The impact of cell dose (number of nucleated donor cells per kilogram recipient weight) on transplantation outcome is controversial and may differ for allogeneic and identical twin (syngeneic) bone marrow transplants. We studied the association between cell dose and outcome in 100 unmanipulated identical twin bone marrow transplantations for leukemia, reported to the International Bone Marrow Transplant Registry between 1985 and 1994, using Cox proportional hazards regression for multivariate analyses. Cell doses ranged from 0.3 to 7.4 x 10(8) nucleated cells/kg (median, 3.0 x 10(8)cells/kg). Median follow-up was 75 months. Five-year cumulative incidences of transplant-related mortality with high (more than 3 x 10(8) cells/kg) versus low (less than or equal to 3 x 10(8) cells/kg) cell doses were 2% (95% confidence interval [CI], 0% to 8%) versus 10% (95% CI, 4% to 20%), respectively. Five-year probabilities of leukemia-free survival were 53% (95% CI, 39% to 67%) and 37% (95% CI, 23% to 52%), respectively. In multivariate analysis, among patients surviving in remission at least 9 months after transplantation, those receiving high cell doses were at significantly lower risk for treatment failure (relapse or death) than those receiving low cell doses (RR, 0.27; 95% CI, 0.12 to 0.6; P =.001). Lower treatment failure resulted from fewer relapses in the high cell dose group (RR for relapse, 0.28; 95% CI, 1.2 to 0.66; P =.003). These findings suggest that outcomes after syngeneic bone marrow transplantation could be improved by transplanting more than 3 x 10(8) nucleated cells per kilogram. The benefit of high cell dose on relapse may represent a delayed graft-versus-leukemia effect.