Analysis of prognostic factors in ewing sarcoma family of tumors

Review of St. Jude Children's Research Hospital studies




Advances in systemic and local therapies have improved outcomes for patients with the Ewing sarcoma family of tumors (ESFT). As new treatments are developed, a critical review of data from past treatment eras is needed to identify clinically relevant risk groups.


The authors reviewed the records of 220 patients with ESFT who were treated on protocols at St. Jude Children's Research Hospital from 1979 to 2004. Two treatment eras were defined. Factors predictive of outcome were analyzed to identify distinct risk groups.


The median age at diagnosis was 13.7 years (range, 1.1–25.2 years). Metastatic disease was associated with tumors measuring >8 cm (P = .002) and axial location (P = .014). The 5-year overall survival (OS) estimate (63.5% ± 3.5%) did not appear to differ by protocol. Tumor stage and size were found to be the only independent predictors of outcome. Treatment era and type of local control therapy were found to influence the outcome of patients with localized disease. Four risk groups were defined: favorable risk (age <14 years with localized, nonpelvic tumors), intermediate risk (localized, age ≥14 years, or pelvic tumors), unfavorable-pulmonary (isolated lung metastases), and unfavorable-extrapulmonary (extrapulmonary metastases). The 5-year OS estimates for these groups were 88.1% ± 4.4%, 64.9% ± 5.2%, 53.8% ± 9.4%, and 27.2% ± 7.3%, respectively (P < .001). The incidence of therapy-related leukemia was significantly higher during the second treatment era, when more intensified regimens were used (6.1% ± 2.7% vs 0% ± 0%; P = .005).


Risk stratification schemes such as this should be used to prospectively evaluate novel risk-based therapies. Studies of biologic pathways may help to refine this model. Cancer 2007. © 2007 American Cancer Society.

The Ewing sarcoma family tumors (ESFT) are comprised of small round cell neoplasms of neuroectodermal origin that vary in their neurogenic differentiation. ESFT comprise 3% of all pediatric malignancies and are the second most common pediatric malignant bone tumor.1

The past 30 years have witnessed great improvements in the outcome of patients with ESFT, largely through multidisciplinary approaches tested in cooperative trials. American and European studies have defined active agents and their optimal schedules and combinations; new agents and improved supportive measures have allowed treatment intensification. Improved local control (better radiotherapy [RT] planning and more aggressive surgery) also has contributed.2 The disease-free survival of patients with localized disease now approaches 70%, and overall survival (OS) may exceed 80%.2, 3–5 However, the outcome of metastatic patients continues to be very poor.

Although most protocols treat all patients with localized disease similarly, they all may not require intensive therapy. Conversely, not all patients with metastatic disease have a dismal prognosis. Known risk factors such as age, tumor size and site, metastatic pattern, and histologic response to chemotherapy may define subgroups requiring differing treatment intensities.2 To identify clinically relevant risk groups, we analyzed St. Jude studies conducted over the past 25 years.


Patients and Treatment

We identified 222 patients with ESFT who were treated on 5 consecutive studies at St. Jude Children's Research Hospital between February 1979 and January 2004. Two patients with undifferentiated neuroectodermal and desmoplastic small round cell tumors were excluded from analysis; 220 patients were included. This retrospective study was approved by the Institutional Review Board.

The ES79 protocol (1979–1986) used vincristine (V), dactinomycin (A), fractionated cyclophosphamide (C) (at a dose of 150 mg/m2/day for 7 consecutive days), and doxorubicin (D).6 The ES87 protocol (1987–1991) tested ifosfamide (I) and etoposide (E) in the pretreatment window and added it to continuation therapy, alternating with VCD.7 The EWI92 protocol (1992–1996) evaluated the feasibility of aggressive early induction with VCDIE and prolonged intensified maintenance therapy with the same agents.3 After 1996, patients with localized ESFT were treated on the Pediatric Oncology Group (POG)-Children's Cancer Group (CCG) cooperative POG9354/CCG7942 study5 until September 1998; these patients were not included in our analysis. After September 1998, localized disease was treated as on the standard arm of the cooperative study (St. Jude Best Clinical Management [SJBCM]), with alternating courses of VCD plus IE. Between 1996 and 2000, the high-risk sarcoma (HIRISA)-1 and HIRISA-2 protocols used high-dose, short-term intensification regimens followed by myeloablative chemotherapy and autologous stem cell rescue for patients with high-risk disease (tumors measuring ≥8 cm, pelvic primary tumor, or metastatic disease). Therapy was comprised of 6 courses of VCDIE.8 Table 1 summarizes the cumulative doses per protocol.

