Clinicopathologic prognostic factors in childhood atypical teratoid and rhabdoid tumor of the central nervous system

A multicenter study




The objective of this study was to describe the clinical and pathologic features and to identify prognostic factors in patients with atypical teratoid/rhabdoid tumors (AT/RT) of the central nervous system (CNS).


Patients aged <18 years with newly diagnosed CNS AT/RT who were treated in France between 1998 and 2008 were retrospectively identified. The study included all patients who had a diagnosis of AT/RT confirmed by pathologic review, including immunostaining for INI 1, tumor protein 53 (p53), β-catenin, claudin-6, and Ki-67 and analysis for SMARCB1/hSNF5/INI1 mutation.


Fifty-eight patients with confirmed AT/RT were eligible for the current analysis. The median age at diagnosis was 1.4 years (range, 14 days to 8.5 years). The site of the primary tumor was supratentorial in 26 patients, infratentorial in 28 patients and spinal in 4 patients. Loss of INI1 nuclear expression was observed in 49 of 50 evaluable tumors. Positive claudin-6 was observed in 37 of 42 assessed tumors and, in 12 of those tumors, the staining was strong and diffuse. Positive nuclear immunoreactivity for β-catenin was observed in 24 of 44 tumors, and P53 was overexpressed in 31 of 44 tumors. Primary adjuvant therapy included chemotherapy in 47 patients and radiotherapy in 16 patients. The median follow-up was 58 months (range, 9-125 months), and the median survival was 9 months. Multivariate analysis identified age <2 years (P = .01), metastasis at diagnosis (P = .03), and strong immunopositivity for claudin-6 (P = .03) as prognostic factors for the risk of death.


AT/RT tumors in children carry a dismal prognosis. Age <2 years, metastasis at diagnosis, and strong claudin-6 positivity appeared to be independent prognostic factors for outcome. Cancer 2012. © 2011 American Cancer Society.


Atypical teratoid/rhabdoid tumors (AT/RT) are very rare, aggressive, and highly malignant tumors of the central nervous system (CNS).1-3 The main histologic characteristics of rhabdoid tumor cells are abundant cytoplasm with juxtanuclear, eosinophilic inclusions and nuclei that display a single, prominent nucleolus in clear, uncondensed chromatin.4 In addition, polyphenotypic immunohistochemical expression is common, in which mesenchymal, epithelial, and/or neuronal components may be present within the same tumor. However, differential diagnosis of AT/RT from other CNS tumors, such as primitive neuroectodermal tumor (PNET), choroid plexus carcinoma, and medulloblastoma, remains very difficult based on histopathology alone. In the 1990s, it was observed that AT/RT, malignant rhabdoid tumor of kidneys, and other soft tissue tumors often demonstrate a loss of all or part of chromosome 22 (monosomy 22).3, 5 It was discovered subsequently that deletion or mutation of SMARCB1/hSNF5/INI1 gene located at chromosome 22q11.2 was a specific genetic event in the pathogenesis of these rhabdoid tumors and was documented in >80%.6, 7 Multiples studies have demonstrated that a loss of INI1 protein expression caused by homozygous deletions or truncating mutations of INI1 are associated with rhabdoid tumors. Consequently, immunohistochemical staining for INI1 is now used for the diagnosis of AT/RT. Complicating the issue, a small subset of these tumors display loss of INI1 but have the atypical histopathologic features of AT/RT.8 Debate continues over how these tumors should be classified. In addition, another small subset of these tumors have the classic histopathologic features of AT/RT but retain INI1 staining, raising the possibility that a second locus mat exist that can lead to AT/RT pathology. Recently, somatic inactivation of SMARCA4/BRG1 gene was identified in a family with rhabdoid tumor predisposition.9 Thus, the identification of additional positive markers associated with AT/RT is crucial. Recently, claudin-6 was identified as a positive marker in AT/RT,10 and glucose transporter-1 protein (Glut1) was identified as a useful marker to define the embryonal nature of CNS neoplasms.11 In addition, β-catenin nucleopositive cells have been described in AT/RT.12

The outcome of patients with AT/RT is poor compared with the outcome of patients with other pediatric cancers. The best treatment modality for this disease has not yet been established, and prognostic factors remain unknown. The objective of the current study was to describe the clinical and pathologic features of AT/RT and to identify prognostic factors in patients with these tumors.



