Gene Expression Profiling of Extramedullary Disease-Related Toward Identification of a Terminal Disease Pathway in Multiple Myeloma

Introduction

Extramedullary disease (EMD) is a primary disease manifestation of MM, which while not seen frequently at presentation increases in incidence at relapse where its incidence seems to be increasing following the introduction of novel agents.  Patients with EMD have a shorter overall survival as well as an increased incidence of anemia, thrombocytopenia, elevated serum lactate dehydrogenase, cytogenetic abnormalities, and high-risk features as determined by gene expression profiling.  There is also an increased incidence of the high risk MAF subtypes t (14:16 or 14; 20).  Understanding the biology of EMD and identifying its present could give important information about how to improve the outcome of this group.  In this work we have used GEP analysis of bone marrow derived plasma cells to predict the presence of EMD so that we can identify the genomic risk factors that define the features of a plasma cell clone, which can develop the capacity to metastasize outside the BM.

Materials and Methods

We focused on patients treated on TT protocols, at the UAMS, Myeloma Institute between 1989 - 2010, a total of 1154 patients, of which 46 developed EMD before the start of therapy (EMD-1), and 91 developed EMD after registration to UAMS for MM treatment  EMD-2.

Results

We show that most   EMD2 cases (57.14%) develop within 3 years after initiation of therapy  at the UAMS with few cases developing after this time.

 

Predicting the risk of EMD

Combining patients with EMD1 and EMD2 diagnosis within 3 years gave a total of 98 EMD cases. We used 824 samples from 1017 myeloma patients who never developed EMD and had follow up at least 3 years as a comparator group. The data were divided into training (n=619 with 66 EMD cases and 553 controls) and  test sets (n=303 with 32 EMD cases and 271 controls).  Using the training set, we identified 5 significant gene probes (with a q value < 0.001) and made a score to predict cases and controls. The sensitivity and specificity turned out to be 74.24% and 77.40% in the training set, and 56.25% and 76.75% in the test set, respectively.

Predicting the time to EMD2

We tested whether we could predict time to EMD2 based on using baseline GEP samples. In this analysis, all EMD2 cases and controls were included. We divided the data into training (n=743 with 61 EMD2 and 682 controls) and  test sets (n=365 with 30 EMD2 and 335 controls). By fitting a uniform Cox regression model to each gene in the training set, we identified 68 gene probes that are associated with time to EMD2 (with a q-value <0.1). We then created a score based on the 68 gene probes and identified an optimal cutoff based on the training set. Applying the optimal cutoff to both training and test sets, we found that the new 68-gene high/low risk model is a good predictor on the cumulative incidence of EMD2 (p value < 0.0001).

 

Conclusion

We show that EMD2 cases mostly occur within 3 years of diagnosis and a 68 gene based risk score that can predict a cumulative incidence of EMD. Of the 68 genes that are used to develop the prognostic score for EMD, 6 genes are also part of the 70-gene risk score developed by our group.  GEP studies can help us identify EMD-specific gene signature that can further help develop target agents.