Hepatocellular adenoma management and phenotypic classification: The Bordeaux experience^†

Patients.

From January 1984 to October 31, 2008, 135 cases were retrieved from our pathological files (eight before 1990, 36 from 1990 to 1999, and 91 from 2000 to October 31, 2008). Since 2006, immunohistochemistry was performed routinely.4 Thereafter, ancient cases have been retrospectively analyzed. Briefly, taking into account good genotype/phenotype correlations, we classified our cases as follows: (1) nodules lacking liver fatty acid binding protein (LFABP) expression as a presumably HNF1α-mutated subtype; (2) nodules with aberrant nuclear β-catenin expression as presumably β-catenin–mutated; and (3) nodules expressing inflammatory proteins (serum amyloid A [SAA] and/or C-reactive protein [CRP]) as IHCAs. Some of these IHCAs also expressed aberrant nuclear β-catenin; therefore, this group was identified as presumably β-catenin–mutated IHCAs; nodules without any of these above mentioned phenotypic markers were reported as unclassified HCAs. Finally, in six cases, nodules were not suitable for immunohistochemistry analyses because of massive hemorrhage and/or necrosis. In 80% of the cases where genotyping had been performed, the immunohistochemical classification was confirmed.

Methods.

A total of 135 cases constituted the basis of this study. The gold standard for the diagnosis of HCA is the pathological examination of the resected specimen using classical criteria. We used immunohistochemical markers such as LFABP, SAA, CRP, glutamine synthetase, and β-catenin for diagnosis purposes, as well as for the classification of HCA (see below). For differential diagnosis with focal nodular hyperplasia (FNH), we sometimes used other antibodies such as anti-cytokeratin 7 (CK7) and anti-CK19 to highlight the characteristic ductular reaction at the interface between parenchyma and fibrotic bands, and analyzed overall glutamine synthetase staining, which exhibited the characteristic distribution pattern of FNH.5 For the differential diagnosis with HCC, we used classical criteria and, if necessary, glypican 3 and MIB-1, particularly when HCA was associated with HCC.

Clinical, surgical, radiological, and pathological data were reviewed and analyzed. The following items were collected (Table 1): age, sex, pregnancy, drugs, contraception, disease known to be related to HCA, body mass index (BMI), mode of discovery (hemorrhage/shock, abdominal pain, discomfort, mass, abnormal liver function tests, by chance), as well as overt consumption of alcohol (more than 30 g/day for women and 60 g/day for men) and tobacco (≥10 cigarettes/day for women or ≥1 pack/day for men for at least 5 years).

We recorded the number of HCAs (or potential HCAs) vizualized by imaging and classified them as solitary or multiple (≤5 or >5) hepatocellular nodules without absolute distinction between HCA, FNH, or HCC, as well as the size of the largest HCA. We arbitrarily defined a 1 cm cutoff to identify nodules by imaging. We also recorded the year and type of surgery. On resected specimens, we searched for hematoma (intranodular and extranodular, rupture of Glisson's capsule), hemorrhagic necrosis, peliosis, sinusoidal dilatation/telangiectasia, and steatosis. In the nontumoral liver, when available, we evaluated steatosis (>20%) and the presence of undetected additional nodules (HCA, FNH, hemangioma). Immunohistochemistry was then used to classify newly discovered nodules when necessary. In six cases, we did not have access to the whole liver but to paraffin blocks only (first pathological examination outside our department).

Follow-Up.

Imaging data reports (ultrasound and/or computed tomography scan, magnetic resonance imaging) were reviewed. The surgical cases performed after October 2007 were not analyzed for the follow-up; however, number and size of nodules <1 cm were recorded, because imaging techniques were recent. Only two patients were lost for follow-up and this occurred shortly after surgery. Finally, among the 135 cases, seven cases with rare or specific conditions were analyzed separately. Statistics were performed in the large series of 128 cases.

Statistics.

