Prognostic significance of p16INK4a alteration for Ewing sarcoma

A meta-analysis

Authors

  • Kanya Honoki MD, PhD,

    Corresponding author
    1. Department of Orthopedic Surgery, Nara Medical University, Nara, Japan
    2. Center for Clinical Epidemiology and Biostatistics, School of Medicine and Public Health, the University of Newcastle, Newcastle, New South Wales, Australia
    • Department of Orthopedic Surgery, Nara Medical University, 840 Shijo-cho, Kashihara, Nara 6634-8521; Japan
    Search for more papers by this author
    • Fax: (011) 81-744-25-6449

  • Elizabeth Stojanovski PhD,

    1. Discipline of Statistics, School of Mathematics and Physical Science, the University of Newcastle, Newcastle, New South Wales, Australia
    Search for more papers by this author
  • Mark McEvoy PhD,

    1. Center for Clinical Epidemiology and Biostatistics, School of Medicine and Public Health, the University of Newcastle, Newcastle, New South Wales, Australia
    Search for more papers by this author
  • Hiromasa Fujii MD,

    1. Department of Orthopedic Surgery, Nara Medical University, Nara, Japan
    Search for more papers by this author
  • Toshifumi Tsujiuchi PhD,

    1. Laboratory of Cancer Biology and Bioinformatics, Department of Life Science, Faculty of Science and Technology, Kinki University, Kinki, Japan
    Search for more papers by this author
  • Akira Kido MD,

    1. Department of Orthopedic Surgery, Nara Medical University, Nara, Japan
    Search for more papers by this author
  • Yoshinori Takakura MD,

    1. Department of Orthopedic Surgery, Nara Medical University, Nara, Japan
    Search for more papers by this author
  • John Attia MD, PhD

    1. Center for Clinical Epidemiology and Biostatistics, School of Medicine and Public Health, the University of Newcastle, Newcastle, New South Wales, Australia
    Search for more papers by this author

Abstract

BACKGROUND.

Despite findings from individual studies regarding prognostic factors for Ewing sarcoma, no conclusive results have been produced, partly because of small sample sizes. The objective of the current study was to evaluate whether the presence of p16INK4a alteration is associated with a poorer prognosis in patients with Ewing sarcomas.

METHODS.

A review was conducted of publications that assessed associations between p16INK4a status and 2-year survival among patients with Ewing sarcoma. The association between metastatic disease at initial diagnosis and 2-year survival was evaluated by synthesizing data in the form of risk ratios.

RESULTS.

Of 11 studies that were identified in the initial search strategy, 6 studies, representing 188 patients, met the inclusion criteria and, consequently, were pooled for quantitative analyses. The estimated pooled risk ratio of p16INK4a aberration was 2.17 (95% confidence interval [95% CI], 1.55–3.03; P < .001), whereas the estimated pooled risk ratio of metastasis at diagnosis among the 164 eligible patients was 2.60 (95% CI, 1.71–3.97; P < .001). There was no statistically significant difference in the pooled estimated risk ratios of p16INK4a aberration for a poor prognosis between patients with and without metastasis at diagnosis (1.86 and 2.21, respectively; P > .59).

CONCLUSIONS.

The presence of p16INK4a alteration was a statistically significant predictor of prognosis for patients with Ewing sarcoma. Along with other prognostic factors, such as metastasis, the p16INK4a alteration may be a potential candidate for improving the risk-stratifying strategy for patients with these tumors. © 2007 American Cancer Society.

Ewing sarcoma is the second most common type of bone tumor in childhood and is associated with a poor prognosis.1 Recent advances in multidisciplinary treatments have dramatically improved the prognosis for patients with Ewing sarcoma. However, even among patients with localized extremity tumors at diagnosis, between 30% and 40% of patients develop recurrent disease within 4 years and die of the disease despite current aggressive protocols.2 Risk-adapted chemotherapy regimens for patients with Ewing sarcoma include EURO-EWING-99 and other modifications.3 These regimes include the type of risk-adapted treatment strategies that depend on tumor location (pelvic vs nonpelvic), tumor volume (<200 mL vs >200 mL), age (<18 years vs >18 years), metastatic pattern (pulmonary vs extrapulmonary), type of EWS-FLI1 fusion gene, and histologic response to chemotherapy.

