Diagnostic accuracy of clathrin heavy chain staining in a marker panel for the diagnosis of small hepatocellular carcinoma



The American Association for the Study of Liver Diseases guidelines recommend the use of all available markers for improving the accuracy of the diagnosis of small hepatocellular carcinoma (HCC). To determine whether clathrin heavy chain (CHC), a novel HCC marker, is effective in combination with glypican 3 (GPC3), heat shock protein 70, and glutamine synthetase, we compared the performances of a three-marker panel (without CHC) and a four-marker panel (with CHC) in a series of small HCCs (≤2 cm) and nonsmall HCCs by core biopsy with a 20- to 21-gauge needle. The series included 39 nonsmall HCCs and 47 small HCCs (86 in all); the latter showed a well-differentiated histology [small grade 1 (G1)] in 30 cases (63.8%). The panel specificity was analyzed with the adjacent/extranodular cirrhotic liver (n = 30) and low-grade (n = 15) and high-grade dysplastic nodules (n = 16) as a control group. Absolute specificity (100%) for HCC was obtained only when at least two of the markers were positive (which two markers were positive did not matter). The addition of CHC to the panel increased the diagnostic accuracy for small HCCs (from 76.9% to 84.3%), and there was an important gain in sensitivity (from 46.8% to 63.8%). The four-marker panel had lower rates of accuracy (67.4%) and sensitivity (50%) for small G1 HCCs versus nonsmall G1 HCCs (93.9% and 88.2%, respectively). In seven cases (including six small G1 HCCs), there was no staining with any of the markers. Cirrhotic control livers were stained for CHC in four cases (13.3%) and for GPC3 in one case (3.3%). Conclusion: The addition of CHC to the panel supports the diagnosis of small HCCs in challenging nodules on thin core biopsy samples. Small G1 HCCs include a group of earlier tumors characterized by a more silent phenotype and the progressive acquisition of the markers under study. The search for additional markers for early HCC diagnosis is warranted. (HEPATOLOGY 2011;)

See Editorial on Page 1427

Hepatocellular carcinoma (HCC) is one of the most common malignancies worldwide, and its incidence is growing in association with viral (hepatitis C virus) and nonviral (nonalcoholic steatohepatitis) chronic liver diseases.1 HCC is a lethal cancer, and improved survival relies on the detection of small (≤2 cm) and early tumors, which are less likely to have disseminated. Radiology is the main technique used to detect HCC in the setting of cirrhosis; typical imaging shows HCCs > 2 cm in more than 90% of cases. However, when the radiological features of hepatic liver nodules in cirrhosis are not typical, the American Association for the Study of Liver Diseases (AASLD) guidelines recommend the use of liver biopsy. The potential risks of bleeding and seeding due to liver biopsy can be minimized with very thin, 20- to 21-gauge needles, so this invasive technique is recommended now and will be recommended in the future for approximately 50% of dubious hepatic nodules between 1 and 2 cm in size. Indeed, the cumulative experience of several internationally recognized hepatology centers2-5 has shown that 30% of 1- to 2-cm nodules are not malignant. Recently, Leoni et al.5 have reported that the application of international guidelines, even with three imaging techniques, leads to a false-negative rate of 20% for nodules ultimately shown to be HCC, whereas Sangiovanni et al.4 have documented in a single-technique scenario that 55% of patients will need to undergo biopsy for a final diagnosis. Thus, biopsy has a very crucial role when radiology fails and pathologists are asked to document malignancies in lesions that are also the most difficult to ascertain.1 To improve the ability of pathologists to diagnose HCC, several histopathological features have been proposed; for example, stromal invasion is definitive proof of malignancy, but it is hardly detectable in a liver biopsy sample.6, 7 More recently, the translation of data from gene expression profiles to clinical practice8-12 has revealed that the combination of a panel of immunomarkers, such as glypican 3 (GPC3), heat shock protein 70 (HSP70), and glutamine synthetase (GS), may be very useful in proving malignancy and increasing the accuracy of diagnoses of small and well-differentiated HCCs in both surgical specimens13 and liver biopsy samples.6 Indeed, the use of this panel has been recently endorsed by the AASLD guidelines for the diagnostic workup of small HCCs.14 Seimiya et al.15 have found that the endothelial marker clathrin heavy chain (CHC) is overexpressed in HCCs, and it has been suggested to be diagnostically useful, particularly in combination with GPC3; however its value was tested only in surgical specimens and not in liver biopsy samples.

