Thirty-six patients with Philadelphia chromosome-positive acute lymphoblastic leukemia (Ph+ ALL) were studied for the presence of the bcr-abl fusion mRNA transcript after an allogeneic matched related (N = 12), partially matched related (N = 4), matched unrelated (N = 14), autologous (N = 5), or syngeneic (N = 1) bone marrow transplant (BMT). Seventeen were transplanted in relapse, and 19 were transplanted in remission. Twenty-three patients had at least one positive bcr-abl polymerase chain reaction (PCR) assay after BMT either before a relapse or without subsequent relapse. Ten of these 23 relapsed after a positive assay at a median time from first positive PCR assay of 94 days (range, 28 to 416 days). By comparison, only 2 relapses occurred in the 13 patients with no prior positive PCR assays; both patients had missed at least one scheduled follow-up assay and were not tested 2 months and 26 months before their relapse. The unadjusted relative risk (RR) of relapse associated with a positive PCR assay compared with a negative assay was 5.7 (95% confidence interval 1.2 to 26.0, P = .025). In addition, the data suggest that the type of bcr-abl chimeric mRNA detected posttransplant was associated with the risk of relapse: 7 of 10 patients expressing the p190 bcr-abl relapsed, compared with 1 of 8 who expressed only the p210 bcr-abl mRNA (P = .02, log-rank test). The RR of p190 bcr-abl positivity compared to PCR-negative patients was 11.2 (confidence interval 2.3-54.8, P = 0.003), whereas a positive test for p210 bcr-abl was apparently not associated with an increased relative risk. In separate multivariable models, PCR positivity remained a statistically significant risk factor for relapse after separately adjusting for donor (unrelated and partially matched v matched, autologous, and syngeneic), remission status at the time of transplant, the presence of acute graft-versus-host disease (GVHD), and type of conditioning regimen (total body irradiation dose of < or = 1,200 cGy v > 1,200 cGy). The PCR assay appears to be a useful test for predicting patients at high risk of relapse after BMT and may identify patients who might benefit from therapeutic interventions. The finding that the expression of p190 bcr-abl may portend an especially high risk of relapse suggests a different clinical and biologic behavior between p190 and p210 bcr-abl.