Baseline Predictors of Mortality in Patients with Relapsed or Refractory Acute Myeloid Leukemia Treated with Vosaroxin Plus Cytarabine or Placebo Plus Cytarabine in the Phase 3 VALOR Study

Introduction: The rate of early mortality in patients treated for acute myeloid leukemia (AML) reflects a balance of efficacy and safety outcomes. In the randomized phase 3 VALOR trial in patients with first relapsed or refractory AML, rates of 30-day (7.9% vs 6.6%, respectively) and 60-day (19.7% vs 19.4%, respectively) mortality were comparable between vosaroxin plus cytarabine and placebo plus cytarabine arms.

The treatment-related mortality (TRM) score is a prognostic scoring system to predict risk of 30-day mortality with intensive treatment protocols in patients with newly diagnosed AML (Walter, 2011, J Clin Oncol 29:4417-4424). The "simplified TRM" score includes age, performance status (PS), platelet count, serum albumin, type of AML (secondary vs primary), white blood cell count, blast percentage in the peripheral blood, and serum creatinine. Here we present a multivariate analysis to identify baseline predictors of 30-day and 60-day mortality and a post hoc analysis of the 30-day mortality by TRM score to inform the assessment of treatment risk in VALOR patients.

Methods: In VALOR, patients were randomized 1:1 to receive vosaroxin (90 mg/m² intravenously [IV] over 10 min in cycle 1 and 70 mg/m² in subsequent cycles, d 1 and 4) plus cytarabine (1 g/m² IV, over 2 h, d 1-5) or placebo plus cytarabine. Baseline factors associated with 30-day mortality and 60-day mortality in patients treated with vosaroxin/cytarabine were identified from a univariate logistic regression analysis. Factors with significant associations (P < 0.10 by Type 3 P values) that were considered clinically and biologically relevant were included in a multivariate logistic regression model in the overall safety population. A stepwise selection procedure with entrance and retention criteria set to 0.05 was employed to reduce the risk factors to a parsimonious set. Simplified TRM scores were calculated retrospectively according to Walter 2011.

Results: A total of 705 patients were included in the safety population (355 treated with vosaroxin/cytarabine and 350 treated with placebo/cytarabine). Based on multivariate analysis, vosaroxin/cytarabine treatment was not predictive for increased 30-day or 60-day mortality. Eastern Cooperative Oncology Group PS > 1 was predictive (P < 0.05) of 30-day mortality (odds ratio [OR] = 3.2) and 60-day mortality (OR = 2.1). Additional predictive factors identified for 30-day mortality were hemoglobin < 10 g/dL (OR = 3.8) and bilirubin > 1.0 mg/dL (OR = 3.3). For 60-day mortality, other predictive factors were albumin ≤ 3.6 g/dL (OR = 2.0), intermediate or high bone marrow blasts (10% to < 30% [OR = 2.6] or ≥ 30% [OR = 6.3]), prior myelodysplastic syndrome (OR = 2.2), and achievement of complete remission with first-line therapy (OR = 0.66). Age (< 65 y vs 65-69 y vs 70-74 y vs ≥ 75 y) was not a significant predictor of early mortality in the vosaroxin arm in the univariate logistic regression and therefore was not selected for multivariate analysis, although interestingly, age was predictive of 30-day mortality in the control arm. When added to the final multivariate model, age was of borderline significance (P = 0.0606).

The median simplified TRM score at baseline was 2.6 (range 0-21) in the vosaroxin/cytarabine arm and 2.6 (range 0-33) in the placebo/cytarabine arm. Most patients (473/554, 85%) for whom a simplified TRM score could be calculated had a score < 8. Patients with lower TRM scores had lower rates of 30-day mortality (Table).

Conclusions: Several individual baseline factors independently predicted risk of early mortality in patients with relapsed/refractory AML treated with vosaroxin or placebo plus cytarabine. In addition, the previously validated TRM score for predicting early mortality in newly diagnosed AML also predicted mortality in this relapsed/refractory population. In future studies of vosaroxin (and other intensive regimens), patient selection based upon these predictors of early mortality should be considered.