Prognostic Value of Serum Biological Markers in Patients with Hepatocellular Carcinoma¹ Free

The prognosis of patients with HCC³ is highly variable and is affected by both the cancer and the underlying liver cirrhosis that is present in most cases of HCC (1). Therefore, a clinical staging system for HCC should include the assessment of both residual liver function and tumor characteristics. The Okuda staging system (2) was the first to include both tumor and liver function factors. However, recently the CLIP group devised a new prognostic score, the CLIP score (3), which is based on the assessment of clinical and instrumental parameters. This score has been shown to have a better predictive power than the Okuda (2) or Child-Pugh (4) classifications and has also been prospectively validated in separate cohorts (5, 6).

In the past years, serum levels of a number of biological variables have been found to be increased in patients with HCC and associated with different outcomes. Specifically, HCC patients with serum levels of sICAM-1 <1000 ng/ml showed significantly longer survival than those with higher levels of the molecule (7). Increased circulating levels of the soluble form of sIL-2R have been demonstrated in HCC patients, and the highest levels were associated with a poorer prognosis (8). Furthermore, higher levels of IL-6 (9) and IL-10 (10) were observed in the serum of HCC patients, and increased IL-10 values were associated with a worse outcome in HCC patients undergoing surgical resection (11). Finally, the presence of serum anti-p53 antibodies in patients with HCC correlated with a shorter overall survival compared with those without anti-p53 antibodies (12).

Whether the addition of any of the above biological indicators improves the prognostic ability of clinical features of HCC patients is presently unknown. Therefore, our aim in this study was to confirm the apparent prognostic role of the serum levels of sICAM-1, sIL-2R, IL-6, and anti-p53 and to assess whether the addition of any of the above serum markers could further improve the prognostic power of the CLIP score.

All of the patients who were diagnosed with HCC at the National Cancer Institute of Naples, Italy between January 1996 and December 1998 were considered eligible for this study. Diagnosis was either cytohistological or based on AFP levels of >400 ng/ml and ultrasonographic findings consistent with HCC.

Staging procedures included contrast-enhanced computer-assisted tomography and a serum AFP measurement performed by standard RIA; AFP levels were categorized (≤400 or >400 ng/ml) according to the criteria of the CLIP (3). Data were collected regarding the HCC morphology (uninodular, multinodular, or massive) and portal vein thrombosis (absent or present). The degree of liver function was classified according to the Child-Pugh stage (4); patients without underlying liver cirrhosis were operatively categorized as stage A. Patients were further classified according to the CLIP score, which accounts for both liver function and tumor extension (3, 5) and is calculated by summing up scores assigned to four variables: Child-Pugh stage; tumor morphology; AFP; and portal vein thrombosis. The higher the total score, the worse the prognosis of the HCC patients. The choice of treatment was left to the physicians responsible for the patients’ care.

Blood samples were collected at the time of diagnosis, before any treatment was performed, and after an overnight fast. Serum was immediately separated by centrifugation, aliquoted, stored at −80°C and thawed only at the time of the assay.

sICAM-1 levels were determined by using a sandwich ELISA test kit (Cellfree ICAM-1 test kit; Endogen, Inc., Woburn, MA). Briefly, each sample was diluted 1:100 in sample diluent (a buffered solution containing bovine proteins and thimerosal); 25 μl of the diluted sample were added to 75 μl/well of sICAM-1 conjugate reagent in the plate wells that had previously absorbed anti-ICAM-1 monoclonal antibody and then incubated at room temperature for 2 h on a rotator set at 150 ± 10 rpm. sICAM-1 that was present in the sample or in the standard solutions bound to the antibody-coated wells, whereas unreactive components were removed by washing three times with wash buffer. A peroxidase substrate diluent (100 μl) was added to all of the wells, and the plate was incubated for 30 min at room temperature. The reaction produced a colored reagent that was proportional to the amount of sICAM-1 present in the sample. After the addition of a stop solution, the absorbance of each well was determined at 490 nm. The experiment was performed in duplicate; the arithmetic mean of the two determinations was considered to be the final value and expressed in ng/ml.

