The medical literature was searched using PubMed and EMBASE from 1970 to 2007 for reports of children with SCU type of HB. All known final reports of clinical trials of HB conducted by the major pediatric cooperative groups were reviewed, as were abstracts from the American Society of Clinical Oncology and American Society of Pediatric Hematology and Oncology meetings over the past 10 years.

The small cell component of HB is illustrated in Figure 1A. These cells are undifferentiated in routine histologic preparations, have minimal cytoplasm and round lightly chromatic nuclei with inconspicuous nucleoli, fail to form tubules or to cohere, infiltrate the liver while sparing bile ducts, and often invade veins. When comprising a portion of an otherwise typical HB with epithelial and possibly mesodermal derivatives, they are identified as discrete small nests most often associated with embryonal elements. Immunohistochemically, they share with rhabdoid tumors biphenotypic expression of intermediate filaments typical of epithelial and mesenchymal cells, cytokeratin and vimentin, which facilitates their detection. INI1 antibody produces a variable degree of nuclear staining, unlike rhabdoid tumor cells, which are always negative.

The treatments these patients received varied, although most treatments included cisplatin and doxorubicin, with or without vincristine and 5-fluorouracil in some combination. Of the eleven patients studied, none survived. Ten patients died of disease. One patient died of sepsis.

Our literature review identified another 29 cases with initially incompletely resected SCU HB. Demographics are shown in Table 2 [9–22]. Patients ranged in age from 2 to 138 months (median 15 months). The stage distribution was 5, 9 and 3 patients for stages II, III, and IV, respectively in the intergroup system and 2, 7 and 3 patients for PRETEXT II, III and IV. There were 19 males and 10 females. Twenty-two patients’ serum AFP results were reported: 10 had values in the normal range, 8 had values only minimally elevated for age (10.4 –307 ng/mL), and 4 had results reported as markedly elevated. Twenty-three patients were described as anaplastic or SCU, and 6 patients had SCU histology in combination with embryonal and/or fetal histology. The reports did not include descriptions of the degree or percentage of SCU histology.

The treatments these patients received varied significantly. Of those reported, all but one patient received chemotherapy. Most treatments included cisplatin, doxorubicin, vincristine and 5-fluorouracil in some combination. Four patients underwent liver transplantation, two of whom are long term survivors.

Only 4 of these 29 patients are alive without evidence of disease at the time of the reports. Twenty-two patients died of disease, including one patient reported to have died of disease and treatment-related complications. One patient died of complications following liver transplantation, and one patient died of a surgery-related death. One patient was lost to follow up after developing recurrent disease (#14).

The 4 patients alive without evidence of disease were older at diagnosis (ages 27.6 months, 28 months, 5 years and 11 years). In addition, two of the three with AFP data had levels said to be markedly elevated at diagnosis. Both of these features are distinctly atypical compared to other patients in this review. We were unable to obtain slides on these cases to independently confirm their stated histology. One of these patients was reported to have a combination of embryonal and anaplastic histology, and the remaining three had anaplastic features. Three were stage III and one was PRETEXT stage II. Two of the survivors underwent liver transplantation after chemotherapy. The agents used prior to transplant were not identified.

The stage distribution for patients with SCU histology is not significantly different from patients with other histologic types of HB. However, the SCU HB patients with confirmed genetic alterations and/or INI1 negative immunostaining (patients 3, 4, 7, 8, 9, 11, 15, 19) more often presented with advanced stage compared with patients with other histologic types of HB (12.5% stage II, 25% stage III and 62.5% stage IV compared with 5% stage II, 64% stage III and 31% stage IV on CCG 8881 [1]). The male predominance of patients with SCU HB in our study is slightly more pronounced than in hepatoblastoma overall: 73% (29 of 40 patients) compared with 62% male (113 of 182 patients) on CCG 8881 [1]. In the group of SCU HB patients with INI1 alterations and/or negative immunostaining, this slight increase in male gender is similar (75% male) to the male predominance in our study. Though they had poorer outcomes than patients with Stage I HB without SCU histology, the patients with SCU histology and Stage I disease fared better than those with Stage II to IV SCU HB in our study. Of the 16 patients with Stage I SCU HB reported by Haas et al, 6 patients survived event-free, 5 patients survived following recurrence, and 5 patients died of recurrent disease [7].

