While CHOP has been the standard treatment for DLBL, the low cure rate of approximately 40% implies that there is a need for improvement. Several hypotheses have been tested to optimize treatment. The Goldie and Coldman hypothesis sustains that drug resistance is a function of tumor burden and, accordingly, third generation regimens were designed. A randomized trial comparing CHOP with them showed no difference in the outcome. The myeloablative dose-intensity has become established as the best salvage therapy currently available for chemosensitive relapse of DLBL. However when tested as a frontline therapy clear results have not been obtained. The Norton and Simon hypothesis is based in Gompertzian kinetics whereby the growth rate of smaller tumors is higher. Thus, the dose-dense concept emerges, shortening the intervals between cycles and lowering tumor regrowth with CHOP-14 has recently been shown to improve the outcome in patients with DLBL. Addition of Rituximab-based immunotherapy is being tested in ongoing trials. Other approaches to optimize the treatment are based on a dynamic decision following early responses based on new imaging techniques. Finally, the analysis of tissue arrays and genomic profiling, which provide insights into the oncogenic pathways involved, is allowing the development of new targets.