Patients aged <18 years with newly diagnosed CNS AT/RT who were treated in France between January 1998 and July 2008 were identified retrospectively through the hospital files. A standardized data sheet was sent to treating physicians who were members of the French Pediatric Cancer Society to collect clinical data. Metastatic disease was assessed using the staging system described by Chang et al,13 in which M1 indicates microscopic tumor cells present in cerebral spinal fluid; M2 indicates nodular seeding demonstrated in the cerebellar, cerebral subarachnoid space, or in the third or lateral ventricles; M3 indicates nodular seeding in the spinal subarachnoid space; and M4 indicates extra-CNS seeding.

The main chemotherapy regimens administered were a protocol comprising 7 cycles of 3 pairs of drugs (carboplatin/procarbazine, cisplatin/etoposide and vincristine/cyclophosphamide) (BB-SFOP)14; a protocol with 2 courses of carboplatin and etoposide followed by 5 high-dose chemotherapy courses with stem cell support (2 courses of melphalan, 2 courses of cisplatin, and 1 course of thiotepa) (PNET-HR)15; and the ATRT-04 protocol, which included 3 courses of cyclophosphamide, doxorubicin, and vincristine and 3 courses of ifosfamide, carboplatin, and etoposide followed by a course of carboplatin-thiotepa with stem cell support. Radiation therapy was planned in all strategies except for the BB-SFOP regimen.

Pathologic and Molecular Analysis

A single neuropathologist reviewed all available tumor material. The histologic review included morphologic assessment of AT/RT characteristics according to the 2007 World Health Organization (WHO) classification.4 Each specimen was reviewed using light microscopy and initial immunohistochemistry. The presence or absence of classic histologic features, such as rhabdoid cells, a PNET component, a fusiform cell component, extensive necrosis, microcalcification, lymphoid infiltration, and hemorrhage, was recorded for each tumor. Immunolabeling was performed on the most representative tumor block (the largest amount of viable, malignant-appearing tumor with a tumor-adjacent brain margin if possible), including monoclonal Ini1 (1:50 dilution; clone BAF47; BD Biosciences, San Jose, Calif), glial fibrillary acidic protein (GFAP) (1:200 dilution; clone 6F2; Dako, Glostrup, Denmark), smooth muscle actin (SMA) (1:750 dilution; clone 1A4; Dako), cytokeratin 18 (CK18) (1:200 dilution; DC10; Zymed, South San Francisco, Calif), epithelial membrane antigen (EMA) (1:25 dilution; clone E29; Dako), vimentin (Vim) (1:400 dilution; clone V9; Dako); and, when needed for the purposes of making a differential diagnosis, Olig2 (1:50 dilution; polyclonal; R&D Systems, Minneapolis, Minn), synaptophysin (Syn) (1:50 dilution; clone SY38; Progen, Heidelberg, Germany), neuronal nuclei (NeuN) (1:500 dilution; clone A60; Chemicon, Billerica, Mass), and neurofilament (1:50 dilution; clone 2F11; Dako).

