Introduction

The prognosis in hepatocellular carcinoma (HCC) patients is considered to be a reflection of both liver damage and tumor factors [1], and either or both are causes of death in these patients. Thus, HCC patients general have two diseases in the liver, namely hepatitis or cirrhosis, as well as their HCC. Although the cirrhosis has been understood to precede and be involved in HCC causation, they have generally been thought of as separate, if linked, processes. These two processes have both been included in most modern HCC staging systems [2]. The idea that liver damage might be closely involved in HCC phenotype has received recent support from clinical analysis of HCC datasets [3], as well as considerations of the underlying liver as representing a complex microenvironment, with many cell types and processes contributing to HCC development [4].

The four parameters of HCC clinical behavior have recently been combined as an HCC aggressiveness index [5, 6], involving the sum of the scores for maximum tumor diameter (MTD), tumor multifocality, portal vein thrombosis (PVT), and blood alpha-fetoprotein (AFP) levels. This aggressiveness index has in turn been shown to relate to a composite liver index (LI), composed of the sum of scores for the levels of blood total bilirubin, albumin, gamma glutamyl transpeptidase (GGTP), and platelet levels, as a cirrhosis surrogate [7].

The current work was undertaken to validate the LI in an independent and large HCC patient cohort. We found that the LI in this cohort also significantly related to the aggressiveness index as well as to trends in individual HCC parameters.

Methods

Liver Index

The liver index was constructed, as previously reported [7] as the sum of the scores for blood GGTP + total bilirubin + albumin + platelet levels. The scores were assigned as follows:

  • GGTP IU/ml (cut-off): GGTP<100; 100≤GGTP≤200; GGTP>200 levels were assigned a score of 1, 2, or 3, respectively.

  • Bilirubin mg/dl (cut-off): bilirubin<1.5; 1.5≤bilirubin≤2.5; bilirubin>2.5 levels were assigned a score of 1, 2, or 3, respectively.

  • Albumin g/dl (cut-off): albumin>3.5; 2.5≤albumin≤3.5; albumin<2.5 levels were assigned a score of 1, 2, or 3, respectively.

  • Platelets ×109/l (cut-off): platelets<100; 100≤platelets≤150; platelets>150 levels were assigned a score of 1, 2, or 3, respectively.

The score points were derived from dividing the full range of each parameter into terciles, each increasing tercile then being given an increased point of 1, 2, or 3.

The liver index score was divided into three groups for the association with the aggressiveness index (Table 1) and the trends analysis (Fig. 1): a, score = 4 points; b, 4<score≤8; and c, score >8 points.

Table 1 Relationship between liver index score categories (a), (b), and (c) and aggressiveness index score in the total cohort
Full size table
Fig. 1
figure 1

Trends between liver index score categories and a maximum tumor dimension (cm), b alpha-fetoprotein (ng/ml), c portal vein thrombosis (%+ve), and d number nodules (% >3). Octothorpe indicates z test for trend; circumflex accent indicates Mann-Whitney test; psi indicates chi-square test for trend; section sign indicates multiple comparisons of proportions

Full size image

Patients and Data Collection

Data Collection

We retrospectively analyzed prospectively collected data of 4139 HCC patients, as previously reported [8] who had full baseline tumor parameter data, including CT scan information on maximum tumor diameter (MTD), number of tumor nodules and presence of PVT and plasma AFP levels; blood counts; routine blood liver function tests, (total bilirubin, GGTP, albumin). Database management and this study conformed to the ethical guidelines of the Declaration of Helsinki. Approval for this retrospective study on de-identified and deceased HCC patients was obtained by the Institutional Review Board.

An aggressiveness index was calculated as the sum of scores, as previously reported [5, 6]:

  • MTD (cm, in tertiles): MTD<4.5; 4.5≤MTD≤9.6; MTD>9.6; scores 1, 2, 3, respectively;

  • AFP ng/ml (cut-off): AFP<100; 100≤AFP≤1000; AFP>1000; scores 1, 2, 3, respectively;

  • PVT (No/Yes): PVT(No); PVT(Yes); scores 1, 3, respectively;

  • Number of tumor nodules: nodules≤3; nodules>3; scores 1, 3, respectively.

Statistical Analysis

Mean and SD for continuous variables were used as indices of centrality and dispersion of the distribution.

It was necessary, for non-normally distributed values, for the continuous variables, to use non-parametric methods. We used the Kruskal-Wallis rank test for differences of the parameters among the three categories of the liver index, and the Wilcoxon rank-sum (Mann-Whitney) test for the comparisons of the aggressiveness index between two categories at a time of the liver index score.

The test for trend was used to evaluate the trend of the aggressiveness index means among liver index score categories.

A linear regression and a multiple linear regression model were used for the association of the aggressiveness index score on the liver index score Table 2 (A), and on each serum variable, gamma glutamyl transpeptidase (GGTP), total bilirubin (bilirubin), albumin (Alb), and platelets (Plt), included together in the model Table 2 (B). The results were presented as coefficients (β) with 95% C.I.

Table 2 Linear regression model of aggressiveness index score on liver index score (A), and multiple linear regression model on all parameters of the liver index: platelets, GGTP, total bilirubin, and albumin, together in the model (B)
Full size table

In the linear regression, the β represents the variation of the dependent variable, for one-unit variation of the predictor variable considered as continuous variable.

For trends in the relationship between liver index score categories and the individual HCC parameters (Fig. 1), we used the z test for trend, Mann-Whitney test, chi-square test for trend, and multiple comparisons of proportions.

