Molecular Measurable Residual Disease in Patients with Newly Diagnosed mIDH1 Acute Myeloid Leukemia Treated with Ivosidenib + Azacitidine

Background: Ivosidenib (IVO) + azacitidine (AZA) improved complete remission (CR) rates and overall survival (OS) relative to placebo (PBO) + AZA in patients (pts) with newly diagnosed mutant isocitrate dehydrogenase 1 (mIDH1) acute myeloid leukemia (AML) in the pivotal AGILE study (Montesinos et al. NEJM 2022). Prior analyses showed clinical responses to IVO+AZA were associated with deep clearance of mIDH1 (limit of detection [LOD]: 0.02–0.04% variant allele frequency [VAF]), as well as clearance of baseline co-mutations below the threshold of conventional next-generation sequencing (NGS) (LOD: 2.0% VAF) (Döhner et al. Blood 2022; Daigle et al. Blood 2021).

Measurable residual disease (MRD), a negative prognostic marker, following chemotherapy is often detected via multiparametric flow cytometry or quantitative polymerase chain reaction although both techniques have their limitations. (Heuser et al. Blood 2021). NGS can detect a wide range of AML mutations and studies show that NGS-assessed MRD is highly predictive of both relapse and survival (Thol et al. Blood 2018; Tsai et al. Blood Adv 2021). However, a limitation of NGS is difficulty in distinguishing between somatic leukemia-associated mutations and clonal hematopoiesis of indeterminate potential (CHIP) (Bacher et al. Blood Cancer J 2018). We aimed to perform an NGS analysis of MRD responses in the AGILE study.

Methods: All AGILE pts who had a best overall response (BOR) of CR, CR with incomplete count recovery, or CR with incomplete platelet recovery, and had ≥1 on-treatment bone marrow mononuclear cell (BMMC) sample available were included. Suggested timepoints for MRD testing included day 1 of treatment cycles 3, 5, 7, 9, 11, 14, 20, 26, and 32.

Baseline and on-treatment BMMC DNA samples were analyzed using a diagnostic 51-gene myeloid NGS panel (LOD: 3% VAF) and a 26-gene AML MRD panel, respectively. For variants detected at baseline, the LOD was 0.1% VAF; for variants not detected at baseline (either emerging mutations or if baseline samples were unavailable), the LOD was 0.5% VAF.

All variants above the applicable LOD with known or potential clinical significance were considered evidence of positive MRD, aside from DNMT3A, TET2, and ASXL1 (“DTA”) mutations to reduce the risk of false-positive MRD due to CHIP.

Results: The analysis set comprised 173 samples from 33 IVO+AZA-treated pts and 10 PBO+AZA-treated pts (Fig. 1). Clinical response data were based on a March 18, 2021 primary analysis data cut; OS data were based on an updated June 30, 2022 data cut. The median number of baseline mutations was 4 (range: 1–10) per patient, median number of MRD assessments was 3 (range: 1–9) and median follow-up period was 189 days (range: 49–875).

Ten (30.3%) IVO+AZA-treated pts became MRD-negative (MRD_neg), compared with 2 (20.0%) PBO+AZA-treated pts. All pts attaining an MRD_neg response had a BOR of CR (10 [34.5%] of 29 IVO+AZA-treated pts in CR; 2 [22.2%] of 9 PBO+AZA-treated pts in CR). Seven (70.0%) of the 10 MRD_neg IVO+AZA-treated responding pts converted to an MRD_neg response by day 1 of cycle 7 (C7D1). Fewer pts with an MRD_neg response had an Adverse risk at baseline according to European LeukemiaNet 2022 genetic risk classification versus those pts with an MRD-positive (MRD_pos) response (26.3% vs 73.7%) (Table 1).

In the IVO+AZA arm, rates of event-free survival at 12 months and OS at 24 months were numerically higher in MRD_neg vs MRD_pos responders (80.0% vs 60.6% and 88.9% vs 72.3%, respectively) but did not reach statistical significance.

All 5 (15.2%) IVO+AZA-treated pts with confirmed relapses had detectable MRD at the last completed assessment prior to relapse. One pt converted from an MRD_neg to an MRD_pos response (“MRD relapse”) 169 days prior to overt clinical relapse. Among the 23 IVO+AZA-treated pts without an MRD_neg response, 16 were alive as of the updated data cut; OS duration ranged from 20.3–48.9 months.

Conclusions: IVO+AZA induced molecular MRD negativity in approximately one third of responding pts with newly diagnosed mIDH1 AML; MRD_neg responses were most often observed on or before C7D1. Interestingly, several pts had durable clinical responses and OS despite ongoing MRD positivity, potentially due to persistent CHIP rather than true MRD, illustrating the complexity of interpreting NGS MRD data. Additional analyses, including associations between specific molecular alterations, MRD response, and OS are ongoing and will be presented.