Background and aims Conventional pathological analysis fails to achieve sufficient sensitivity and specificity for the diagnosis of hepatocellular carcinoma (HCC) in small nodules. Immunohistochemical staining for glypican 3 (GPC3), heat shock protein 70 (HSP70) and glutamine synthetase (GS) has been suggested to allow a confident diagnosis but no prospective study has established the diagnostic accuracy of this approach. The aim of this study is to assess prospectively the diagnostic accuracy of a panel of markers (GPC3, HSP70, GS) for the diagnosis of HCC in patients with cirrhosis with a small (5–20 mm) nodule detected by ultrasound screening.
Methods Sixty patients with cirrhosis with a single nodule 5–20 mm newly detected by ultrasound were included in the study. Contrast-enhanced ultrasound, magnetic resonance and fine needle biopsy of the nodule (gold standard) were performed; the biopsy was repeated in case of diagnostic failures. Three consecutive sections of the first biopsy sample with meaningful material were stained with antibodies against GPC3, HSP70 and GS.
Results Forty patients were diagnosed with HCC. The sensitivity and specificity for HCC diagnosis were: GPC3 57.5% and 95%, HSP70 57.5% and 85%, GS 50% and 90%, respectively. The sensitivity and specificity of the different combinations were: GPC3+HSP70 40% and 100%; GPC3+GS 35% and 100%; HSP70+GS 35% and 100%; GPC3+HSP70+GS 25% and 100%. When at least two of the markers were positive (regardless of which), the sensitivity and specificity were 60% and 100%, respectively. Conventional pathological analysis yielded three false negative results, but the addition of this panel only correctly classified one of these cases as HCC.
Conclusion These data within a prospective study establish the clinical usefulness of this panel of markers for the diagnosis of early HCC. However, the panel only slightly increases the diagnostic accuracy in an expert setting.
- Hepatocellular carcinoma
- glypican 3
- heat shock protein 70
- glutamine synthetase
- hepatocellular carcinoma
- hepatobiliary cancer
- abdominal mri
- liver transplantation
- liver failure
Statistics from Altmetric.com
- Hepatocellular carcinoma
- glypican 3
- heat shock protein 70
- glutamine synthetase
- hepatocellular carcinoma
- hepatobiliary cancer
- abdominal mri
- liver transplantation
- liver failure
Significance of this study
What is already known about the subject?
A conclusive diagnosis of hepatocellular carcinoma (HCC) can be obtained if a liver nodule within a cirrhotic liver displays a specific vascular profile.
A large proportion of small HCC does not meet the non-invasive diagnosis criteria and a biopsy should be requested for HCC confirmation.
In some cases it is unfeasible to obtain a reliable differential diagnosis between high-grade dysplastic nodules and very well-differentiated HCC.
The usefulness of a panel of three markers—GPC3, HSP70 and GS—has recently been proposed. However, all these studies were performed on retrospectively collected tissue samples with a conclusive diagnosis of HCC after conventional pathological analysis.
What are the new findings?
In this prospective study which included nodules <2 cm, this panel had 100% specificity when at least two of the markers (regardless of which) were positive.
In addition, the sensitivity of the panel in our cohort was comparable to that recently reported by other authors in nodules <2 cm.
However, most diagnoses were done by an expert assessment of conventional pathological analysis and this panel only slightly increased the diagnostic accuracy.
How might it impact on clinical practice in the foreseeable future?
The cornerstone for histological diagnosis of HCC should rely on an expert morphological analysis by conventional pathological analysis and the use of this panel should be reserved for reinforcing the suspected HCC diagnosis, particularly in those biopsies with scarce sampling or for less experienced pathologists in the field of liver cancer.