The most representative paraffin-embedded specimens from each case were analyzed and compared systematically with a panel of 5 immunohistochemical markers, which included KI-67 antigen (Mib-1; 1:200 dilution; Dako), P53 (1:20 dilution; clone DO-1; Immunotech, Marseille, France), βCatenin (prediluted 1:1; clone 14; Ventana, Tucson, Ariz), Glut1 (1:1 dilution; polyclonal; Abcam, Cambridge, United Kingdom), and claudin-6 (1:10 dilution; clone C20; Santa Cruz Biotechnology, Santa Cruz, Calif). Immunohistochemical staining of paraffin sections was carried out using an automated immunohistochemistry instrument (Discovery XT; Ventana Medical Systems) for Ini1, Mib-1, P53, βCat, Glut1, claudin-6, GFAP, Vim, CK18, and SMA antibodies; and a semiautomated method was used for the other antibodies. For the automated protocol, the standard pretreatment included CC1 buffer (a Tris-based buffer; Ventana Medical Systems) and a primary antibody incubated for 8 minutes at room temperature. Antibody binding was observed with an Omni-UltraMapHRT Kit (Ventana Medical Systems). Diaminobenzidine tetra hydrochloride (DAB) (Ventana Medical Systems) was used as the chromogen. Concerning the semiautomated protocol, the sections were exposed to 30-minute microwave treatments at 98°C in citrate buffer, pH 6.00, for antigen retrieval and were treated with 3% hydrogen peroxide in distilled water to block endogenous peroxidase activity. Sections were incubated with primary antibodies for 1 hour at room temperature and analyzed using an Immunotech peroxidase kit (Kit 1476), and DAB was used as the chromogen associated with light hematoxylin counterstaining.

β-Catenin and P53 protein accumulation was evaluated semiquantitatively. For both proteins, immunostaining was classified as positive if ≥10% cells expressed β-catenin or P53. Blood vessel cytoplasmic staining for Glut1 was used as an internal control and as a parameter for staining intensity, as described by Loda et al11 (Fig. 1A,B). Claudin-6 cytoplasmic immunoreactivity was scored as negative, focally positive, or widespread positive (Fig. 1C-E). Finally, the Mib-1 proliferation index was calculated as the percentage of labeled nuclei (excluding vascular and inflammatory elements) per total tumor nuclei in the highest labeled areas.

Figure 1.

These photomicrographs illustrate (A,B) immunohistochemical analysis and scoring for (A,B) glucose transporter-1 protein (Glut1) and (C-E) claudin-6 staining. (A,B) Glut1 is expressed in tumor cells and endothelial cells (A, arrow), whereas endothelial proliferation remains negative (B, arrow). Photomicrographs reveal (C) claudin-6 negativity; (D) focally positive claudin-6 expression; and (E) strong and diffuse, positive claudin-6 expression (original magnification, ×400).

SMARCB1/hSNF5/INI1 molecular screening also was performed when the material was available. DNA was extracted from frozen tumors and blood lymphocytes. All coding exons and splice site regions were sequenced using the Sanger method and an ABI automated fluorescent sequencer (Applied Biosystems, Foster City, Calif). We searched for large size deletions using the multiplex ligation-dependent probe amplification (MLPA) assay Salsa MLPA (KIT P258-B1 SMACB1; MRC-Holland, Amsterdam, the Netherlands). For patients with whole hSNF5/INI1 deletion in the germline, the deletion borders were studied using Affymetrix ANP6-arrays (Affymetrix, Santa Clara, Calif). The current study included all patients for whom the reference pathologist confirmed the diagnosis of AT/RT based on rhabdoid features or loss of INI1 expression assessed by immunostaining and those who had no histologic material available for review but who had a specific mutation of the SMARCB1/hSNF5/INI1 gene.

