Prognostic variables for response and outcome in patients with extragonadal germ-cell tumors.

BACKGROUND: This investigation evaluates prognostic variables in patients with seminomatous and non-seminomatous extragonadal germ-cell tumors (EGCT) in order to identify relevant factors for long-term outcome following cisplatin-based chemotherapy. PATIENTS AND METHODS: Patients from six countries treated at 11 centers in Europe and the USA from 1975 to 1996 were evaluated retrospectively. Uni- and multivariate analyses of prognostic variables for survival and for response to chemotherapy were performed. RESULTS: Data were available for 635 EGCT patients, 104 with seminomatous and 524 with non-seminomatous EGCT (n = 7 not specified). For non-seminomatous EGCT the following independent adverse factors were identified: presence of either liver, lung or central nervous system metastases, primary mediastinal tumor or elevation of pretreatment beta-human gonadotropin; for extragonadal seminoma (only univariate) adverse factors were: presence of liver metastases, two or greater metastatic sites or International Germ Cell Cancer Collaborative Group (IGCCCG) grouping (intermediate versus good). Integration of these variables produced the following prognostic risk groupings: 'excellent prognosis', all seminomatous EGCT (89% 5-year survival rate); 'intermediate low', 'intermediate high' and 'poor', all non-seminomatous EGCT with a 69, 55 and 17% 5-year survival rate, respectively. The decreased survival among the different groups was due to a lower rate of favorable objective remissions and a higher rate of relapses. Classification and regression tree (CART) modeling confirmed histology and location of primary tumor as the major prognosticators. For the subgroup of patients with mediastinal non-seminoma, the 2-year survival rate ranged from 34 to 84%. Multivariate testing for the probability to respond to chemotherapy revealed non-seminomatous histology, primary mediastinal tumor site, and the presence of liver, lung or CNS metastases as independent adverse factors. CONCLUSIONS: In EGCT, prognostic variables for the outcome and for the response to chemotherapy could be identified, which in part differ from gonadal GCT. The proposed model might help to better understand the specific prognosis of EGCT and to tailor risk-adapted treatment strategies. In addition, CART analysis demonstrated the heterogenous prognosis of patients with mediastinal non-seminoma.