This work was undertaken to address two issues raised in an editorial about our previous article16: (1) testing the accuracy of HCC immunomarkers in a homogeneous series of HCCs up to 2 cm in size and (2) improving the accuracy of the panel with additional markers. To this end, we retrospectively evaluated a series of HCCs consecutively diagnosed on core biopsy samples with a 20- to 21-gauge needle; with this material, we tested the diagnostic accuracy of a refined panel of markers (CHC, GPC3, HSP70, and GS). The performance of the panel was also evaluated according to HCC grading [grade 1 (G1) versus grade 2 (G2)/grade 3 (G3)] and sizes (≤2 versus >2 cm).


3M, three-marker; 4M, four-marker; AASLD, American Association for the Study of Liver Diseases; CHC, clathrin heavy chain; G1, grade 1; G2, grade 2; G3, grade 3; GPC3, glypican 3; GS, glutamine synthetase; H&E, hematoxylin and eosin; HCC, hepatocellular carcinoma; HGDN, high-grade dysplastic nodule; HSP70, heat shock protein 70; LGDN, low-grade dysplastic nodule.

Materials and Methods

Cases Under Study.

The series under study was composed of 20- to 21-gauge needle core biopsy samples from 86 HCCs with a cirrhotic background. They were obtained from the files of the Policlinico General Hospital (Milan, Italy) and Melegnano General Hospital (Melegnano, Italy) and were collected from 2005 to 2009. The diagnosis of HCC was made in all the cases according to AASLD guidelines.17 The diagnostic process included routine laboratory tests, serum alpha-fetoprotein measurements, and abdominal ultrasound, contrast-enhanced spiral computed tomography, or magnetic resonance imaging. The diagnosis of cirrhosis was based on histology or concordant laboratory and imaging findings. The tumor size was the largest diameter measured by imaging.

The histopathological diagnosis of HCC was originally made mostly after hematoxylin and eosin (H&E) staining supplemented by routine histochemical stains such as Gomori staining for reticulin, Perls' staining for iron, and Masson trichrome staining. All the slides were preliminary revised by two expert pathologists (M.R. and L.D.T.), and the diagnosis of HCC was confirmed after accurate morphological analysis in all cases. HCC grading was based on the available material according to Edmondson and Steiner,18 and cases were divided into two groups: well-differentiated histology (G1) and moderately to poorly differentiated histology (G2/G3). The main pathological criteria for identifying well-differentiated HCCs and distinguishing them from high-grade dysplastic nodules (HGDNs) are reported in Supporting Table 1. The series included only cases with a tumor core and material available for immunocytochemical analyses (at least five recuts from the original block). Figure 1 shows a paradigmatic G1 HCC with an extralesional sample, which well represents the material under study.

Figure 1.

Representative fragments of well-differentiated hepatocellular nodules (≤2 cm) stained with H&E. In comparison with the extranodular cirrhotic parenchyma (Cirr.) on the left, the increased cell density and pseudoglands in the right fragment (marked by asterisks) are consistent with a diagnosis of well-differentiated (G1) HCC. The use of specific immunomarkers may be helpful in this fine diagnosis (see Fig. 3).

The series samples were divided according to the tumor size: small HCCs ≤ 2 cm (n = 47) and nonsmall HCCs > 2 cm (n = 39). An extralesional nontumoral sample was available in 30 cases; in addition, cases of low-grade dysplastic nodules (LGDNs; n = 15) and HGDNs (n = 16) were collected. All dysplastic nodules were confirmed to be nonmalignant because during follow-up (at least 12 months), they did not show evidence of malignant changes. The median ages, gender distributions, HCC grading, and chronic liver disease etiologies of the two HCC groups are reported in Table 1.

