More than 90% of all Ewing's Sarcoma Family of Tumors (ESFT) exhibit specific chromosomal rearrangements between the EWS gene on chromosome 22 and various members of the ETS gene family of transcription factors. The gene fusion type and other secondary genetic alterations, mainly involving cell cycle regulators, have been shown to be of prognostic relevance in ESFT. However, no conclusive results have been reported. We analyzed the clinicopathological significance of relevant cell cycle regulators in genetically confirmed ESFT. A total of 324 cases were analyzed for the immunohistochemical expression of p53, p21Waf1/Cip1, p27Kip1 and Ki67 and the chromosomal alterations of the p53 and 9p21 locus by fluorescent in situ hybridization. We observed that expression of p53 (p = 0.025), p21Waf1/Cip1 (p = 0.015) and p27Kip1 (p = 0.013) was higher in disseminated than in localized disease. Furthermore, a cohort of 217 patients with localized disease was considered for studying the prognosis involvement of these factors on patient follow-up. The median follow-up was 39 months (range: 0.17–452) with an overall survival (OS) of 55%. Ki67 was expressed in 34% of cases and constituted an independent prognostic factor for progression free survival and OS independently of the type of treatment [hazard ratio of 2.0 (95% CI: 1.3–3.1; p = 0.003) and 1.9 (95% IC: 1.3–2.9; p = 0.007) for progression free survival and OS, respectively, being especially relevant in the group of patients which incorporated radiotherapy in their regimen schedules. In conclusion, this study demonstrates that Ki67 expression constitutes a valuable indicator of poor prognosis in localized ESFT.
The Ewing's Sarcoma Family of Tumors (ESFT) are comprised of small round cell neoplasms most frequently found in adolescents and young adults between the ages 10 and 20 years, comprising 3% of all pediatric malignancies and being the second most common pediatric malignant bone tumor.1
ESFT are histologically similar to other small round cell tumors, but they have a distinct clinical behavior and therefore require different therapeutic management. For this reason, a precise diagnosis of EFST is crucial, requiring not only light microscopy but other ancillary techniques such as immunohistochemistry and molecular biology.2–4 More than 90% of all ESFT exhibit specific chromosomal rearrangements between the EWS gene on chromosome 22 and various members of the ETS gene family of transcription factors.5, 6 These specific gene fusions are considered a diagnostic feature of these tumors, and their gene products are believed to play an important role in ESFT development and biology.7 However, induction of tumorigenicity by EWS-FLI1 itself is dependent on the cellular background, indicating that additional genetic alterations are necessary for tumor progression. In this regard, the additional genetic alteration of p53 or p16INK4a/p14ARF loss as well as other cell cycle regulators may facilitate subsequent cellular transformation, which may lead to further aggressive tumor progression.8
A subpopulation of patients with ESFT has genetic alterations of p16INK4a and p14ARF (both located and codified by the same locus gene at 9p21.3) or p53 (at 17p13.1). Recent studies have indicated the presence of the p53 gene alteration in 4% to 13% of patients9–14 and the p16INK4a gene alteration in 12% to 26% of patients13–19 with ESFT. These data indicated the presence of the p53 gene alteration among 11% of patients, the p16INK4a gene alteration among 17% and either alteration among 24% of patients.8–19 It has been suggested that the presence of these alterations is potentially linked with a poor prognosis for patients with this disease.8, 11–15, 19
The presence of metastasis at diagnosis defines an advanced stage of the disease and has been considered the most unfavorable prognostic factor.20 However, there have been discrepancies in the observed associations between p16INK4a or p53 alterations and advanced-stage disease.16, 18
Other studies have suggested that genes related to cell cycle regulation; invasion and metastasis were associated with a pour outcome.21 However, clinical and biologic characteristics alone fail to accurately classify patients with ESFT, mainly because of the limited sample sizes on which the underlying studies are based. To further improve risk-stratification and treatment options for patients with localized ESFT, there is a need for novel molecular prognostic parameters in addition to the conventional factors.
This study was undertaken to better understand the clinicopathological significance of the alterations of a series of cell cycle regulator markers in a series of 324 genetically confirmed ESFT. We also evaluated the prognosis impact of these alterations in a cohort of 217 patients with localized disease.
