The Revised European-American Lymphoma (REAL) classification has been criticized for its emphasis on the unproven clinical relevance of immunophenotype. A worse prognosis for peripheral T-cell non-Hodgkin's lymphomas (PTCLs) has been inconsistently reported in part because the definition of PTCL has been imprecise (eg, T-cell-rich B-cell non-Hodgkin's lymphomas [TCRBCLs] have been misdiagnosed as PTCLs in the past) and because its correlation with other known prognostic factors has not been studied by multivariate analysis. We analyzed six protocols from 1984 to 1995 with Working Formulation intermediate grade and immunoblastic lymphomas (exclusive of mantle cell) and selected only those cases in which immunophenotyping was performed and was conclusive. Of a total of 560 evaluable patients, 68 were PTCLs (12%) and the remaining 492 (88%) were B-cell non-Hodgkin's lymphomas, including 16 TCRBCLs (3% of total). The 5-year failure-free survival (FFS) for PTCLs and B-cell large-cell lymphomas (BCLCLs) is 38% and 55%, respectively (P < .0001) and the 5-year overall survival (OS) is 39% and 262%, respectively (P < .001). The M.D. Anderson prognostic tumor score (MDATS) and International Prognostic Index (IPI) for all patients was calculated. With MDATS of less than 3 (good prognosis), the 5-year FFS for PTCL and BCLCL is 56% and 69%, respectively (P = .01), and the 5-year OS is 64% and 77%, respectively (P = .06). With MDATS of greater than 2 (poor prognosis), 5-year FFS for PTCL and BCLCL is 26% and 38%, respectively (P = .03), and the 5-year OS is 24% and 41%, respectively (P = .02). With an IPI of less than 3 (good prognosis), the 5-year FFS for PTCL and BCLCL is 49% and 64%, respectively (P = .001), and the 5-year OS is 55% and 71%, respectively (P = .013). With an IPI greater than 2 (poor prognosis), the 5-year FFS for PTCL and BCLCL is 11% and 35%, respectively (P = .044), and the 5-year OS is 10% and 40%, respectively (P = .011). Multivariate analysis shows that MDATS, IPI, and T-cell phenotype are totally independent and are the most significant predictors of FFS and OS. The 68 PTCLs include 45 PTCLs unspecified, 10 Ki-1 anaplastic (ALCL), 8 angioimmunoblastic, and 5 angiocentric lymphomas. Angiocentrics were usually refractory (1 of 5 remissions only). ALCL rarely relapsed late. We conclude that the immunophenotypic basis of the REAL classification is clinically relevant and that, although other prognostic features also influence outcome, the T-cell phenotype still remains an independent and significant prognostic factor.