Résumé : Autologous transplantation using peripheral blood stem cells (PBSCs) collected after chemotherapy, followed by growth factor administration (ASCT), is increasingly used in the treatment of non-Hodgkin's lymphoma (NHL). However, quantitative data regarding contaminating malignant cells in the harvests are still scarce. We prospectively investigated 37 diffuse large-cell lymphomas (DLCLs) in complete remission (CR) that were treated according to multicentric protocols at our centre. DNA was extracted from the diagnostic lymph node. The complementarity-determining region (CDR) III was sequenced and a patient-specific oligomer synthesized. Contamination was evaluated semiquantitatively by polymerase chain reaction (PCR) and was confirmed by a limiting dilution analysis. PBSCs collected at regeneration after administration of granulocyte colony-stimulating factor (G-CSF), steady-state bone marrow (BM) and peripheral blood samples at CR were compared. DNA was available in 37 patients, from which 22 rearrangements could be sequenced. Patients (n = 15) who had both the required follow-up samples and a suitable clonal marker were investigated. In two cases, the patient-specific PCR assay set up at diagnosis later gave false-negative results in samples in which clonal DNA was still detectable by other sets of primers. PBSC contamination was highly variable: 7 out of 15 patients showed a PBSC/BM ratio of NHL cells greater than 1 log, whereas 8 out of 15 patients showed no difference and could vary from one apheresis to another. Eight ASCTs were performed, five of which used highly contaminated PBSCs: four patients relapsed early, three with disseminated lymphoma. Thus, 50% of DLCLs in CR seem to mobilize significantly malignant cells at regeneration under G-CSF. Considering the higher numbers of cells reinfused, this translates into a much higher number of lymphoma cells reinfused when compared with autologous bone marrow transplantation (ABMT). However, their clonogenic potential remains unknown and, despite concerning observations in certain cases, it is still unclear whether this has an impact upon the outcome of ASCT.