Hepatocellular carcinoma (HCC) is a highly prevalent and deadly neoplasia that appears mainly in patients with cirrhosis of the liver. In these patients, HCC has become the main cause of death.1 The only means of achieving long-term survival is diagnosing the disease at an asymptomatic early stage when potentially curative treatments are feasible. Surveillance for early diagnosis of HCC is therefore recommended and should rely on abdominal ultrasonography (US) every 6 months.2 The correct characterisation of the nodules detected by US is a major clinical challenge, particularly in those <2 cm. Non-invasive diagnosis by the detection of a specific vascular pattern by CT scanning or magnetic resonance imaging (MRI) is feasible, has repeatedly shown its value3–5 and is accepted in most clinical practice guidelines.2 ,6–9 Regrettably, a large proportion of small HCC does not meet the non-invasive diagnosis criteria and a biopsy should be requested for confirmation of HCC.2 Although a biopsy is usually considered the gold standard, it is also flawed by an excessive rate of false negative results. In some cases sampling is inadequate and/or not sufficiently representative to allow a confident pathological diagnosis and, in other cases, it is unfeasible to perform a reliable differential diagnosis between high-grade dysplastic nodules and very well-differentiated HCC.10 Accordingly, refinement of the pathological diagnosis of HCC at this very early stage has become a major clinical challenge. The development of validated tumour markers that could help to improve the diagnostic accuracy of pathology, allowing the diagnosis of HCC before it develops an overt malignant phenotype, is urgently needed.11
Several markers have been proposed for use in the diagnosis of HCC including glypican 3 (GPC3), heat shock protein 70 (HSP70) and glutamine synthetase (GS). GPC3 is a member of the glypican family of glycosyl-phosphatidylinositol-anchored cell surface heparan sulfate proteoglycans and has a cytoplasmic and/or membranous staining pattern. Several groups have suggested that GPC3 could be a specific tumour marker for the diagnosis of HCC,12–18 and its detection may be associated with a poor prognosis.19 HSP70 is a housekeeping gene implicated in oncogenesis, regulation of cell cycle progression and protection against apoptosis.20–22 HSP70 expression increases gradually with the evolution of hepatocarcinogenesis; its gene is significantly overexpressed in advanced HCC compared with early HCC or precancerous lesions and its immunoreactivity is reported in most types of HCC including early and well-differentiated forms, but not in non-malignant nodules.23 HSP70 immunoreactivity is nucleocytoplasmic and mostly focal. GS catalyses the synthesis of glutamine, which is a source of energy for tumour cells. Accumulation of GS was first found by analysing increased ubiquitinated protein in HCC, and its stepwise increase in expression from precancerous lesions to early and advanced HCC was shown by immunohistochemistry.24 It is noteworthy that GS has been reported as a target of β-catenin signalling, which is implicated in the development of HCC.25 ,26 GS immunoreactivity is shown to be increased from precancerous lesions to early or advanced HCC.24 To increase its specificity as a marker of malignancy, GS immunostaining has to be diffuse and with strong intensity, a pattern that can be seen in 50% of HCC and in early forms as well.27
Di Tommaso et al recently reported the usefulness of a panel of these three markers in surgically removed nodules.27 They showed that when at least any two of the three markers were positive, the specificity for HCC diagnosis was 100% with a sensitivity of 73%. More recently, the same authors have externally validated their results in samples obtained by percutaneous biopsy28 and improved the diagnostic accuracy by adding clathrin heavy chain to the panel.29
Regrettably, all these studies were done in retrospectively collected tissue samples with a conclusive diagnosis of HCC after conventional pathological analysis. These biopsy samples therefore may not represent cases of poor sampling or with an overlapping pattern that may be difficult to characterise correctly. Moreover, it can be argued that, in those cases, conventional pathological analysis theoretically has an absolute sensitivity and specificity and therefore this panel of markers cannot be considered as superior to conventional pathological analysis. Validation of this panel should preferably be performed in tissue samples prospectively recruited and correctly followed until a confident diagnosis is obtained.30
Since November 2003 we have been prospectively recruiting asymptomatic patients with Child–Pugh A–B cirrhosis with no history of HCC in whom a new solitary well-defined solid nodule ≤20 mm was detected by ultrasound screening. Repeated biopsies were performed until the diagnosis was confirmed and, in those with no HCC, a close imaging follow-up was done to assure the absence of incipient HCC without an overt malignant phenotype.