Table 1. Cumulative Doses Per Protocol
  1. SJBCM indicates St. Jude Best Clinical Management; HIRISA, high-risk sarcoma.


Local control varied with the protocol and the patient. In the ES79 and ES87 protocols, patients with completely resected tumors received no RT. In the ES79 protocol, a dose of 30 to 35 grays (Gy) of RT was given to patients with either unresected lesions with no evidence of a macroscopic residual soft-tissue tumor on computed tomography, masses that were proven by biopsy not to contain active tumor, or microscopic residual disease after surgery. Patients with macroscopic residual soft-tissue tumors or evidence of active tumor received 50 to 55 Gy of RT. In the ES87 protocol, patients with unresectable small tumors (measuring <8 cm) or microscopic residual disease after surgery received 35 Gy; patients with unresectable large tumors received 60 Gy. In the EWI92 study, the RT dose was determined by tumor location (soft tissue or bone), size (≥8 cm or <8 cm), response to chemotherapy, and extent of resection. Patients with completely resected tumors received either no RT (bone) or 36 Gy (soft tissue). Patients with microscopic residual disease after surgery received either 36 Gy (after at least a partial response to chemotherapy) or 60 Gy, hyperfractionated RT (after a lesser response). Patients with unresectable tumors received either 68.4 Gy hyperfractionated RT or 55.8 Gy (tumors measuring <8 cm and at least a partial response). The SJBCM protocol called for a dose of 55.8 Gy for patients with unresectable tumors or macroscopic residual disease and 45 Gy for patients with microscopic residual disease or inadequate tumor margins. The primary site of the tumor received 36 to 68 Gy in the HIRISA trials, and metastatic sites were also irradiated.

Statistical Analysis

The duration of OS was defined as the interval between diagnosis and death from any cause or last follow-up. The duration of event-free survival (EFS) was defined as the interval between diagnosis and disease progression, recurrence, second malignancy, death from any cause, or most recent follow-up. The OS and EFS distributions were estimated using the Kaplan-Meier method. Factors were examined as predictors of OS or EFS using the Mantel-Haenszel (log-rank) test or the exact log-rank test. The Cox proportional hazards model was applied when evaluating continuous variables (ie, age) and multiple factors as predictors of OS and EFS.

The time of local failure was defined as the interval between diagnosis and local or regional disease recurrence or progression. The cumulative incidence (CI) of local and distant recurrence, second malignancy, and second leukemia was estimated using the methods of Kalbfleisch and Prentice.9 The Gray test was used to compare the CI among groups.10

The interrelation of factors was examined using the Fisher exact test, chi-square test, Wilcoxon rank-sum test, or Kruskal-Wallis test. When an expected count was <5, exact methods were used. No adjustment for multiple comparisons was made in this exploratory study. SAS software (version 9.1; SAS Institute Inc, Cary, NC) and StatXact software (version 5; StataCorp, College Station, TX) were used for statistical analysis.


Patient Characteristics and Treatment

Patient, disease, and treatment characteristics are shown in Table 2. Most patients had axial primary tumors (n = 141 patients; 64%). The majority of tumors (61%) measured ≥8 cm, and most (71%) were localized at the time of diagnosis. The 64 patients who presented with metastatic disease had metastases in the lung (45 patients [lung only, 26 patients]), bone (36 patients), bone marrow (15 patients), or other sites (3 patients).

Table 2. Patient, Disease, and Treatment Characteristics (n = 220)
CharacteristicNo. (%)
  1. BM indicates bone marrow; HIRISA, high-risk sarcoma; SJBCM, St. Jude Best Clinical Management; Gy, grays.

