We assessed tumor cell DNA content (ploidy) and N-myc gene copy number as predictors of long-term disease-free survival in 298 children with neuroblastoma. Diploid tumor stem lines were identified in 101 patients (34%), clonal hyperdiploid abnormalities in 194 (65%), and hypodiploid stem lines in three (1%). In children with widely disseminated tumors at diagnosis (stage D), ploidy had a highly age-dependent influence on prognosis. Among infants (less than 12 months) treated with cyclophosphamide-doxorubicin, hyperdiploidy was closely associated with long-term disease-free survival (greater than 90% of cases), while diploidy invariably predicted early treatment failure (P less than .001). Similarly, in children 12 to 24 months of age who were treated with cisplatin-teniposide and cyclophosphamide-doxorubicin, diploidy uniformly predicted early failure, whereas half of the children with hyperdiploidy achieved long-term disease-free survival (P less than .001). There was no relationship between ploidy and treatment outcome in children older than 24 months with stage D tumors who had a very low probability of long-term disease-free survival (less than 10%). N-myc gene amplification was detected in 37 (25%) of the 147 tumors tested, with the remainder showing single-copy levels of the gene. N-myc gene amplification was more frequent in diploid than in hyperdiploid tumors (23 of 57 v 14 of 87, P = .001) and predicted a high likelihood of early treatment failure. In children younger than 2 years with disseminated neuroblastoma, tumor cell ploidy and N-myc gene copy number provide complementary prognostic information that will distinguish patients who can be cured on current regimens from those who require new treatment strategies.