The Pediatric Oncology Group (POG) phase 3 trial 9404 was designed to determine the effectiveness of high-dose methotrexate (HDM) when added to multi-agent chemotherapy based on the Dana-Farber backbone. Children with T-cell acute lymphoblastic leukemia (T-ALL) or advanced lymphoblastic lymphoma (T-NHL) were randomized at diagnosis to receive/not receive HDM (5 g/m² as a 24-hour infusion) at weeks 4, 7, 10, and 13. Between 1996 and 2000, 436 patients were enrolled in the methotrexate randomization. Five-year and 10-year event-free survival (EFS) was 80.2% ± 2.8% and 78.1% ± 4.3% for HDM (n = 219) versus 73.6% ± 3.1% and 72.6% ± 5.0% for no HDM (n = 217; P = .17). For T-ALL, 5-year and 10-year EFS was significantly better with HDM (n = 148, 5 years: 79.5% ± 3.4%, 10 years: 77.3% ± 5.3%) versus no HDM (n = 151, 5 years: 67.5% ± 3.9%, 10 years: 66.0% ± 6.6%; P = .047). The difference in EFS between HDM and no HDM was not significant for T-NHL patients (n = 71, 5 years: 81.7% ± 4.9%, 10 years: 79.9% ± 7.5% vs n = 66, 5 years: 87.8% ± 4.2%, 10 years: 87.8% ± 6.4%; P = .38). The frequency of mucositis was significantly higher in patients treated with HDM (P = .003). The results support adding HDM to the treatment of children with T-ALL, but not with NHL, despite the increased risk of mucositis.