Based on their own experience and knowledge of the literature, the authors review the pathobiological characteristics of peripheral T-cell lymphomas (PTCLs), focusing on the available prognostic indicators. The International Prognostic Index (IPI), which is based on age, performance status, lactate dehydrogenase [LDH], stage, and extranodal involvement, appears to be efficient as a prognostic index for PTCLs, at least in part and especially for certain PTCL subtypes. However, it is not so satisfactory for the two commonest PTCLs, PTCL not otherwise specified (PTCL/NOS) and angioimmunoblastic T-cell lymphoma (AITL), for which novel scores, possibly based on the biologic features of the tumors, have been explored. An Italian cooperative group proposed a revision of the IPI for PTCL unspecified (PTCL-U), the Prognostic Index for PTCL-U (PIT), which includes age, performance status, LDH, and bone marrow involvement. The PIT apparently offered some advantages, but they were not confirmed in subsequent studies. A clinical-biological score (the Bologna score) was then proposed, including tumor proliferation and clinical features (age, LDH, and performance status). This score appears promising and offers the intriguing advantage of integrating biological and clinical elements, but independent validation on a large series is still warranted. More recently, gene expression profiling has been used to identify novel molecular prognostic factors. In particular, inactivation of the NFκB pathway, high expression of proliferation-associated genes, and cytotoxic molecular phenotype seem to be associated with a worse outcome. So far, however, none of these indicators has been validated in an independent series. Finally, various reports have dealt specifically with the prognostication of NK-derived tumors, including nasal and nasal-type lymphomas. Both the IPI and dedicated models have turned out to be of prognostic relevance for these tumors. In conclusion, although the IPI is somewhat effective for PTCL prognostication, novel scores that are more refined and possibly disease-specific are warranted. The validation process for several models, including clinical-pathological and molecular models, is now ongoing.