BACKGROUND Current staging systems for unresectable or metastatic neuroblastoma do not reliably predict responses to chemotherapy in infants under 1 year of age. Previous studies have indicated that the DNA content, or ploidy, of malignant neuroblasts can discriminate between good and poor responders in this group of patients, but the clinical utility of ploidy assessment has remained in question. OBJECTIVE We tested, in a prospective nonrandomized study, the hypothesis that neuroblast ploidy could be used as the sole guide for treatment selection in infants with unresectable or metastatic tumors and could differentiate between those who would respond to our previous standard regimen and those who would benefit from an immediate switch to another therapy. METHODS One hundred seventy-seven infants were enrolled in this trial. Five of these infants were subsequently excluded (two ineligible, two lacking ploidy information, and one protocol violation); therefore, 172 patients were included in the study. One hundred thirty infants with hyperdiploid tumors (DNA index > 1.0; better prognosis in retrospective studies) were treated with a well-tolerated regimen of cyclophosphamide (150 mg/m2 per day orally or intravenously on days 1-7) and doxorubicin (35 mg/m2 intravenously on day 8). Forty-two infants with diploid tumors (DNA index = 1.0; worse prognosis in retrospective studies) received cisplatin (90 mg/m2 intravenously on day 1) and teniposide (100 mg/ m2 intravenously on day 3) after an initial course of cyclophosphamide plus doxorubicin. Statistical end points were response and long-term survival. In addition, we assessed within each ploidy group (i.e., patients with hyperdiploid tumors and those with diploid tumors) the prognostic significance of NMYC gene copy number, tumor stage, and other variables commonly measured in this disease. RESULTS Of the 127 assessable infants with hyperdiploid tumors, 115 (91%) had complete responses--85 after receiving five courses of cyclophosphamide plus doxorubicin and 30 after receiving further therapy including cisplatin plus teniposide. The 3-year survival estimate for the entire hyperdiploid group was 94% (95% confidence interval [CI] = 89%-98%). Nineteen (46%) of 41 assessable infants with diploid tumors were complete responders. The overall 3-year survival estimate for this group was 55% (95% CI = 39%-70%). Prognostic factor analysis indicated that NMYC gene amplification and an elevated serum lactate dehydrogenase level were statistically significant markers of higher risk disease within the diploid group (two-sided P values of .005 and .003, respectively). Only NMYC was predictive in the hyperdiploid group (P = .003). CONCLUSIONS Use of a prognostic staging system based on tumor cell ploidy, augmented with the NMYC gene copy number and serum level of lactate dehydrogenase, would very likely improve the treatment of infants with unresectable or metastatic neuroblastoma. Patients with diploid tumors characterized by an amplified NMYC locus represent a particularly unfavorable risk group that may benefit from innovative new therapies.