At the molecular level, at least 90% of Ewing sarcomas involve fusion genes that comprise EWS-FLI1, whereas another 5% to 10% involves the EWS-ERG gene.4 These alterations have been recognized for their diagnostic role in detecting Ewing sarcoma. Recent studies have unveiled some direct contribution of the EWS-FLI1 fusion transcript to Ewing sarcoma oncogenesis. The EWS-FLI1 fusion protein, which results from chromosomal translocation t(11;22)(q24;q12), behaves as an aberrant transcriptional regulator5 that induces oncogenic proteins, such as MYC6 and MMP-3,7 while repressing cell cycle regulators p21/CIP1/WAF1,8p57KIP,9 and transforming growth factor β2 receptor (TGFβ2).10 The suppression of the EWS-FLI1 fusion protein results in reduced cell growth in vitro and tumor-forming capacity in vivo.11 However, induction of tumorigenicity by EWS-FLI1 itself is dependent on the cellular background. Thompson et al. demonstrated that EWS-FLI1 elicited enhanced tumorigenicity in immortalized NIH3T3 cells,12 whereas others demonstrated an induction of p53-dependent growth arrest in primary human fibroblasts13 and apoptosis in mouse embryonic fibroblasts.14 Recent studies demonstrated that transfection of the EWS-FLI1 fusion gene induced tumorigenicity, which produces the small, round, blue cell tumors from primary bone marrow-derived mesenchymal progenitor cells. Those studies also demonstrated that the disruption of either the p53, p16INK4a, or p19(p14)ARF gene was unnecessary for the induction of tumorigenicity by EWS-FLI1.15 Although it is presumed that these fusion genes are the initiating oncogenic events in the development of Ewing sarcoma, there is no established link between this genetic alteration and either prognosis or response to treatment with chemotherapy or radiation.16 The additional genetic alteration of p53 and p16INK4a loss may facilitate subsequent cellular transformation, which may lead to further aggressive tumor progression.

Furthermore, a subpopulation of patients with Ewing sarcoma has genetic alterations of p53, p16INK4a, or p19(p14)ARF. Recent studies have indicated the presence of the p53 gene alteration in 4% to 13% of patients17–23 and the p16INK4a gene alteration in 12% to 26% of patients23–29 with Ewing sarcoma. 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.17–29 It has been suggested that the presence of these alterations potentially is linked with a poor prognosis for patients with this disease.21–23, 27, 29 In addition, the presence of metastasis at diagnosis defines an advanced stage of the disease and has been considered the most unfavorable prognostic factor.30 However, there have been discrepancies in the observed associations between p16INK4a or p53 alterations and advanced-stage disease.25, 26

Ohali et al., in a study of 31 patients with Ewing sarcoma, identified telomerase activity as a significant prognostic variable that was superior to the clinical prognostic variables they examined.31 A recent study on gene expression profiling suggested that genes related to cell cycle regulation (such as CDK2) were associated with a poor prognosis.32 Adding to this finding, genes related to invasion and metastasis (such as MTA1) were up-regulated, whereas tumor suppressor genes (such as FHIT) and apoptosis inducers (such as TGFβ1) were down-regulated among patients with who had a poorer prognosis.32 However, clinical and biologic characteristics alone fail to classify patients with Ewing sarcoma accurately 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 Ewing sarcoma, there is a need for novel molecular prognostic parameters in addition to the conventional factors (tumor volume and chemotherapy-induced necrosis, among others). Thus, a quantitative synthesis is warranted, and a meta-analysis is proposed in the current report of the available studies that correlate the p16INK4a alteration with responses to chemotherapy and/or survival.

MATERIALS AND METHODS

Search Strategy

A literature search was conducted to identify publications relating to the association of p16INK4a alterations, including deletion, mutation, and lack of RNA or protein expression, with Ewing sarcoma family of tumors (ESFT). The MEDLINE and PubMed databases were searched for the period up to and including March 2006 to locate related publications. We used the following search terms: “Ewing's sarcoma,” “Ewing's sarcoma family of tumors,” and “ESFT” combined with “p16,” “p16INK4a,” “p16 deletion,” “p16 mutation,” and “9p21 gene.” All searches were limited to the English language. Citations that included the key terms in either the title, abstract, or article or Medical Subject Heading (MeSH) terms were retained. References of retrieved articles also were screened for relevance.

