Molecular monitoring of the BCR-ABL transcript in chronic myelogenous leukemia (CML) using quantitative RT-PCR provides clinicians with important diagnostic and prognostic information. To determine whether molecular detection and monitoring of CML is comparable using peripheral blood (PB) and bone marrow (BM) aspirate samples, we performed a prospective study using quantitative real-time RT-PCR (QRT-PCR) of paired PB and BM samples from 41 patients with CML entered onto a single Cancer and Leukemia Group B (CALGB) treatment study. QRT-PCR analysis of PB and BM samples was performed prior to initiation of, and during, treatment with homoharringtonine and cytarabine on a CALGB study for previously untreated CML. Statistical analyses demonstrated good agreement of PB and BM pre-treatment samples. However, using the Bland-Altman statistical method that measures true agreement between PB and BM values, we found that there was only modest agreement of BCR-ABL measurements in PB and BM for samples obtained during treatment. PB values obtained during treatment tended to be lower than the corresponding BM values [average difference = -0.37 (p<0.001) in 36 paired samples] and the 95% limits of agreement ranged from -1.23 to 0.48. Nevertheless, our study demonstrates that BM and PB QRT-PCR values followed a similar trend during treatment (Spearman correlation coefficient, 0.83; 95% CI, 0.70, 0.96). Our data suggest that, quantitatively, PB and BM measurements of BCR-ABL are frequently disparate. Since BM values tended to be higher than PB values, BM sampling provides the most accurate assessment of minimal residual disease (MRD). Based on these results, we caution against interchanging BM with PB sampling for MRD monitoring during treatment of CML since this may lead to misinterpretation of treatment results.