Purpose Our aim was to determine the role of end-of-induction (EOI) minimal residual disease (MRD) assessment in the identification and stratification of induction failure in patients with pediatric acute lymphoblastic leukemia (ALL) and to identify genetic abnormalities that drive disease in these patients. Patients and Methods Analysis included 3,113 patients who were treated in the Medical Research Council UKALL2003 multicenter randomized trial (NCT00222612) between 2003 and 2011. MRD was measured by using standardized real-time quantitative PCR. Median follow-up was 5 years 9 months. Results Fifty-nine patients (1.9%) had morphologic induction failure with 5-year event-free survival (EFS) of 50.7% (95% CI, 37.4 to 64.0) and 5-year overall survival of 57.7% (95% CI, 44.2 to 71.2). Of these, a small proportion of patients with M2 marrow (6 of 44) and a low EOI MRD level (< 0.01%) had 5-year EFS of 100%. Conversely, among patients with morphologic remission 2.3% (61 of 2,633) had high MRD (≥ 5%) and 5-year EFS of 47.0% (95% CI, 32.9 to 61.1), which was similar to those with morphologic induction failure. Redefining induction failure to include morphologic induction failure and/or MRD ≥ 5% identified 3.9% (120 of 3,133 patients) of the trial cohort with 5-year EFS of 48.0% (95% CI, 39.3 to 58.6). Induction failure (morphologic or MRD ≥ 5%) occurred most frequently in T-ALL (10.1%; 39 of 386 T-ALL cases) and B-other ALL, that is, lacking established chromosomal abnormalities (5.6%; 43 of 772 B-other cases). Genetic testing within the B-other group revealed the presence of PDGFRB gene fusions, particularly EBF1-PDGFRB, in almost one third of B-other ALL cases. Conclusion Integration of EOI MRD level with morphology identifies induction failure more precisely than morphology alone. Prevalence of EBF1-PDGFRB fusions in this group highlights the importance of genetic screening to identify abnormalities that may be targets for novel agents.