Development of a Prognostic Model That Predicts 3- and 5-Year Overall Survival in Multiple Myeloma Using the Connect^® MM Patient Registry

Introduction: Novel therapies have improved the median overall survival (OS) for newly diagnosed multiple myeloma (NDMM). In an unselected population of NDMM patients, the estimated rates of 5-year OS were 77%, 53%, and 19% for revised international staging system (R-ISS) Stage I, II, and III, respectively (Kastritis, 2017). However, the improvements have not been uniform, and prognosis varies considerably based on several baseline characteristics. Because of this clinical heterogeneity it is helpful to have a clear idea of risk factors involved in OS of patients with NDMM. Connect MM is the first and largest US-based, multicenter, prospective observational registry of patients with NDMM. Using data from the Connect MM registry, an unselected patient population reflective of the heterogeneity seen in routine clinical practice, we developed a prognostic tool to identify factors predictive of 3- and 5-year OS.

Methods: The Connect MM registry is designed to describe baseline characteristics, patterns of care, and outcomes for NDMM in clinical practice. Patients were enrolled in Cohort 1 (n = 1493) between Sept 2009 and Dec 2011. Enrollment for Cohort 2 (n = 1518) began Dec 2012 and was completed April 2016. Data cut-off was July 7, 2016. Multivariate analyses of patient- and disease-related characteristics were conducted using Cohort 1 data to identify key predictors of OS. A series of univariate Cox regression models were used to identify variables significantly associated with 3-year and 5-year OS in patients ≤75 years and >75 years, as 75 years was determined to be the most discriminating age cut point, using Connect MM data, in identifying independent factors significantly associated with 3-year and 5-year OS. Prediction matrices for OS were constructed from the Cox model. Models were evaluated using the concordance (C)-index, a goodness of fitness measure. Internal validation of the logistic model in predicting the probability of survival was performed on data from Cohort 1. External validation of the model was performed on data from Cohort 2 and the Celgene-sponsored randomized phase III NDMM trials, MM-015 (N = 459) and FIRST (N = 1623).

Results: Characteristics associated with late survival identified by multivariate analyses (P < 0.05) were: age, EQ-5D mobility, del(17p), ISS stage, platelet count, solitary plasmacytoma, Eastern Cooperative Oncology Group performance status, history of diabetes, creatinine category, and triplet therapy use (Figure). The C-index results were robust and consistent in the external validations: Connect MM Cohort 2 (C-statistic = 0.698), MM-015 (C-statistic = 0.649) and MM-020 (C-statistic = 0.647), thereby confirming model validity.

Conclusions: This is the first assessment of long-term survival using data from the Connect MM patient registry. Our study highlighted factors known to impact OS, including, mobility, del(17p) and ISS stage. We have developed a model that allows for robust prediction of long-term survival in patients with NDMM based on a variety of baseline characteristics. For clinicians, identifying patients with an increased likelihood of longer survival may be a useful tool to identify patient specific risk factors to better inform treatment decisions.

View largeDownload slide

View largeDownload slide

Close modal