Material and Methods
Patients and samples
Within the context of two European Translational Research projects [PROTHETS (http://www.prothets.org) and EuroBoNet (http://www.eurobonet.eu)], we retrospectively analyzed a total of 324 paraffin-embedded samples from ESFT patients treated at Instituto Ortopedici Rizzoli (IOR), Bologna, Italy, and at the Department of Pathology of the University of Valencia Estudi General (UVEG), Spain, from the period comprised between April 1971 and May 2007.4 All cases were genetically confirmed as belonging to the ESFT by molecular biology and/or fluorescent in situ hybridization (FISH). Approval for data acquisition and analysis was obtained from the Ethics Committee of the institutions involved in the study. The clinical data were reviewed and stored within a specific database.
Of the 324 tissues analyzed, 272 (84%) corresponded to primary localized tumors, 22 (7%) to disseminated primary tumors, 9 (3%) to recurrences, and 21 (6%) to metastases (6%). For statistical considerations, disseminated primary tumors, recurrences and metastasis were grouped into the same category (disseminated disease).
For the prognostic evaluation, we analyzed a cohort of 217 patients (198 from IOR and 19 from UVEG) with localized disease. The median age of the series was 17 years (range: 1–99 years), and the male–female ratio was 1.6. Sixty percent of tumors were appendicular skeleton. Seventy-six patients (35%) were treated with surgery and chemotherapy, whereas the remaining cases were treated with radiotherapy, chemotherapy and/or surgery. The median follow-up was 39 months (range = 0.17–452 months) with an overall survival of 55%.
Histological ESFT subtypes together with other structural parameters were reviewed to obtain supplementary criteria for prognosis as previously studied.4
Table 1 shows the most relevant clinical information from our series.
Table 1. Characteristics of the cohort with localized disease
Tissue microarray construction
A total of 24 tissue microarrays (TMAs) containing two representative cores for each case (1 mm in diameter) were constructed to carry out the immunohistochemical and FISH analysis.
The immunohistochemical analysis was performed in the TMA sections, and 4-μm thick sections from FFPET were processed. The antibodies used for this study were: anti p53 (clone DO7, Novocastra, Newcastle upon Tyne, UK), anti Ki-67 (MIB-1, Dako, Glostrup, Denmark) and anti p27Kip1 (Dako, Glostrup, Denmark) at 1/50 dilution and anti p21Waf1/Cip1 (Dako, Glostrup, Denmark) at 1/25 dilution with specific pretreatment. Antigen retrieval was performed by pressure cooker boiling at 1.2 atmospheres for 3 min in 10 mmol/L citrate buffer (pH 6.0). The LSAB method (Dako) was performed, followed by revelation with 3,3′-diaminobenzidine as conventional protocols. Nuclear staining was considered positive for the four antibodies. Immunoreactivity was defined as follows: negative, fewer than 5% of tumors cells stained; poorly positive (+), between 5% and 10% of tumors cells stained; moderately positive (++), between 10% and 50% of tumors cells stained and strongly positive (+++), when more than 50% of the tumors cells were stained. All sections were independently evaluated by 3 pathologists (IM, SN and ALLB). The agreement of staining intensity scoring by all was recorded, and in cases of disagreement, intensity and score was determined by consensus. For the statistical analysis, the cases were scored as negative when no staining or less than 5% of tumor nuclei were stained and positive when the staining was observed in more than 5% of tumor cells.
Molecular biology and fluorescence in situ hybridization for the assessment of EWS-ETS gene fusion status
Ribonucleic acid was extracted and tested in all paraffin-embedded tumors with reverse transcription polymerase chain reaction to detect any of the following gene fusions: EWS/FLI1, EWS/ERG and EWS/FEV as previously described.22 In addition, a FISH analysis was carried out with the EWSR1 break-apart probe on the TMAs. Further details are described in Machado et al.3
FISH for 9p21 deletions and p53 alterations
Two 4-μm thick sections of TMA were used for FISH studies as previously reported.3, 23 The probes used to analyze the 9p21 locus status were the critical region (9p21/red) with control region (SE9 region/green) and to show p53 integrity, we used the critical region (17p13/red) with control region (SE17 region/green), both from Kreatech Diagnostics (Amsterdam, The Netherlands). We designed our own system to detect genetic alterations in FFPE. We applied the system on control tissue incorporated into the TMA to determine the proportion of cells with cutting effects. We defined 2 groups: (i) cells without genetic alteration, having the same number of gene signals and control probe signals; (ii) cells with cutting effects, including nuclei fragments without probe signals and those with less control signals than gene signals. Then a similar scoring scheme was applied to the tumor samples. We added Group 3, cells with possible genetic alteration, having a lower number of gene probe signals than control probe signals; and Group 4, cells with genetic alteration, subtracting the total percentages of cells with cutting effects (Group 2) from the cellular population belonging to Group 3, thus eliminating false-positives. We considered a genetic alteration when the percentage of tumor cells in Group 4 was higher than 15%. Using these parameters, for the 9p21 status we defined two groups of cases as: nondeleted (when the same number of 9p21 locus red signals as control probe green signals is observed) and deleted [including heterozygous deletion (when less red than green signals are observed, minimum one red signal) and homozygous deletion (no red signals with at least two green signals are observed)]. To define the p53 status, we define also two groups: no alteration (when the same 17p13-red and control-green signals were observed) and alteration (which includes polysomies, monosomies and heterozygous deletion of the p53 locus).
