Previous reports of Ewing sarcoma cohorts suggested that there is a difference in incidence according to racial origin. However, to the authors' knowledge, this finding has never been tested in a population-based database, and the impact of race on clinical outcome and the significance of known risk factors stratified to racial groups have not been reported.
Patients who had Ewing sarcoma diagnosed between 1973 and 2005 were identified in the Surveillance, Epidemiology, and End Results database. Patient demographic and clinical characteristics; incidence; year of diagnosis; tumor location, tumor size, and disease stage at diagnosis; treatment(s); cause of death; and survival were extracted. Kaplan-Meier, log-rank, and Cox regressions were used to analyze the significance of prognostic factors.
Race-specific incidence indicated that Caucasians have the highest incidence (0.155), followed by Asians/Pacific Islanders (0.082), and African Americans (0.017). The difference in incidence between Caucasians and African Americans was 9-fold and significant (P < .001). The incidence of Ewing sarcoma increased over the past 3 decades among Caucasians (P < .05). Survival was not impacted by race. Local disease stage, primary tumor location in the appendicular skeleton, and tumor size ≤8 cm conferred a significant survival benefit. Women demonstrated improved survival among the Caucasian patients (P < .03).
The Ewing sarcoma (ES) family of tumors includes ES of bone, peripheral primitive neuroectodermal tumor, extraosseous ES, and Askin tumor (ES of the chest wall).1, 2 In the era when only local disease control could be addressed with surgery and radiation therapy, the survival of patients with ES was devastating. Despite local tumor control, >90% of patients died because of systemic tumor burden within 2 years of diagnosis.3 It was only after the introduction of combination chemotherapy (vincristine, actinomycin D, and cyclophosphamide) in the 1970s that an improvement in survival was reported.4 The need to establish multi-institutional cooperation was realized and lead to the establishment of the Cooperative Ewing Sarcoma Study, Children's Oncology Group, European Ewing Tumor Working Initiative of the National Group (EURO-EWING 99) across Europe and the United States.5-8 These multi-institutional groups contributed toward the validation of newer chemotherapeutic regimens and strategies, such as the report about the addition of ifosfamide and etoposide under National Cancer Institute intergroup protocol INT-0091 (Children's Cancer Group [CCG]-7881 and Pediatric Oncology Group [POG]-8850), which was opened to all member institutions of the CCG and the POG.9 Treatment strategies combining adjuvant and neoadjuvant chemotherapy with local resection and/or radiation therapy continue to evolve, and individual centers have reported an improvement in the overall survival rate from 10% to 65%.10 The small number of patients in individual studies limited the ability of those studies to identify risk factors for survival and did not permit an assessment of improved outcomes over the previous 4 decades.11-15 However, it has been determined that Caucasians are affected much more frequently than Asians or Africans.16-18 To our knowledge, there are no reports from a population-based cancer registry verifying this observation, and no investigation has addressed the potential impact of race on outcome. This striking discordance between racial proclivities implores the stratification of prognostic factors with respect to race.
Previously, the population based Surveillance, Epidemiology, and End Results (SEER) database was used to describe outcomes for breast, colorectal, prostate, lung, ovarian, and neuroectodermal cancers and was validated with regard to accuracy.19-25 A recent study described the incidence and survival of patients with ES using the SEER data but lacked analyses of several important prognostic factors.26 The population that was studied was limited to the pediatric cohort, and no analysis was done to elucidate racial disparities in prognostic factors. For the current study, we investigated racial differences in the incidence of ES using a population-based cancer registry. Race as a potential risk factor in ES was assessed, and the significance of known risk factors was stratified by racial origin.
