The application of DLI is limited by the potential development of GVHD. Results of single-arm trials suggest that CD8+ depletion of DLI may reduce the incidence of GVHD while still inducing pathologic and cytogenetic remissions. To test the impact of CD8 depletion on GVHD, we initiated a randomized trial comparing outcome among patients receiving unselected donor lymphocytes or CD8+-depleted cells. DLI was administered to patients with disease in remission to prevent relapse 6 months after T-cell-depleted allogeneic BMT. CD8 depletion was performed with monoclonal antibody and rabbit complement. Donor lymphocytes obtained from the original donor were infused fresh without cryopreservation. Infusions were adjusted so that all patients received 1.0 x 10(7) CD4+ cells/kg. Patients randomized to CD8 depletion received a median of 0.7 x 10(5) versus 32.0 x 10(5) CD8+ cells/kg in the unmanipulated cohort. Six (67%) of 9 patients receiving unselected DLI developed acute GVHD compared with 0 (0%) of 9 recipients of CD8-depleted DLI (P = .009). In the unselected group, 2 patients died while the disease was in remission, and 3 patients had relapses. In the CD8-depleted cohort, there were no toxic deaths and only 1 relapse. Measures of immunologic reconstitution by T-cell receptor excision circle analysis and T-cell receptor spectratyping demonstrated similar patterns of T-cell recovery in both the CD8-depleted and the unselected cohorts. Both groups converted from mixed to full donor hematopoietic chimerism after DLI. Our results indicate that CD8 depletion reduces the incidence of GVHD associated with DLI without adversely affecting conversion to donor hematopoiesis or immunologic recovery.