Statistical Analysis

The distribution of initial characteristics and treatment modalities was compared between age groups (<2 years vs ≥2 years) using chi-square tests or exact tests, as appropriate. We also compared the main pathologic features between localized and metastatic disease at diagnosis. The main endpoint to evaluate prognostic factors was overall survival (OS), which was calculated from the date of diagnosis to the date of death, whatever the cause. We also estimated progression-free survival (PFS), which was calculated from the date of diagnosis to the date of progression or relapse or to the date of death. Survival curves (OS and PFS) were estimated using the Kaplan-Meier method with Rothman 95% confidence intervals (95% CIs). The median follow-up was estimated using the Schemper method. The cutoff date for the analysis was September 2009. Hazard ratios (HR) for death were estimated with their 95% CIs using Cox regression. The variables studied were age (<2 years vs ≥2 years), tumor site (supratentorial vs others), metastatic status, histologic and immunohistochemical features (rhabdoid component; tumor; and/or endothelial Glut1 positivity, P53 positivity, strong claudin-6 positivity, β-catenin positivity), and the extent of resection assessed by postoperative radiologic imaging when available or according to the surgeon's report in other cases (complete or subtotal resection vs partial resection or biopsy alone). Resection was deemed complete when the neurosurgeon and neuroradiologist confirmed the absence of residual tumor at the end of the operative procedure. Resection was considered subtotal when ≥90% of the tumor had been removed. All other resections were considered partial. The variables associated with a P value < .20 in the univariate analysis were tested in the multivariate analysis using a stepwise procedure. The final model included the variables associated with a P value < .05. All the prognostic factor analyses were stratified on treatment (radiotherapy, yes vs no). Data were analyzed with the SAS statistical software package (version 9.1; SAS Institute, Cary, NC).


Patient Characteristics

Overall, 78 patients with a local diagnosis of CNS AT/RT were identified by the French Pediatrics Cancer Centers. Central histologic review was available for 58 patients: Some archival cases either could not be located or did not contain qualitatively viable tumor. The diagnosis of AT/RT was confirmed histologically for 51 patients (88%) and was rejected by the reference pathologist for 7 of 58 patients: the rejected tumors included 2 PNETs, 2 glioblastomas, 2 WHO grade III gliomas, and 1 unclassified malignant tumor. Among the 20 children who did not have material available for central review, the diagnosis of AT/RT was retained in 7 patients because molecular analysis confirmed the specific gene mutation. Thus, 58 patients with AT/RT were eligible for the current analysis. Their characteristics are listed in Table 1. The median age was 1.4 years (range, from 14 days to 8.5 years), and 35 patients (60%) were male. The primary tumor site was supratentorial in 26 patients (45%), infratentorial in 28 patients (48%), and spinal in 4 patients. A bifocal tumor was diagnosed in 4 patients (CNS and kidney tumor; details available on request). Seventeen patients (30%) had metastases at diagnosis. According to the staging system described by Chang et al,13 there were 38 patients with M0 disease at diagnosis, 1 patient with M1 disease, 6 patients with M2 disease, 9 patients with M3 disease, and 1 patient with M4 disease. In 3 patients, no information about metastasis at diagnosis was available.

Table 1. Clinical Characteristics and First-Line Treatment Strategy Overall and by Age Group
 No. of Patients (%) 
CharacteristicOverall, N = 58Age ≤2 Years, N = 38Age >2 Years, N = 20P
  1. Abbreviations: Glut1, glucose transporter-1 protein; MD, missing data; p53, tumor protein 53; RT, radiotherapy.

Sex   1.00
 Male35 (60)23 (60)12 (60) 
 Female23 (40)15 (40)8 (40) 
Tumor site   .28
 Supratentorial26 (45)15 (40)11 (55) 
 Infratentorial28 (48)22 (58)6 (30) 
 Other4 (7)1 (2)3 (15) 
Metastases at diagnosis, n = 3 MD   .55
 Yes17 (30)10 (28)7 (37) 
 No38 (70)26 (72)12 (63) 
Rhabdoid cell, n = 8 MD   .13
 Yes41 (82)23 (26)18 (95) 
 No9 (18)8 (74)1 (5) 
Endothelial Glut1, n = 16 MD   .008
 Negative17 (40)6 (23)11 (69) 
 Positive25 (60)20 (77)5 (31) 
p53, n = 14 MD   .64
 Negative13 (29)9 (30)4 (29) 
 Positive31 (70)21 (70)10 (71) 
Claudin-6, n = 16 MD   .29
 Negative or mildly positive30 (71)21 (78)9 (60) 
 Strongly positive12 (29)6 (22)6 (40) 
β-Catenin, n = 14 MD   .12
 Negative20 (45)10 (36)10 (62) 
 Positive24 (55)18 (64)6 (38) 
Surgical resection   .57a
 Complete27 (47)18 (47)9 (45) 
 Subtotal9 (15)7 (18)2 (10) 
 Partial15 (26)9 (24)6 (30) 
 Biopsy alone7 (12)4 (11)3 (15) 
RT   .06
 Yes16 (27)7 (18)9 (45) 
 No42 (73)31 (82)11 (55) 
Chemotherapy   .49
 Yes47 (81)32 (84)15 (75) 
 No11 (19)6 (16)5 (25) 
Treatment modality   .07
 No Chemotherapy, no RT9 (16)6 (16)3 (15) 
 RT alone2 (3)0 (0)2 (10) 
 Chemo alone33 (57)25 (66)8 (40) 
 Chemo + RT14 (24)7 (18)7 (35) 