Qualitative Data Extraction (Inclusion and Exclusion Criteria)

Citations that included abstracts were used to select publications that assessed the relation between p16INK4a status and chemotherapy response and/or survival at 2 years postdiagnosis, because all studies were based on at least 2 years of follow-up data. Methods for the detection of p16INK4a alterations included Southern blot analysis, fluorescence in situ hybridization (FISH), and differential polymerase chain reaction (PCR) analysis for genomic deletion; reverse transcriptase (RT-PCR) and immunohistochemistry (IHC) for messenger RNA and protein expression; single-strand conformational polymorphism for mutation analysis; and methylation-specific PCR for methylation status, all of which result in loss of function for p16INK4a.

A quality checklist form was developed (Table 1), and articles that were identified from the literature search were checked against this from to ensure that they contained the required information, which related directly to the assessment of a molecular association with the prognosis for patients with Ewing sarcoma. Studies that had total scores <3 (ie, scores of 0, 1, or 2) were considered ineligible and, thus, were excluded from further analysis.

Table 1. Scale for Quality Assessment of the Studies
CriteriaScore
Representativeness of patients
 Consecutive/randomly selected from patient population with clearly defined sampling frame2
 Consecutive/randomly selected from patient population without clearly defined sampling frame or with extensive inclusion/exclusion criteria1
 No method of selection described0
Ascertainment of prognostic outcome for Ewing sarcoma
 Clearly described objective criteria for prognostic diagnosis of Ewing sarcoma1
 Not described0
Genotyping examination
 Genotyping done under “blinded” condition1
 Unblinded or not mentioned0
Association assessment
 Assess association between genotypes and prognosis with appropriate statistics and adjustments for confounders2
 Assess association between genotypes and prognosis without appropriate statistics and adjustment for confounders1
 Inappropriate statistics used0
Total6

Quantitative Data Extraction

For each article that was identified, the author name(s), journal, and year; the number of patients enrolled and the number analyzed; the disease stage and grade of Ewing sarcoma; demographic characteristics; the types of chemotherapy and surgery used; the type of measurements used to determine p16INK4a status and the presence of metastasis at diagnosis; the antibodies used for IHC; the exons analyzed with RT-PCR; and definitions used for positive tests were recorded, as shown on the data-extraction form (Fig. 1). Two investigators (K.H. and H.F.) extracted data from eligible studies independently, discussed discrepancies, and reached at least 90% agreement on all criteria used. Data regarding the main outcomes were then tabulated in 2 × 2 tables demonstrating the occurrence/nonoccurrence of death within 24 months according to p16INK4a status and metastasis at diagnosis.

Figure 1.

The data-extraction form.

Statistical Analysis

All statistical analyses were performed using Stata for Windows statistical software (version 8.0).33 Data concerning the predictive ability of p16INK4a alterations were combined across the 6 studies using risk ratios (RRs). RRs were calculated with 2-year mortality as the outcome by comparing the group with p16INK4a alterations against the group without p16INK4a alterations. The prognostic significance of the presence of metastasis also was assessed across the 5 studies that used RRs with 2-year mortality as the outcome among patients with metastasis versus patients without metastasis. The homogeneity of separate study RR estimates was assessed using Q-statistics and the Cochran chi-square value.34 Homogeneity indicates that variation among effect sizes (in this instance, RRs) is caused by sampling error alone. In the absence of significant heterogeneity, at the .05 significance level (which occurs if the associated P value is >.05), RRs were combined using a fixed-effects model.35 If the homogeneity of separate study estimate assumptions was rejected, then a random-effects model would be used to combine study estimates of the RR. The overall pooled RR estimates, along with 95% confidence intervals (95% CIs), were calculated for each outcome. Study estimates, along with pooled estimates, are presented as forest plots. The effect of publication bias on the reported outcomes was assessed graphically using funnel plots and empirically using regression tests according to the method reported by Egger et al.36 Publication bias indicates that there is a greater possibility that a study with statistically significant results will be published relative to a study with results that are not statistically significant. This type of bias is present if the studies that are published are not representative of all studies that have been conducted, which would indicate an inference that is inaccurate.

