Identification of prognostic subgroups among patients with metastatic `IGCCCG poor-prognosis' germ-cell cancer: An explorative analysis using cart modeling

Objectives:The IGCCCG classification has identified threeprognostic groups of patients with metastatic germ-cell tumors. `Poorprognosis' is based on primary tumor localization, the presence of visceralmetastases, and/or high tumor-marker levels. The overall survival rate ofthese patients is about 45%–55%. The present analysisattempts to identify subsets of patients with a more or less favorable outcomeamong the `poor-prognosis' group.Patients and methods:We retrospectively explored prognosticsubgroups in 332 patients with `IGCCCG' poor-risk GCT using theclassification-and-regression-tree model (CART). The following variables wereincluded: primary tumor localization, presence of visceral or lung metastases,presence of an abdominal tumor, number of metastatic sites, serum levels of-HCG, AFP and LDH. All patients had been treated withcisplatin–etoposide-based chemotherapy within controlled clinical trialsbetween 1984 and 1997.Results:Patient characteristics: gonadal/retroperitoneal (G/R)primary tumor 260 patients (78%), mediastinal primary tumor 72 patients(22%), visceral metastases 205 patients (62%) including 33patients with CNS metastases, lung metastases 247 patients (74%),abdominal tumor 241 patients (72%), elevated AFP, -HCG or LDHlevels 235 (71%), 253 (76%) and 275 (83%) of patients,respectively. Patients with primary mediastinal disease plus lung metastasesexhibited the worst two-year PFS (28%), whereas patients with a primaryG/R tumor and without visceral metastases showed the highest chance oftwo-year PFS (75%). The latter group of patients without visceralmetastases and with a primary G/R tumor also had the most favourable two-yearOS (84%). In contrast, patients with a primary mediastinal tumor andvisceral metastases displayed the worst two-year OS (49%).Conclusions:Different prognostic subsets of patients can beidentified among the group of `poor-prognosis' GCT patients. The CART analysismodel results in a hierarchy of prognostic factors which may allow to moreprecisely estimate the individual patient's prognosis. Identifying subgroupsof `very poor-prognosis' among `poor-prognosis' patients may allow to test fornew treatment strategies in selected subgroups.