Central Histopathologic Review

Loss of INI1 nuclear expression was observed in 49 of 50 evaluable tumors, and only 1 tumor with classic rhabdoid features expressed INI1. In 1 reviewed tumor with PNET features, INI1 staining was not available because of the absence of a positive endothelial control; however, a mutation was identified in the SMARCB1/hSNF5/INI1 gene by molecular analysis.

Rhabdoid cells were present in 41 tumors. Thirty-five tumors had an undifferentiated small cell component that mimicked PNET/medulloblastoma. Sixteen samples contained a fusiform cellular component. The tumor exhibited a major astrocytoma-like component in 1 sample, major ependymal differentiation in 1 sample, and 2 samples had multiphenotypical features.

All but 1 tumor that exhibited positivity for Glut1 had heterogeneously strong, diffuse cytoplasmic staining with intense, membrane-bound reactivity. The extent of immunopositivity varied from widespread to focal positivity (Fig. 1A,B). Clusters of cohesive, positive cells were observed in either rhabdoid cells or PNET components. In the adjacent brain parenchyma, there was no staining of astrocytes, oligodendrocytes, or neurons; whereas nearly all microvessels and erythrocytes were intensively positive (Fig. 1A,B). It is noteworthy that intratumor vessels were negative for Glut1 in 17 of 42 samples (40%), particularly when an endothelial proliferation was present.

Positive immunoreactivity for claudin-6 was observed in 37 of 42 assessed tumors (88%). In 12 tumors, staining was strong and diffuse and occurred frequently in clusters of large cells or rhabdoid cells (Fig. 1). In contrast, 25 tumors manifested only strong, focal immunopositivity, which occurred more often in large cells. The pattern of immunostaining was cytoplasmic and membranous. Endothelial cells and adjacent brain tissue revealed negative immunoreactivity for claudin-6. Positive nuclear immunoreactivity for β-catenin was observed in 24 of 44 tumors (55%), and P53 was overexpressed in 31 of 44 tumors (70%).

A SMARCB1/hSNF5/INI1 gene mutation was identified in 17 of 17 screened tumors and in constitutional DNA from 7 of 11 tested patients. Among the 17 patients who had somatic hSFN5/INI1 inactivation, a germline genetic analysis was performed in 9 patients and was positive in 5 patients. In 2 patients, the somatic mutation was not investigated, and only a constitutional mutation was identified.

Relation Between Pathologic Features, Age, and Metastatic Status at Diagnosis

Among the 31 children aged <2 years in this study, rhabdoid cells were absent in 8 patients versus 1 of 18 older patients (26% vs 5%; P = .13). Of 26 children aged <2 years, endothelial Glut1 staining was positive in 20 patients versus 5 of 11 older children (77% vs 45%; P = .008). P53, β-catenin, and claudin-6 staining was not correlated with age. P53 immunostaining was correlated with metastatic status: Only 6 of 29 patients who had metastatic disease at diagnosis were negative for P53 versus 7 of 12 patients who had localized disease at diagnosis (21% vs 58%; P = .03). No correlation was observed between metastatic status and the presence of rhabdoid cells, β-catenin, or endothelial Glut1 immunoreactivity.