All P values < 0.05 were considered statistically significant. Pearson χ² test was used to assess the possible correlation between the HCA subtype (IHCA versus LFABP) and the following criteria: (1) BMI > 25, (2) tumoral peliosis, (3) tumoral steatosis, (4) nontumoral liver steatosis, (5) alcohol exposure, (6) risk of bleeding, (7) association with FNH, and (8) recurrence after complete resection.

Global Analysis (128 Cases)

The cases included 116 women (106 taking OCs of whom 100 had been for more than 2 years) and 12 men (Table 1). None in this group had gross evidence for diseases specifically linked to HCA or adenomatosis such as glycogenosis, etc. The median age was 41 years. Among women, 86 had children, 30 were smokers, and six drank alcohol in excess. Among men, nine drank alcohol in excess and five were heavy smokers. One woman had surgery in two steps and another had two operations. Forty-two women and seven men showed a BMI > 25. The most common mode of discovery was bleeding, abdominal pain, or by chance. Two patients were hospitalized in internal medicine with fever and an inflammatory syndrome with anemia (one case has been previously published6).

Seventy-eight nodules were solitary and 50 were multiple (38 ≤ 5 cm, 12 > 5 cm). The median size of the largest nodule was 7 cm. In 94 cases, the size of the excised nodules exceeded 5 cm. The great majority of patients had a laparotomy, and 46 had a major hepatectomy. Resection was complete in 95 cases. In one case, a solitary large HCA could not be completely excised. A total of 220 nodules were resected in the 128 patients of the cohort. Nodule pathology differed from one case to another. Considering the global analysis, 57.8% of HCAs were steatotic and 30.5% had peliotic areas. Hematoma and hemorrhagic necrosis was observed in each subgroup, globally in 14.8% and 17.2%, respectively. Cytological abnormalities could be found near necrotic areas, but were mainly obvious in the β-catenin–activated subgroup. HCC or borderline lesions between HCC and HCA were also found only in the latter subgroup. There were five HCCs (all in women: in three cases HCC was associated with HCA; in one case there were multiple HCCs on the resected specimen, and in this patient the size of the main nodule decreased after stopping Danazol; in one case HCC was discovered several years after surgery [see below, case 1]) and one borderline lesion in a man. In addition to HCA > 1 cm, microadenomas (<1 cm, ≥1 mm) were found in 32 cases. The volume of nontumoral liver was highly variable from one case to another, and it was totally absent in 21 cases. Additional FNH and hemangioma were resected in 22 and 12 cases, respectively.

The samples displayed the following characteristics upon immunohistochemical analysis: LFABP-negative (n = 46), SAA/CRP-positive (n = 59), SAA/CRP-positive and β-catenin–positive (n = 9), β-catenin–positive (n = 7), and without specific marker (n = 6). In six cases, the hemorrhage or necrosis was too extensive to rely on the results. Eighty-two cases with complete resection were followed for more than 1 year (median = 82 months; range = 12-245 months). In 71 cases, there were no new nodules. HCAs or nonidentified nodules were observed in eight and three cases, respectively. Among the 27 cases with incomplete resection (nodules left in surrounding liver) during follow-up (median = 49 months; range = 12-163 months), nodules were no longer visible in five cases and were still present in 22 cases. In the great majority of cases, the size remained stable. Fourteen female patients gave birth after surgery (11 after complete resection of their HCA). Two patients are presently pregnant. A woman (case 1) with incomplete resection of one single large HCA died of HCC 12 years after surgery of her HCA. The retrospective analysis of this case showed it was corresponding to a β-catenin–activated IHCA. In our series, four patients died, two of HCC (see below) and the two other deaths being unrelated to HCC.

Analysis According to Subgroups

Rare and Specific Etiologies

There were five men (one child) out of seven cases (Table 3). Among the etiologies known to be linked to HCA, there were three glycogenosis (two underwent transplantation; one did not undergo transplantation with HCC), one had familial adenomatous polyposis coli, one child was treated for aplastic anemia with Danazol, and one who had a portacaval shunt in infancy had hepatoportal sclerosis. Finally, one case (a woman) with Cushing syndrome was classified in this group; she had a β-catenin–activated HCA and died of HCC 5 years after surgery.