Age at diagnosis, y
 Range1.1 – 25.2
 <14114 (51.8)
 ≥14106 (48.2)
 Male134 (60.9)
 Female86 (39.1)
 White205 (93.2)
 Other15 (6.8)
 Localized156 (70.9)
 Metastatic64 (29.1)
Site of metastasis(N = 64)
 Lung only26 (40.6)
 Lung/BM/bone/other38 (59.4)
Tumor size, cm(N = 218)
 <886 (39.4)
 ≥8132 (60.6)
Primary site
 Axial141 (64.1)
 Extremities79 (35.9)
Primary site
 Pelvic49 (22.3)
 Nonpelvic171 (77.7)
Protocol treatment
 ES7980 (36.4)
 ES8746 (20.9)
 EWI9250 (22.7)
 HIRISA11 (5.0)
 SJBCM33 (15.0)
Type of local control(N = 218)
 Surgery alone42 (19.3)
 Radiation alone121 (55.0)
 Surgery plus radiation55 (25.2)
Dose of radiation, Gy(N = 175)
 <4096 (54.9)
 ≥4079 (45.1)

Nearly all patients (218 patients) received local control therapy initially or after early disease progression; 42 patients (19%) received surgery alone, 121 patients (55%) were treated with RT alone, and 55 patients (25%) received both. Thus, 176 patients received radiation. Data regarding the dose of radiation were available for 175 patients: 96 received <40 Gy and 79 patients received ≥40 Gy. Histologic tumor response was available for only 52 patients (24%) and therefore could not be analyzed.

Distribution of Characteristics by Protocol

There were significant differences across protocols with regard to race, stage of disease, and type of local control. The HIRISA protocols had the largest proportion of nonwhite patients (27% vs <15%; P = .008) and, not surprisingly, of patients with metastatic disease (64% vs <37%; P = .017). When HIRISA patients were excluded, the frequency of metastatic disease demonstrated no significant difference among protocols (P = .137).

The type of local control also differed significantly across protocols (P = .006) (Table 3). At least half of the patients on the ES79, ES87, and EWI92 protocols received RT alone, compared with approximately one-third of patients on the HIRISA and SJBCM regimens. On the most recent regimen (SJBCM), nearly two-thirds of patients received surgery or surgery plus RT.

Table 3. Local Control Per Protocol
No.No. (%)P
  • SJBCM indicates St. Jude Best Clinical Management.

  • *

    Two patients did not undergo local control.

  • Overall P value.

  • P value of comparing surgery ± radiation among protocols.

      (N = 218)
Surgery alone16 (20)5 (11)6 (12)3 (27)12 (36) 
Radiation alone53 (66)27 (61)25 (50)4 (36)12 (36).006
Surgery plus radiation11 (14)12 (27)19 (38)4 (36)9 (27).046

Interrelation of Disease Factors

As shown in Table 4, axial tumor site was associated with a greater likelihood of metastasis at the time of diagnosis (P = .014), as were pelvic tumor site (P = .008) and larger (≥8 cm) tumor size (P = .002). Age was found to be only marginally associated with tumor size (P = .071) but was not found to be associated with tumor site (P = .61) or stage of disease (P = .82).

Table 4. Association Between Clinical Factors
FactorsNo. of patients (%)P
 Localized92 (65)64 (81) 
 Metastatic49 (35)15 (19).014
 Localized129 (75)27 (55) 
 Metastatic42 (25)22 (45).008
 Size < 8 cmSize ≥ 8 cm 
 Localized70 (82)84 (63) 
 Metastatic15 (18)49 (37).002
Age, y
 Range1.7–21.31.1 – 25.2.071

Follow-up and Outcome

Of the 220 patients studied, 123 (56%) were still alive at the time of last follow-up, a median of 11.7 years (range, 3.6 months to 25.5 years) after diagnosis. First adverse events included second malignancy (8 patients), disease recurrence/progression (95 patients), and death (8 patients, 2 of whom died of treatment-related toxicities). Ninety-five patients had disease recurrence/progression that was distant (37 patients;, 39%), locoregional (41 patients; 43%), or a combination (17 patients; 18%) at a median of 1.6 years (range, 50 days to 11.9 years) after diagnosis. The 5-year and 10-year OS estimates were 63.5% ± 3.5% and 56.0% ± 4.1%, respectively. The 5-year and 10-year EFS estimates were 55.1% ± 3.6% and 50.3% ± 4.3%, respectively. (Fig. 1). The OS and EFS rates demonstrated no significant differences between protocols (Table 5).