The present study was conducted to validate the diagnostic accuracy of this panel of markers (HSP70, GPC3 and GS) in a large series of liver biopsies of nodules <2 cm prospectively obtained in our institution. In addition, we compared the accuracy of the panel with the final diagnosis to assess the additional benefit of this panel over conventional pathological analysis for diagnosing HCC at a very early stage.
Materials and methods
Between November 2003 and August 2006 we prospectively included patients with cirrhosis without a prior diagnosis of HCC in whom a single nodule of 5–20 mm was detected by ultrasound screening. Patients with severe comorbidities, poor liver function or candidates for liver transplantation even without a diagnosis of HCC were excluded because a diagnosis of HCC would not have changed the clinical decision/treatment. Patients with severe alterations of the coagulation parameters or contraindications for fine needle biopsy (FNB) were also excluded. Contrast-enhanced ultrasound (CEUS), magnetic resonance (MR) and FNB of the nodule were performed in all patients.
A diagnostic algorithm, explained in detail elsewhere,3 was applied. Briefly, upon detection of the target small solitary nodule, patients were examined by dynamic MR and CEUS with second-generation contrast agent and then submitted to FNB. The biopsy result was considered the gold standard. If a conclusive diagnosis of HCC by pathological analysis was not achieved, a second FNB was done. If the report was again negative for malignancy and did not produce a conclusive diagnosis, the policy was decided according to the imaging pattern. If any imaging technique showed evidence of arterial hypervascularisation, a third FNB was indicated. All other cases and those with a third negative biopsy for malignancy were followed-up with CEUS every 3 months and MR every 6 months. If growth or hypervascularisation was detected, a new FNB was performed.
Fine needle biopsy
FNB was performed using a 20-gauge spinal needle (Yale Spinal; BD Medical, New Jersey, USA). Several back and forth passages were done after insertion of the needle. When technically feasible because of location and accessibility, a core biopsy was additionally performed using an 18-gauge needle biopsy (Monopty; Bard Inc, Covington, UK).
Specimens were routinely processed and stained with H&E. Stains for reticulin and CD34 were applied when necessary. Diagnosis of HCC was in all cases based on pathology and was made according to the International Working Party criteria.31 Tumour grading was assessed on H&E- stained sections according to WHO criteria.32 Immunohistochemistry staining samples were independently registered by two authors with more than 10 and 20 years of experience in the assessment of liver biopsies (LB and MS). We report the diagnoses after conventional pathological analysis of the stained samples and the final diagnoses obtained after repeated biopsies.
Immunohistochemistry (IHC) was performed on serial 4 μm sections of the formalin-fixed paraffin-embedded first biopsy with meaningful material using specific monoclonal antibodies to GPC3 (Cat# B0134; BioMosaics, Burlington, Vermont, USA), HSP70 (Cat# ab5442; Abcam, Cambridge, UK) and GS (Cat# 610517; BD Transduction Laboratories, Franklin Lakes, New Jersey, USA). Sections were dehydrated and treated with hydrogen peroxide/methanol for 10 min at room temperature (RT) to quench endogenous peroxidase activity. To perform antigen retrieval, sections were placed in citrate buffer (pH 6.0, 0.01 M) and boiled in a pressure cooker for 10 min (for GPC3) or 3 min (for HSP70 and GS) and allowed to cool for 20 min at RT. After antigen retrieval, non-specific antibody binding was blocked by incubating the slides with UltraPlus Ab Diluent (Cat# TA-125-UDX; Thermo Scientific, Fremont, California, USA) for 30 min at RT. Incubation with GPC3 (dilution 1:100), HSP70 (dilution 1:100) and GS (dilution 1:200) was carried out for 1 h at RT followed by a 30 min incubation with anti-mouse secondary antibody-labelled polymer (Cat# K4000; EnVision+ System-HRP, Dako). Diamino benzidine (DAB, Cat# K3468, Dako, Glostrup, Denmark) detection was followed by Mayer's haematoxylin (Cat# MHS1; Sigma, Sant Louis, MO, USA) nuclei counterstaining.