A FileMaker database was constructed and implemented to collect the main clinical, histopathological, immunohistochemical and genetic data of the patients included. For the statistical analysis, we used binary variables reflecting the positivity status of the measures (yes or no). Association with histopathological parameters, all categorical, was also assessed using a chi-square test to determine homogeneity or linear trend for ordinal variables. The significance level was set at 5%. To study the impact of the histological, immunohistochemical and molecular factors on progression-free survival (PFS) and disease-specific survival (DSS), the Kaplan-Meier proportional risk test (log rank) was used.24, 25 PFS is considered as the proportion of patients with no evidence of local or distant recurrence during follow-up, whereas DSS reflects the proportion of patients alive (with or without disease) at the time of the last clinical review. Only cancer-specific deaths were considered for the statistical analysis of the DSS. Evidence of the relative risk for each patient was also provided by means of a Cox proportional hazards model using stepwise selection to identify the independent predictors of poor outcome.26 The variables used in the prognostic evaluation included: clinical (age, sex, location of the tumor, and type of treatment), histological (histological varieties, necrosis, lobular pattern, filigree pattern, and mitosis), immunohistochemical (Ki67, p53, p21Waf1/Cip1 and p27Kip1), and molecular parameters (fusion gene type, FISH of p53, FISH of the locus 9p21). All the statistical analyses were carried out using the SPSS (version 15.0.1, SPSS, Chicago, IL).
Immunohistochemical expression of Ki67, p53, p21Waf1/Cip1 and p27Kip1
All the antibodies employed in this study specifically stain the nuclei of tumor cells (Fig. 1). The immunohistochemical analysis was performed in the 324 ESFT samples, but only informative cases were considered. Table 2 shows the frequencies of the immunohistochemical parameters of our series.
Table 2. Frequencies of immunohistochemical cell cycle and genetic markers in the ESFT series
In the immunohistochemical analysis, we included a variable that we denominated p53mut as a surrogate of those cases with a possible p53 point mutation. In this variable we considered that p53 is mutated when p53 is overexpressed and p21Waf1/Cip1, a direct effector of p53, is negative, as other authors have considered previously.12 According to this categorization, 15% of our cases presented a mutated p53 (Table 2). A direct association between the expression of Ki67 and mitotic count per 10 high power fields (HPF) was observed (Fig. 2).
No association between p21Waf1/Cip1 and p53mut and the ESFT subtypes was observed. However, p53 was more frequently expressed in PNET (45%) compared with conventional (22%) and atypical (23%) ES (p = 0.009). In a similar trend, Ki67 was more frequently expressed in PNET (47%) and atypical (39%) ES than in the conventional subtype (29%) (p = 0.05). In contrast, p27Kip1 expression was frequently observed in conventional (25%) and atypical (27%) ES than in PNET (8%) (p = 0.051).
FISH for 9p21 deletions and p53 alterations
One hundred and sixty-nine of the 324 ESFT cases were informative for the study of the 9p21 locus deletion, from which 34 (20%) presented homozygous (3 cases) or heterozygous (31) deletion (Table 2; Figs. 3a and 3b). No association with the immunohistochemical parameters was observed.
In the case of p53, 191 cases were informative for the analysis of p53 alterations by FISH. From these, 32 (17%) presented some genetic alteration, including: 26 cases with polysomy and 4 with monosomy of chromosome 17, and 2 cases with heterozygous deletion of the p53 locus. All these alterations were grouped in one group for statistical proposes (Table 2; Figs. 3c and 3d).