RESULTS

Eligible Studies

There were 11 studies identified that examined the role of p16INK4a status among patients with Ewing sarcoma. Five studies were excluded because of a lack of information for the purposes of the current review; 2 of those studies were review articles,37, 38 and the other 3 studies were excluded because of insufficient clinical data.25, 39, 40 There remained 6 independent eligible studies, which were based on 188 patients with available data on 2-year survival status. The primary outcome, 2-year survival status, is referred to below as death of disease (DOD).

The modalities that were used for the detection of p16INK4a aberrations were PCR in 2 studies,24, 28 Southern blot analysis in 4 studies,23, 24, 26, 27 methylation-specific PCR in 2 studies,26, 29 FISH,23 and immunohistochemistry28 in the remaining study (Table 2). The prevalence of p16INK4a alteration was between 13.3% and 31% among these studies.

Table 2. Characteristics of Eligible Studies
StudyYearTotal no. of patientsNo. of patients with p16 alterationsMethod for detecting p16INK4a alterationOutcome of DOD
No. of patients with normal p16No. of patients with altered p16
  1. DOD indicates died of disease; FISH, fluorescence in situ hybridization; SB, Southern blot analysis; PCR, polymerase chain reaction; MSP, methylation-specific PCR; IHC, immunohistochemistry; SSCP, single-strand conformational polymorphism; RT, reverse transcriptase.

Huang et al., 2005232005608FISH and SB14/524/8
Lopez-Guerrero et al., 2001292001196Differential PCR, MSP6/124/6
Maitra et al., 2001282001204IHC4/163/4
Wei et al., 2000272000397SB9/326/7
Tsuchia et al., 2000262000244SB, MSP9/202/3
Kovar et al., 1997241997278SSCP, SB, RT-PRC5/196/8

Table 3 presents the demographic characteristics of patients from the eligible studies. The median patient age was 16.5 years (range, 0–72 years). Information regarding tumor location was available in 4 studies: Tumors were located in the axis in 70 patients (51%) and in the peripheral region in 67 patients (49%). Metastatic disease at diagnosis was present in 27% of patients (range, 10–78%) among the 5 studies that had information on metastasis. Details of the treatment (chemotherapy and surgery) are presented in Table 3, but some studies lacked information on either or both of those variables.

Table 3. Demographic Data of the Patients From the Eligible Studies
StudyMedian age (range), yTumor locationNo. of patients with metastasis [%]ChemotherapySurgery
  1. A indicates axis; P, peripheral; V, vincristine; D, doxorubicin; C, cyclophosphamide; I ifosfamide; E, etoposide, NA, not available; B, bleomycin.

Huang et al., 20052317 (1–72)A, 32; P, 286 [10]60/60 (VDCIE)NA
Lopez-Guerrero et al., 20012916.5 (9–38)A, 6; P, 1214 [78]18/18 (details NA)18/18
Maitra et al., 20012812.5 (2–16)A, 12; P, 85 [25]NANA
Wei et al., 20002722 (3–51)A, 20; P, 197 [18]39/39 (VDCIE or VDCB)38/39
Tsuchia et al., 20002615 (0–30)NANANANA
Kovar et al., 199724NANA12 [44]27/27 (VACD, VAID)NA

Summary RRs

DOD outcome by p16INK4a

For the current meta-analysis, DOD was compared between patients with and patients without the p16INK4a alteration. Figure 2 displays the RRs and associated 95% CIs for the eligible studies. The heterogeneity test for RRs among these studies was not statistically significant (chi-square value = 4.09; P = .536), supporting a fixed-effects summary estimate. The fixed-effects Mantel-Haenszel pooled RR of DOD for these comparisons was 2.17 (95% CI, 1.55–3.03; P < .001). This indicates that death is statistically significant more likely and, hence, the prognosis is poorer for patients with p16INK4a alteration. Publication bias was not evident based on theregression analysis described by Egger et al. (r = 1.3844; P = .4831). This finding is supported by the funnel plot based on all 6 studies that is displayed in Figure 3.

