Factors associated with tumor volume and primary metastases in Ewing tumors : Results from the (EI)CESS studies

Tumor volumes of more than 100 ml and the presence of primary metastases have been identified as determinants of poor prognosis in patients with Ewing tumors. We sought to assess the prevalence of critical tumor size and primary metastases in a large national sample of patients at the time of first diagnosis and to identify factors that are associated with their occurrence. The present report is based on data of 945 German patients who were enrolled into the (EI)CESS therapy studies between 1980 and 1997. It is assumed that registration of German patients with Ewing tumors under the age of 15 years was almost complete since around 1985. Diagnoses of primary tumors were ascertained exclusively by biopsies. Analyses were restricted to patients with Ewing tumors of bone due to the few occurrences in soft tissues. Tumor volume data as assessed by radiography, computed tomography or nuclear magnetic imaging were available for 821 patients. The diagnosis of primary metastases was based on thoracic computed tomography or on whole body bone scans in 936 patients. Suspicious lesions had to be confirmed by bone marrow biopsies. We explored how year of first diagnosis, age at first diagnosis, sex, histological subtype and site of the primary tumor related to tumor size and presence of metastases by univariate and multivariate statistical techniques. Sixty-eight percent of the patients (n = 559) had a volume above 100 ml with smaller tumors being more common in childhood than in late adolescence and early adulthood. Extensive volumes were observed in almost 90% of the tumors located in femur and pelvis while they were less common in other sites (P < 0.001). On average, 26% of all patients presented with clinically apparent primary metastases. The detection rate of metastases was markedly higher in patients diagnosed after 1991 (P < 0.001). Primary metastases were also significantly more common for tumors originating in the pelvis and for peripheral neuroectodermal tumors (PNET; P < 0.01). Tumors greater than 100 ml were positively associated with metastatic disease (P < 0.001). Multivariate analyses, which included simultaneously all univariate predictors in a logistic regression model, indicated that most of the observed associations were essentially unconfounded. The adjusted odds ratios (OR) for the presence of tumor volumes > or = 100 ml were OR = 1.5 per age rise of 10 years, and OR = 5.8 for pelvis and OR = 7.1 for femur as primary tumor site (all P < 0.001). Th