Staining for GPC3 (cytoplasmic staining) and HSP70 (nucleocytoplasmic staining) were classified blindly as positive in samples with more than 5% immunoreactive lesional hepatocytes. External positive controls were always included in the batch of slides. GS was considered positive when staining was diffuse and unrelated to vascular areas. Non-lesional immunoreactive pericentral hepatocytes were used as an internal control. Figures 1–3 show examples of the staining with the three markers.
Data are shown as number (%) or median (range) as appropriate. Comparison between patients with HCC and patients with non-HCC nodules was done using the Student t test or the Mann–Whitney test for continuous variables and the χ2 test/Fisher exact test for categorical variables. A p value of 0.05 was considered statistically significant. Sensitivity was calculated as the proportion of HCC biopsies resulting in positive tests. Specificity was calculated as the proportion of non-HCC biopsies resulting in negative tests. The positive predictive value was calculated as the proportion of positive tests that correctly identified HCC lesions. The negative predictive value was calculated as the proportion of negative tests that correctly identified non-HCC nodules. The likelihood positive ratio was calculated as sensitivity/(1 − specificity) and the likelihood negative ratio as (1 − sensitivity)/specificity. Diagnostic accuracy was calculated as the proportion of correctly identified biopsies of the total of biopsies. Calculations were performed using the PASW Statistical Package V.18 (SPSS Inc).
Characteristics of patients
A total of 60 patients with liver cirrhosis were included. In 47 cases the first biopsy was analysed. In 11 cases the second biopsy was used due to insufficient material in the first one. Only in two cases was a third sample used. In cases in which we used the second or third biopsy sample, the median interval between the first and the analysed biopsy sample was 1.7 months. Their characteristics are summarised in table 1. In 33 cases the sample was obtained by FNB, in five cases by cutting needle biopsy and, in the remaining 22 patients, the sample was obtained by both FNB and core biopsy. The median age was 64 years and in most cases the cirrhosis was due to infection with hepatitis C virus (n=47, 78.3%) with preserved liver function (Child–Pugh class A=52, 86.7%). Median baseline alpha-fetoprotein (AFP) was 6.5 ng/ml (range 1–1154). The median size of the nodules was 14 mm (range 7–20); 6 (10%) were ≤10 mm, 30 (50%) were 11–15 mm and 24 (40%) were 16–20 mm. Table 2 summarises the final diagnosis of the 60 nodules included in the study. At the end of the follow-up period in May 2010 the final diagnoses were: HCC in 40 cases (66.7%), which was well differentiated in 19 cases (31.7%), moderately differentiated in 17 cases (28.3%) and poorly differentiated in four cases (6.7%); regenerative nodules in 19 cases (31.7%); and high-grade dysplastic nodule in one case (1.7%). Patients with non-malignant nodules were followed up for a median of 526 days (range 134–1145) to ensure their benign nature. Three HCC nodules were initially classified as non-HCC by conventional pathological analysis of the analysed specimen but the HCC diagnosis was confirmed by a repeated biopsy. Twenty-five of the 40 HCC nodules and none of the non-HCC lesions showed contrast arterial enhancement followed by washout with MRI.
There were no significant differences between patients diagnosed with HCC and those with non-HCC nodules. The median AFP was 9 ng/ml (range 1–1154) and 5 ng/ml (range 2–128) in the HCC and non-HCC cohorts, respectively (non-significant difference). HCC nodules were significantly larger than non-HCC nodules (median 15.5 mm and 12 mm for HCC and non-HCC lesions, respectively; p=0.008).