Figure 2.

The study-specific risk ratios (black squares) with the 95% confidence interval (95% CI) of p16INK4a alteration for 2-year survival in each eligible study and the pooled estimate (diamond). The size of the black square in each study indicates the weight of the study in the meta-analysis.

Figure 3.

Funnel plots for the 6 studies that were eligible for the category of 2-year survival by p16INK4a status. SE indicates standard error.

DOD outcome by metastasis

There were 5 studies that were eligible for the analysis of the relation between metastasis at diagnosis and DOD (n = 162 patients) (Table 4). Between-study heterogeneity of RRs was not observed among these studies (chi-square value = 4.06; P = .398). Figure 4 presents the RRs of DOD and associated 95% CIs for these comparisons among the eligible studies. The fixed-effects Mantel-Haenszel pooled RR was 2.60 (95% CI, 1.71–3.97; P < .001), suggesting a statistically significant, higher death rate or a poorer prognosis for patients who had metastasis compared with patients who did not have metastasis at diagnosis.

Figure 4.

The study-specific risk ratios (black squares) with the 95% confidence interval (95% CI) of metastasis at diagnosis for 2-year survival in each eligible study and the pooled estimate (diamond). The size of the black square in each study indicates the weight of the study in the meta-analysis.

Table 4. The Proportion of Patients With and Without Metastasis at Diagnosis Who Died of Disease
StudyOutcome of DOD: no. of patients
With metastasis at diagnosisWithout metastasis at diagnosis
  1. DOD indicates died of disease.

Huang et al., 2005235/613/54
Lopez-Guerrero et al., 20012910/140/4
Maitra et al., 2001283/54/15
Wei et al., 2000276/79/32
Kovar et al., 1997246/125/13

DOD outcome by p16INK4a and metastasis

The correlation between p16INK4a alteration and DOD in patients with and without metastasis also was assessed. The individual study RRs and the pooled RRs, along with the associated 95% CIs, of DOD (ie, poorer prognosis) in patients with and without metastasis are presented in Figures 5a,b, respectively. The Mantel-Haenszel pooled RR of p16INK4a alteration for DOD in patients with metastasis was 1.86 (95% CI, 1.15–3.01; P = .041); whereas, for patients without metastasis, the pooled RR was 2.21 (95% CI, 1.23–3.98; P = .008). There was no statistically significant difference observed in the pooled RRs between patients with and without metastasis at diagnosis (chi-square value = 0.28; P = .5945), indicating that the presence of metastasis is not a confounder of the association between p16INK4a and DOD.

Figure 5.

The study-specific risk ratios (black squares) with 95% confidence intervals (95% CI) and the pooled estimate (diamond). The size of the black square in each study indicates the weight of the study in the meta-analysis. (a) The risk ratio of p16INK4a for 2-year survival in patients with metastatic disease at diagnosis. (b) The risk ratio of p16INK4a for 2-year survival in patients without metastatic disease at diagnosis. (c) The risk ratio of metastasis at diagnosis for 2-year survival in patients without p16INK4a alterations. (d) The risk ratio of metastasis at diagnosis for 2-year survival in patients with p16INK4a alterations.

Outcome by metastasis at diagnosis and p16INK4a status

Finally, the significance of the presence of metastasis at diagnosis for DOD outcome in each group of patients with and without p16INK4a alteration was assessed. The Mantel-Haenszel pooled RR of metastasis for the DOD outcome in patients without p16INK4a alteration was 2.43 (95% CI, 1.35–4.36; P = .003), as shown in Figure 5c. This indicates a statistically significant, poorer prognosis for patients with metastasis at diagnosis without p16INK4a alteration. In contrast, the Mantel-Haenszel pooled RR of metastasis for DOD outcome in patients with the p16INK4a alteration was 1.92 (95% CI, 0.99–3.25; P = .055), as shown in Figure 5d, suggesting a borderline, nonstatistically significant relation between metastasis at diagnosis and a poorer prognosis among patients with the p16INK4a alteration.

