Background: IGCCCG risk classification is used since 1997 to allocate metastatic GCT patients (pts). Recently, its applicability and survival estimates have been called into question. Guidelines recommend 4 cycles of BEP or VIP for both intermediate (int) and poor risk disease. We sought to revisit the outcomes and treatments in a population of int and poor risk NSGCT from Indiana and INT Milano. Methods: Data on consecutive pts receiving first-line chemotherapy (CT) from 1990 to 2014 were collected. Kaplan-Meier method was used to estimate relapse-free survival (RFS) and overall survival (OS). Cox regression analyses were used to evaluate potential prognostic factors of RFS and OS in an univariable model. Forward stepwise selection was used to construct a multivariable (MV) model. A risk factor (RF) model was then constructed and compared with IGCCCG classification using the concordance statistics (CS). Results: 647 pts were identified; median age was 27 years (range 13-60), 115 had a mediastinal primary (PMNSGCT), 60 (9.3%) brain, 131 (20.3%) liver, and 37 (5.7%) bone metastases (mets). 437 (67.5%) had received BEP CT. RF in the MV model were PMNSGCT (HR = 3.23, 95% CI = 2.28-4.59), brain mets (HR = 2.29, 95% CI = 1.48-3.55), lung mets (HR = 1.75, 95% CI = 1.22-2.50) and age at diagnosis ( ≥ 30 vs < 30, HR = 1.67, 95% CI = 1.22-2.29). CS were improved based on the RF model compared to IGCCCG classification (RFS: 0.63 vs 0.58; OS: 0.65 vs 0.59). For int risk, there were no differences between 3 (n = 26) and 4 cycles BEP (n = 160) or BEPx3+EPx1 (n = 31) for both RFS (p = 0.31) and OS (p = 0.055). Conclusions: In this contemporary cohort, the outcomes of int risk NSGCT have improved compared to historical estimates. Many int risk pts would not require BEPx4 to attain a cure. Using new RF, improved risk stratification was observed, which gives further evidence that a reclassification of GCT is needed.