MATERIALS AND METHODS
The SEER Program of the National Cancer Institute was established as a direct result of the National Cancer Act of 1971. Currently, SEER collects data from 17 population-based registries covering approximately 26% of the US population. It is the only comprehensive source of population-based data in the US that includes the stage of cancer at the time of diagnosis and follow-up of all patients for survival data. In addition, each registry collects data on patient demographics, primary tumor site and morphology, and first course of treatment (within 4 months of diagnosis).27-39 The SEER Program currently is regarded as the standard of quality among cancer registries around the world with case completeness of 98%.40
The SEER database was used to identify all patients who were diagnosed with ES from 1973 to 2005 using International Classification of Disease for Oncology, third edition.41 In total, 1631 patients were identified, and information regarding patient demographics, stage at diagnosis, primary tumor site and size, year of diagnosis, surgical and radiation treatment (if provided within 4 months of diagnosis), and length of survival (in months) until death or loss to follow-up were extracted. Because of reporting omissions, data on tumor size were available for only 789 patients (48.4%). Percentages were based on available data for each individual variable. Patients with missing data were excluded from each respective univariate and multivariate analysis.
Patient age was converted into a categorical variable (ages 0-24 years, 25-59 years, and >60 years) for the purpose of analysis. The appendicular skeleton included long and short bones of the limbs and associated joints and the scapula. The axial skeleton included vertebra, ribs, sternum, clavicle and associated joints, bones of the skull, face and associated joints, mandible, and pelvic bones. Staging categories of local, regional, and distant were used, and size was converted into a categorical variable (≤8 cm vs >8 cm) according to the American Joint Committee on Cancer (AJCC) staging system.42 In the SEER database, local disease is defined as tumor within the confines of the periosteum, regional disease is defined as tumor confined within a fascial compartment (eg, fascia lata) but breaching the integrity of the periosteum, and systemic stage is defined as metastatic disease. The staging information was reported by the individual cancer registry and was recorded under “SEER historic staging A,” “SEER modified AJCC staging,” and “SEER summary staging.” The staging information is explained further in the SEER Summary Staging Manual 2000.43 Year of diagnosis was categorized in 4 categorical variables: 1973 to 1975, 1976 to 1985, 1986 to 1995, and 1996 to 2005. The results reported herein are in compliance with the Health Insurance Portability and Accountability Act of 1996.
Incidence rates were age adjusted and normalized using the 2000 US standard population.44 The annual percentage change (APC) was calculated using the weighted least-squares method described in the SEER Cancer Statistics Review (1973-2005).45 Statistical analysis was performed using SPSS version 16.0 (SPSS Inc., Chicago, Ill). The chi-square test was used to identify correlations between categorical variables. Overall and disease-specific survival (DSS) from the time of initial diagnosis to the date of last contact (or the date of death) was calculated using the Kaplan-Meier method. The effects of demographic, clinical, pathologic, and treatment variables were tested using the log-rank test for categorical values. A multivariate analysis was carried out to identify independent prognostic factors using the Cox proportional hazards model. All prognostic factors that were identified as significant in the univariate analysis (namely, sex, stage, primary site, size, and surgical therapy) were included in the multivariate analysis.
In total, 1631 cases of ES were reported in the SEER database from 1973 to 2005. More than half of the cases (52.9%) were diagnosed during the decade from 1996 to 2005. The demographic and clinical characteristics of the entire cohort are summarized in Table 1. The most common age group at diagnosis was ages 0 to 24 years: This group comprised 76.3% of the cases, and men comprised 60.3%.; Race was predominantly Caucasian (91.6%), and ethnicity was predominantly non-Hispanic (86.3%). Among cases for which tumor size was identifiable, 56.1% of tumors measured >8 cm in greatest dimension at the time of diagnosis. Overall, 40.2% of patients had regional disease at the time of diagnosis. The most common tumor location was the axial skeleton (44.1%). Twenty-seven percent of patients underwent surgical resection alone, 26% underwent surgery and also received radiation, 26% received radiation without surgery, and 16% received no attempts at local disease control.
Table 1. Demographic and Clinical Characteristics of Entire Cohort
No. of Patients
Valid % of Total
Year of diagnosis
The overall incidence for ES in 2005 was 0.128 per 100,000 population and had not changed significantly since 1973 (P > .05) with an APC of 0.568% (Fig. 1a). Compared with other races, Caucasians had a significantly higher incidence of ES, and this incidence had increased over the past 3 decades. The population-adjusted incidence by race was 0.155 (Fig. 1b) for Caucasians and 0.017 for African Americans (0.083 per 100,000 population). Caucasians had a significant increase in incidence over time (P < .05) with an APC of 0.844% (Fig. 1b). APC values for the other 2 race categories could not be determined, because those groups had fewer patients (difference not significant [dns]). An age analysis of incidence indicated that most patients fell into the group ages 10 years to 19 years (Fig. 1c).