Concerning the 7 patients who had germline mutations, their median age was 0.6 months (range, 0.05-2.0 months), and 4 children had disseminated disease at diagnosis. Five of these patients received chemotherapy. All patients died, and their median survival was 6.4 months (range, 0.69-13.02 months).


Treatment is described in Table 1. Nine children received no treatment after initial surgery or biopsy because of rapid disease progression. Forty-seven patients (81%) received various chemotherapy regimens. Nine patients received the BB-SFOP protocol, 11 patients received the PNET-HR regimen, 24 patients received the ATRT-04 protocol, and 3 patients received other chemotherapy regimens. Sixteen children (27%) received radiotherapy as part of their primary therapy. Eleven patients received high-dose chemotherapy followed by stem cell support. No treatment-related deaths occurred. Because of the study design (nonrandomized) and the absence of standardized treatment in this series, the efficacy of the different treatment strategies could not be evaluated.


The median duration of follow-up was 58 months (range, 9-125 months). At the last update, 8 patients remained alive without evidence of disease, including 3 patients in first complete remission 17 months, 35 months, and 77 months after the diagnosis. An early death occurred in 9 children at a median of 0.7 months (range, from 2 days to 4.4 months), because of progressive disease. In addition, recurrence or disease progression occurred in 46 patients, including metastatic sites in 30 patients. Fourteen children received no treatment at relapse. Salvage treatment was chemotherapy in 20 children, radiotherapy in 3 children, and radiotherapy combined with chemotherapy in 9 children. Five children were still in second complete remission at the last follow-up. All received chemotherapy combined with radiation therapy as salvage treatment, 3 of these 5 children underwent surgical resection of the relapsed tumor.

The median OS was 9 months (95% CI, 6.5-14.2 months). The 1-year PFS and OS rates were 17% (95% CI, 10%-29%) and 41% (95% CI, 30%-54%), respectively (Fig. 2).

Figure 2.

Progression-free survival and overall survival.

Prognostic Factor Analysis

Table 2 provides details of the multivariate analysis, in which age <2 years (HR, 3.1; 95% CI, 1.30-7.5; P = .01), metastasis at diagnosis (HR, 2.7; 95% CI, 1.10-6.4; P = .03), and strong immunopositivity for claudin-6 (HR, 2.7; 95% CI, 1.09-6.5; P = .03) were identified as prognostic factors for the risk of death (Table 2). The quality of resection had a significant impact on OS in the univariate analysis but was no longer significant when adjusted on the other prognostic factors (HR, 1.32; 95% CI, 0.59-2.9; P = .50). The risk of death was not associated significantly with primary tumor site, the absence of rhabdoid cells, endothelial Glut1 protein, or P53 and β-Catenin positivity. The results of the multivariate analysis were relatively stable when the analysis was done without any stratification, when the factor radiation therapy was included in the model (adjusted analysis), and within each stratum of the factor radiation therapy.

Table 2. Prognostic Factor Analysis of the Risk of Death
  Univariate AnalysisMultivariate Analysisa
CharacteristicNo. of Patients1-Year OS: Mean ± SE, %HR (95% CI)PHR (95% CI)P
  • Abbreviations: AT/RT, atypical teratoid/rhabdoid tumor; CI, confidence interval; Glut1, glucose transporter-1 protein; HR, hazard ratio (the hazard of death); MD, missing data; OS, overall survival; p53, tumor protein 53; SE, standard error.

  • a

    The multivariate model included 3 variables that were associated significantly with the risk of death in multivariate analysis: age, metastatic status at diagnosis, and positive Claudin-6 status.

  • b

    Reference category.

  • c

    When included in the multivariate model, the quality of resection was no longer associated significantly with the risk of death (HR, 1.32; 95% CI, 0.59-2.9; P = .50); all analyses were stratified on treatment (radiotherapy as first-line treatment: yes vs no).