The levels of sIL-2R were determined by using an enzyme immunometric test kit (Milenia IL-2 kit; Diagnostic Products Corp., Los Angeles, CA). Undiluted blood samples (25 μl) were added to ligand-coated wells, along with ligand-labeled monoclonal IL-2R antibody and horseradish peroxidase-labeled monoclonal antibody specific for IL-2R. During the initial 60-min reaction, endogenous IL-2R reacted with the antibodies in the liquid phase. Then, 25 μl/well of multivalent anti-ligand were added, and the plate was rotated for 1 h on a plate mixer; the addition of this anti-ligand created a bridge between the IL-2R/anti-IL-2R complexes and the ligand-coated wells. The microplate was washed four times with a buffered wash solution to remove unreacted material. The chromogenic substrate (TMB) reactive with the enzyme label was added. After a 30-min reaction, the wells were read at 450 nm, and the absorbance was directly related to endogenous soluble IL-2R concentrations. The experiment was performed in duplicate, and the arithmetic mean of the two determinations was considered to be the final value and expressed in units/ml.

Similarly, IL-6 levels were performed by using an enzyme immunometric test kit (Milenia IL-6 kit; Diagnostic Products Corp.). Undiluted patient sample (100 μl) was added to each well, which was coated with monoclonal antibodies specific for IL-6, and incubated for 2 h. After a washing step, a horseradish peroxidase-labeled polyclonal antibody directed against another epitope of the IL-6 molecule was added. During a 2-h incubation, a sandwich complex consisting of the two antibodies and IL-6 was formed. Unreacted antibody was removed by using a washing step, and a chromogenic substrate (TMB) reactive with the enzyme label was added. The resulting color that developed after adding stop solution was read at 450 nm, and the final value was expressed in pg/ml.

Anti-p53 antibodies were detected with an ELISA test kit (Pharmacell, Paris, France) by using microtiter plates coated with recombinant wild-type human p53 protein (to detect specific anti-p53 antibodies) or with a control protein (to detect nonspecific anti-p53 interactions). A peroxidase-conjugated goat antihuman IgG bound to anti-p53 antibodies. The specific p53/anti-p53-conjugated complexes were revealed by the addition of a peroxidase substrate (TMB), resulting in a colorimetric reaction. The absorbance was read at 450 nm, and the anti-p53 levels were expressed in units/ml and categorized as positive when >0.90 units/ml and negative otherwise (13).

Nonparametric procedures were applied to test the independence between sICAM-1, sIL-2R, IL-6, anti-p53, and other baseline prognostic factors. Because of the natural ordering of polychotomous characteristics (i.e., age, Child-Pugh stage, and CLIP score), Jonckheere-Terpstra and χ² for trend tests (14) were used to investigate differences among the groups with the null hypothesis of equality of distribution being tested against an ordered alternative hypothesis. When only two categories are compared, the Jonckheere-Terpstra test reduces to Wilcoxon’s rank-sum test. Two-sided Ps are always reported.

To assess their prognostic role, sICAM-1, sIL-2R, and IL-6 were also dichotomized. For sICAM-1, we used the cutoff value of 1000 ng/ml, as reported previously (7); for sIL-2R and IL-6, we used as cutoff values the median observed values of 950 units/ml and 82 pg/ml, respectively. Survival was defined as the amount of time that elapsed between the HCC diagnosis and the patient’s death or the date of the last known follow-up information. Survival curves were drawn by using the Kaplan-Meier product-limit method (15) and compared by using the log-rank test (16). Multivariate survival analysis was performed by using the Cox proportional hazards model (17).

Because of the low number of events, four separate models were analyzed, with each model involving one biological variable, and coded as dichotomous covariates with the CLIP score as a continuous covariate. Multivariate analysis was also repeated using sICAM-1, sIL-2R, and IL-6 as continuous variables. Because the results of such analyses are equivalent to those with dichotomous variables, these data are not reported. Results are expressed as relative hazard of death with 95% CIs.

