The continued improvement in the prognosis of childhood acute myeloid leukaemia (AML) has been paralleled by the use of increasingly intensive therapy. This has led to attempts to develop risk-directed strategies in which the most intensive treatment is reserved for those at highest risk of relapse. Unfortunately, current approaches, which rely on cytogenetic sub-grouping and morphological assessment of response to therapy, are inaccurate. New prognostic factors are needed. This annotation proposes that the introduction of protocols based on the measurement of minimal residual disease (MRD) holds the key to progression from an era of 'cure at all costs' to a more individualised approach. However, the full potential of MRD technologies will only be realised through properly designed studies with scrupulous attention to logistics and quality assurance. The article illustrates which children may benefit most from MRD analysis in AML and explores practical issues that should be addressed in the design of clinical trials.