Figure 1.

Overall survival and event-free survival for 220 patients with the Ewing sarcoma family of tumors. Solid line indicates overall survival; dashed line, event-free survival.

Table 5. Univariate Analysis of Factors Predictive of Outcome
FactorsNo.Overall survival, SE (%)Event-free survival, SE (%)
  • SE indicates standard error; RR, relative risk; 95% CI, 95% confidence interval; NA, not applicable; SJBCM, St. Jude Best Clinical Management; HIRISA, high-risk sarcoma.

  • *

    P values were obtained by the log-rank test except when indicated.

  • P value was obtained by Cox regression analysis.

  • Comparison was based on 212 patients; 6 patients who received local control therapy after disease progression were excluded.

Age RR (95% CI).107*RR (95% CI).093
 Continuous2201.034 (0.993–1.078) 1.034 (0.995–1.074) 
 <14 y11469.9 (4.6)62.3 (5.6) 59.4 (4.9)53.6 (5.8) 
 ≥14 y10656.6 (5.1)49.2 (5.9).07150.3 (5.2)46.7 (6.1).131
 Female8667.2 (5.4)57.0 (6.4) 55.3 (5.8)50.3 (6.7) 
 Male13461.2 (4.5)55.4 (5.3).9754.9 (4.6)50.3 (5.5).85
 Nonwhite1555.6 (16.6)55.6 (21.4) 50.8 (17.8)50.8 (25.2) 
 White20564.1 (3.5)56.4 (4.2).5655.4 (3.7)50.5 (4.3).69
 ES798065.0 (5.3)57.4 (5.5) 56.3 (5.5)48.6 (5.6) 
 ES874652.2 (7.2)45.5 (7.3) 41.3 (7.1)41.3 (7.3) 
 EWI925066.0 (6.6)57.9 (9.4) 54.0 (6.9)49.8 (9.8) 
 HIRISA1163.6 (15.7)NA 63.6 (15.7)NA 
 SJBCM3374.7 (12.5)74.7 (37.6).3675.6 (13.2)75.6 (37.3).163
Treatment era (1)
 ES79/ES8712660.3 (4.3)53.0 (4.5) 50.8 (4.4)45.9 (4.5) 
 EWI92/SJBCM/HIRISA9468.1 (5.6)61.7 (9.5).2360.8 (6.0)56.7 (10.3).062
Treatment era (2)
 ES798065.0 (5.3)57.4 (5.5) 56.3 (5.5)48.6 (5.6) 
 ES87/EWI92/SJBCM/HIRISA14062.4 (4.5)55.2 (6.2).8054.1 (4.8)51.8 (6.5).57
 Localized15674.2 (3.8)67.9 (4.6) 66.0 (4.1)60.9 (4.9) 
 Metastatic6438.2 (6.1)27.4 (6.7)<.00128.7 (5.7)24.9 (7.2)<.001
Primary site
 Axial14161.2 (4.3)52.0 (5.0) 53.5 (4.4)48.5 (5.1) 
 Extremity7967.6 (5.8)64.1 (6.9).09557.7 (6.3)53.6 (7.5).27
Primary site
 Nonpelvic17168.2 (3.8)59.8 (4.8) 59.4 (4.1)55.5 (5.0) 
 Pelvic4947.5 (7.3)43.0 (7.4).05239.5 (7.2)32.6 (6.9).012
Tumor size (N = 218)
 <8 cm8575.9 (4.9)64.3 (6.2) 62.6 (5.6)61.2 (6.4) 
 ≥8 cm13355.6 (4.5)50.7 (5.4).01049.8 (4.6)42.8 (5.5).004
Site of metastasis (N = 64)
 Lung only2653.8 (9.4)36.3 (11.0) 38.5 (9.1)28.8 (12.2) 
 Other3827.2 (7.3)20.7 (7.5).13521.8 (6.8)21.8 (7.9).66
Local control therapy (N = 218)
 Surgery only4273.9 (7.9)70.0 (9.9) 71.2 (8.3)71.2 (10.2) 
 Radiation only12152.9 (4.6)44.5 (5.1) 44.1 (4.7)37.9 (5.1) 
 Surgery plus radiation5583.0 (5.4)75.8 (7.5).00375.4 (6.2)71.1 (8.0).001

