We reviewed the records of 408 patients who were less than 21 years old at the time of diagnosis of acute myeloid leukemia (excluding patients with Down syndrome or acute promyelocytic leukemia) treated on five consecutive institution protocols from 1980–2002 to investigate prognostic factors for attainment of second complete remission (CR2) and overall survival (OS) after first relapse. Of the 320 (78%) patients who achieved CR, 158 patients suffered hematologic relapses at a median of 11.9 months (range, 2.9–119.9 months) from the time of diagnosis. Forty-one patients relapsed on therapy and 117 relapsed after the completion of all planned therapy. For patients who relapsed off therapy, 38 were diagnosed with relapse because of symptoms suggestive of recurrent leukemia and 78 were diagnosed at the time of a routine follow up (information not available for one patient). After relapse, 20 patients received palliative care, 82 received chemotherapy alone, 36 received chemotherapy followed by hematopoietic stem cell transplant (SCT), and 20 proceeded directly to SCT. Eighty-five (54%) patients attained CR2. In univariate analyses, factors associated with the achievement of CR2 include initial therapy (chemotherapy alone, 56%; autologous SCT, 71%; allogeneic SCT, 20%; p=0.008) and time from the diagnosis of AML to relapse (≤ 1 year, 44%; > 1 year, 65%; p=0.010). Logistic regression analysis demonstrated that patients with male gender (odds ratio [OR], 2.46; 95% confidence interval [CI], 1.14–5.28; p=0.021) and greater time from diagnosis to relapse (OR, 1.05; CI, 1.01–1.09; p=0.016) were more likely to achieve CR2, whereas patients with M7 morphology (OR, 0.11; CI, 0.01–1.04; p=0.054) and allogeneic SCT in first remission (OR, 0.17; CI, 0.03–0.85; p=0.031) were less likely to achieve CR2. At the time of last follow up, 19 patients were alive, 115 died of progressive disease, and 24 died of regimen-related toxicity. The 2-year OS estimate ± SE for the entire cohort of 158 patients was 15.8% ± 2.8%. For patients who relapsed off therapy, there was no significant difference in outcome between those whose relapse was diagnosed at the time of a routine follow up and those diagnosed because of symptoms. Cox proportional-hazards regression modeling indicated that M7 morphology (hazard ratio [HR], 3.06; CI, 1.44–6.51; p=0.004) and allogeneic SCT in first remission (HR, 2.17; CI, 1.22–3.87; p=0.008) were associated with significantly worse OS after relapse. In fact, there were no survivors in these two groups of patients. In contrast, age 1–10 years (HR, 0.53; CI, 0.36–0.77; p=0.001), M2 morphology (HR, 0.57; CI, 0.39–0.85; p=0.006), allogeneic SCT after relapse (HR, 0.34; CI, 0.22–0.53; p<0.001), and greater time from diagnosis to relapse (HR, 0.97; CI, 0.95–0.99; p=0.003) had significantly favorable effects on OS. However, outcome was poor even among patients who underwent allogeneic SCT after relapse (2-year OS, 34.5% ± 6.1%) and among patients who relapsed greater than one year from diagnosis (2-year OS, 19.2% ± 4.3%), suggesting that novel therapies are warranted for all patients with relapsed AML.