DISCUSSION

Results from the current meta-analysis demonstrate that the presence of p16INK4a alterations is indicative of a poorer 2-year survival among patients with Ewing sarcoma. Loss of p16INK4a is the second most common molecular genetic alteration observed in Ewing sarcoma, occurring in 20% to 25% of patients.27 The prevalence of patients with p16INK4a alteration complied from the current study was 19.7%, a proportion similar to that reported in previous studies.23–29 The sample size considered for the current analysis was 188 patients (including 38 patients with p16INK4a alteration) and was sufficient to detect an effect size of 2.0 at 80% statistical power and 5% significance level. The estimated pooled RR of p16INK4a alteration for 2-year survival was statistically significant (2.17; 95% CI, 1.55–3.03), suggesting that p16INK4a alteration is a strong predictor of poor prognosis among patients with Ewing sarcoma. This supports findings from previous studies.23, 24, 27 However, those studies had limited statistical power, because they were based on relatively small samples.

The pooled estimate RR of metastasis at diagnosis on 2-year survival was 2.60 (95%CI, 1.71–3.97), indicating that metastasis at diagnosis also was predictive of a poorer prognosis. The relation between p16INK4a alteration and disease stage could not be assessed because of the insufficient information presented in separate studies. However, there were no statistically significant differences in estimated RRs of p16INK4a alteration for disease outcomes between patients with and without metastasis at diagnosis (pooled estimated RRs, 1.86 and 2.21, respectively; P > .05). The p16INK4a alteration appears to be a significant and independent predictor of a poorer prognosis in patients with Ewing sarcoma, irrespective of whether patients have advanced-stage disease with metastasis.

The results of this meta-analysis appear reasonably robust. First, the pooled estimates do not indicate any evidence of heterogeneity. Second, the regression analysis described by Egger et al. and the funnel plot do not suggest the presence of substantial publication bias. Conversely, there was some unavoidable variability in defining methods, measurements, and outcomes in each study, such as the variety of treatment strategies and modalities for detecting p16INK4a alteration. Nevertheless, the results from alterations detected using any of the methods can be considered reasonably accurate in terms of the loss of function of p16INK4a. Finally, the estimates that were obtained were not adjusted for other parameters that may relate to Ewing sarcoma outcomes, such as tumor location (axis vs peripheral), tumor volume, treatment regimens, and response to chemotherapy, which classically have been proposed as predictive factors.2, 3 In particular, the lack of detailed information regarding the types of chemotherapy and surgical resectablity may be a strong confounder in the current meta-analysis. Multivariate models should be considered in future studies to include such potential predictors, and their prognostic significance should be confirmed in a prospective cohort setting with identical treatment strategy. However, because Ewing sarcomas are extremely rare on a population basis, the sample size of the current investigation is among the largest to date in the field.

The role of p16INK4a alteration in the tumorigenesis of Ewing sarcoma and its biologic contribution to a poorer prognosis remains unclear. The p16INK4a product interacts negatively with CDK4; thus, deletion or loss of function of p16INK4a would be equivalent to amplification of CDK4, which eventually accelerates the cell cycle progression and cell proliferation.41, 42 In addition, the simultaneous loss of p14ARF, which is a biproduct from the same locus of p16INK4a, leads to MDM2 activation, subsequently resulting in acceleration of cell cycle progression through p53 inactivation. It is speculated that the p16INK4a alteration in Ewing sarcoma is a rather late event that may accelerate tumor aggressiveness through inactivation of both the RB and p53 pathways and may link to a poor disease outcome or a poor treatment response. Further studies are needed to clarify this issue.

In conclusion, the presence of the p16INK4a alteration is a significant and independent predictor of a poorer outcome in patients with Ewing sarcoma, and this alteration may be one of the key candidates for the assessment of patient prognosis. However, it will be necessary to define the prognostic assessment system in combination with possible multivariate risk factors to determine the best risk-adapted strategies against Ewing sarcoma to improve the prognosis for subpopulations of patients who have generally poor outcomes.

Acknowledgements

We thank Drs. Marc Ladanyi and Yasuhiko Kaneko for kindly providing detailed information through personal communication.

Ancillary