Change in proportion of metaphases with a Ph chromosome is often used to measure effect of treatment in chronic myelocytic leukemia, but this method requires analysis of dividing cells. Dual-color fluorescence in situ hybridization (D-FISH) analysis of bone marrow (BM) or blood (PB), can be used to establish the proportion of interphase nuclei with BCR/ABL fusion. We studied the correlation of these three measures with each other and with response classification categories. In a clinical trial to test rIFN(-) versus rIFN(-) +Ara-C (N = 65 patients), we defined for each cytogenetic and FISH parameter, the patient's "best response" over time and at any given time. The strength and nature of agreement among the three measurements were determined by regression, correlation, and kappa statistic. Correlations between cyBM versus D-FISH-BM, cyBM versus D-FISH-PB and D-FISH-BM vs. D-FISH-PB were fairly strong. Regression showed cyBM values were significantly higher than those of D-FISH-PB or D-FISH-BM. No difference between D-FISH-PB and D-FISH-BM was apparent. When numerical responses (BR%) were classified into response categories (CR, PR, MR, NR), levels of agreement were weak to moderate. We observed a trend toward disagreement between cyBM and each D-FISH method, with cyBM tending to "downgrade" response category. Whether analyzed using actual numerical values or classified into response categories, cyBM differed from D-FISH-BM or D-FISH-PB. The results indicate that mixing data of different methods to measure treatment response within or between patients may be misleading clinical index.