Efficacy, Safety, and Biomarkers of Response to Azacitidine and Nivolumab in Relapsed/Refractory Acute Myeloid Leukemia: A Nonrandomized, Open-Label, Phase II Study

Abstract

Preclinical models have shown that blocking PD-1/PD-L1 pathways enhances antileukemic responses. Azacitidine upregulates PD-1 and IFNγ signaling. We therefore conducted this single-arm trial, in which patients with relapsed/refractory (R/R) acute myeloid leukemia (AML) were treated with azacitidine 75 mg/m² days 1 to 7 intravenously or subcutaneously with nivolumab 3 mg/kg intravenously on days 1 and 14, every 4 to 6 weeks. For the seventy patients who were treated, the median age was 70 years (range, 22-90) and the median number of prior therapies received was 2 (range, 1-7). The overall response rate (ORR) was 33%, including 15 (22%) complete remission/complete remission with insufficient recovery of counts, 1 partial response, and 7 patients with hematologic improvement maintained >6 months. Six patients (9%) had stable disease >6 months. The ORR was 58% and 22%, in hypomethylating agent (HMA)-naïve (n = 25) and HMA-pretreated (n = 45) patients, respectively. Grade 3 to 4 immune-related adverse events occurred in 8 (11%) patients. Pretherapy bone marrow and peripheral blood CD3 and CD8 were significantly predictive for response on flow cytometry. CTLA4 was significantly upregulated on CD4⁺ Teff in nonresponders after 2 and 4 doses of nivolumab. Azacitidine and nivolumab therapy produced an encouraging response rate and overall survival in patients with R/R AML, particularly in HMA-naïve and salvage 1 patients. Pretherapy bone marrow aspirate and peripheral blood CD3 percentage may be biomarkers for patient selection. SIGNIFICANCE: Azacitidine in combination with nivolumab appeared to be a safe and effective therapy in patients with AML who were salvage 1, prior hypomethylator-naïve, or had increased pretherapy CD3⁺ bone marrow infiltrate by flow cytometry or IHC. Bone marrow CD3 and CD8 are relatively simple assays that should be incorporated to select patients in future trials. This article is highlighted in the In This Issue feature, p. 305.

Figure 1.

Swimmers plot illustrating the clinical course of study patients (N = 70). The best response, on or off study status, alive or dead status, and allogeneic stem cell status for the 70 patients enrolled on study is shown in this swimmers plot.

Figure 2

A. Overall survival in the 70 patients treated with azacitidine and nivolumab. Figure 2B. Event free survival in the 70 patients treated with azacitidine and nivolumab. Figure 2C. Duration of response among the 23 patients with a response (CR, CRi, PR, HI) on azacitidine with nivolumab. Figure 2D. Overall survival in patients who had response/stable disease (CR, CRi, PR, HI, SD) versus patients who had no response with azacitidine with nivolumab (N = 70). Figure 2E. Overall survival by the best response to therapy (N = 70) (P value <0.0001). Figure 2F. Event free survival by the best response to therapy (N = 70) (P value <0.0001).

Figure 3

A and B. Bone marrow T-cell profile and checkpoint expression in responders versus non-responders. Bar graphs indicating frequency of CD3+, CD4+Teffector, and CD8+ T cells in total live cells (Fig 5A) and CTLA4+ CD4+Teffectors and CTLA4+CD8+ T cells (Fig 5B) in BMA of responders (CR/CRi/PR/HI) (n=19) and NRs (n=23) at pretherapy, EOC1 and EOC2 as analyzed by flow cytometry. Figure 3C. Phenograph based clustering approach of T cell subsets by mass cytometry (CyTOF). t-SNE map of 10,000 randomly selected CD3+ cells colored by distinct clusters (–24) in responders (Fig 5C left), NRs (Fig 5C middle) and heatmap showing normalized expression of different immune markers on CD3+ metaclusters including Cluster 2 (C2) and cluster 14 (C14) (Fig 5C right). Figure 3D. CD45RA+PD1loTbethiEomeslo (C2) cells were significantly higher in the pretherapy BMAs of patients with CR/CRi (n=5) than NRs (n=5), and with a trend toward expansion in patients with CR/CRi but not in NRs particularly after 8 doses of nivolumab (EOC4), by mass cytometry (Fig 5D left). By contrast, CD4+PD1+(RORgThi) (C14) which was suggestive of Th17-like T-cell population is higher in NR compared to responders (4·0% versus 1·5%; p=0·02). Th17 cells were reported to negatively correlate with prognosis in AML. Each shape/structure in the plot represents an individual patient at baseline and followed over time for this analysis.