Analyses were performed by using the StatXact-Turbo (CYTEL Software Corp., Cambridge, MA) and BMDP (BMDP Statistical Software, Los Angeles, CA) statistical software packages.

During a 3-year period (1996–1998), 80 consecutive HCC patients were enrolled in the study. Baseline characteristics of the 80 patients are reported in Table 1.

Seventeen patients did not receive any kind of locoregional treatment, mostly because of poor general conditions. About one-fourth of the patients (n = 21; 26.3%) were treated with percutaneous ethanol injection, and 15 patients (18.8%) were treated with surgical resection; 10 patients (12.5%) were treated with radiofrequency interstitial tumor ablation. Only 4 patients received systemic chemotherapy, whereas 8 received transcatheter arterial chemoembolization with epidoxorubicin and lipiodol (TACE); in 5 cases, two types of locoregional treatment were combined. Mean values (SDs) of sICAM-1, IL-2R, and IL-6 were 881 (404) ng/ml, 1260 (1058) units/ml, and 107 (98) pg/ml, respectively. As shown in Table 2, distribution of both sICAM-1 and sIL-2R correlated with Child-Pugh categories, CLIP score, and portal vein thrombosis. In addition, sIL-2R correlated with age and tumor size and morphology.

No statistically significant correlation was found between IL-6 levels and all of the clinical baseline characteristics (Table 2). Similarly, anti-p53 antibodies were not affected by any of the clinical baseline characteristics (Table 3).

As of May 30, 2000, 39 patients (48.7%) had died. The Kaplan-Meier estimated median survival was 28 months, and their survival probabilities at 1-, 2-, and 3-year follow-ups were 75, 57, and 34%, respectively.

By univariate analysis, patients with low levels of sICAM-1 (≤1000 mg/ml) had no statistically significant (P = 0.10) longer survival as compared with those with higher serum levels (Fig. 1,A). Similar results were observed for IL-6 (P = 0.15; Fig. 1,B) and anti-p53 (P = 0.10; Fig. 1,C). On the contrary, patients with low serum levels of sIL-2R (≤950 units/ml) survived longer (P = 0.01) than those with higher values (Fig. 1,D). However, by multivariate analysis sIL-2R did not confirm its predictive role when tested in Cox models containing the CLIP score as covariate (Table 4). Indeed, the relative risk of death was 1.09 (95% CI, 0.51–2.32; P = 0.82) for patients with sICAM-1 values of >1000 ng/ml; 1.51 (95% CI, 0.76–3.01; P = 0.24) for patients with sIL-2R values of >950 units/ml; 1.18 (95% CI, 0.62–2.27; P = 0.61) for patients with IL-6 values of >82 pg/ml; and 1.36 (95% CI, 0.69–2.67; P = 0.37) for patients with anti-p53 antibody positivity, respectively. On the other hand, the relative risk of death increased with an increasing CLIP score, with highly significant Ps in each model (P = 0.0001, P = 0.0002, P < 0.0001, and P < 0.0001, respectively; Table 4).

Our data show that the addition of any of the four biological variables does not improve the prognostic ability of the CLIP score in HCC patients; it has been validated both prospectively (5) and in independent populations (5, 6) and performed better than the Okuda classification (5).

