Résumé : We prospectively investigated minimal residual disease (MRD) in 51 children with B-lineage acute lymphoblastic leukaemia (ALL) treated according to the Fralle 93 protocol. PCR follow-up was performed in children in morphological and cytogenetic complete remission, provided an immunoglobulin (IgH) gene rearrangement could be detected using FR 3/J(H) amplimers. MRD was studied according to our previously described methodology, with a few modifications including the use of a consensus J(H) probe to control for PCR efficiency variations. Out of the initial 51 patients, 34 were assessable for MRD at the end of induction at the time of analysis. MRD levels were as follows: > 1/10(3) in 26%, 1/10(3) to 1/10(4) in 50% and < 1/10(4) or not detectable in 24%. With a median follow-up of 20 months there were five relapses, all of which occurred in the group of patients with MRD > 1/10(3). To date, none of the patients with MRD < or = 1/10(3) (good molecular responder) has relapsed. Classification according to molecular response at the end of induction did not correlate with the conventional risks groups: there were no statistically significant differences between good and bad molecular responders. Of particular interest is the absence of correlation between WBC at diagnosis and MRD level at the end of induction. We conclude that classification of patients into good and bad molecular responders using PCR seems to be a better prognostic indicator than conventional risk factors in childhood B-lineage ALL. Patients with MRD level > 1/10(3) have a particularly poor outcome and should always be considered for alternative therapeutic strategies in the future, whereas in good molecular responders belonging to poor or intermediate risk categories, treatment de-escalation might be contemplated.