No association between the FISH results for p53 and the protein expression of p53 and p21Waf1/Cip1 was observed. However, those cases presenting a p53 FISH alteration expressed a p53mut in a higher proportion (22%) than those with no genetic alterations (12%), although these differences were not statistically significant (p = 0.140). Interestingly, a direct association between the FISH of p53 and Ki67 expression was observed. Hence, those cases with the highest expression of Ki67 showed a greater proportion of FISH alterations (24%) compared with the cases (13%) with no Ki67 expression (p = 0.044).
As some authors have reported previously,14, 15 the alterations of p53 and/or 9p21 locus defines a subgroup of cases with a more aggressive behavior. In our series, 58 (38%) out of the 154 FISH-informative cases showed an alteration of 9p21 or p53 locus. A concurrence of both alterations was reported in only 8 cases (5%). No association of this variable with the immunohistochemical markers was observed.
The association between the FISH analysis with the ESFT histological subtypes showed that PNET and atypical ES presented a higher proportion of p53 alterations, 28% and 21%, respectively, than the conventional ES (12%) (p = 0.080).
Association with tumor progression
Of the 324 tissues analyzed, 272 (84%) corresponded to primary localized tumors, 22 (7%) to disseminated primary tumors, 9 (3%) to recurrences and 21 (6%) to metastases (6%). Comparing between the analyzed parameters for the localized (primary tumors) and disseminated (disseminated primary tumors, recurrences and metastasis) cases, we observed a significant direct association between p53, p21Waf1/Cip1 and p27Kip1 expression in the disseminated cases (Table 3). In the case of FISH, only deletion of the 9p21 locus was higher in disseminated disease (30% vs. 18%), although this association was not statistically significant (p = 0.105).
Table 3. Frequencies of the expression of cell proliferation markers and genetic alterations of p53 and 9p21 locus in localized and disseminated disease
Association with follow-up
For the prognostic evaluation, we analyzed a cohort of 217 patients4 with localized disease, the main features of which are listed in Tables 1 and 3. Tables 4 and 5 show the results of the univariate and multivariate analysis for both PFS and DSS, respectively. The multivariate analysis revealed that Ki67 expression (p = 0.003) and the type of treatment (p = 0.003) constituted independent prognostic factors for PFS (Table 4; Fig. 4). The same trend was observed for DSS, where both variables remained as independent prognostic factors (Table 5; Fig. 4). The patients treated with radiotherapy represented the group with the worst prognosis, probably due to the presence of inoperable cases (n = 63).
Table 4. Log Rank and Cox regression tests for PFS
Abbreviations: NS, not significant; HR, hazard ratio; CI, confidence interval.
Table 5. Log Rank and Cox regression tests for DSS
Abbreviations: NS, not significant; HR, hazard ratio; CI, confidence interval.
To analyze prognostic behavior with respect to treatment, two groups of patients were considered. A surgery group composed of 76 patients who received surgery with chemotherapy and a radiotherapy group comprising 141 patients for whom radiotherapy was included in their treatment regimen (78 patients with surgery and 63 patients without surgery).
In the surgery group, only the histological variety of ES/PNET constituted a unique independent prognostic factor for both PFS and DSS. The atypical ES/PNET being the variety with the worst behavior (see reference Llombart-Bosch et al.,4 2009 for more details).
In the case of the group treated with radiotherapy, Ki67, p53 alterations detected by FISH and the presence of alterations in p53 or 9p21 locus provided prognostic information (Table 6; Fig. 5). However, only Ki67 expression constituted an independent unfavourable prognostic indicator in the multivariate analysis for both PFS and DSS (Table 6).
Table 6. Log rank and Cox regression tests for PFS and DSS in the group of patients treated with radiotherapy
Abbreviations: NS, not significant; HR, hazard ratio; CI, confidence interval.
