P-glycoprotein (P-gp) is a crucial factor in the development of chemotherapy resistance in malignant disorders. Between 1989 and 1995, P-gp expression was studied in bone marrow blast cells of 322 (239 AML; 83 ALL) acute leukemia patients. 166 AML patients with the AML-6 protocol (EORTC), containing daunorubicin, vincristine and conventional-dose cytarabine (ara-C), and 63 AML patients treated with intermediate-does Ara-C plus amsacrine. Further 71 ALL patients were treated according to a German standard polychemotherapy protocol (BMFT04/1989). P-gp was determined by using monoclonal antibodies C219 and 4E3, and the cutoff point for P-gp overexpression was set at >/= 10%. A significant (P < 0.06) difference in P-gp overexpression was demonstrated between AML (21.6%) and ALL (10.2%) patients at primary diagnosis and between primary diagnosis and relapse/refractoriness in AML (21.6%; 51.0%) and ALL (10.2%; 27.2%) patients. According to FAB classification P-gp overexpression was detected in AML patients significantly (P < 0.05) more frequently in classes M4, M5a and M5b and less frequently in M3, as compared to other types. For AML patients with P-gp overexpression at primary diagnosis or early relapse/refractoriness, the predictive value for nonresponse to the AML-6 protocol was 91 and 95%, respectively, while late-relapsed AML patients with P-gp overexpression had a significantly (P < 0.05) lower predictive value of 73% for nonresponse. Additionally, in refractory and late-relapsed P-gp--overexpressing AML patients treated with intermediate-dose ara-C plus amsacrine the predictive values for nonresponse were 44 and 39%, respectively, significantly (P < 0.05) lower as compared to AML-6 protocol-treated refractory or late-relapsed AML patients. In P-gp-overexpressing treated ALL patients the predictive values of 50 and 55% for non-response were calculated at primary diagnosis and late relapse, respectively. We conclude that P-gp overexpression is a common phenomenon in AML patients at primary diagnosis or relapse, has an inverse influence on AML-6 treatment outcome and should be taken into consideration in the development of new therapy strategies.