Age, years   .12 .01
 ≤23834 ± 81.67 (0.88-3.2) 3.1 (1.30-7.5) 
 >22055 ± 111.00b 1.00b 
Primary tumor site   .91  
 Supratentorial2635 ± 91.00b  
 Other3247 ± 91.03 (0.59-1.8)  
Metastasis at diagnosis, n = 3 MD   .05 .03
 No3853 ± 81.00b 1.00b 
 Yes1724 ± 101.91 (1.05-3.6) 2.7 (1.10-6.4) 
Rhabdoid cells, n = 8 MD   .43  
 No933 ± 161.00b  
 Yes4141 ± 80.74 (0.34-1.6)  
Endothelial Glut1, n = 16 MD   .68  
 Negative1747 ± 121.00b  
 Positive2532 ± 91.15 (0.59-2.3)  
p53, n = 14 MD   .26  
 Negative1315 ± 101.00b  
 Positive3145 ± 90.67 (0.33-1.34)  
Claudin-6, n = 16 MD   .07 .03
 Negative or mildly positive3043 ± 91.00b 1.00b 
 Strongly positive1225 ± 132.02 (0.95-4.3) 2.7 (1.09-6.5) 
β-Catenin, n = 14 MD   .47  
 Negative2045 ± 111.00b  
 Positive2438 ± 101.28 (0.68-2.5)  
Surgical resection, n = 1 MD   .07  
 Complete or subtotal3653 ± 81.00b c 
 Partial or biopsy2223 ± 91.72 (0.95-3.1)  


The current review provides one of the largest analysis to date of the clinical and pathologic risk factors for childhood AT/RT. Our data confirm the prognostic value of age <2 years and metastatic status at diagnosis, but the data also suggest a relation between a strong immunopositivity for claudin-6 and outcome.

Age at diagnosis has been reported previously as an important prognostic factor.16, 17 Similarly, in this review, children aged <2 years had a statistically significant worse outcome than older patients. Of note is that the treatment administered to infants and young children differed from that received by older patients, because only 7 young children had received radiotherapy as part of their initial treatment. Although the difference in treatment received may have introduced a confounding factor, the prognostic factor analysis was stratified for whether or not radiotherapy was received. Tekautz et al postulated that radiotherapy was an important treatment with respect to tumor control on the basis that younger patients who received radiotherapy early during the course of their treatment were the only ones who emerged as long-term survivors.17 The poorer prognosis potentially may be explained by the presence of a germline mutation, which was present more frequently in younger patients. Kordes et al suggested that patients with a germline mutation had a younger median age at diagnosis and a worse prognosis compared with patients without a detectable germline mutation (5.5 months vs 13 months; P = .001).18 In our study, it was difficult to investigate the real impact of the presence of a germline mutation on outcome, given the small size of the population studied, the low number of survivors irrespective of their germline status, and the heterogeneity of the treatment received. Furthermore, germline genetic analyses were performed mostly in the younger cohort of patients (median age at diagnosis, 0.6 months vs 1.5 months for the group in which germline analyses were not performed). Because the presence of a germline mutation was assessed more frequently in the younger cohort, it is possible that the constitutional mutation rate may be overestimated in our study.

The frequency of metastases of 30% in our cohort is consistent with previously published retrospective experiences,16, 17, 19 as is the increased likelihood of younger children having disseminated disease at diagnosis.16, 17, 19 In addition, in support of previous reports, the current review demonstrates that the presence of metastases at diagnosis had a significant impact on survival (P = .03),16, 17, 19, 20 a finding that has led to a more aggressive treatment approach in this subgroup of patients. The incidence of metastasis at diagnosis, however, did not differ statistically between children aged <2 years and >2 years.

