Résumé : We sequentially studied bone marrow (BM) samples of 25 patients in complete remission of an acute lymphoblastic leukemia (ALL) using a simplified polymerase chain reaction (PCR) strategy (direct use of the PCR product as a clonogenic probe recognizing rearranged Ig heavy chain sequences) as a first approach. BM aspirates were serially investigated after obtention of a complete response. When sensitivity was less than 1:10(4), the PCR fragment was sequenced and a specific oligonucleotide was synthetized and used as a probe (five cases). Cases in which minimal residual disease (MRD) became undetectable were cross-controlled using either TCR delta rearrangement or a specific translocation to circumvent the problem of false-negative results arising from clonal evolution. The median follow-up was 30 months (3 to 51 months). Within the first 3 months of complete remission, MRD was detectable in 22 of 23 investigated patients and remained so in 19 of 21 patients examined at 6 months, regardless of the long-term clinical outcome. In patients remaining in complete remission at 30 months or more, two patterns of MRD emerged during the follow-up. Either it continuously decreased to ultimately become undetectable (five patients) or remained detectable (five patients) with an increase after discontinuation of treatment in two. In the eight patients who relapsed, MRD persisted throughout the clinical course, and eventually increased 3 to 12 months before relapse was clinically detectable. In one case, clonal evolution of the VDJ heavy chain region was observed and recurrence of MRD shown by the use of TCR delta rearrangement as a control. We conclude that the use of this simplified methodology is a valuable tool for the follow-up of MRD in a majority of ALL patients, though in a few cases, sequencing needs to be performed to achieve a relevant sensitivity. The possibility of clonal evolution requires a cross-control of any sample becoming negative whatever the initial rearrangement used to generate a probe. In patients on therapy, sequential search for MRD seems to be a good tool for predicting the long-term outcome. In addition, patients remaining positive at the time treatment is discontinued or with a high tumor burden after a few months therapy may be at a higher risk of subsequent relapse, although a longer follow-up is needed to answer this question.