As is the case for all HB histologic types, aggressive attempts for full surgical resection is essential for cure, and patients in whom a full resection appears to be problematic should be referred early to centers with expertise with liver transplantation in HB patients. Extent of surgical resection was reported for 34 of the 40 patients in the overall cohort examined, although margin status and/or gross tumor spillage was explicitly stated in only 9 patients. Of the 34 patients, 14 did not undergo a resection (one due to parental refusal; otherwise details were not reported). These 14 patients all died of progressive disease. All 4 patients who survived had a complete resection. Two of these four survivors had complete resection in conjunction with liver transplantation. These 2 patients received chemotherapy before and after transplant. One of the 4 surviving patients had a complete resection with negative margins after 3 cycles of cisplatin and doxorubicin, and the fourth surviving patient had a complete resection after 4 months of vincristine and cyclophosphamide. No margin status was reported. The patients who underwent liver transplantation tended to fare better, but this occurred for only four of the patients in this series. Earlier attempts at full resection may be more critical for patients with SCU histology, given the poor response of these tumors to the standard chemotherapy used for HB.

In the complete cohort of patients reported here, AFP levels were normal or only mildly increased in 24 of the 28 patients with data. Patients with low or negative AFP have been reported to fare worse than those with elevated AFP levels [23]. More extensive pathologic review of the entire tumor of these patients with non-elevated AFP may reveal areas of SCU histology, because small biopsies at diagnosis may easily miss one component of a large heterogeneous tumor (Figure 1A).

Some of these tumors may be more appropriately classified as rhabdoid or rhabdoid-like tumors. Immunostaining for INI1 was negative for six patients (as would be expected in rhabdoid tumors), and it is known that rhabdoid tumor of the liver has been initially misdiagnosed as HB [24]. Both tumors share the simultaneous expression of intermediate filaments typical of epithelial and mesenchymal cells, cytokeratin and vimentin. Some of the clinical and biologic factors of both HB and rhabdoid tumors are also similar, but not identical (e.g., gender and proportion metastatic at diagnosis). It is difficult to classify the tumors as rhabdoid tumors definitively given the occasional INI1 positivity in some of the SCU, especially in the setting of an otherwise typical HB.

When SCU elements are found within otherwise typical HBs having fetal and embryonal histology, immunostaining for INI1 varies, with some nuclei positive, in contrast to cells with rhabdoid features. When otherwise indistinguishable small undifferentiated cells form the entirety of a neoplasm, it typically occurs in infancy, has negative nuclear staining for INI1, fails to secrete alpha-fetoprotein, and behaves like the rhabdoid tumor. Both cell types display biphenotypic expression of the intermediate filaments, cytokeratin and vimentin, but in the rhabdoid cell those filaments are more abundant and in routine stains are larger with eosinophilic cytoplasm, an eccentric nucleus and prominent nucleolus. On rare occasions, the two morphologically distinct cell types may be found within the same neoplasm. A thorough comparative molecular genetic analysis will provide ultimate clarification of the similarities and differences between these cell types. We are using laser capture microdissection to isolate the various cell types for this specific purpose.

Cytogenetic and molecular studies on the tumors of 4 patients in our report revealed translocations or deletion of the 11 q band on chromosome 22. Rhabdoid tumors are associated with cytogenetic abnormalities involving chromosome 22 [25–28]. Compared with other SCU HBs, tumors with cytogenetic, molecular and clinical characteristics of rhabdoid tumors should be differentiated from HB and treated like rhabdoid tumors. Although non-resected rhabdoid tumors have heretofore fared poorly, 5 patients with metastatic rhabdoid tumors of the kidney or liver have been treated successfully with ifosfamide, carboplatin and etoposide alternating with vincristine, doxorubicin and cyclophosphamide [29–32]. Patients with malignant rhabdoid tumors of the liver have also responded to other treatment regimens [33–34].

Further evaluation of patient characteristics and outcomes of HB patients with SCU histology should be reported by all the large pediatric cooperative groups. In the meantime, we suggest that HB patients with incompletely resected tumors containing SCU elements have careful cytogenetic, molecular and immunohistochemical evaluation to ascertain rhabdoid features and receive treatment that is more aggressive than provided for other HB patients. If SCU HB is indeed similar biologically to rhabdoid tumors, patients with SCU tumors of the liver should be included in future trials for rhabdoid tumors of other sites.

Drawbacks to this review are its retrospective nature, lack of thorough and expert pathology review on many of the cases to verify the presence and estimate the extent of the SCU component, and lack of clinical and surgical details in many cases. In addition, there may be a bias in that patients with small amounts of SCU histology not recognized in their tumors would not be reported as SCU HB. Nevertheless, this report constitutes by far the largest series of patients with SCU HB reported to date, and provides important clues as to the molecular pathogenesis of at least some of these tumors.

Supp Appendix