Abstract

Gene expression profiling has proven useful for identification of gene sets predicting response and/or prognosis in various hematological malignancies. So far expression profiling studies in CML predicting response to imatinib therapy suffered from major flaws, as previous retrospective studies relied on various technical platforms using rather small patient populations. Here, we present the interim results of an expression profiling study performed within a prospective multicenter CML trial of the Central European Leukemia Study Group(CELSG) comparing standard (SD) with high-dose (HD) imatinib as second line treatment for chronic phase CML. Blood samples from at total of 102 CML patients with early or late chronic phase disease were collected prior to and six weeks after treatment with either 400 mg/day for 12 months or 800 mg/day for 6 months followed by 6 months of 400 mg imatinib therapy. Whole blood samples preserved by PAX gene technology were further processed and evaluated centrally at the expression profiling core facility at Innsbruck Medical University. For expression profiling samples were subjected to hemoglobin mRNA reduction before target preparation for hybridization to Affymetrix hGU133 Plus 2 genechips detecting ∼47000 transcripts thereby allowing a whole genome profiling approach. For response prediction we used both the linear discriminatory analysis (LDA) and a random Forest (RF) decision tree algorithm provided in Bioconductor software packages for the open source statistical language “R”. A 100- or 10-fold crossvalidation was performed using a robust leave-10-out setting for LDA and RF respectively. The planned interim analysis of clinical data from 76 evaluable patients undergoing 12 months of therapy with daily doses of 400 mg or 800 mg imatinib revealed 50 (66%) patients with and 26 (34%) patients without a major cytogenetic response (<35% Ph+ metaphases). We found, irrespective of imatinib dosing, a predictive gene expression signature for achieving a major cytogenetic response after 12 months of imatinib therapy. This predictive gene set comprises five genes (ENTPD4, F5, SCL22A4, AIM2, TLR5, PXK) achieving a correct prediction rate of 87% within the 6-week-treatment patient subgroup and an overall prediction rate of 76%. In conclusion, this is the first prospective gene expression profiling study in CML indicating a gene expression profile predicting major cytogenetic response, which still represents a robust response parameter for long-term outcome during imatinib therapy of patients with early or late chronic phase CML.

Author notes

Disclosure:Research Funding: Novartis.