The 5-year and 10-year DSS for patients with ES is summarized in Table 2. Kaplan-Meier prediction of survival of patients with ES was not affected by race (Fig. 2a). The overall DSS rate was 55% at 5 years, 53% at 10 years, and plateaued thereafter (dns). Women had significantly higher 5-year and 10-year survival rates (5-year survival: 59% vs 52% for men; P = .004) (Fig. 2b). Patients who were diagnosed with local disease had significantly better 5-year survival rate of 71% compared with 60% for patients who were diagnosed with regional disease and a mere 35% for patients with distant disease (P < .001). Tumors ≤8 cm carried a significantly better prognosis with a 5-year survival rate of 66% compared with 48% for tumors >8 cm. Axial lesions had a worse prognosis with a 5-year survival rate of 48% compared with 60% for appendicular lesions (P < .001) and 66% for extraskeletal lesions (P = .004). Patients who underwent surgery fared significantly better than patients who did not undergo resection (5-year survival rate, 64% and 45%; P < .001). No significant improvement in survival could be observed over time when the analysis was stratified by decade for the past 2 decades (P = .786).
Table 2. Disease-specific Survival According to Demographic and Clinical Characteristics (Proportion Surviving)
P values shown were determined by using the log-rank test between variables.
Ages 0-24 years versus aged >60 years only, P = .666; ages 0-24 years versus ages 25-59 years, P = .116; ages 25-59 years versus aged >60 years, P = .930.
Caucasian versus other only, P = .434; Caucasian versus black, P = .334; black versus other, P = .195.
Extraskeletal versus axial only, P = .004; extraskeletal versus appendicular, P = .786; appendicular versus axial, P < .001.
For 1973-1975 versus 1996-2005 only, P < .001; 1973-1975 versus 1976-1985, P = .220; 1973-1975 versus 1986-1995, P = .001; for 1976-1985 versus 1986-1995, P = .003; 1976-1985 versus 1996-2005, P = .002; 1986-1995 versus 1996-2005, P = .786.
Table 3 illustrates the step-wise multivariate analysis using the Cox proportional hazards model to ascertain the independently significant variables for the entire cohort. The parameters distant disease stage, primary location in the axial skeleton, and size >8 cm were all independent predictors of lower overall survival.
Table 3. Cox Proportional Hazards Model for the Risk of Death From Ewing Sarcoma: Multivariate Analysis
The same analysis was performed segregating patients into Caucasians, African Americans, and Asians/Pacific Islanders. The Cox proportional hazards model did not reveal any variable that was significant prognostically for the non-Caucasian groups (data not shown). The small numbers of patients in each of these 2 latter categories limited the statistical power of analysis.
The demographic and clinical characteristics for Caucasians followed trends similar to those described for the entire cohort. It is noteworthy that, only among Caucasians, female sex emerged as an independent predictor of improved survival (P = .031), whereas the prognostic significance of the other parameters (size, stage, and location) were unchanged from their significance for the entire patient population.
The current study is the first to our knowledge that demonstrates a significant racial difference in the incidence of ES in a large, US population-based cancer database. Older, nonpopulation-based studies have appreciated that Caucasians are affected much more frequently than Asians and Africans.16-18 To our knowledge, there are no reports from a population-based cancer registry verifying this observation, and no investigations have addressed the potential impact of race on the outcome of patients with ES. A recent study described incidence and survival of ES patients using SEER data but lacked the analysis of several important prognostic factors.26 The population that was studied was limited to the pediatric cohort, and no analysis was done to elucidate racial disparities in prognostic factors. Recently, Damron et al. attempted to give an account of descriptive epidemiology and survival data for ES using the National Cancer Data Base of the American College of Surgeons.46 Despite an unmatched number of cases (4870), their study lacked information in some key areas, such as patient demographics and other clinical information. In addition, the registry was not population based and, thus, was not representative of the average cross section of patients treated for ES. Epidemiologic studies comparing SEER areas with non-SEER areas in the United States concluded that the age and sex distributions were comparable, except that SEER areas tended to be more affluent and more urban than non-SEER areas. Compared with the National Program of Cancer Registries (NPCR) and United States Cancer Statistics, incidence rates for all sites combined were lower in SEER. However, for the category of interest, bones and joints, the differences were small (0.6 per 100,000 population in SEER vs 0.8 per 100,000 population in the NPCR among black men was the largest reported difference).