Immunohistochemical Classification/Molecular Analysis

Immunohistochemical classification is based on the correlation between genotype and phenotype. The correspondence between pathological description and immunophenotyping classification was highly concordant most of the time, particularly for the very homogeneous LFABP-negative subgroup with marked steatosis and lobulated contours, as well as for the SAA/CRP-positive HCA subgroup which exhibited inflammatory infiltrate, telangiectasia, thickened arteries, and more or less obvious ductular reaction, as previously described.4 However, if the LFABP-negative phenotype is excellent for correlation with HNF1α inactivation, it does not identify HNF1α constitutional (MODY3) versus somatic mutations. We have to also highlight that some characteristic HNF1α-inactivated HCAs may be difficult to identify by immunohistochemistry when the expression of LFABP in the background liver is low, resulting in a low LFABP ratio in the tumor versus surrounding liver. Another difficulty can occur for some typical IHCAs when the expression of SAA or CRP is faint inside the tumor with little contrast with the surrounding liver or when the nontumoral liver may express SAA or CRP after massive bleeding or portal embolization (unpublished data).

Global Analysis (128 Cases)

There is a strong relationship between the occurrence of HCA and use of OCs. The best arguments are the time concordance between the beginning of OC marketing and the first publications about HCA, and the almost absence of reported cases in Japan where the use of OC pills is limited to medical indications and not for contraception. The number of HCAs has not decreased as predicted with the use of third-generation OCs that contain less estrogen. The increasing use of imaging techniques probably explains in part the lack of expected decline. This study does not bring new data regarding age, predominance of females, the role of OCs, or the rate of complications of either hemorrhage or HCC.

The Phenotypic Classification (Major Subgroups)

This study confirms that LFABP-negative HCA and IHCA are the two main subgroups which differ on clinical and pathological grounds.4 Schematically, LFABP-negative HCAs are more steatotic, and IHCAs are more peliotic and occur more often in overweight patients with liver steatosis. It also confirms that β-catenin–activated HCAs are associated with the risk of HCC.2 The individualization of the two main subgroups, which represents more than 80% of the cases, helps to provide better interpretation of the results than the global analysis.

Comparison between the groups revealed the following interesting data:

Phenotypic Classification (Minor Subgroups)

The other phenotypic subgroups were small. It is, however, important to outline that β-catenin activation (present also in the IHCA subgroup) represents a major potential risk of HCC.2 Unfortunately, we do not have the tools to solve this issue without a liver biopsy yet. The issues that remain unclear concern unclassified HCAs without known phenotypic markers. Thus far, this group appears independent. One patient in six had an elevated BMI. Two patients with a BMI near 25 had liver steatosis. Finally, the unexplored hemorrhagic/necrotic HCA could belong to any of the above groups. It is likely that some cases of this necrotic subgroup belong to the IHCA group; indeed, three cases had a raised BMI, but liver steatosis was observed in one case only.

HCA Within a Specific Context

In this small group of our whole series, we encountered the well-described glycogenosis context,17 children with aplastic anemia exposed to Danazol,19 but also rare entities such as familial adenomatous polyposis coli24 and vascular disorders.18 We did not observe cases in body builders. From this very short series of seven particular cases, we noted again that men were often affected and that β-catenin mutation was a frequent observation associated with a high risk of HCC.

During the past 10 years, great progress has been made concerning HCA.1-4, 6, 7 The organization of clinical data according to phenotypic classification presented in this large series from a single center, if not likely to change our management strategy, still offers new paradigms about the disease. This has led our multidisciplinary team to adopt a novel management decision tree (Fig. 1). The main driving ideas are the role of imaging techniques (magnetic resonance imaging, contrast agent ultrasound)25, 26 combined with clinical and biological data,4 the size and number of nodules, along with biopsy and immunohistochemistry (biopsy and resection).

We definitely need additional data from different centers before modifying our global management strategy, particularly concerning the risk (bleeding and HCC) independent of the size of the nodule. The present data indicate that HCA is a complex entity; as a consequence, we believe that management of HCA needs to be adapted to the phenotypic classification.