THE AUTOMATED MOLECULAR TECHNIQUE “ULTRA” ALLOWS A SENSITIVE AND ACCURATE BCR-ABL1 QUANTIFICATION IN PATIENTS AFFECTED BY CHRONIC MYELOID LEUKEMIA

Background: The chronic myeloid leukemia (CML) is characterized by the presence of the Philadelphia chromosome and the BCR-ABL1 fusion gene. The introduction of tyrosine kinase inhibitors (TKIs) significantly improved the survival, but 15% of patients don’t reach the optimal responses at the defined end-points or develop secondary resistance. The 2013 ELN guidelines identified as fundamental the early molecular response (BCR-ABL1/ABL1 % ≤10% IS), the MR3 (<0.1%) and the deep molecular response (MR4<0.01%, MR4.5<0.032%, and MR5 <0.001%). Consequently, the molecular monitoring plays a crucial role in the clinical management of CML patients, with a consequent research of sensitive and standardized molecular techniques. The automated methods offer advantages in terms of reduced time for analysis, decreased manual steps, and reduction of possible errors and contamination. Aims: We compared the automated technique GeneXpert “Ultra” (Cepheid, Maurens-Scopont, France) with the “Labnet” method that includes the “classical” manual real-time PCR techniques standardized in the Italian network (57 centers), according to the European guidelines [Cross N, 2015]. We compared the sensitivity of the two methods (based on the number of ABL1 detected copies), the classification of molecular responses, with particular attention to the deep molecular one. Methods: We assessed the BCR-ABL1 transcript in 86 patients afferent to laboratories of Pisa, Napoli, Torino, and Bologna (Italy). For statistical analysis, the t-, the Pearson’s and the Cohen’s K test were adopted. Because our patients presented different transcript levels (from the >10% to the 0% (MR4, MR4.5, MR5) the two techniques have been compared in the different molecular subgroups. Results: Firstly we compared the number of detected ABL1 copies, that are fundamental for definition of the molecular response categories, especially for defining the degree of deep response (32,000 for MR4.5, 100,000 for MR5). By the “LabNet” method, 51 (81%) samples exceeded the 100,000 copies of ABL1, while by the automated method 81 samples (94.2%) reached >100,000 ABL1 copies. Then, we compared the two methods in discriminating positive and negative samples (K Cohen=0.690; p <0.02): 77 samples were concordant (89.5%) and only 9 (10.4%) were discordant. Of the 18 negative samples with the “LabNet” method, 2 (11.1%) were in MR4.0, 10 (55.5%) in MR4.5 and 6 (33.4%) in MR5.0. On the other hand, of the 19 negative samples with the method “Ultra”, 1 (5.3%) was in MR4.5 and 18 (94.7%) in MR5.0, confirming the higher sensitivity of the automated method. In the cohort of positive cases by the two methods, the median values of transcript expression were superimposable (p=0.55) and the linear regression coefficient was very satisfying (Pearson’s r=0.9399; p-value <0.0001). Finally we compared the results produced by the two methods according to the “molecular classes” (MR1 vs MR2+MR3 vs MR4+MR4.5 vs MR5). This comparison showed a good concordance between the two methods (Cohen’s k=0.78–good correlation). Variation analysis demonstrated high concordance between “Ultra” and “LabNet” methods using assay comparison criteria proposed by Müller et al. [Leukemia 2009]