Purpose: Interim positron emission tomography (PET) using the tracer, [18F]fluorodeoxyglucose, may predict outcomes in patients with aggressive non-Hodgkin lymphomas. We assessed whether PET can guide therapy in patients who are treated with cyclophosphamide, doxorubicin, vincristine, and prednisone (CHOP). Patients and Methods: Newly diagnosed patients received two cycles of CHOP - plus rituximab (R-CHOP) in CD20-positive lymphomas - followed by a PET scan that was evaluated using the DSUVmax method. PET-positive patients were randomly assigned to receive six additional cycles of R-CHOP or six blocks of an intensive Burkitt's lymphoma protocol. PET-negative patients with CD20-positive lymphomas were randomly assigned or allocated to receive four additional cycles of R-CHOP or the same treatment with two additional doses rituximab. The primary end point was event-free survival time as assessed by log-rank test. Results: Interim PET was positive in 108 (12.5%) and negative in 754 (87.5%) of 862 patients treated, with statistically significant differences in event-free survival and overall survival. Among PET-positive patients, 52 were randomly assigned to R-CHOP and 56 to the Burkitt protocol, with 2-year event-free survival rates of 42.0% (95% CI, 28.2% to 55.2%) and 31.6% (95% CI, 19.3% to 44.6%), respectively (hazard ratio, 1.501 [95% CI, 0.896 to 2.514]; P = .1229). The Burkitt protocol produced significantly more toxicity. Of 754 PET-negative patients, 255 underwent random assignment (129 to R-CHOP and 126 to R-CHOP with additional rituximab). Event-free survival rates were 76.4% (95% CI, 68.0% to 82.8%) and 73.5% (95% CI, 64.8% to 80.4%), respectively (hazard ratio, 1.048 [95% CI, 0.684 to 1.606]; P = .8305). Outcome prediction by PET was independent of the International Prognostic Index. Results in diffuse large B-cell lymphoma were similar to those in the total group. Conclusion: Interim PET predicted survival in patients with aggressive lymphomas treated with R-CHOP. PET-based treatment intensification did not improve outcome.