With 5-year event-free survival of 80% now commonplace for paediatric acute lymphoblastic leukaemia (ALL), and 50% for paediatric acute myeloid leukaemia (AML), recent efforts have focused on optimum risk-directed treatment. Because cranial irradiation can cause many acute and late complications (eg, second cancers, neurocognitive deficits, endocrine disorders, and growth impairment), it has been largely replaced by intensive intrathecal treatment and systemic chemotherapy. Prophylactic cranial irradiation (12-18 Gy) is given to 2-20% of patients with ALL who have an increased risk of CNS relapse (such as T-cell immunophenotype, overt CNS disease, high-risk cytogenetic features, or poor response to remission induction treatment), and for the estimated 2% of patients with AML who have overt CNS disease at diagnosis. Although this strategy restricts CNS relapse to 3-8% of patients, several challenges remain. Methods need to eliminate relapse without the use of cranial irradiation in very high-risk patients. Effective retrieval treatment is needed for patients with relapsed or refractory CNS leukaemia, and intrathecal treatment with improved efficacy and reduced side-effects remains a long-term objective. Perhaps the most formidable challenge is to treat children with CNS relapse after a short initial remission or cranial irradiation.