In the past years, a prognostic value has been claimed in HCC patients for a number of biological variables, such as sICAM-1 (7), sIL-2R (8), IL-6 (9), and anti-p53 (12). Mutations of the p53 gene in HCC tissue specimens are consistently found in a percentage ranging from 20 to 30% of patients and show a poorer prognosis compared with HCC patients not bearing p53 mutations (18, 19, 20). The use of p53 gene mutations for predicting the prognosis of HCC patients is, however, difficult in clinical practice because of the need for liver tissue samples, which are not frequently available. Thus, an attempt to use serum anti-p53 antibody as a marker of p53 mutation was made with conflicting results. In one study, anti-p53 antibodies were detected in 32% of 86 patients with HCC, which showed a shorter overall survival compared with p53 antibody-negative patients (12). The presence of p53 antibody was also an independent prognostic variable in multivariate analysis (12). In contrast, in another study on 130 HCC patients, anti-p53 serum antibodies were found in a much lower percentage of cases (7%) and did not correlate with p53 gene mutation in the liver (21). In addition, in this study patients with serum p53 antibody showed a better survival rate than patients without serum p53 antibody (21). Our results indicate that the presence of serum p53 antibody is not a predictor of a specific outcome, either better or worse, and therefore does not appear to be useful in clinical practice. Similarly, our data failed to confirm that serum IL-6 correlates with HCC patients’ stage or prognosis and are consistent with the results of Chau et al. (11).

ICAM-1 belongs to the immunoglobulin superfamily (22) and plays an important role in the regulation of immune response, particularly in the antigen-presenting mechanism. It has been shown that binding of this adhesion molecule to its physiological ligand, the lymphocyte function-associated antigen-1 expressed on leukocytes (23), activates antigen-specific CTL recognition and consequently induces cytolysis of particular target cells (24).

ICAM-1 expression has been demonstrated on the surface of malignant cells such as melanoma, adenocarcinoma, bladder cancer, and small cell lung cancer (25, 26) as well as in HCC (27). In addition, its soluble form (sICAM-1) can be detected in the serum of patients with such malignancies (28).

The prognostic role of sICAM-1 was addressed in 50 HCC patients by Shimizu et al. (7). They found a correlation between sICAM-1 values and HCC stage, a finding consistent with our observation that sICAM-1 levels are correlated directly with CLIP score, considering that the factors included in their stage definition are almost all accounted for in the CLIP scoring system (3), although some of them with a different coding. However, we failed to confirm that sICAM-1 is predictive of patients’ prognosis in univariate analysis. To explain these contrasting results, it should be noted that Shimizu et al. (7) found that sICAM-1 was an independent prognostic factor in a Cox model containing 11 covariates, which is too high a number of covariates, considering the low number of enrolled patients (n = 50) and recorded deaths (not precisely reported but apparently ∼45 from the survival curves). Although we had a higher number of patients (n = 80), the number of events in our study (n = 39) was even lower than that in the Shimizu study. Thus, we decided to perform multivariate analysis with only two covariates to allow a good performance of the statistical model. Unfortunately, sICAM-1 did not add statistically significant prognostic information to a Cox model containing the CLIP score as covariate.

IL-2 is a cytokine produced by T cells as a response to inflammatory stimuli. It induces the surface expression of IL-2 receptor and, consequently, the production of its soluble form, sIL-2R. The excess of sIL-2R is capable of binding IL-2 and causes the inhibition of an appropriate immune response (29).

Hepatocellular injury has also been associated with high levels of sIL-2R, and serum levels correlate with the histological severity of hepatocellular damage and with a high risk of transforming into HCC (30). High levels of sIL-2R have been demonstrated in the serum from patients with a variety of malignancies (31) including HCC (8). Our findings do partially support previous results. Indeed, we found that sIL-2R is significantly increased in patients with worse liver function, as measured by the Child stage, and also increases with higher CLIP scores. In addition, sIL-2R also has prognostic value in univariate analysis. However, as with sICAM-1, in a multivariate analysis with the CLIP score sIL-2R did not significantly improve the predictive power of the CLIP score.

Thus, we conclude that neither sICAM-1 nor sIL-2R is useful as a prognostic factor in clinical practice, because neither one improves the predictive ability of the CLIP score. Still, it remains interesting to investigate which mechanisms produce the correlations found because such mechanisms could be implicated in the pathogenesis of HCC.

We thank Giuliana Canzanella, Federika Crudele, Fiorella Romano, and Susanna Caiazzo for secretarial assistance.