Metastatic germ cell tumors: Modeling for response to chemotherapy

Purpose: This study sought to determine factors that predict response to chemotherapy for patients with advanced germ cell tumors (GCTs), to evaluate different methods by which serum tumor markers can be used to predict response independent of assay method, and to develop criteria to guide allocation to clinical trials. Methods: Pretreatment data from 796 patients treated with platinum-based chemotherapy on Memorial Hospital protocols from 1975 to 1990 were analyzed. Multivariate analyses were performed on a developmental data set (n = 597) to assess for independently significant predictors of response. Predictive models developed using these methods were confirmed on the validation set (n = 199). Results: Independently significant factors for response included a mediastinal primary tumor (.015), pure seminomatous histology (.002), metastases to non-pulmonary visceral sites (bone, liver, and brain; .0001), and the pretreatment values of lactate dehydrogenase (LDH; .0001) and human chorionic gonadotropin (HCG; .0001). The likelihood of complete response (CR) was increased by seminomatous histology and decreased by the other factors. Serum levels of HCG and LDH could predict outcome independent of assay methodology. A model to predict good-, intermediate-, and poor-risk categories with CR rates of 92%, 76%, and 39%, respectively, was developed. Conclusion: The independent prognostic factors of serum tumor markers LDH and HCG, the sites of metastases, and the primary origin of GCTs can predict outcome in patients with advanced GCT. These factors should be considered in the allocation of patients with advanced disease to clinical trials and have been included in the new tumor-node-metastasis (TNM) staging systems developed for the American Joint Committee on Cancer (AJCC) and the Union Internationale Contre le Cancer (UICC). (C) 1998 by American Society of Clinical Oncology.