Table 1. Clinicopathological Features of the HCCs Under Study
 Small HCC ≤ 2 cm (n = 47)Nonsmall HCC > 2 cm (n = 39)
Age, years [median (range)]70.5 (40-85)71.5 (24-86)
Gender [n (%)]  
 Male30 (63.8)29 (74.4)
 Female17 (36.2)10 (25.6)
Size, cm [median (range)]1.6 (1-2)4.2 (2.1-10)
Grading [n (%)]  
 G130 (63.8)17 (43.6)
 G2/G317 (36.2)22 (56.4)
Etiology [n (%)]  
 Hepatitis C virus36 (76.7)23 (59.0)
 Hepatitis B virus2 (4.2)5 (12.8)
 Hepatitis C virus/hepatitis B virus1 (2.1)0
 Alcohol3 (6.4)4 (10.3)
 Other5 (10.6)7 (17.9)


Five consecutive recuts from the original paraffin blocks were obtained in all cases and were stained with antibodies against GPC3, HSP70, GS, and CHC. Supporting Table 2 details the reagents used in this study, the working dilutions, and the detection system. Immunocytochemical analysis was performed according to standard procedures. Samples stained for GPC3 and HSP70 were scored as positive when at least 10% of the tumoral population showed undisputed cytoplasmic/membrane (GPC3) or nucleoplasmic/cytoplasmic (HSP70) staining. GS immunoreactivity was scored as positive when at least 50% of the neoplastic cells showed nucleoplasmic/cytoplasmic staining. Positive controls included an HCC case as an external standard for GPC3 and nonneoplastic bile duct epithelium and perivenular nontumoral hepatocytes as internal standards for HSP70 and GS, respectively.

CHC-positive cases were considered to be those showing undisputed cytoplasmic antigen overexpression in more than 10% of the malignant hepatocytes in comparison with the surroundings, the latter being the extralesional sampling or nontumoral parenchyma adjacent to the core HCC. Nonparenchymal cells such as endothelial cells were used as internal standards for CHC staining.

The staining assessment was made by two pathologists (M.R. and L.D.T.) and was always based on antigen overexpression in the lesion versus the surroundings. These pathologists independently evaluated the specimens and applied the designated criteria. The results were then discussed between them, and concordance on agreed scores was achieved with a high k coefficient value (>0.80). To further address the issue of interobserver variability in the evaluation of CHC immunostaining, we trained both a junior pathologist and a senior pathologist with a small set of surgical specimens and liver biopsy samples. Then, the junior and senior pathologists independently evaluated all the cases of HCC and dysplastic nodules, and the results were compared to the previously agreed scores.

Evaluation of the Panel.

The specificity of the panel was evaluated through a comparison of HCC immunostaining against staining observed in the cirrhotic parenchyma adjacent to tumors (this was available for 30 cases: 14 small HCCs, 13 nonsmall HCCs, and 3 HCCs with recut material available only for the nontumoral counterpart) and in core biopsy samples of LGDNs (n = 15) and HGDNs (n = 16) obtained from the files of the aforementioned institutions and from the Department of Pathology of Istituto Clinico Humanitas (Istituto di Ricovero e Cura a Carattere Scientifico, Rozzano, Milan, Italy). In all, 61 cases of cirrhotic and dysplastic nodules were used as negative controls.

The sensitivity, specificity, and diagnostic accuracy of the panels composed of three immunomarkers (without CHC) or four immunomarkers (with CHC) were then evaluated and compared in the two HCC groups (small HCCs and nonsmall HCCs) so that we could determine whether the addition of CHC improved the diagnostic performance.

Statistical Analysis.

The data are reported as numbers and percentages or as medians and ranges. The sensitivity was calculated as the proportion of affected biopsy samples resulting in positive tests. The specificity was calculated as the proportion of unaffected biopsy samples resulting in negative tests. The accuracy was calculated as the proportion of biopsy samples that were correctly identified. Calculations were performed with Stata 10 (http://www.stata.com). Interobserver variability was assessed with the kappa index.