We analyzed the impact of the incorporation of IE into the ES87 protocol and of intensified local control and chemotherapy in the EWI92 protocol. We first compared the ES79 protocol with those that incorporated IE and found no significant difference with regard to outcome (5-year EFS of 56.3% ± 5.5% vs 54.1% ± 4.8% [P = .57] and 5-year OS of 65.0% ± 5.3% vs 62.4% ± 4.5% [P = .8]). We then compared the ES79 and ES87 protocols with the more intensive EWI92, SJBCM, and HIRISA protocols. Intensified chemotherapy and more aggressive local control measures were found to improve patient outcomes, but not significantly (5-year EFS of 50.8% ± 4.4% vs 60.8% ± 6.0% [P = .062] and 5-year OS of 60.3% ± 4.3% vs 68.1% ± 5.6% [P = .23]).

Analysis of Prognostic Factors

Table 5 summarizes the univariate analysis of predictive factors. Tumor size <8 cm (P ≤ .010) and localized disease at the time of diagnosis (P < .001) were found to be significantly associated with improved OS and EFS. Survival estimates decreased in a linear manner with increasing age at diagnosis (Fig. 2). Patients who received treatment with surgery or surgery plus RT were found to have better outcomes than those who were treated with RT only. Among patients receiving definitive RT for local control, outcome was a function of tumor size and RT dose.

Figure 2.

Overall survival and age.

A multiple Cox regression model that included all variables found to be significant at a nominal level of .20 analyzed age (≥14 years vs 14 years), protocol, tumor site (extremity vs axial; pelvic vs nonpelvic), tumor size (≥8 cm vs <8 cm), and stage (metastatic vs localized disease) for OS and age, protocol, disease stage, and tumor size for EFS. Only tumor size and disease stage remained significant predictors of both OS and EFS. Large tumor size was found to be associated with a lower OS (P = .042) and EFS (P = .023) after adjustment for stage of disease. Similarly, higher OS (P < .001) and EFS (P < .001) rates were found to be associated with localized disease, independent of tumor size.

We then analyzed prognostic factors separately for patients with localized versus metastatic disease. Neither subset demonstrated a significant difference in OS or EFS by protocol. In patients with localized disease, younger age was associated with a better outcome; the 5-year OS for patients aged <14 years and those aged ≥14 years was 82.7% ± 4.6% versus 65.2% ± 5.8% (P = .020). Only the type of local control was found to be associated with both OS and EFS. The 5-year EFS for patients treated with RT alone, surgery alone, and combined surgery and RT was 55.3% ± 5.6% versus 71.8% ± 9.0% versus 83.9% ± 6.4% (P = .005). The OS rate was 65.0% ± 5.4% versus 77.1% ± 8.3% versus 92.1% ± 4.7% (P = .044). Because lower doses of RT were given in the earlier studies (ES79 and ES87 protocols), we also compared outcomes in those 2 studies with later studies in which more aggressive local control was used. The 5-year EFS and OS rates were better on the later protocols (74.8% ± 6.6% vs. 59.8% ± 5.1% [P = .031] and 83.1% ± 5.6% vs 68.5% ± 4.8% [P = .068], respectively).

In 64 patients with metastatic disease, the OS and EFS rates did not appear to differ significantly by age, protocol, or treatment era. Patients with isolated lung metastases fared somewhat better than patients with extrapulmonary metastases. The 5-year EFS rate for patients with localized disease, isolated lung metastases, and extrapulmonary metastases was 66.0% ± 4.1% versus 38.5% ± 9.1% versus 21.8% ± 6.8% (P < .001). The OS was 74.2% ± 3.8% versus 53.8% ± 9.4% versus 27.2% ± 7.3% (P < .001).

