SCCA was quantified in matching sera and tissues of 82 patients, 27 LC and 55 HCC. In these latter patients tissue expression of SCCA was investigated in neoplastic and paired peritumoral tissue. Patients were stratified according to nodule size. The mean (95%CI) size of the HCC lesion was 4.07 ± 2.08 cm; 36.4% (20/55) patients had a single nodule smaller or equal to 3 cm (s-HCC) and 63.6% (35/55) a nodule larger than 3 cm or multifocal (l-HCC). A statistically significant difference (p = 0.01) was observed in the percentage of Child-Pugh stage A among s-HCC (55%), l-HCC (82%) and LC (88%). The most frequent etiology was HCV alone or with HBV, with no significant difference among groups (p = 0.108): 65% (15/20) in s-HCC, 71.4% (25/35) l-HCC and 92.6% (25/27) for LC patients (Table ).
Characteristics of the patients
SCCA was detected by immunohistochemistry in all the patients, although with some differences, in both tumoral and peritumoral tissues of small and large HCC (Figure ). In particular, the mean (95% CI) expression in neoplastic tissue was 1163.2 (863.6–1566.8) μm2 in s-HCC and 625.8 (534.5–732.6) μm2 in l-HCC. There was a statistically significant difference among the groups (F = 17.45, p = 0.002) and Tukey grouping showed a significant difference (p < 0.05) between s-HCC vs l-HCC (Figure ). To further investigate the tissue expression of SCCA, paired peritumoral tissues were analyzed as well as LC samples used as proper control. The mean (95% CI) of SCCA tissue expression was 263.8 (176.6–394.01) μm2 in s-HCC peritumoral tissue, 345.49 (263,98 – 452,16) μm2 in l-HCC and 232.63 (163.3–331.38) μm2 in LC, Figure . The model did not result statistically significant (F = 1.84, p = 0.16). In conclusion, SCCA was more strongly expressed in the neoplastic tissue of smaller compared to larger HCC. In addition, it is noteworthy that the ratio between tumoral and peritumoral SCCA shows a trend ranging from 4.4 in s-HCC to 1.8 of l-HCC.
Figure 1 Tissue expression of SCCA. In Figure 1A, immunohistochemistry of SCCA in small and large HCC at low (A, C) and high (B, D) magnification. Paired peritumoral tissue at low magnification of small (E) and large (F) HCC. In A, C, E and F scale bar = 100 μm, (more ...)
95% Confidence intervals of non neoplastic peritumoral tissue SCCA in the different HCC patients groups.
In the same patients, we measured the serum concentrations of SCCA. As reported in Figure , concentrations were 1.6 (1.02–2.6) ng/ml in s-HCC, 2.2 (1.28–2.74) ng/ml in l-HCC, 0.41 (0.31–0.55) ng/ml in LC patients (Figure ). The model resulted statistically significant (F = 20.81, p < 0.0001), and Tukey grouping displayed a significant difference (p < 0.05) between each HCC group and LC. In conclusion, serum SCCA levels were similar in the different HCC groups, but statistically different in the LC group.
95% Confidence intervals of serum SCCA in the different HCC patients groups.
The underlying liver disease does not seem to affect SCCA tissue expression levels among patients with different Child-Pugh stages (p = 0.5) as shown by the generalized linear model, while the only significant factor remains the diagnosis of HCC as compared to LC (p = 0.049).
The regression model with dependent peritumoral SCCA shows that in the subgroup of single nodule HCC, nodule size was not statistically significant as a means of predicting peritumoral SCCA (b = 0.064; F = 1.55; p = 0.219), whereas it was a statistically significant predictor of tumoral SCCA (b = -0.099; F = 5.47; p = 0.024; R2 = 0.117).
Linear regression between serum SCCA and nodule size did not show a statistically significant relation between these two variables (b = 0.023; F = 0.14; p = 0.71; R2 = 0.003).
In conclusion, tumoral SCCA depends on nodule size, and there was an inverse trend between nodule size and higher SCCA values. Peritumoral and serum SCCA, instead, do not show any relation with tumor size.
The area under the ROC curve for serum SCCA was 0.897 (CI95% 0.81–0.953), with a suggested cut-off value of 1.1 ng/ml, showing 72.7% sensitivity, 100% specificity and a 72.7%Youden index. Assessing only s-HCC and LC patients, an analogous accuracy of serum SCCA was obtained: AUC 0.873 (CI95% 0.743–0.952), cut-off 1.1 ng/ml, with 70% sensitivity, 100% specificity and 70% Youden index.
However, in HCC no statistically significant correlation was observed between SCCA levels present in the tumoral and/or peritumoral tissue and in the serum (Table ). Pearson's correlation coefficient resulted statistically significant only to evaluate relations between serum levels vs cirrhotic liver tissue expression of SCCA in LC patients: 0.39 (p = 0.04). In other words, the higher the tissue values the higher the serum values, but only in LC patients.
Pearson correlation coefficient for evaluation of relation between markers in HCC and LC patients
Finally, serum SCCA values could be expressed as a function of tumoral or peritumoral values of this marker, so a regression analysis was performed; results are shown in Figure . The models evaluated did not result statistically significant except in LC patients (F = 5.02, p = 0.034). In this group the slope resulted 0.33 (p = 0.034), suggesting increased SCCA serum values with stronger tissue expression. In HCC patients, neither SCCA tumoral nor peritumoral tissue expression resulted predictive of the serum levels, and this finding was consistent in both groups of patients. Furthermore, in smaller nodules we calculated that 711 μm2 of SCCA antigen were necessary to measure one unit in the serum, while in larger tumors only 268 μm2 were enough.
Scatter plot of serum SCCA in tumoral and peritumoral tissue in HCC patients, and tissue SCCA in LC patients. Ln = natural logarithm.