Preliminary Evaluation of CHC Immunostaining.

CHC immunoreactivity was preliminarily assessed in 15 neoplastic lesions (8 HCCs and 7 HGDNs) and in 5 hyperplastic lesions (focal nodular hyperplasia); all had been surgically removed. As shown in Fig. 2, CHC immunoreactivity mostly decorated the endothelial sinusoidal lining of the cirrhotic parenchyma and dysplastic nodules without hepatocyte staining; inflammatory cells within septa and the luminal surfaces of interlobular bile ducts were also stained. In a single HGDN case, CHC immunoreactivity was focally seen in neoplastic hepatocytes. In contrast, HCC hepatocytes were CHC-immunoreactive with diffuse staining (>50% of the cells) in five of eight cases and with focal staining (10%-50% of the cells) in three of eight cases. CHC immunoreactivity in malignant hepatocytes was optimally evaluated, even at a low magnification, as staining overexpression in comparison with the adjacent nonmalignant cirrhotic parenchyma. Focal nodular hyperplasia never showed CHC immunostaining.

Figure 2.

Representative surgical specimens from cases with (A,B) HGDNs (marked by double asterisks) and (C,D) well-differentiated HCCs (marked by asterisks) stained with H&E and CHC against a cirrhotic background (Cirr.). The hepatocytes of the cirrhotic and dysplastic nodules were immunonegative, whereas the hepatocytes of the well-differentiated HCCs showed CHC overexpression in comparison with the surroundings (for additional comments, see the main text).

Interobserver Variability of CHC Immunostaining.

As for the interobserver variability of the junior and senior pathologists in the evaluation of CHC immunostaining, the results showed substantial agreement between the junior (k = 0.86) and senior pathologists (k = 0.92) and the previously agreed scores for the entire set of liver biopsy samples of HCC and dysplastic nodules.

Panel Staining in Small and Nonsmall HCCs.

Figure 3 shows the immunoreactivity for GPC3, HSP70, GS, and CHC in an HCC and extralesional sample, which well represents the material under study.

Figure 3.

Case from Fig. 1 after immunostaining with the 4M panel. The overexpression of (A) GPC3, (B) HSP70, (C) GS, and (D) CHC in malignant hepatocytes (marked by asterisks) is documented in comparison with cirrhotic parenchyma (Cirr.).

Six of 47 small HCCs (≤2 cm; 12.8%) and 1 of 39 nonsmall HCCs (>2 cm; 2%) were not stained with any of the markers (Table 2). These seven unreactive cases all involved G1 HCCs. Cirrhosis control cases (n = 30) were negative for the panel in 25 cases (83.3%) and were focally positive with one marker in 5 of 30 cases (16%; 4 cases were positive for CHC, and 1 case was positive for GPC3) but never with two markers. LGDNs were positive for CHC in 1 of 15 cases (6.7%); HGDNs were positive for one marker in 3 of 16 cases (18.8%; 2 cases were positive for CHC, and 1 case was positive for GPC3), but they were never positive for two markers.

Table 2. Immunohistochemical Features of the Different Lesions From the Whole Series With the 4M Panel
Positive MarkerNonmalignant Nodule (n = 61)HCC ≤ 2 cm (n = 47)HCC > 2 cm (n = 39)
Cirrhosis (n = 30)LGDN (n = 15)HGDN (n = 16)G1 (n = 30)G2/G3 (n = 17)G1 (n = 17)G2/G3 (n = 22)
All four0004548
At least three000811914
At least two00015151521
At least one51324171622

As shown in Table 3, absolute specificity (100%) for HCC was obtained when staining with at least two markers was taken into account. With respect to the performance of staining with at least two markers in the detection of small and nonsmall HCCs, a four-marker (4M) panel with CHC was superior to a three-marker (3M) panel without CHC. In particular, a gain in sensitivity was seen for small HCCs with 4M staining (63.8%) versus 3M staining (46.8%). In small HCCs, the 4M panel showed an accuracy of 84.3%, which was superior to the accuracy of the 3M panel (76.9%).