Second Malignancies

Eight patients had a second malignancy as their first event, a median of 3.3 years after diagnosis (range, 1.8–19.6 years). One additional patient developed a second malignancy after recurrence of ESFT. The CI of a second malignancy was 2.4% ± 1.1% at 5 years and was 3.7% ± 1.4% at 10 years. Three patients had solid tumors (2 patients had osteosarcoma within the radiation field, and 1 patient had carcinoma of the uterine cervix after RT of a pelvic tumor).

Six patients had leukemia after ESFT. Patients treated on the ES79, ES87, and HIRISA protocols did not develop treatment-related myelodysplastic syndrome or acute myeloid leukemia (t-MDS/AML), but 3 cases were reported in the EWI92 protocol and 2 cases were observed in the SJBCM protocol. One additional patient who was treated on the ES79 protocol developed acute lymphoblastic leukemia (ALL) 19.6 years after the diagnosis of ESFT. The estimated CI of t-MDS/AML at 5 years and 10 years was 2.4% ± 1.1%. When the treatment eras were compared, the 5-year CI of t-MDS/AML was 0.0% ± 0.0% for the ES79 and ES87 protocols, and 6.1% ± 2.7% for the EW92, SJBCM, and HIRISA protocols (P = .005).

Investigation of Models for Risk Stratification

We explored models of risk stratification (Table 6). We first included all patients with metastatic disease in the unfavorable risk group, and we used the results of prognostic factor analysis for patients with localized disease to distinguish between favorable-risk and intermediate-risk groups. In Model 1, we defined the favorable-risk group by localized disease and age <14 years, the intermediate-risk group by localized disease and age ≥14 years, and the unfavorable-risk group by metastatic disease. OS was found to differ significantly between the favorable-risk and intermediate-risk groups (P = .020), but the EFS did not (P = .121). We then introduced tumor size (<8 cm vs ≥8 cm) into the model (Model 2) and subsequently added tumor site (pelvic vs nonpelvic) (Model 3). Model 2 demonstrated no significant difference between the favorable-risk and intermediate-risk groups, but this distinction was improved in Model 3. Building on Model 3, we analyzed patients with isolated lung metastases separately, thus defining 4 risk groups (Model 4). The outcome of patients with isolated lung metastases was intermediate between patients with extrapulmonary metastases and intermediate-risk patients (Table 6) (Fig. 3). The outcome of the favorable-risk groups did not differ significantly among the protocols (data not shown).

Figure 3.

Overall survival by risk group.

Table 6. Investigation of Risk Groups
FactorsSubgroupSubgroup no.Overall survival (SE), %Event-free survival (SE), %
  • SE indicates standard error; P, pulmonary; EP, extrapulmonary.

  • *

    All overall P values were significant (<.001). P values in Models 1–3 compare outcome distribution between favourable-risk and intermediate-risk patients.

  • Student t test for paired data: 2 vs. 1.

  • Student t test for paired data: 3 vs. 2.

  • §

    Student t test for paired data: 4 vs. 3.

Model 1
 FavorableLocalized < 14 y8082.7 (4.6)76.5 (5.9) 71.2 (5.5)64.4 (6.7) 
 IntermediateLocalized ≥14 y7665.2 (5.8)58.8 (6.8).02060.4 (6.0)57.2 (6.9).121
 HighMetastatic6438.2 (6.1)27.4 (6.7) 28.7 (5.7)24.9 (7.2) 
Model 2
 FavorableLocalized <14 y, <8 cm4183.7 (6.3)74.0 (8.2) 71.1 (7.8)68.0 (8.8) 
 IntermediateLocalized ≥14 y or ≥8 cm11370.6 (4.5)65.4 (5.4).2863.8 (4.8)58.0 (5.7).23
 HighMetastatic6438.2 (6.1)27.4 (6.7) 28.7 (5.7)24.9 (7.2) 
Model 3
 FavorableLocalized <14 y, non-pelvic6388.1 (4.4)79.9 (6.4) 76.3 (5.9)69.7 (7.5) 
 IntermediateLocalized ≥14 y or pelvic9364.9 (5.2)59.8 (6.0).01259.0 (5.5)54.9 (6.1).024
 HighMetastatic6438.2 (6.1)27.4 (6.7) 28.7 (5.7)24.9 (7.2) 
Model 4
 FavorableLocalized <14 y, non-pelvic (1)6388.1 (4.4)79.9 (6.4) 76.3 (5.9)69.7 (7.5) 
 IntermediateLocalized ≥14 y or pelvic (2)9364.9 (5.2)59.8 (6.0).01259.0 (5.5)54.9 (6.1).024
 High-PMetastatic – lung only (3)2653.8 (9.4)36.3 (11.0).05438.5 (9.1)28.8 (12.2).009
 High-EPMetastatic – extrapulmonary (4)3827.2 (7.3)20.7 (7.5).135§21.8 (6.8)21.8 (7.9).66§


