Gene Expression Profiling of Myeloma Cells at Diagnosis Can Predict Response to Therapy with Thalidomide and Dexamethasone Combination.

Background: Thalidomide and dexamethasone combination has been shown to be an effective therapy for multiple myeloma (MM). In newly diagnosed patients with MM, the combination results in overall response rates of over 70% of patients. However, a considerable number of patients do not respond and will need to proceed to other therapies. The ability to identify patients who are unlikely to respond to particular therapies will allow tailoring of therapeutic approaches and in turn will reduce unnecessary toxicity and morbidity from lack of rapid disease control. With the availability of new technology such as gene expression profiling (GEP), this has become a possibility. We employed this approach to identify genes that can reliably predict lack of response to the thalidomide dexamethasone combination.

Methods: Patient samples from the Eastern Co-operative Oncology Group clinical trial (E1A00), that compared thalidomide-dexamethasone combination with dexamethasone, and patient samples from the phase II Mayo Clinic study of thalidomide and dexamethasone, both for newly diagnosed MM, were used for this study. Thirty newly diagnosed patients who were evaluable for response were included in the analysis. GEP was performed using the Affymetrix U133A microarray platform as per manufacturer’s recommendation. The Affymetrix output (CEL files) was imported into Genespring 7.2 (Agilent Technologies) microarray analysis software, normalized across chips using GCRMA followed by per gene normalization to median. For the purposes of this study, responses were defined as reduction in the serum paraprotein (or the urinary M protein in the absence of a serum M protein) of >=50% (PR), 25 to 49% (MR), increase of >=50% (PD) and stable disease for the remaining.

Results: Five of the 30 patients had no response to the thalidomide Dexamethasone therapy. Using the class prediction tool available in Genespring (Support Vector Machines), we identified 25 genes that reliable predicted non responders (PD, NC) from the responders (MR, PR, and CR). See table for a list of predictive genes with identified genes.

Conclusion: Using a combination of two datasets we have identified a set of genes that can be used to reliably predict lack of response to thalidomide and dexamethasone in patients with newly diagnosed MM. We think that this represents a step towards creation of custom microarrays spotted with genes that are capable of predicting lack of response (or response) for the purpose of tailoring therapy to the patient.