Diagnostic accuracy of the panel
The immunohistochemical features of the nodules included in the analysis are summarised in table 3. GPC3 immunoreactivity was seen in 23/40 HCC nodules and in 1/20 non-malignant nodules (a 7 mm regeneration nodule); overall, the sensitivity, specificity and positive and negative predictive values of GPC3 for HCC detection were 57.5%, 95%, 95.8% and 52.8%, respectively. HSP70 immunoreactivity was seen in 23/40 HCC nodules and in 3/20 non-malignant nodules (regeneration nodules, 12, 18 and 19 mm); overall, the sensitivity, specificity and positive and negative predictive values of HSP70 for HCC detection were 57.5%, 85%, 88.5% and 50.0%, respectively. Finally, GS immunoreactivity was seen in 20/40 HCC nodules and in 2/20 non-malignant nodules (regeneration nodules, 11 and 12 mm); overall, the sensitivity, specificity and positive and negative predictive values of GS for HCC detection were 50%, 90%, 90.9% and 47.4%, respectively.
The different combinations of two of these positive markers are summarised in table 3. The combination GPC3+HSP70+GS (all positives) was seen in 10/40 HCC nodules but in none of the non-malignant nodules. Accordingly, the sensitivity, specificity and positive and negative predictive values for this combination were 25%, 100%, 100% and 40%, respectively. When at least two of these markers were positive (regardless of which), the sensitivity, specificity and positive and negative predictive values were 60%, 100%, 100% and 55.6%, respectively. Table 4 shows the performance of the panel in the detection of HCC nodules with respect to the grade (G1 vs G2/G3) when at least two markers or at least one positive marker was considered. As expected, the diagnostic accuracy was slightly lower in G1 HCC nodules; the sensitivity for detecting G1 HCC was 52.6%, leading to a diagnostic accuracy of 77%.
Diagnostic gain using the immunohistochemical panel
The conventional pathological analysis of the 60 samples diagnosed 37 of 40 cases of HCC, which translates into a false negative rate of 7.5%. The main tumour characteristics of the three HCC nodules not identified by conventional pathological analysis are summarised in table 5. Using a combination of the markers and applying the criteria of at least two positives, we would have reached a conclusive diagnosis of HCC in only one case previously considered to be a non-HCC lesion by conventional pathological analysis.
A confident diagnosis of cancer is a critical step prior to treatment. Early small HCC nodules are usually composed of well-differentiated hepatocytes,11 and this turns a confident diagnosis by examination of FNB samples into a pathological challenge. Their reading requires major expertise and, even so, it is usual to assume a high rate of false negative reports.3 Morphological criteria alone still pose problems for the differential diagnosis of high-grade dysplastic nodules versus early HCC, especially in needle biopsies where stromal invasion, one of the most reliable features of malignancy, can be absent. For this reason, the development of reliable tumour markers is urgently needed.2 ,11
In the last few years, several tumour markers have been suggested, most of which were initially evaluated in explants.12–18 20–22 ,24 ,33 In these specimens there are usually no diagnostic challenges as the pathologist is able to analyse the whole tumour, looking for the specific histological features for a confident diagnosis of HCC. Accordingly, the diagnostic accuracy obtained in these studies cannot be extrapolated to tissue obtained by percutaneous biopsy where the sampling is scarce and not always representative of the tumour. In addition, studies evaluating the diagnostic capability of any technique including patients with already diagnosed HCC are at risk of selection bias since atypical specimens without a confident diagnosis are systematically excluded.
Our study was intended to externally validate the use of a panel of markers (GPC3, HSP70 and GS) in samples obtained by percutaneous biopsies from nodules <2 cm. This panel showed an absolute specificity when initially assessed in surgically removed nodules,27 and the same authors were able to confirm its utility in two retrospective studies evaluating tissue obtained by percutaneous biopsy.28 ,29 We have intentionally used the same reagents and evaluation criteria to make the comparison reliable.
