Summary. Flow-cytometric detection of minimal residual disease (MRD) identifies patients with high relapse risk in childhood acute lymphoblastic leukaemia (ALL). We studied the efficacy of this method in adult T-ALL treated with the Italian co-operative GIMEMA (Gruppo Italiano Malattie Ematologiche dell'Adulto) LAL0496 protocol. Bone marrow samples from 53 patients were taken at fixed treatment time points and MRD was analysed using a leukaemia-specific immunophenotype (cytoplasmic-CD3/nuclear-terminal desoxynucleotidyl transferase). The median follow-up was 17 months (range 3–61) and a median of 4·5 analyses/patient was performed (range 3–12). Six out of 53 (11·3%) patients were refractory to treatment, 30/53 (56·6%) relapsed and 17/53 (32·1%) remain in continuous complete remission. The probability of relapse at 2 years for MRD-positive patients at preconsolidation was 81·5%vs 38·9% for MRD-negative patients (P = 0·00078). This risk was still 54·5% for MRD-positive vs 15·8% for MRD-negative patients pre-third reinduction (P = 0·0098) and 50·0% for MRD-positive vs 16·4% for MRD-negative patients pre-sixth reinduction (P = 0·032). The relapse-predicting value of MRD did not depend on features at diagnosis such as age, sex and leucocyte count. Our data suggest that immunophenotypic MRD monitoring in the first year of treatment is a useful outcome predictor for adult T-ALL patients.