Table 3. Diagnostic Accuracy of the 4M and 3M Panels in Nonmalignant Nodules and Malignant Nodules (≤ or >2 cm)
Positive MarkerNonmalignant Nodule (n = 61)Malignant Nodule ≤ 2 cm (n = 47)Sensitivity (%)Specificity (%)Accuracy (%)Malignant Nodule > 2 cm (n = 39)Sensitivity (%)Specificity (%)Accuracy (%)
4M panel         
 All four0919.1100.064.81230.8100.073.0
 At least three01940.4100.074.12359.0100.084.0
 At least two03063.8100.084.33692.3100.097.0
 At least one94187.285.286.13897.485.290.0
3M panel         
 All three01021.3100.065.71230.8100.073.0
 At least two02246.8100.076.92564.1100.086.0
 At least one23268.196.784.33692.396.795.0

Table 4 reports the accuracy in the detection of small G1 HCCs; HGDNs that did not transform into HCCs during follow-up were used as a negative control group. Absolute specificity was obtained when at least two markers were scored as positive, with 50% sensitivity for the 4M panel versus 33.3% sensitivity for the 3M panel and with 67.4% accuracy for the 4M panel versus 56.5% accuracy for the 3M panel.

Table 4. Diagnostic Accuracy of the 4M and 3M Panels in Well-Differentiated Hepatocellular Lesions: HGDNs Versus G1 HCCs (≤2 cm)
Positive MarkerHGDN (n = 16)G1 HCC ≤ 2 cm (n = 30)Sensitivity (%)Specificity (%)Accuracy (%)
4M panel     
 All four0413.3100.043.5
 At least three0826.7100.052.2
 At least two01550.0100.067.4
 At least one32480.081.380.4
3M panel     
 All three0413.3100.043.5
 At least two01033.3100.056.5
 At least one11653.393.867.4

Table 5 reports the performance of the 4M panel in the detection of HCCs with respect to the grade (G1 versus G2/G3) and the size (small versus nonsmall) when at least two markers or at least one positive marker was considered; HGDNs were used as a negative control group. When the staining involved two of the four immunomarkers (regardless of which ones), the accuracy of the panel was excellent in both small and nonsmall G2/G3 HCCs (93.9% and 97.4%) and in nonsmall G1 HCCs (93.9%; Table 5). In contrast, the accuracy of the same panel (two-marker staining) decreased in small G1 HCCs (67.4%) because the sensitivity of HCC detection dropped to 50%, although absolute specificity was retained. In the same group of tumors, the sensitivity and accuracy were partly restored (80% and 80.4%, respectively) when at least one immunomarker was considered (regardless of which one), but absolute specificity was not maintained (Table 5).

Table 5. 4M Panel: A Comparison of at Least Two Positive Markers and at Least One Positive Marker for HCCs Subdivided According to Their Grades and Sizes
 At Least Two Positive MarkersAt Least One Positive Marker
Sensitivity (%)Specificity (%)Accuracy (%)Sensitivity (%)Specificity (%)Accuracy (%)
G1 HCC≤2 cm50.010067.480.081.380.4
>2 cm88.210093.994.181.387.9
G2/G3 HCC≤2 cm88.210093.910081.390.9
>2 cm95.510097.410081.392.1

The performance of the individual markers in the detection of small G1 HCCs is shown in Supporting Fig. 1. In this HCC subpopulation, CHC and GS appeared to be the most sensitive markers, whereas GS and HSP70 were the most specific markers.


Pathologists today are asked to provide timely and conclusive diagnostic reports for the management and therapy of radiologically equivocal hepatocellular nodules found in small biopsy samples. Although the traditional H&E-based morphology remains the milestone, integration with biological information is required to make biopsy interpretation more objective and reproducible.

