Paper Details

ABSTRACT

Introduction: Retrospective studies have shown that measurable residual disease (MRD) positivity at morphologic complete remission (CR) in patients with acute myeloid leukemia (AML) prior to allogeneic hematopoietic cell transplantation (alloHCT) is associated with higher risk of relapse. Various molecular MRD approaches have been validated, but these methods are restricted to specific genetic subsets and require knowledge of tumor-derived variants present at AML diagnosis. Methylation-based profiling offers a potentially tumor-agnostic alternative for pre-alloHCT MRD assessment in AML by interrogating differentially methylated regions to detect tumor-derived signal at very low tumor fractions. Here, we evaluated the performance of a plasma-based tumor-agnostic targeted methylation cell-free DNA (cfDNA) platform (GRAIL Inc., Menlo Park, CA) for pre-alloHCT AML MRD assessment using samples from the prospective multicenter MEASURE study (Molecular Evaluation of AML Patients After Stem Cell Transplant to Understand Relapse Events, NCT05224661).

Methods: MRD status was assessed using a plasma-based tumor-agnostic targeted methylation cfDNA assay on remission plasma samples collected in EDTA tubes a median of 14 days prior to alloHCT. A machine-learning classifier of hematological malignancies was optimized specifically for AML detection to interrogate informative methylation regions for signals indicative of cancer. This classifier (AML MRD classifier) was locked prior to evaluating the MRD status of the MEASURE cohort. The classification (positive vs negative) was performed without knowledge of clinical or disease characteristics. AML MRD classifier calls were based on a threshold pre-specified in the statistical analysis plan, corresponding to a 99% target specificity in a non-cancer population. As hypomethylating agents (HMA) may confound the MRD signals, additional analyses were prespecified for the HMA-unexposed population. The primary endpoint was cumulative incidence of relapse (CIR) with non-relapse mortality as competing risk; overall survival (OS) was a key secondary endpoint. 

Results: The parent MEASURE protocol enrolled 304 participants; 299 (98%) had a stored plasma sample from pre-alloHCT CR available for this study, with 278 evaluable based on QC criteria.

For the analysis population (n=278), the median age was 59yrs (range 19-79yrs); 53% were male; 73% were non-Hispanic White. The median follow-up time was 18 months. The CIR was 15% at 1yr. 

AML MRD classifier identified 126 (45%) positive patients, who had higher risk of relapse (HR=3.1, 95% CI 1.7-5.5, p<0.001; 1yr CIR 25% vs 7%) and worse OS (HR=3.2, 95% CI 1.8-5.7, p<0.001; 1yr OS 79% vs 91%) compared to those testing negative. The sensitivity and specificity for detecting relapse were 74% and 60% at 1yr.

Clinical standard of care multiparameter flow cytometry (MFC) at pre-alloHCT CR was also reported for 98% of these patients (n=272); testing positive (n=31, 11%) was significantly associated with worse clinical outcomes (pos vs neg: relapse HR=2.7, 95% CI 1.4-5.3, p=0.003, 1yr CIR 33% vs 12%; OS HR=3.1, 95% CI 1.6-6.0, p<0.001, 1yr OS 69% vs 89%). The sensitivity and specificity for detecting relapse were 25% and 91% at 1yr.

Among the MFC MRD negative patients, 100 (41%) were classified as positive by AML MRD classifier; they had worse clinical outcomes compared to those classified as negative by both tests (relapse HR=2.4, 95% CI 1.3-4.6, p=0.007, 1yr CIR 20% vs 7%; OS HR=2.9, 95% CI 1.5-5.7, p=0.002, 1yr OS 84% vs 92%).

Patients treated with HMA (n=71) in this non-randomized cohort had a higher median age (70yrs, range 39-79yrs) than those not treated with HMA (n=207; 55yrs, range 19-77yrs; p<0.001); CIR at 1yr was 23% vs 13% (p=0.1). In subgroup analysis, AML MRD classifier pre-transplant results stratified patients for post-alloHCT outcomes in both HMA-treated (pos 76%; CIR p=0.049, 1yr 28% vs 6%; OS p=0.03, 1yr 72% vs 94%) and non-HMA-treated patients (pos 35%; CIR p=0.004, 1yr 22% vs 7%; OS p=0.005, 1yr 84% vs 91%).

Conclusions: A tumor-agnostic plasma cfDNA methylation assay identified, in pre-transplant samples from a prospective multicenter study, those at increased risk of relapse and death following alloHCT. These findings support further optimization of the AML MRD classifier with intended-use samples to enable blood-based methylation profiling as a broadly applicable, biopsy-free approach for AML MRD.