Recognition of early mortality in multiple myeloma by a prediction matrix

Abstract

Early mortality (EM; death ≤ 6 months from diagnosis) has been reported in several newly diagnosed multiple myeloma (NDMM) trials. Before the era of novel agents, the incidence was 10%-14%. Causes of death included infections/pneumonia, renal failure, refractory disease, and cardiac events. Staging systems, such as the revised International Staging System (r-ISS), and prognostic factors including cytogenetics, lactate dehydrogenase levels, and myeloma-specific factors, are useful to assess overall prognosis; however, they cannot predict EM. We evaluated patients treated with novel agents in the Connect MM^® Registry and identified risk factors of the EM cohort. Eligible patients were enrolled in the registry within 60 days of diagnosis. Univariate and multivariate analyses were conducted to evaluate associations between baseline characteristics and EM. Prediction matrices for EM were constructed from a logistic model. Between September 2009 and December 2011, 1493 patients were enrolled in the registry and had adequate follow-up. Of these patients, 102 (6.8%) had EM and 1391 (93.2%) survived for > 180 days. Baseline factors significantly associated with increased EM risk included age > 75 years, higher Eastern Cooperative Oncology Group performance status, lower EQ-5D mobility score, higher ISS stage, lower platelet count, and prior hypertension. Renal insufficiency trended toward increased EM risk. These risk factors were incorporated into a prediction matrix for EM. The EM prediction matrix uses differential weighting of risk factors to calculate EM risk in patients with NDMM. Identifying patients at risk for EM may provide new opportunities to implement patient-specific treatment strategies to improve outcomes.

Figure 1

EMPM with estimated probability of mortality within 180 days (A) and external validation model using data from previous trials (B, C, and D). (A) Green, yellow, and red shading represent lower, intermediate, and higher probabilities of EM, respectively. As an example in the practical use of the EMPM, consider a patient with NDMM enrolled in the Connect MM Registry. This patient answered “I have some problems walking about” to the EQ‐5D mobility question, and his platelet count was 150 × 10⁹/L. This patient had an ECOG PS of 2, had a history of hypertension, and was 56 years old. He had ISS stage III disease and a serum creatinine level of 3.19 mg/dL. Entering these baseline factors into the EMPM showed that this patient had a 57% chance of mortality within the first 180 days on study and, in actuality, the patient died at 26 days. Creat indicates serum creatinine and PC, platelet count. (B) Data from MM‐015, (C) data from the FIRST trial, and (D) data from the second cohort of the Connect MM registry. Triangles represent observations in groups of 30 (groups ordered from most probable to least probable) for whom the actual probabilities (from the MM‐015, FIRST trial, or Cohort 2 data) are plotted against the predicted probabilities (based on the Connect MM logistic model). The solid line and the curved dotted line (nonparametric curve) are the fitted curves for the plot of actual and predicted probabilities. The C‐index or concordance probability is the probability that a randomly selected pair of patients in the independent external data set, one with a poorer survival outcome than the other, will be correctly differentially identified based on inputting the 2 patients’ baseline prognostic characteristics in the fitted model obtained from the Connect MM Registry.