Continuous genomic monitoring of multiple myeloma patients to identify patients of high risk for poor prognosis.

person Yu Liu Sema4 Genomics, A Mount Sinai Venture, Stamford, CT info_outline Yu Liu, Haocheng Yu, Alessandro Lagana, Samir Parekh, Eric Schadt, Li Wang, Jun Zhu

Authors person Yu Liu Sema4 Genomics, A Mount Sinai Venture, Stamford, CT info_outline Yu Liu, Haocheng Yu, Alessandro Lagana, Samir Parekh, Eric Schadt, Li Wang, Jun Zhu Organizations Sema4 Genomics, A Mount Sinai Venture, Stamford, CT, Icahn School of Medicine at Mount Sinai, New York, NY, Tisch Cancer Institute, Icahn School of Medicine at Mount Sinai, New York, NY, Mount Sinai School of Medicine Tisch Cancer Institute, New York, NY Abstract Disclosures Research Funding No funding received None Background: Multiple myeloma (MM) is heterogeneous from symptoms, mechanisms of tumorigenesis and progression, to the treatment outcome. Besides age and tumor stage, prognostic models based on the genomic or transcriptomic profiles were developed to stratify patients into risk groups at diagnosis. Even though mutation burden at diagnosis is not associated with overall survival (OS), its change during the course of treatment, which might reflect the emergence of treatment resistant subclones, could provide insights about the MM evolution and hence associate with patient outcome. Methods: In this study, we developed a new parameter referred to as “tumor evolution rate” and investigated its association withclinical outcomes. The tumor evolution rate was measured as the rate of acquiring new somatic mutations among the sequential samples from the same patients. We utilized genomic data generated on 140 MM patients with two sequential samples from the MMRF CoMMpass database. The prognostic power of evolution rate was compared with other clinical parameters as well as gene expression-based markers. Results: There was significant heterogeneity in the evolution rate among MM patients (median, 0.114; 0.011 – 2.34 new mutations per day). The evolution rate was significantly associated with OS in a univariate Cox regression model (Wald test p = 5.0×10 -7 ). When stratified into high, intermediate and low-rate groups, they had a median OS of 316, 743 and 998 days after the collection date of 2 nd samples, respectively (logrank test p = 1.0×10 -6 ). With age and ISS stage as covariates in the multivariate Cox regression model, the evolution rate remained significantly associated with OS (p = 6.3×10 -6 ). The mutation burden itself at either sampling time was not associated with OS. For patients with two sequential samples collected less than 2 years apart (N =82), the evolution rate was significantly associated with OS (Wald test p = 2.0×10 -6 ). However, for patients with two sequential samples collected more than 2 years apart, the evolution rate was not significantly associated with OS anymore, likely due to the inaccurate evolution rate estimates over long intervals. In addition, the patient stratified by the evolution rate shared some similarity with those stratified via a gene expression-based prognostic model (previously developed by us). The majority of patients in the low or intermediate risk group based on gene expression were in the low or intermediate evolution rate group. For patients with a high-risk prognosis (n = 21), 57% (12 out of 21) had tumors with a high evolution rate during treatment, suggesting alternative treatment options may be needed for this group of patients. Conclusions: Our study demonstrated that the tumor evolution rate as derived from continuous monitoring of tumor genomic changes provided extra prognostic information for identifying MM patients of poor prognosis after drug treatment.