To support the morphological criteria, additional and more objective criteria of malignancy, such as stromal invasion and the composite expression of a number of tissue biomarkers (translated to clinical practice from expression studies of human hepatocarcinogenesis8-13), have been proposed. Some have already been validated, as recently emphasized by Roskams and Kojiro,19 and their use in clinical practice has been suggested in the recent update of the AASLD practice guidelines.14 Indeed, the use of these tools can make pathologists, even those not specializing in HCC, more confident in the fine diagnostics of this challenging field.

This is particularly true for small HCC, which is the most curable form and is particularly difficult to recognize with imaging. Forner et al.3 reported that concordant noninvasive imaging techniques were successful in 1- to 2-cm HCC detection in patients with cirrhosis in only 33% of cases. We previously reported that a panel of three markers was able to detect 2- to 5-cm G1 HCCs in 49% of cases (with 72.9% accuracy) when at least two of the three markers were positive.6 We conducted the present study with a homogeneous series of small HCCs (≤2 cm) and, for comparison, nonsmall HCCs sampled by a fine-needle approach (20-21 gauge) with the aim of determining whether the addition of a novel marker (CHC) to the previously validated panel could maintain or even increase the panel's diagnostic accuracy in the detection of small HCC. Notably, the series was preliminary divided into HCC cases (including small G1 HCCs) and non-HCC cases (LGDNs and HGDNs) according to the diagnosis of malignancy or dysplasia made by expert pathologists; the “uncertain for HCC” category, which could be optimally evaluated only in a prospective study, was omitted. We intentionally challenged the new panel with a retrospective series collected with fine needles (20-21 gauge) because this mini-invasive approach may minimize the risks of bleeding and seeding and thus be more acceptable in clinical practice. CHC was chosen because it is an endothelial marker, works well as an internal standard for nonparenchymal liver cells, and, as already suggested in a surgical series, is overexpressed in the cytoplasm of malignant hepatocytes.15 In contrast, most nonmalignant hepatocytes were reported by Seimiya et al.15 to be negative for staining or to have weak to moderate staining intensity. We had the same experience in our preliminary study of CHC immunoreactivity in HCC and non-HCC tissues, and we concluded that only CHC overexpression, which is optimally evaluated by a comparison to adjacent nontumoral tissue (which is mostly negative), can be taken as supportive proof of malignancy. In the same article, Seimiya et al. endorsed the use of this marker in combination with GPC3 to improve its efficacy. However, CHC has not been validated in routine core biopsy samples of HCC; this is the real diagnostic challenge for pathologists.

With the new panel, absolute specificity (100%) for HCC detection was obtained only when staining with at least two markers (regardless of which ones) was seen (66/86, 76.7%). Thus, having tried to improve the strength of the panel through the addition of a new marker, we evaluated its accuracy only in cases in which at least two out of the four markers were positive in malignant hepatocytes.

Overall, the performance of the 4M panel was superior to the performance of the same panel without CHC (the 3M panel). With staining by at least two markers, the accuracy was 97% and 84.3% in nonsmall and small HCCs, respectively, and this was superior to the accuracy of the panel without the addition of CHC (86% and 76.9%, respectively). For small HCCs, the addition of CHC to the panel consistently increased the sensitivity from 46.8% to 63.8%. Interestingly enough, for nonsmall HCCs, even though the material was sampled with 20- to 21-gauge needles, the accuracy of the novel panel (97%) was better than the accuracy that we previously reported (78.4%) with a 3M panel in an analogous HCC series sampled with 16- to 18-gauge needles.6 This means that the addition of CHC not only counterbalances the putative loss of sensitivity of thinner core materials but also increases the diagnostic accuracy. Although the use of a 4M panel is more elaborate and time-consuming for pathologists, the unitary cost of an additional immunoreaction to the panel (approximately $15-20) is much less expensive than confirmatory additional imaging4 or repeat biopsy.

