A randomized phase 2 trial of azacitidine with or without durvalumab as first-line therapy for older patients with AML

Abstract

Evidence suggests that combining immunotherapy with hypomethylating agents may enhance antitumor activity. This phase 2 study investigated the activity and safety of durvalumab, a programmed death-ligand 1 (PD-L1) inhibitor, combined with azacitidine for patients aged ≥65 years with acute myeloid leukemia (AML), including analyses to identify biomarkers of treatment response. Patients were randomized to first-line therapy with azacitidine 75 mg/m2 on days 1 through 7 with (Arm A, n = 64) or without (Arm B, n = 65) durvalumab 1500 mg on day 1 every 4 weeks. Overall response rate (complete response [CR] + CR with incomplete blood recovery) was similar in both arms (Arm A, 31.3%; Arm B, 35.4%), as were overall survival (Arm A, 13.0 months; Arm B, 14.4 months) and duration of response (Arm A, 24.6 weeks; Arm B, 51.7 weeks; P = .0765). No new safety signals emerged with combination treatment. The most frequently reported treatment-emergent adverse events were constipation (Arm A, 57.8%; Arm B, 53.2%) and thrombocytopenia (Arm A, 42.2%; Arm B, 45.2%). DNA methylation, mutational status, and PD-L1 expression were not associated with response to treatment. In this study, first-line combination therapy with durvalumab and azacitidine in older patients with AML was feasible but did not improve clinical efficacy compared with azacitidine alone. ClinicalTrials.gov: NCT02775903.

© 2022 by The American Society of Hematology. Licensed under Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND 4.0), permitting only noncommercial, nonderivative use with attribution. All other rights reserved.

Figures

Graphical abstract

Figure 1.

Maximal blast reduction from baseline for responders. The waterfall plot shows the percentage of blast reduction for all responders. Each responder’s blast percentage is illustrated as a bar along the x-axis. The y-axis shows the change from baseline in the percentage of blasts.

Figure 2.

Analysis of methylation patterns during treatment. The figure shows changes in GDMS in PB of patients with AML upon treatment with azacitidine + durvalumab or azacitidine.

Figure 3.

Changes in surface PD-L1 expression on BM cells during treatment. (A) This plot represents the mean (circles) and 90% CIs (error bars) for surface PD-L1 (CD274) abundance (MESF) at screening and C3D22 on 3 different cell types: granulocytes, monocytes, and tumor blasts. PD-L1 is notably higher at C3D22 compared with screening on monocytes for Arm A patients because the CIs are not overlapping. (B) This plot represents the mean (circles) and 90% CIs (error bars) of the C3D22 minus screening values for patients who had both time points measured. In this plot, we also observed that PD-L1 has an increased density on monocytes for Arm A patients because the CIs do not span the dashed line at zero. However, PD-L1 on tumor cells is not increased beyond the CI in either plot. Combo, combination therapy; Mono, monotherapy.

Figure 4.

RNA-sequencing analyses. (A) Increase or decrease of gene expression from screening to C3D22, split by treatment arm. CD3D (T-cell gene), PDCD1 (PD-1), CD274 (PD-L1), interferon-γ signature (mean of 4 genes [CD274, LAG3, IFNG, CXCL9]), CD34 (gene expressed on tumor cells), and PNMA5 (cancer-testis antigen). The x-axis is the log2 of the fold change between C3D22 and screening for patients with samples at both time points. (B) Interferon-γ signature at screening and at C3D22 in responders vs nonresponders to treatment with azacitidine and durvalumab (Arm A) or azacitidine monotherapy (Arm B). The plot on the left shows the screening samples and C3D22 samples separately, with lines connecting those which are from the same patient. The y-axis is the expression score. The right plot shows the fold change for those patients with both time points. The y-axis is the log2 of the fold change between C3D22 and screening. For all plots in A and B, the line in the middle of the box is the mean, and the edges of the boxes are the 90% CIs.

Figure 5.

Effect of mutation status (WT vs mutated) on overall response. (A) The bar graph summarizes the mutation frequency for genes analyzed in this study. The x-axis presents the number of patients with a mutation, and the y-axis presents a bar for the genes of interest. Red bars, Arm A; blue bars, Arm B. (B) The graph summarizes probabilities of overall response among patients stratified by mutation status. The x-axis presents the odds ratio (circles) with 80% CIs (error bars) for overall response, and the y-axis presents the genes of interest. Red circles and error bars, WT gene status; blue circles and error bars, mutated gene status.