Abstract
Multiple myeloma is the second most common hematological malignancy and remains incurable despite high dose chemotherapy. Patients experience survival periods of few months to more than ten years. We used gene-expression profiles of malignant plasma cells obtained at diagnosis to develop and validate biomarkers to predict disease prognosis. Gene-expression profiling was realized on 250 bone marrow samples obtained from patients less than 66 years, treated according to the Intergroupe Francophone du Myelome (IFM) 99 clinical trials, based on double intensive treatments. The median follow-up time was 3 years and 55 patients died from their disease during this period. Supervised methods were performed to identify expression signatures associated with survival in a training group of 182 patients. From these genes, a classifier of prognosis was developed in the training group and validated in an independent validation group of 68 patients. Univariate Cox proportional hazards model at a significant level of p<0.05 and selection procedure from BRB ArrayTools at a significant level of p<0.001 were used to select 50 genes associated with survival. Resampling and permutation tests were performed in an rigorous internal validation procedure and 28 genes with a permutation p<0.005 and a resampling mean p<0.01 were retained. To reduce gene list instability, multiple training/test partitions (100) were performed and the 15 most stable genes were used to build a survival predictor. For each patient, a score based on the first component of principal component analysis was calculated. The patients were ranked according to their score and divided according to the median (low-risk and high-risk). The model was highly significantly associated with survival in the training group (p<0.001) and it was confirmed in the validation group (p<0.001). The relative risk of death for high-risk patients was 8.46 (95% confidence interval (CI) = 3.3 to 21.6) in the training group and was 15.3 (95% CI = 2 to 116.5) in the validation group. Various biological variables including serum beta2-microglobulin, serum albumin, combined in the International Staging System (ISS), the deletions/monosomy of chromosome 13 (del13) and the t(4;14)(p16;q32) have emerged as relevant prognostic factors in MM. The 15-gene prognostic classifier was independent of ISS (p<0.001) with an adjusted hazard ratio of death =7.5 (95% CI = 3.09 to 18.2), of del13 (p<0.001) with an adjusted hazard ratio of death =8.8 (95% CI = 3.7 to 21), and of t(4;14) (p<0.001) with an adjusted hazard ratio of death =9.3 (95% CI = 3.9 to 22.1). These results demonstrate that this molecular predictor is very powerful to identify different categories of patients based on expected survival, and thus may be clinically useful to design therapeutic trials.
Figure
Disclosure: No relevant conflicts of interest to declare.