Caucasians had a 9-fold higher incidence of ES (0.155) compared with African Americans (0.017). An increase in the incidence of ES over time (P < .05) also was noted only among whites. This increase was not demonstrated in other racial groups. The reason for the increase in incidence remains unclear. Viral and environmental etiologies have been discussed in the literature as potential etiologies for ES.47-49 The increase in incidence is affected by time and race. Environmental factors and exposures to potential mutagens or viruses change over time, and Caucasians could carry a genetic propensity to respond to specific kinds of mutagens that cause ES. Further sex-related mutagenic and viral exposure studies will have to clarify this relation.
The large, population-based database that we used allowed us to make a detailed analysis of survival risk factors in ES. Aside from well known and multiply validated, independent risk factors that warrant improved survival, including local disease stage, primary tumor location in the appendicular skeleton, and size ≤8 cm, this large, population-based database identified male sex as a previously unidentified and significant risk factor for limited survival for Caucasian patients with ES.10, 12, 50-60 To our knowledge, this has never been demonstrated in a large, population-based study. Damron et al. reported a predilection of ES for male sex.46 The current analysis revealed a very similar male-to-female ratio of 6:4. Male sex has been associated with a poor outcome, whereas others have reported conflicting evidence.12, 61 Female sex had borderline significance as an independent prognostic factor for improved survival on multivariate analysis for the entire cohort (P = .051) (Table 3). This association became stronger when we analyzed the Caucasian population (P = .031) (Table 4). It is unclear why male sex puts patients at risk for limited survival. Further genetically based studies will have to clarify this matter. Male sex is determined by the Y chromosome and by the lack of a second X chromosome. Genes on these 2 chromosomes may lead to sex-related differences in outcome among patients with ES.
Table 4. Cox Proportional Hazards Model for the Risk of Death From Ewing Sarcoma Specific for Caucasians: Multivariate Analysis for the White Cohort
No. of Patients
HR indicates hazard; CI, confidence interval.
No tumor size data were available for 1973-1975, resulting in censoring of all patients in this bracket.
Many previous studies attempting to identify the prognostic factors for ES were based on data from a single large center. Most of them actually included only a particular subset of patients with respect to clinical features, resulting in bias and conflicting evidence for various prognostic factors.11-15 Metastasis at the time of diagnosis has been recognized widely as a predictor of worse prognosis.12, 62 Other factors like age, sex, tumor site and size, various hematologic indices, factors related to chemotherapy, and type of local treatment were not associated consistently with prognosis in the literature.12
Scurr and Judson comprehensively reviewed conflicting evidence in the literature regarding age as an independent prognostic factor.63 They concluded that it is uncertain whether adult patients have worse outcomes compared with their younger counterparts. The current analysis revealed no statistically significant difference in outcome between the 3 age groups neither for the entire cohort nor for the Caucasian population (Table 2). Thus, we support the recommendation of Scurr and Judson of using treatment protocols for adult patients that are as aggressive as the treatment protocols used for the pediatric population. The current results are in conformity with this commonly held belief, and patients with local disease had significantly better outcomes than patients with distant disease both among Caucasians and in the entire cohort.
Investigators have reported that a poor prognosis in ES with an axial location, and this is 1 of the well established prognostic factors in the literature.11, 64 The current results support these observations. Appendicular location was recognized as an independent determinant of improved survival on multivariate analyses that were carried out for the entire cohort (P = .011) and for Caucasians (P = .005). This can be correlated with an improved ability to achieve wide margins in lesions with an appendicular location compared with an axial location.
