Expression of the multidrug resistance efflux pump P-glycoprotein (Pgp) was measured in a series of AML patients using two flow cytometry methods. Transport function was assessed by measuring the modulating effect of the Pgp inhibitor cyclosporin A (CsA) on the cellular accumulation of daunorubicin, and Pgp antigen expression by surface immunofluorescence using the MRK-16 antibody. Both methods showed a wide range of values for Pgp expression between individual patients, but in contrast to a series of cell lines expressing Pgp there was no correlation between antigen expression and transport function in the clinical samples. As previously reported for chronic lymphocytic leukemia (CLL), pretreatment with neuraminidase markedly improved MRK-16 staining in some cases, indicating that abnormal glycosylation can cause epitope masking in AML blasts. Because experience with cell lines shows that Pgp expression is a continuous variable which correlates with the level of drug resistance, rather than the 'positive' or 'negative' which are frequently reported by clinical flow cytometry laboratories, we used a calibration procedure to estimate the actual number of Pgp molecules expressed in the AML samples. Despite the additional refinements of neuraminidase treatment and antigen quantification, the correlation between Pgp antigen expression and daunorubicin accumulation remained extremely weak (r = 0.11; P = 0.63). It is suggested that the assay for transport function can detect molecules that affect daunorubicin accumulation but are antigenically distinct from classical P-glycoprotein. Heterogeneity of multidrug resistance efflux pumps might in part explain the relatively weak prognostic significance of immunofluorescence detection of Pgp in AML patients.