Background: Point mutations in the kinase domain (KD) of BCR-ABL that impair drug binding are a common cause of imatinib resistance. In some cases the mutations present at relapse were detected in pretherapeutic samples, suggesting selection in the presence of imatinib and a possible association with disease progression. To test this hypothesis, we studied the presence of KD mutations in imatinib-naive patients in various disease phases. Methods and patients: we developed fluorescent allele-specific RT-PCR assays for 8 common KD mutations (Q252H, Y253F, Y253H, E255K, E255V, T315I, M351T, F359V), comprising approximately 75% of mutations detected in patients, using universal BCR and mutation-specific ABL primers. Mutation-complementary oligonucleotides were further modified to favor amplification of mutant over wild type. Sensitivity and specificity were optimized in serial dilutions of mutant in wild type, using plasmid DNA or cDNA extracted from BaF3 cells expressing BCR-ABL constructs. The assays reproducibly detected mutations with a sensitivity of at least 10 −4 – 10 −5 . To avoid false-positive results cDNA was prepared in a physically separate lab that never handled recombinant mutant BCR-ABL. Negative controls included “blanks”, K562 cells (positive for wild type BCR-ABL) and HL60 cells (BCR-ABL-negative). Pretherapeutic samples from 20 patients in first chronic, 24 in accelerated and 21 in blastic phase were studied. Results: Mutations were detected in 13 samples from blastic phase (5) or from accelerated phase (8) but in none from chronic phase. Mutations were: T315I (7), Y253F (2), 253H (2), Q252H (1) and F359V (1). Sequencing of the same material detected F359V in one sample (consistent with the allele specific PCR), and K247R in another (positive for T315I by allele-specific PCR) but was wild type in the remaining 10 samples with good quality readings. Multivariate analysis of baseline factors revealed significant and independent correlations between mutation detection and disease phase ( P = 0.0008) and clonal cytogenetic evolution ( P = 0.0004). In contrast, mutation detection was not an independent predictor of failure to achieve complete hematologic response (CHR) or major cytogenetic response (MCR) and was not correlated with overall survival. Four patients with mutations achieved CHR and 3 MCR; 4 patients failed to achieve CHR, 1 died from neutropenic sepsis and 1 discontinued because of skin toxicity. Thus far, follow-up specimens during imatinib therapy have been sequenced in 6 patients. In 2 cases (T315I and Q252H) the identical mutation was detected, one case (T315I) showed K247R (detected prior to therapy by sequencing but not included in the allele specific assays), and 3 cases were wild type BCR-ABL. Conclusion: We provide direct evidence that the presence of BCR-ABL KD mutantions is correlated with disease stage. Moreover, the strong association with clonal evolution suggests that genomic instability may be responsible for both phenomena. However, not all mutants that were detected prior to therapy were subsequently selected in the presence of imatinib, suggesting that some mutations may occur in cells without self-renewal capacity or that additional factors may be required to confer a fully resistant phenotype.