The treatment of ESFT has evolved over the last decades; systemic treatments have become more intensive, and local control measures more aggressive. As we advance toward the new generation of studies, a critical evaluation of our current understanding of the treatment of this malignancy must be performed, and the relative contribution of each of the therapeutic components should be analyzed. In this review of the experience of the St. Jude Children's Research Hospital with the treatment of ESFT over the last 20 years, we have analyzed the advances made with each consecutive study.

Advances in the Treatment of ESFT

The 5 consecutive institutional studies are representative of the advances identified internationally in the management of this neoplasm. Treatments built on the efficacy of the 4-drug regimen that was used in the ES79 protocol,6 first by adding the IE pair (the ES87 protocol),7 and then by intensifying therapy and increasing cumulative doses with the addition of granulocyte–colony-stimulating factor (G-CSF) and improvements in support measures (the EWI92, SJBCM, and HIRISA protocols).3 As systemic approaches evolved, so did local control techniques. Although nearly two-thirds of patients in the earlier studies received definitive RT for local control, the majority of patients currently are treated with either surgery or a combination of surgery and RT.

The ES79 protocol was a variant of the 4-drug VACD regimens that were used by American and European cooperative groups.11–16 Similar to the ES79 protocol, surgery was seldom used, and RT alone was used in >75% of the patients.11, 12, 15, 17–19 The ES87 protocol built on the evidence indicating the efficacy of IE,20 and incorporated this pair into a VACD backbone.7 However, using similar local control approaches to the ES79 protocol, the addition of IE in the ES87 protocol did not appear to improve the outcomes. Contemporary to the ES87 protocol were 2 multi-institutional randomized studies that investigated the impact of adding etoposide to the VACD and the VAID (in which ifosfamide replaced cyclophosphamide) regimens.4, 21 In the POG8850/CCG7881 trial, all patients were randomized to receive VACD with or without IE.21 Among patients with metastatic disease, the addition of IE did not prove to be advantageous. However, for patients with nonmetastatic disease, the VACD/IE regimen was found to be superior to the standard VACD regimen (5-year EFS of 69% ± 3% vs 54% ± 4% respectively [P = .005]).21 The ES87 protocol demonstrated slightly inferior results for patients with localized disease compared with the experimental arm of the cooperative study (5-year EFS rate of 58.1% ± 8.6% vs 69% ± 3%, respectively), but likewise, the outcome for patients with localized disease on the ES79 protocol was superior to that reported on the standard treatment arm of the cooperative group study (5-year EFS rate of 60.7% ± 6.2% vs 54 % ± 4%, respectively).

ESFT are very sensitive to alkylating agents, which have a very steep dose-response curve, and the next generation of studies aimed at increasing the total cumulative doses of the active agents, as well as intensifying therapy by increasing the number of doses per cycle and per unit of time, with G-CSF support (Table 1). This approach was evaluated by the EWI92 protocol.3 The results of this study were superior to the prior 2 regimens; however, only 66% of patients completed therapy.3 The importance of dose intensification was also evaluated in the POG9354/CCG7942 trial, in which patients were randomized to receive alternating courses of VCD and IE over either 48 weeks or 30 weeks. The early results of this trial demonstrated no significant difference in outcome between the standard and the dose-intensified arms (3-year EFS of 76% ± 4% vs 74% ± 4%, respectively [P = .57]).5