In this study, we have observed that the cellular proliferation marker Ki67, also known as MKI67,27 was expressed in approximately one-third of the ESFT cases. Ki67 is strictly associated with cell proliferation. During interphase, Ki67 can be exclusively detected within the cell nucleus, whereas in mitosis, most of the protein is relocated to the surface of the chromosomes. Ki67 protein is present during all active phases of the cell cycle (G1, S, G2, and mitosis) but is absent from resting cells (G0). Furthermore, we have demonstrated that Ki67 constitutes an independent prognostic factor for PFS and DSS in localized, genetically confirmed, ESFT, independently of the treatment regime. Hence, those cases expressing Ki67 in more than 5% of nuclei presented a hazard ratio of 2.0 [1.3–3.1] and 1.9 [1.3–2.9] for PFS and DSS, respectively, being especially relevant in the group of patients with radiotherapy incorporated in their treatment schedules (Table 6). Although the prognostic implication of Ki67 has been well established in other sarcomas,28–30 it has not been confirmed so far in ESFT. Only one study performed by Sollazzo et al.31 on a series of 38 cases demonstrated that patients with tumors expressing Ki67 had a PFS and DSS of 25% and 50%, respectively, whereas compared with the negative cases, this was 68% and 90%, respectively. Furthermore, in our series, Ki67 is associated with alterations in the cell cycle regulators such as p53 whose clinicopathological involvement in the ESFT has been more extensively studied as discussed later.
Genetic alterations in genes involved in cell cycle regulation target at least one of two critical pathways implicated in cell cycle control: the pRb pathway that regulates G1 to S-phase transition and the p53 pathway that induces growth arrest or apoptosis in response to either DNA damage or inappropriate mitogenic stimuli.32 In ESFT, these alterations constitute part of a multistep process with equivalent cellular effects, of which the EWS-ETS genetic fusion seems to be the initiating mechanism of ESFT pathogenesis.33
We found expression of p53 in 25% of ESFT cases, being higher in PNET subtype (45%) than in conventional (22%) and atypical ES (23%). This finding, together with the fact that PNET and atypical ES express a higher proportion of Ki67 than conventional ES, confirms our previous observations that ESFT histological subtypes constitute differential entities defined not only at morphological but also at phenotypic and prognostic level.4
p53 expression was significantly more frequently observed in disseminated disease than in primary localized tumors (38% vs. 23%) indicating its role in tumor progression. Although this observation would suggest that p53 is associated with a more aggressive behavior, when we analyzed those ESFT patients with localized disease, no association with PFS and DSS was observed, which is in contrast with previous reports in smaller series of cases in which p53 expression constituted an independent prognostic factor for poor prognosis.11, 12 Several studies reported the impact of ESFT prognosis of p53 mutations, which are more consistently associated with poor prognosis in contrast to p53 expression.14–16 To find, among all cases expressing p53, those candidates for carrying a p53 mutation, we defined the variable p53mut as a surrogate marker. In this variable, we considered that p53 is mutated when p53 is overexpressed and p21Waf1/Cip1, a direct effector of p53, is negative as other authors considered previously.12 Although the percentage of cases with p53mut (10–15%) was similar to the frequencies of p53 mutations, no association with tumor progression and prognosis was observed.
We also evaluated the p53 locus status by FISH and found that 17% of cases presented chromosomal alterations of this locus gene, including polysomies and monosomies of chromosome 17 as well as heterozygous deletions of the p53 locus. Alterations of p53 detected by FISH were correlated with p53 immunohistochemistry, and as observed at protein level, were more frequently observed in the PNET subtype. No differences were observed between localized and disseminated disease, indicating that this alteration is not acquired during tumor progression. However, there was a significant association with both PFS and DSS, especially in the group of patients who received radiotherapy within their therapeutic schedule. To our knowledge, this is the first study that reports the prognostic involvement of the p53 locus alterations detected by FISH in ESFT.
