Characteristics and prognosis of pelvic Ewing sarcoma: a SEER population-based study

Ethics statement

This study was based on the data released from an online publicly available SEER database, and data extracted from SEER were identified as nonhuman study. Thus, it was deemed exempt by the Ethics Committee of West China Hospital, Sichuan University (Chengdu, China).

Data selection

SEER*Stat software version 8.3.5 (NCI, Bethesda, MD, USA) was used to obtain data from the SEER database of pelvic ES patients who were diagnosed and treated between 2004 and 2016, as patients diagnosed before 2004 had insufficient information on tumor size in this database. The database is population-based and contains data from 18 states in the USA, with annual updates on the clinical information of cancer patients, including age at diagnosis, sex, race, tumor characteristics, treatment, follow-up, and survival (Doll, Rademaker & Sosa, 2018). As shown in Fig. 1, the selection process was conducted to obtain sufficient eligible data from the database. Briefly, patients who had the International Classification of Diseases for Oncology, third edition histology/behavior code 9260/3-Ewing sarcoma with the primary site C41.4-Pelvic bones were included. Patients were excluded based on the following: (1) those without positive histological confirmations, (2) the pelvic ES was not their first tumor; (3) the tumor size or tumor stage in the database was unknown, or (4) the complete dates of follow-up were unavailable or survival did not exceed 0 days.

Variables

Patients’ clinical characteristics, including age at diagnosis, sex, race, tumor size, stage, treatment, vital status, and survival months were collected. A subset of characteristics was categorized for further analysis. The age at diagnosis was categorized as ≤9, 10–17, and ≥18 years (Ogura et al., 2015). Race was categorized as white or others (Black, American Indian/Alaskan Native, Asian/Pacific Islander). Tumor sizes were grouped based on the largest tumor diameter (≤ or >8 cm) from the variable “CS tumor size (2004+)” (Davenport et al., 2016). The tumor stage was divided into three groups based on the variable “SEER historical stage A,” including localized, regional, and distant. According to the 2018 version of the Summary Stage Manual provided by SEER (https://seer.cancer.gov/tools/ssm/), localized indicated that the tumor was confined to the pelvis or one to two pelvic segments were involved without known extraosseous extensions. Regional indicated one to two pelvic segments were involved with extraosseous extensions, and without distant metastasis.

Statistical analysis

For each patient variable, the prognostic effect was clarified using SPSS 24.0 (IBM Corp., Armonk, NY, USA). The Kaplan–Meier method was performed to construct cumulative survival curves and was compared using the log-rank test. OS was chosen as the primary survival outcome in this study. OS was defined as the period from diagnosis to death from any causes. Cox proportional hazards regression was used to identify significant prognostic factors, and variables with P-values < 0.05 in univariate analyses were further analyzed in multivariate analyses (Hu et al., 2019; Pan et al., 2019; Wang et al., 2019). Following this, a prognostic nomogram based on the results of the multivariate analyses was constructed using the rms package in R software, version 3.5.1 (R Core Team, 2018). The maximum score for each prognostic factor was set at 10. The performance of the nomogram was evaluated by the concordance index (C-index) ranging from 0.5 (a very poor model) to 1.0 (a perfect model). Generally, a C-index >0.7 indicates a good model (Liang et al., 2015). The calibration curve was based on 1,000 bootstrap replicates and was derived to compare the nomogram-predicted survival to the actual survival. A two-tailed P < 0.05 was defined as a significant difference.

Baseline characteristics of patients

The SEER database contained 2,613 patients with ES, of which 24.7% (646/2,613) were pelvic ES. Following assessments of the aforementioned inclusion and exclusion criteria, a total of 267 patients were included in the study (Fig. 1). The 267 pelvic ES patients were located in California (112, 41.9%), Connecticut (15, 5.6%), Georgia (26, 9.7%), Hawaii (3, 1.1%), Iowa (8, 3.0%), Kentucky (12, 4.5%), Louisiana (10, 3.8%), Michigan (9, 3.4%), New Jersey (32, 12.0%), New Mexico (6, 2.2%), Utah (13, 4.9%), and Washington (21, 7.9%).

The 267 patients had a median age of 16 years and an interquartile range of 12–22 years. As shown in Table 1, the sample set consisted of 174 (65.2%) male and 93 (34.8%) female patients. In the sample population, whites accounted for the majority. Among the patients, 97 (36.3%) had a tumor size of no more than eight cm, while 170 (63.7%) had a tumor size exceeding eight cm. A total of 39 patients were at the localized stage (14.6%), while 102 were at the regional stage (38.2%), and 126 were at the distant stage (47.2%). Additionally, surgery was performed in 80 (30.0%) patients and radiotherapy was performed in 55 (20.6%) patients.

Survival analysis

The OS of the 267 patients ranged from 1 to 142 months, with a median of 28 months. As shown in Fig. 2A, the overall 3- and 5-year survival rates were 61.1% and 50.3%, respectively. And the 3- and 5-year survival rates for different subgroups are shown in Table 1. Based on Kaplan–Meier curves and log-rank analyses, younger patients (P < 0.001, Fig. 2B), whites (P = 0.037, Fig. 2D), patients with localized tumor stages (P < 0.001, Fig. 2F), and those who underwent surgery (P < 0.001, Fig. 2G) were associated with better prognoses, while no significant differences were observed in OS when considering sex (P = 0.802, Fig. 2C), tumor size (P = 0.161, Fig. 2E), and radiotherapy treatment (P = 0.138, Fig. 2H). Besides, surgery showed an association with better prognosis in patients without radiotherapy, but had an unfavorable influence for radiotherapy patients. The details of the correlation between survival and factors mentioned above are shown in Table 1.

To control for confounding variables, the risk factors identified by univariate analyses were further explored in multivariate analyses. As shown in Table 2, Cox proportional hazards regression analyses revealed that age (P = 0.006), race (P = 0.018), tumor stage (P < 0.001), and surgery (P = 0.030) were the independent risk factors for prognosis.

Predictive nomogram

As shown in Fig. 3, the nomogram for predicting 3- and 5-year OS rates of pelvic ES was constructed based on the significant risk factors identified by multivariate analyses. To calculate the 3- and 5-year OS rates, each factor was first evaluated using the points scale at the top of the nomogram, which were then summed for each factor. Following this, the point scale located at the base of the nomogram were used to determine the 3- and 5-year OS rates. Through bootstrap resampling validations, calibration plots were determined (illustrated in Fig. 4) showing excellent agreement with actual survival. The C-index of the nomogram was 0.728, suggesting good prediction accuracies for OS in patients with pelvic ES.