Dose-dense induction and upfront consolidation with autologous stem cell transplantation (ASCT) remain controversial issues when treating high-risk diffuse large B cell lymphoma patients. GELA designed a randomized phase II trial evaluating the efficacy of either R-ACVBP or R-CHOP14 induction and a PET-driven ASCT or standard immunochemotherapy (SIC) consolidation in aaIPI2-3 patients. PET was done at baseline, after 2 (PET2) and 4 induction cycles (PET4) and centrally assessed using international harmonization project (IHP) criteria. PET2-/PET4- patients were assigned SIC, PET2+/PET4- ASCT and PET4+ patients treated with investigator' choice. The primary end-point was the 2007 international working group CR rate after induction. ΔSUVmax PET assessment was explored. 211 patients were randomized to R-ACVBP (n=109) or R-CHOP14 (n=102). PET4-/CR rates were 53/47% with R-ACVBP and 41/39% with R-CHOP14 (CR 95%CI: 38%-67% v 28%-54%; p=0.076). Consolidation in the R-ACVBP and R-CHOP14 arms was SIC in 26% and 23% of patients and ASCT in 28% and 18%, respectively. PET4 positivity was higher with R-CHOP14 (54% v 41%; p=0.08) leading to more salvage therapy (37% v 26%; p=0.07) and lower EFS (4y-EFS= 31% v 43%; p<0.01) but PFS and OS were similar in both arms. PET2-/PET4- and PET2+/PET4- patients had similar outcome. Using ΔSUVmax, 79% patients were PET2-/PET4-. ΔSUVmaxPET0-4>70% was associated with better outcome (4y-PFS: 84% v 35%; 4y-OS: 91% v 57%, p<0.0001) whatever the consolidation. Superiority of R-ACVBP over R-CHOP14 was not established as IHP criteria did not properly reflect disease control. ΔSUVmax may help better select patients needing alternative to SIC, including ASCT.