The frequency of chromosomal alterations of the 9p21 locus detected by FISH in our series (20%) is consistent with the literature.8, 13, 17, 18, 34 In addition, these alterations were more frequently observed in disseminated (30%) than in localized disease (21%), supporting the idea that these alterations constitute a rather late event in ESFT tumorigeneis.8 However, no correlation between the deletions of the 9p21 locus and DSS or PFS was observed. This finding is consistent with Brownhill et al.34 who, in a series of 42 primary ESFT, found 9p21 alterations detected by multiplex ligation-dependent probe amplification (MLPA) in 10% of cases, and no association with the clinical parameters. However, these observations are in contrast to previous studies in ESFT. For example, Tsuchiya et al.,18 in a study of 24 ESFT, reported that cases carrying 9p21 locus alterations had a worse PFS than those without. Similarly, Wei et al.,13 analyzed a series of 39 EFST, finding that those with 9p21 deletions had a worse prognosis. A recent meta-analysis, however, demonstrated that the presence of 9p21 locus alterations is indicative of a poorer 2-year survival among patients with ESFT, constituting an independent predictor or poor outcome.8
For the detection of 9p21 deletions, we used a previously described23 FISH-based method, which can distinguish between homozygous and heterozygous deletion. In our series, the heterozygous deletions were more frequently observed than the homozygous (31 vs. 3 cases, respectively), this is in contrast with the study of Brownhill et al.34 that considers heterozygous deletion to be a rare event (only 5% of primary ESFT). This discrepancy could be due mainly to methodological limitations. For example, Brownhill et al.34 used an MLPA-based method, which is more sensitive and accurate than FISH because it amplifies and quantifies small portions of DNA (some base pairs). However, for the FISH approach, we used a probe spanning a region of 400 Kb, which in combination with the cutting effect of the paraffin-embedded material makes it difficult to identify the true cases with 9p21 homozygous deletion. In this regard, Savola et al.,35 using a comparative genomic hybridization (CGH)-array approach, pointed out that approximately 8% of EFST carry microdeletions (less than 190 Kb) of the 9p21 locus that can not be detected by conventional FISH.
The role of 9p21 alterations in tumorigenesis of ESFT and its biological contribution to poorer prognosis remains unclear. This locus codifies for two key cell cycle regulators (p16INK4a and p14ARF), which are frequently lost in cancer. The p16INK4a product interacts negatively with the cyclin-dependent kinase 4 (CDK4) which in turn phosphorylates pRb leading the G1 to S-phase cell cycle transition.32 Thus, deletion or loss of p16INK4a eventually accelerates cell cycle progression and cell proliferation. In addition, the loss of p14ARF leads to MDM2 activation, subsequently resulting in acceleration of cell cycle progression through p53 inactivation. It is speculated that the 9p21 alteration in ESFT is a rather late event that may accelerate tumor aggressiveness through inactivation of both the pRb and p53 pathways, and may link to a poor disease outcome or a poor treatment response.
When p53 and 9p21 locus alterations were combined as a single factor, we found that those cases carrying a p53 and or 9p21 locus alteration presented poor outcome, mainly in the group of patients who had been treated with radiotherapy alone or in combination with surgery and/or chemotherapy. Although this factor did not constitute an independent prognostic indicator after the multivariate analysis, the tendency herein observed is in accordance with previous studies performed in smaller series.14–16
We also analyzed the expression of other cell cycle regulators, particularly p21Waf1/Cip1 and p27Kip1, both belonging to the same family of cyclin-dependent kinase inhibitors. p21Waf1/Cip1 is activated by p53 to induce cell cycle arrest in response to DNA damage36 and in some studies, as in our case, it has been used as surrogate for identifying cases with p53 mutations.12 Although we have found that this protein is highly expressed in disseminated disease, no association with prognosis has been observed.
On the other hand, Matsunobu et al.,37 in a series of 21 EFTS, reported that p27Kip1expression is associated with a better prognosis. In our series, expression of p27Kip1 was observed in 23% of cases and was associated with tumor progression. However, in contrast to Matsunobu's study, no prognostic information was provided by p27Kip1 in localized disease.
At prognostic level, older age, pelvic primary tumor, large tumors, metastatic disease and gene-fusion type have been associated with worse outcome.38–41 However, the relative importance of some of these factors has diminished with better systemic and local control. The most important prognostic factors for patients with localized disease have been shown to be local and systemic treatment with both surgery and radiotherapy.38 In the series of 217 cases with localized disease, the type of treatment was the most important prognostic factor both for PFS and DSS, the group of patients treated with radiotherapy (± surgery and/or chemotherapy) having a worse prognosis, probably due to the presence of inoperable cases (n = 63). For this reason, we evaluated the prognostic impact of the cell cycle regulators together with other clinicopathological and molecular parameters, including the EWS-ETS gene fusion type, in these two groups, and although parameters such as 9p21 deletions or p53 locus alterations were indicators of poor outcome in the univariate analysis, only Ki67 expression constituted an independent adverse prognostic marker. Hence, and although we accept the retrospective nature of the study and some limitations regarding the availability of some clinical information, we consider that Ki67 constitute a valuable indicator of poor prognosis in localized ESFT and strongly recommend its validation in a prospective series.