The most relevant finding was that this panel of markers showed absolute specificity when at least two markers were positive (regardless of which), and therefore validates the results reported by Di Tommaso et al. In addition, the sensitivity of the panel in our cohort was comparable to that recently reported by Di Tommaso et al in a recently published paper including nodules <2 cm.29
Surprisingly, we also found that GPC3, HSP70 and GS were expressed in non-HCC nodules. This has been extensively shown with GPC3,12–18 27–29 ,33 particularly when weak staining is categorised as positive,16 and with HSP70.27–29 ,33 In addition, there is no biological basis to expect that the expression of these markers in non-neoplastic liver would be mutually exclusive, and a false positive combination of two of them is theoretically possible. Accordingly, expert evaluation of the immunostaining is also needed to avoid misinterpretation.
The value of our study is that the tissue samples were obtained prospectively. This is relevant because other studies were performed in tissue retrospectively retrieved in which the pathological diagnosis was confidently made, so samples with scarce material or non-representative tissue are systematically discarded. Regrettably, it is in these cases when a reliable panel of markers is needed for improving the diagnostic accuracy of biopsy in very early lesions. We included the first biopsy sample with meaningful material from 60 patients with cirrhosis of the liver consecutively recruited in whom a new solitary nodule <20 mm was detected by surveillance ultrasound. Moreover, we compared the immunostaining profile with the final diagnosis obtained after a strict period of follow-up. This is extremely important for nodules finally categorised as non-malignant because a median follow-up period of 17.3 months allows us to be sure of the absence of undetected HCC.
In only one case would a final diagnosis of HCC have been reached at the first biopsy using the criteria of two positive markers. Most diagnoses were done by an expert assessment of conventional pathological analysis and this panel only slightly increased the diagnostic accuracy. Furthermore, in 25 HCC nodules the imaging was conclusive for HCC and therefore, according to the updated AASLD recommendations,2 biopsy had been requested in 15 cases, and only one of the 40 HCC nodules with neither imaging nor conventional pathological diagnosis could be effectively diagnosed by this panel. The cornerstone for a histological diagnosis of HCC is therefore still conventional pathological analysis, and the gain in diagnostic accuracy achieved by the use of this panel is minimal. Accordingly, we should emphasise that this panel cannot be used as a substitute for expert morphological analysis by conventional pathological analysis. The use of this panel should always be carefully considered and dictated by morphology, and its role should be only to reinforce a suspected diagnosis of HCC, particularly in biopsies with scarce sampling.
Our study has some limitations. First, the sample size was small, but it should be kept in mind that they were all patients with cirrhosis in whom a new solitary nodule <20 mm was detected by screening ultrasound, and a well-defined diagnostic investigation was conducted and a long prospective follow-up was performed in those with a non-conclusive HCC diagnosis. For these reasons, our study is unique and delivers novel and valuable information.30 Second, only one out of 20 non-malignant nodules was categorised as dysplastic, and the main diagnostic difficulties appear in high-grade dysplastic nodules. However, the distinction between regenerative and low-grade dysplastic nodules is almost impossible in samples obtained by FNB, therefore an important proportion of our lesions would be categorised as dysplastic by other pathologists.
In summary, our results validate the diagnostic accuracy of this panel of markers and confirm a specificity of 100% when at least two markers (regardless of which) are positive. However, the cornerstone for histological HCC diagnosis should rely on an expert morphological analysis by conventional pathological analysis and the use of this panel should be reserved for reinforcing a suspected diagnosis of HCC, particularly in biopsies with scarce sampling or for pathologists who are less experienced in the field of liver cancer.
The content of this work was presented in abstract form at the ILCA 5th Annual Conference in September 2011 and the 46th Annual Meeting of the European Association for the Study of the Liver (EASL) 2011.
ST and AF contributed equally to this work.
Funding This study was supported by grants from the Instituto de Salud Carlos III (PI 06/132 and PI 08/0146). CIBEREHD is funded by Instituto de Salud Carlos III. ST was partially supported by a grant from the BBVA Foundation. MR was partially supported by a grant from the University of Barcelona (APIF RD63/2006). CR-L was supported by a grant of the Instituto de Salud Carlos III (FI09/00510).
Competing interests None.
Ethics approval This study was approved by the Ethics Committee for Clinical Research, Hospital Clínic Barcelona, Spain and written informed consent was obtained from each patient.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.