Bacci et al. reported significantly shorter survival for patients who had tumor volumes <700 mL on univariate analysis.12 Because of the small number of patients with tumor volumes >700 mL in their cohort, the variable “volume” was not included in the multivariate analysis, and the significance of volume as an independent predictor of prognosis could not be ascertained. Others have reported conflicting results for size as an independent predictor of prognosis.54, 65-68 The current analysis revealed that size with at arbitrary cutoff of 8 cm was an independent predictor of prognosis for the entire cohort as well as for the Caucasian cohort. Further analysis revealed that significantly more patients who were diagnosed with metastatic disease had primary tumors >8 cm (148 of 226 patients). Similarly, a significantly higher percentage of patients who had axial tumors (34.9%), compared with patients who had appendicular tumors (24.3%), presented with distant disease. There were no significant differences among axial, appendicular, and extraskeletal tumors with respect to size.
Patients who underwent any sort of surgical treatment had a significant improvement in prognosis on univariate analysis, but this observation was not significant on multivariate analysis for the entire cohort or for Caucasians. Further analysis revealed that significantly more patients (in the entire cohort) who had primary tumors >8 cm did not undergo surgery (202 of 306 patients; dns), suggesting that this group had a higher incidence of inoperable lesions. An analysis of the entire cohort and of Caucasians revealed that no improvement in DSS as a result of resection was an independent predictor of survival on multivariate analysis.
Limitations of the current study include the lack of any information on specific chemotherapy or any other medical therapy in the SEER database. Thus, we are unable to comment directly on the survival benefit conferred by the use of chemotherapy or the efficacy of a particular regimen in a particular subset of patients. Similarly, no information regarding any medical history, radiologic studies, or serologic workup is provided in the database, which limits our analysis on the prognostic significance of fever, duration of symptoms, tumor volume, anemia, hypoalbuminemia, and high lactate dehydrogenase. The accuracy of staging information can be a potential pitfall in all studies based on databases. In the current study, staging information has been extracted from the SEER database as was recorded at the time of reporting by the registry. The lack of any radiologic record makes it impossible to verify the stage at diagnosis. Having acknowledged these limitations, the current article brings novel observations to the literature.
The current analysis of incidence stratified by race supports the long held belief that ES occurs more commonly among Caucasians than among Asians/Pacific Islanders and is less common among African Americans.62 Despite this striking discordance in incidence of the disease, no survival advantage was exhibited by a particular race on univariate analysis. Prognostic factor analysis with respect to race was limited by the small number of patients in categories other than “Caucasians.” The Caucasian population followed trends similar to the trends observed in entire cohort for most of prognostic factors that were analyzed. A noteworthy difference was the statistically significant survival advantage exhibited by Caucasian women on multivariate analysis. Female sex had only borderline significance as an independent prognostic factor for the entire cohort. A closer look at the survival advantage conferred by sex among the different races revealed that this phenomenon was limited to Caucasians. African-American and Asian/Pacific Islander women exhibited no DSS advantage on a univariate analysis (data not shown); in fact, they exhibited slightly poorer outcomes (statistically insignificant) compared with their male counterparts. The reason for this is unknown, but it explains the difference observed among Caucasians and for the entire cohort with respect to sex as an independent predictor of prognosis.
If we were to accept the idea that the chromosomal translocation characterizing ES is causative in the disease process,69, 70 then explanations for the racial disparities in incidence and sex disparities in outcome could be explained by differences in genetic replication machinery. Furthermore, if different translocation sites carry different clinical prognosis, then it is possible that molecular differences may underlie the observed sex and racial disparities reported.71-73 The underlying molecular basis of this hypothesis remains unknown. In the future, establishment of a prospective national database would be of critical importance for the evaluation of molecular markers. Despite these limitations, the current study constitutes a significant step toward identification of independent demographic and clinical factors associated with improved survival and clarifies some of the associated controversies in incidence patterns that could have an impact on the treatment of patients with ES.
Conflict of Interest Disclosures
None of the authors had any financial disclosures to make.