This multicenter, prospective trial was conducted to develop an effective and safe reinduction regimen for marrow-relapsed pediatric acute lymphoblastic leukemia (ALL) by modifying the dose of idarubicin. Between 2006 and 2009, the trial accrued 44 patients, 1 to 21 years old with first marrow-relapsed ALL. The reinduction regimen comprised prednisolone, vincristine, L-asparaginase, and idarubicin (10 mg/m2/week). The idarubicin dose was adjusted according to the degree of myelosuppression. The second complete remission (CR2) rate was 72.7%, obtained by 54.2% of patients with early relapse < 24 months after initial diagnosis and 95.0% of those with late relapse (P = 0.002). Five patients entered remission with extended treatment, resulting in a final CR2 rate of 84.1%. The CR2 rate was not significantly different according to the idarubicin dose. The induction death rate was 2.3% (1/44). The 5-year event-free and overall survival rates were 22.2% ± 6.4% and 27.3% ± 6.7% for all patients, 4.2% ± 4.1% and 8.3% ± 5.6% for early relapsers, and 43.8% ± 11.4% and 50.0% ± 11.2% for late relapsers, respectively. Early relapse and slow response to reinduction chemotherapy were predictors of poor outcomes. In conclusion, a modified dose of idarubicin was effectively incorporated into the reinduction regimen for late marrow-relapsed ALL with a low toxic death rate. However, the CR2 rate for early relapsers was suboptimal, and the second remission was not durable in most patients.