of the mix of HBV- plus HCV-etiology patients in Hatay.

The relationship between cirrhosis and platelet count is an interesting and complex issue. HBV and HCV can replicate in the platelets and decrease platelet survival. This replication is more commonly seen in HCV infection than in connection with other causes and is related to cirrhosis. Furthermore, there is likely a relationship of thrombocytopenia to the degree of cirrhotic fibrosis, and measures of cirrhotic fibrosis severity were not available for this cohort. Platelet counts may also be related to the Child-Pugh score which indicates liver functional capacity. Patients with high Child-Pugh score may have lower platelet counts because thrombopoietin, which promotes bone marrow, is synthesized in the liver.

This study of a large HCC population at presentation shows certain unusual features, such as a high incidence of HBV, presence of HDV, and a large range of cirrhosis and tumor characteristics. Yet, the underlying biology, such as the relationship of MTD to platelets and to PVT, is similar to other reports. A drawback is the absence of survival data, in part due to many patients being from far-flung and rural areas. However, we show the fascinating variability in HCC presentation in a large country and how biological principles in the relationship between parameters can be reproduced here.

The authors declare no conflicts of interest.

This work was supported in part by NIH grant CA 82723 (B.I.C.).

H.A. and B.I.C. contributed equally to this work.