To increase the graft-vs.-leukemia (GVL) effect while maintaining a low mortality from graft-vs.-host disease (GVHD), we conducted a prospective study of T cell titration for 144 patients (90 related, 54 unrelated) between June 1994 and June 1997. Following infusion of a T cell-depleted marrow graft, predetermined doses of T cells, based on the risk factors for GVHD, were administered up to 3 times if greater than a grade II acute GVHD was not seen. Graft failure occurred in three unrelated recipients (2%). Cumulative grades II-IV acute GVHD were seen in 58 +/- 9% of all recipients; 52 +/- 11% related and 75 +/- 13% unrelated. The incidence of grades II-IV acute GVHD following the third add-back (AB) of T cells 78 median days after marrow infusion was lower than that of the earlier ABs: first AB, 36 +/- 8%; second AB, 32 +/- 11%; third AB, 15 +/- 12% (p < 0.05). Chronic GVHD occurred in 56 +/- 12% of related and 79 +/- 16% of unrelated patients. Six died of acute GVHD and two died of chronic GVHD, with an overall GVHD mortality of 6 +/- 4%. In multivariate analyses, unrelated recipients and patients at low risk for GVHD who received a larger number of T cells were identified as patient groups with significant risk for acute and chronic GVHD (both p < 0.05). Unrelated transplant is also shown to be significant for GVHD-related death (p < 0.01). Relapse-free survival of patients with leukemia was shown to be most dependent on chronic GVHD and grades II-IV acute GVHD (both p < 0.01). Anti-leukemic activity independent of GVHD was not observed.