In our study, endothelial Glut1 staining was associated significantly with age. In previous studies, Glut1 staining was not evaluated in AT/RT. It is noteworthy that claudin-6 has been described as a positive marker in the differential diagnosis of AT/RTs. In the study by Birks et al,10 immunohistochemical staining of 33 tumor specimens revealed positive membrane staining in 7 of 7 AT/RT samples and negativity in 26 of 27 other brain tumor samples. It is noteworthy that none of the 16 medulloblastomas/PNETs were stained for claudin-6. In our study, positive immunoreactivity for claudin-6 was observed in 33 of 37 tumors (89%). In a recent study,21 claudin-6 protein expression was encountered in 17 of 59 AT/RTs (29%) and also was observed in other primary CNS tumors, including 60% of medulloblastomas and 21% of malignant gliomas. The sensitivity and specificity of claudin-6 immunohistochemistry in the diagnosis of AT/RT were poorer than initially believed.10 However, in tumors that had positive claudin-6 staining, this review indicates that strong positivity for this marker was associated with a poorer outcome compared with patients who had negative or positive staining for this marker. In the study by Antonelli et al,21 no significant correlation was observed between the level of claudin-6 expression and survival; however, only 1 patient had a strong immunopositivity for claudin-6. Acknowledging the limitations of this retrospective study related to the small number of patients and the heterogeneity of treatments, further prospective studies are needed to assess the prognostic value of strong immunopositivity for claudin-6. Although the biologic significance of claudin-6 in AT/RT will need to be determined, we suggest that this biomarker should be studied in addition to INI1.

The prognostic significance of resection has been reported previously.19, 22-25 In our study, the quality of resection was not correlated statistically with survival when the analysis was adjusted on other risk factors (P = .50). Other authors have reported similar findings.22 However, treatment like surgery appears to be an independent prognostic factor. In the study by Athale et al, the OS of patients who underwent gross total resection was 21.3 months compared with 12.3 months for the subset that underwent partial resection and 10.2 months for those who underwent biopsy alone (P = .04).22 In a recent trial,19 9 patients who did not have any evidence of metastatic disease underwent complete resection. The authors of that report recommended that an aggressive surgical approach should be used under such circumstances; and, when indicated, second-look surgery should be performed to achieve gross total resection. Surgery is a promising option for patients who have relapsed. In the study by Fidani et al,26 1 patient who relapsed remained alive after second surgery followed by high-dose chemotherapy with stem cell support and radiotherapy. In our analysis, of the 5 patients who remained alive after relapse, 3 underwent second surgery.

The heterogeneity of the chemotherapy regimens administered and the small number of patients in previously published series prohibited any evaluation of the efficacy of individual agents or regimens, including survival. In our study, patients received different chemotherapy protocols, but these mostly were ineffective. However, there was no difference in OS according to the different approaches used in the groups. In this series, the role of radiation therapy could not be determined. Few patients (16 of 58) had received radiotherapy as part of initial treatment.

The prognosis for patients with AT/RT is dismal, and this particularly true in children aged <2 years and in children who have metastasis at diagnosis. Our data suggest that strong immunopositivity for claudin-6 may be a prognostic factor in AT/RT. Further prospective studies on patients who are treated homogenously are warranted to confirm this finding.


We thank the following pathologists for their contribution to the study: Dominique Figarella-Branger, MD, PhD (Department of Neuropathology, University Hospital La Timone, Marseille); Anne Jouvet, MD, PhD (Department of Neuropathology, Civil Hospices of Lyon); Claude-Alain Maurage, MD, PhD (Department of Pathology, University Hospital Lille); Marie-Bernadette Delisle, MD, PhD (Department of Pathology, University Hospital Rangueil, Toulouse); Francois Le Gall, MD, PhD (Department of Pathology, University Hospital Pontchaillou, Rennes); Isabelle Quintin-Roué, MD (Department of Pathology, University Hospital Brest); Jean Michel Vignaud, MD, PhD (Department of Pathology, University Hospital Nancy); Marc Polivka, MD (Lariboisiere Hospital, Paris); Pierre Levillain, MD, PhD (Department of Pathology, University Hospital, Poitiers); and Isabelle Salmon, MD (Department of Pathology, Erasme Hospital, Bruxelles).

We thank Lorna Sainte Ange for editing and Sara Calmanti for her help in editing and submitting the article.


No specific funding was disclosed.


The authors made no disclosures.