Fluorescence in situ hybridization (FISH) can detect minor genetic changes that cytogenetic analysis may miss; however, there are few reports on the kinds of genetic changes that show large discrepancies between results obtained with FISH versus G-banding techniques. To investigate genetic changes that tend to be detected with FISH only, we compared the results of cytogenetic study and FISH analysis in 919 consecutive specimens from 304 patients with hematologic malignancies, covering most of the frequent genetic changes by using 18 types of FISH probes. The genetic changes with especially large discrepancy rates at diagnosis were del(7q) (20.0%), PML/RARA (17.6%), and trisomy 21 (12.5%) and, at follow-up, BCR/ABL (28.2%) and AML1/ETO (24.4%); the latter two showed only small discrepancies at diagnosis (4.7 and 4.8%, respectively). The overall discrepancy rate was 6.0% at diagnosis and 11.9% at follow-up, indicating generally greater discrepancy rates at follow-up. In all but one of the cases with discrepant results, G-banding missed the corresponding chromosomal abnormalities revealed with FISH. Considered by type of leukemia, the discrepancy rate at follow-up was higher in acute biphenoptypic leukemia (38%) and acute lymphoblastic leukemia (24.5%) than in acute myelogenous leukemia (10.6%). Given these results, all patients with known genetic changes should have FISH analysis in follow-up, for an accurate assessment of the likelihood of complete remission or recurrence. If this is not practical, then at a minimum FISH analysis should be done in follow-up for patients with genetic changes of BCR/ABL and AML1/ETO seen at diagnosis.