An analysis of the contributions of each of these consecutive studies appears to demonstrate that improvements were obtained over the years. Although the addition of IE in the ES87 protocol did not add a significant survival advantage to the ES79 VACD regimen, the results of the subsequent studies that explored more intensive therapy would suggest that more aggressive regimens result in better disease control. When the 3 intensive protocols (EWI92, HIRISA, and SJBCM) were compared with the first studies (ES79 and ES87), a trend toward improved survival was found for those patients with localized disease who were treated with the more intensive regimens (5-year OS of 83.1% ± 5.6% vs 68.5% ± 4.8% [P = .068]). However, these improvements must be interpreted in the context of more aggressive and evolving surgical and RT techniques, which resulted in improved local control in the later studies.

In this context of escalating intensity and improved local control, the results of the ES79 protocol compare very favorably with the studies that followed. The ES79 protocol was a simple, relatively nontoxic regimen that used fractionated cyclophosphamide over 7 days, instead of the single high dose that became the norm afterward. A major positive impact has been attributed to the introduction of ifosfamide in current regimens, but accurate comparisons between the antitumor effect of ifosfamide and cyclophosphamide are limited by the differences in their administration. Ifosfamide has been typically administered in fractionated doses because autoinduction of enzyme activity increases the concentration of the active metabolites of alkylating agents. This effect has been demonstrated for both cyclophosphamide22, 23 and ifosfamide.24–26 Single-dose administration of these alkylating agents may result in decreased bioactivation because of saturable kinetic mechanisms, which favor the inactivating elimination pathway.27, 28 Therefore, fractionated administration alters kinetic properties avoiding the elimination pathways and favoring bioactivation of enzyme activity; however, this advantageous phenomenon has been applied to the administration of ifosfamide only, whereas cyclophosphamide continues to be administered as a single dose in the majority of ESFT protocols. Given this pharmacokinetic rationale and the results of the ES79 protocol, we may wonder whether the potential of the VACD regimen has been reached.

Prognostic Factors

As expected, older age, pelvic primary tumor, large tumors, and metastatic disease were associated with worse outcome,21, 29, 30 but only stage of disease and tumor size retained significance on a multivariate analysis. 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 was the local and systemic treatment. Local control with both surgery and RT was found to result in the best outcome, as did more intensive systemic therapy. However, because local and systemic therapy evolved simultaneously, it is difficult to determine the relative benefit of each advance. For patients with metastatic disease, the only prognostic indicator was the pattern of metastasis. Our analysis confirms that patients with pulmonary and extrapulmonary metastases differ with regard to the probability of survival.31, 32

Older age is consistently associated with a worse outcome.21, 29, 30, 33 Survival appears to decrease steadily with increasing age. Furthermore, patients aged >14 years are reported to have a higher proportion of large tumors and pelvic primary tumor sites29 as well as of metastatic disease.33

Second Malignancies

The cumulative incidence of second neoplasms in the majority of series published to date is no higher than 2%,21, 34 and sarcomas arising in the radiation field represent >75% of the cases.35, 36 However, in the current series, hematologic malignancies accounted for approximately two-thirds of second neoplasms. Current protocols that include intensification of alkylators and topoisomerase-II inhibitors have resulted in a significant increase in the incidence of t-AML/MDS.33, 37–39

Risk Stratification in ESFT

The advances of the last 20 years represented in the current series have defined the basic guidelines for the multidisciplinary treatment of patients with ESFT, including better characterization of the active drugs and local control measures. The majority of current protocols have focused on chemotherapy intensification. However, it is possible that subgroups exist in which a less intensive regimen may be used. Modern regimens should take into consideration risk factors that will help in adapting the intensity of the therapy. Although we acknowledge the limitations of exploratory analyses, the results of the current study demonstrated that patients can be risk stratified on the basis of simple clinical factors; these models can be refined further by incorporating additional variables, such as histologic response to chemotherapy30, 40 or the molecular detection of circulating tumor cells or bone marrow micrometastases.41 With a better understanding of the biology of ESFT, molecular markers may help improve these risk stratification approaches.


We thank Sharon Naron for her editorial assistance