Leukemic Phenotype Determines Prognostic Relevance of Limit of Quantification for Flow Cytometry-Based Measurable Residual Disease Assessment in Acute Myeloid Leukemia

Background

The limit of quantification (LOQ) is an objective low-level cutoff for measurable residual disease (MRD) assessment with multi-parametric flow cytometry (MFC) in patients with acute myeloid leukemia (AML). It is a fixed cutoff defined as 50 leukemia-associated immunophenotype (LAIP) positive events in samples with at least 500.000 CD45-expressing white blood cells (WBC). It was introduced as an addition to the current clinically validated MRD threshold (European LeukemiaNet advised <0.1% LAIP of total WBC) to reduce false negative results. Its application showed significantly worse overall survival (OS) for LOQ-positive patients compared to MRD-negative patients in retrospective analyses of the GIMEMA AML-1310 and HOVON132 trials, suggesting that MRD below 0.1% might hold prognostic value. However, clinical implementation is challenging as the survival differences were small, and the number of low-level MRD-positive patients was high. As the LOQ is a low-level threshold, it is highly dependent on the background levels of the assay, which vary between different LAIPs. Therefore, we hypothesized that the prognostic relevance of LOQ varies across AML phenotypes. With more than 100 different LAIPs, however, this poses a challenge for LAIP-based analysis, requiring a large amount of detailed data. Hence, we designed a novel approach that allowed us to combine three datasets of different MFC-MRD panels and gating strategies. This rendered the unique possibility of studying the prognostic relevance of LOQ on a LAIP-based level.

Methods

We included AML patients in complete morphologic remission after two cycles of intensive chemotherapy from the AML-1310 trial (N=153), the pooled HOVON-AML trials (H42A, H92, H102, H103 & H132, N=873), and the AMLSG in Brazil (N=12). To perform a combined analysis per phenotype, we gathered detailed MFC-data on LAIP expression and event numbers for cases with more than 500.000 acquired WBC. We harmonized LAIPs measured by different institutions by classifying them based on expression type and background levels as asynchronous/cross-lineage, under/over-expression, and mature. We classified cases as MRDpos (≥0.1%), MRDnegLOQpos (<0.1% and ≥ 50 LAIP events), and LOQneg (< 50 LAIP events) and assessed OS and relapse-free survival (RFS) using Kaplan-Meier method and log-rank test and cox-regression.

Results

A total of 1038 patients were classified into LOQneg (N=292, 28%), LOQposMRDneg (N=555, 54%), and MRDpos (N=191, 18%), with significant differences in 5-year OS (64%; 60%; 46%; p < 0.001) and 3-year RFS (60%; 55%; 41%; p < 0.001). Direct comparison of MRD negative subgroups showed a non-significant trend towards OS (hazard ratio (HR) (95%CI): 1.23 (0.98-1.55); p = 0.07) and RFS benefit (HR (95% CI): 1.18 (0.96-1.45); p = 0.1) for LOQneg compared to LOQposMRDneg.

LAIPs were categorized as asynchronous/cross-lineage (N=689, 66%), under/over-expression (N=282, 27%) and mature (N=40, 4%); only 27 cases (3%) could not be classified. We found notable survival differences for asynchronous/cross-lineage LAIPs for OS (LOQneg 68% vs LOQposMRDneg 58%; p = 0.03) and RFS (LOQneg 60% vs LOQposMRDneg 53%; p = 0.1), while we found no differences in under/over-expression and mature LAIPs. Except for CD7 LAIPs, we observed no differences in survival in LAIP-specific analysis. For CD7 LAIP-positive cases (N=276), we found a non-significant difference in OS (LOQneg 74% vs 58% LOQposMRDneg; p = 0.2) and RFS (LOQneg 64% vs LOQposMRDneg 52%; p = 0.1), with similar relapse incidence for LOQposMRDneg and MRDpos patients.

Sensitivity analyses that corrected for patients with possible MRD-guided consolidation treatment in the AML1310 and H132 trials showed similar trends. When defining MRD positivity ≥0.035% for patients from the AML1310 trial, the clinical cutoff used in this cohort, the results also remained similar.

Conclusion

This study is the first to combine detailed MFC-MRD data of a large cohort of patients measured in different platforms. Using our unique approach, our analyses revealed a prognostic relevance of LOQ in MRDneg patients only for asynchronous/cross-lineage LAIPs, particularly for CD7-based LAIPs. This approach paves the way for the extension of these data with other platforms, which is warranted to substantiate our findings.