A Prognostic Model Including Pre- and Postsurgical Variables to Enhance Risk Stratification of Primary Mediastinal Nonseminomatous Germ Cell Tumors: The 27-Year Experience of a Referral Center

Primary mediastinal nonseminomatous germ cell tumors comprise a heterogeneous series of neoplasms characterized by limited chemosensitivity and a poor prognosis. We analyzed a large series of patients from our tertiary cancer center, including pre- and postsurgical variables, with the aim to provide a prognostic model that might be suitable for clinical use. The variables identified in the prognostic model were surgical removal of residuals after first-line chemotherapy, histological response, and the presence of lung metastases. Their joint analysis defined distinct overall survival (OS) curves. Background: Primary mediastinal germ cell tumors (PMGCTs) poorly benefit from chemotherapy and half of patients die because of disease progression. Enhancing the risk stratification might result in tailoring a more personalized treatment strategy from the time of diagnosis. Patients and Methods: Between the years 1985 and 2012, 86 patients with PMGCT were treated at our center. Cox proportional hazards regression analysis was conducted in the population of nonseminomas to examine the prognostic effect of candidate factors on progression-free and OS. OS curves were compared using the Kaplan-Meier method and the log-rank test. Results: Mean age was 29.8 years (range, 15-63 years). Twenty-five patients (29.1%) had lung and 8 (9.3%) liver, bone, or brain metastases. Twelve patients (13.9%) received upfront high-dose chemotherapy and 45 patients (52.3%) underwent surgery after chemotherapy. Cox analyses included 61 evaluable primary mediastinal nonseminomatous germ cell tumors (PMNSGCTs). The final model of factors indicating a poor prognosis included the combination of surgery and histological response (overall P = .011) and lung metastases (hazard ratio, 3.03; 95% confidence interval, 1.12-8.15; P = .028). The model showed a bootstrap-corrected Harrel c-statistic for OS of 0.66. A risk stratification model based on the combination of these factors and accounting for a 50% 5-year survival cutoff identified 2 groups (poor prognosis, n = 33 vs. good prognosis, n = 28) with distinct OS curves (P < .001). Preoperative serum tumor marker level was not associated with the final histology (P = .853, chi(2) test). Results were limited by small numbers. Conclusion: Patients with PMNSGCT included 2 subpopulations with distinct prognosis, and therapeutic improvements are needed for patients with poor-risk features.