Great heterogeneity in survival exists for patients newly diagnosed with DLBCL. Three scoring systems incorporating simple clinical parameters (age, lactate dehydrogenase, number/sites of involvement, stage, performance status) are widely used: the international prognostic index (IPI), revised-IPI (R-IPI), and National Comprehensive Cancer Network IPI (NCCN-IPI). We evaluated 2124 DLBCL patients treated from 1998 to 2009 with front-line R-CHOP (or variant) across 7 multicenter randomized clinical trials to determine which scoring system best discriminates overall survival (OS). Median age was 63 years and 56% of patients were male. Five-year OS estimates ranged from 54% to 88%, 61% to 93%, and 49% to 92% using the IPI, R-IPI, or NCCN-IPI, respectively. The NCCN-IPI had the greatest absolute difference in OS estimates between the highest and lowest risk groups and best discriminated OS (c-index = 0.632 vs. 0.626 (IPI) vs. 0.590 (R-IPI)). For each given IPI risk category, NCCN-IPI risk categories were significantly associated with OS (P<0.01); the reverse was not true and the IPI did not provide additional significant prognostic information within all NCCN-IPI risk categories. Collectively, the NCCN-IPI outperformed the IPI and R-IPI. Patients with low NCCN-IPI had favorable survival outcomes with little space for further improvement. In the rituximab era, none of the clinical risk scores identified a patient subgroup with long-term survival clearly below 50%. Integrating molecular features of the tumor and microenvironment into NCCN-IPI or IPI might better characterize a high risk group where novel treatment approaches are most needed.