Significant predictors of treatment outcome are poorly defined for patients with T-lineage acute lymphoblastic leukemia (T-ALL). A high WBC at diagnosis, which has traditionally been a predictor of poor response in T-ALL, has considerably weakened prognostic significance in the face of modern, more intensive chemotherapy. To test the hypothesis that bone marrow stroma-supported leukemic cell recovery might identify children at high risk for relapse, we measured the ex vivo recovery of T-ALL lymphoblasts from 29 newly diagnosed patients using a stromal cell co-culture assay. In all cases the T-ALL lymphoblasts showed an increase in recovery of T-ALL cells (RTC), ranging from 4 to 343%, in comparison to samples maintained without stroma. Since we were blinded to patient outcome in this case-control study, we then correlated patient outcome with RTC. The RTC for 18 patients in complete continuous remission (CCR) for greater than 4 years was stochastically larger than for the 11 patients who eventually relapsed (P = 0.011, by the two-sided Wilcoxon test). Furthermore, 100% of patients with an RTC of more than 26% had a CCR greater than 4 years while 78% of the patients with an RTC of less than 25% relapsed within 4 years. This is the first report to show that higher lymphoblast recovery may predict a more favorable outcome for children with T-ALL. A prospective study is needed to test whether stroma-supported leukemic cell recovery might serve as a basis for assigning risk-adjusted therapy.