When we dissected our HCC series into subpopulations according not only to size but also to grading (G1 versus G2/G3), the panel accuracy remained excellent and greater than 90% for G2/G3 HCCs, regardless of the size (Table 5). This confirmed for us that the performance of the 4M panel is optimal when tumor differentiation is compromised; in other words, the individual markers of the panel cooperatively stain HCCs that have progressed. Unfortunately, these are cases for which the pathological diagnosis can be rendered on morphological grounds without the use of staining beyond H&E. Interestingly, although the tumor size was not an issue in G2/G3 HCCs, it was a major issue in well-differentiated (G1) HCCs. Indeed, in this HCC group, which was the most difficult to evaluate, the accuracy of the panel was still excellent in nonsmall G1 HCCs (93.9%) but dropped to 67.4% in small G1 HCCs (Tables 4 and 5). In the latter, the sensitivity for HCC detection was 50% with 100% specificity, and the performance of the 4M panel was much better than that of the 3M panel (Table 4). In addition, we noticed that a consistent fraction of these tumors showed negative staining (6/30, 20%; Table 2) or one marker only (9/30, 30%; data not shown). The most likely (though speculative) explanation is that G1 HCCs greater than 2 cm and G1 HCCs smaller than 2 cm are not the same disease. An international agreement between Eastern and Western pathologists has recently been obtained for a new HCC entity: very well-differentiated, ≤2-cm HCC (which is also called very early HCC).20 This is the earliest described and well-differentiated form of HCC and is likely the morphological link between HGDN (dysplasia) and HCC that has progressed. This HCC type is very difficult to recognize on imaging because of the immature vascular supply. We believe that a consistent fraction of the small G1 HCC cases of the present series likely belong to this so-called very early type. The phenotypic profile of these cases is clearly distinct from that of other HCCs of the present series, and this provides indirect proof of an earlier disease. Indeed, the small G1 HCCs were less likely to be stained with the combination of the panel markers, their profile being intermediate between dysplasia (usually not staining) and HCC that has progressed (mostly staining). It is, therefore, reasonable to assume that when an HCC is just born, its phenotypic profile is not yet settled (e.g., the vascular support), and these markers are individually and progressively acquired and detectable. In our cases, the most represented marker in small G1 HCCs was CHC (58.8%), which was followed by GS (41.2%), HSP70 (17.6%), and GPC3 (11.8%). This means that in small G1 HCCs, CHC is the most overexpressed marker. Thus, its evaluation, particularly in tumor core biopsy samples, is important, needs attention, and requires preliminary individual training. In particular, as for all the other markers under study, its staining should decorate putative malignant hepatocytes, and it should appear as antigen overexpression in comparison with surrounding, adjacent nonneoplastic parenchymal cells. We believe that the prospective evaluation of nodules that remain diagnostically uncertain after biopsy could be very valuable for assessing the diagnostic strength of the present panel. Clearly, the search for additional and early markers has just started and is far less than completed.

In conclusion, we have shown that in core biopsy specimens of HCCs sampled with a 20-to 21-gauge needle, the addition of CHC to a panel composed of GPC3, HSP70, and GS increases the overall diagnostic accuracy in both small HCCs (from 76.9% to 84.3%) and nonsmall HCCs (from 86% to 97%), and there is an important gain in sensitivity in the detection of small HCCs (from 46.8% to 63.8%). Absolute specificity was obtained only when two of the four markers were positive (regardless of which ones). Accuracy for HCC detection was not affected by the tumor size in G2/G3 HCCs (>90%). In G1 HCCs, tumor size played a major role in discriminating cases, with higher accuracy for nonsmall HCCs (93.9%) and lower accuracy for small HCCs (67.4%); likewise, the sensitivity was 88.2% for nonsmall HCCs and 50% for small HCCs. Our results suggest that small G1 HCCs include early tumors characterized by a relatively silent phenotype and the progressive acquisition of the markers under study. The use of the present panel of markers supports the recognition of both small and nonsmall HCCs in the diagnostic pathology of challenging cases sampled by core biopsy.


The authors acknowledge the statistical expertise of Luca Zamataro, M.D., and the contributions of Tatiana Brambilla, M.D., and Bethania Fernandes, M.D., to the assessment of interobserver variability.