We studied the frequency, causes, and predictors of adverse events in 624 patients who had completed treatment for acute lymphoblastic leukemia (ALL) in three consecutive total therapy studies (VII, IX, and X, 1972 to 1983). Event-free survival in study X was significantly better overall than that in studies VIII and IX (P less than .0001 by the log-rank test). In study X, 75% of the patients were electively taken off therapy, compared with 54% in studies VIII and IX. However, the risks of having an adverse event during the first 5 years after completion of therapy were remarkably similar: 22% (95% confidence interval, 17% to 29%) in study X versus 24% (20% to 29%) in studies VIII and IX. Bone marrow, testicular, and CNS relapses accounted for the majority of failures in both groups (85% in study X and 92% in studies VIII and IX). Late adverse events consisted largely of hematologic relapses and the development of solid tumors. Black race (P = .001) and leukemia without an anterior mediastinal mass (P = .05) were associated with an increased risk of failure after completion of treatment in the two earlier clinical trials, whereas a lower leukemic cell DNA content (DNA index less than 1.16) was the only predictor of late treatment failure in the more recent trial (P = .019). None of the other presenting features that were examined (eg, age, leukocyte count, and sex) had value as predictors of late failure. Thus, improved treatment altered the impact of specific prognostic factors and the distribution of sites of relapse, but it did not significantly affect the risk of delayed failure.