When testing the hypothesis of significant association, p value was ≤0.05, two tailed for all analyses. Statistical analysis was performed with StataCorp. 2007. Stata Statistical Software: release 10. College Station, TX: StataCorp LP.

Results

Liver Index and HCC Aggressiveness

The liver index is shown in the “Methods” section, and comprises scores of 1, 2, or 3 which are assigned to the values in an HCC patient for blood levels of GGTP, total bilirubin, albumin, and platelet levels [7]. These four blood parameters were chosen since they had the highest odds ratios (OR) in a multiple regression analysis for the HCC aggressiveness index [6]. They were divided into three categories, each category being given a point score of 1, 2, or 3, respectively. The total liver index scores were then divided into three groups of a, b, or c, for the purposes of further analysis. A score of four was represented as group “a”; a score of 4<score≤8 was represented as group “b”; and a score of >8 was represented as group “c”.

Relationships Between Liver Index and Aggressiveness Index

The relationship between the aggressiveness index scores and the liver index scores (Table 1) was examined. We found a positive association between the aggressiveness index and the liver index, p = 0.0001. Considering the relationship between the aggressiveness index and liver index as shown in Table 1, we found that there was a positive and significant linear trend between them. The average index values of the aggressiveness index increased with each increase in the liver index categories and the relative differences between these average values were all significant (p < 0.001, test for trend). Thus, there was a statistically significant association between the aggressiveness index, as an expression of tumor aggressiveness, and the liver index, as an indicator of liver function.

A linear regression model was then constructed of the aggressiveness index score on the liver index score (Table 2 A). We found a positive association between the aggressiveness index and the liver index. For each liver index unit of change, the aggressiveness index increased by 0.76 units. A multiple regression model was then calculated for all the four parameters of the liver index (GGTP, total bilirubin, albumin, and platelets) together in the model (Table 2 B). We found that three of the individual components used for the construction of the liver index are positively associated with the aggressiveness index, namely, total bilirubin, GGTP, and platelets. By contrast, only the blood albumin showed an inverse association with the aggressiveness index. Thus, with an increase in albumin levels, the aggressiveness index decreased. For all of these components, there was a statistically significant association (for all, p < 0.001).

Trends in Relationships

Trends in the relationships between the liver index score categories and the individual components of the aggressiveness index (Fig. 1) were then examined. For each tumor parameter, namely MTD, PVT, tumor multifocality, or AFP, there was a significant trend between the liver index score and the tumor parameter measures.

Discussion

The twin influences of HCC behavior (biology) and severity of the underlying liver disease are understood to influence HCC patient prognosis and thus have been incorporated into most HCC classification systems [2, 9]. The four liver parameters of blood albumin, total bilirubin, GGTP, and platelet levels were incorporated into our liver index, as having the highest hazard ratios when an HCC aggressiveness index was studied in relation to liver function parameters [6]. Liver disease severity assessment has also recently been simplified by use of the quantitative ALBI (albumin and bilirubin) score [10]. Furthermore, serum albumin levels have been incorporated into the 2-parameter Glasgow score (together with C-reactive protein) as an independent prognosticator for HCC and other tumors, based on its reflection of systemic inflammation [11]. Serum albumin, in addition to being a measure of hepatic synthetic function, may also limit HCC growth [12, 13].

Elevated levels of serum bilirubin are a predominant marker of liver damage. Platelet levels, in addition to their role in hemostasis, have been shown to be a surrogate for cirrhosis [14] and were thus incorporated into our index and they may be involved in HCC growth and resistance to drug actions [15]. GGTP, in addition to being a liver function test, may also reflect HCC biology with specific HCC isoforms and GGTP levels may be a useful marker in HCC patients with low alpha-fetoprotein levels [16, 17].

Cirrhosis (simultaneous regeneration and fibrosis) is considered as a proximate cause (preneoplastic) of HCC. It has been assumed that cirrhosis and HCC, while having a precursor/product relationship, may not be directly linked after the HCC is formed, since tumors are thought of having independent growth, based upon endogenous factors such as oncogene expression, gene copy number or mutations. However, this concept has recently been modified as it has become appreciated that liver microenvironmental factors (vascularity, inflammation, other immune cells and factors, matrix changes, underlying liver molecular signatures as markers of HCC biology) may also be important in determining HCC characteristics and outcomes [18, 19]. Thus, the changes in liver function that are reflected in the liver index may either result from non-specific liver damage, caused by a growing and liver parenchyma-destroying HCC mass, or conversely, they may reflect interactions between bilirubin, albumin, GGTP, and platelets (or cirrhosis, for which platelets are a surrogate marker) and HCC cells. Albumin has already been shown to be protective against HCC growth [12, 13], while platelets can produce HCC mitogens and inflammatory cytokines [15], and GGTP has been recognized as a marker for HCC [16, 17] and may also be involved in HCC growth and drug resistance [20]. Bilirubin has been thought of as an index of residual liver function and might also relate to alpha-fetoprotein levels [18, 21].

Thus, individual components of the liver index, in addition to being reflective of the liver damage of cirrhosis, might also be involved in the characteristics of HCC biology (growth, invasion, spread). Liver function parameters might thus be involved mechanistically in the processes of HCC human biology.

In conclusion, the liver index correlated significantly with HCC aggressiveness and may thus be involved in human HCC biology and is likely another example of microenvironment influencing HCC behavior. These results confirm the previously published liver index, using a different patient dataset.