Complemental Prognostic Impact of Flow Cytometric and Molecular Quantification of Minimal/Measurable Residual Disease in Acute Myeloid Leukemia

Background: Minimal/measurable residual disease (MRD) correlates with outcome in acute myeloid leukemia (AML). Flow cytometry (FC) and quantitative PCR (QPCR) both are recommended by ELN (Schuurhuis et al., Blood 2018) and NCCN guidelines. Comparative MRD data on FC and QPCR are limited.

Aim: To assess the relative contribution of FC and QPCR in generating prognostically relevant MRD data in AML.

Methods: 936 bone marrow samples of 305 AML patients in cytomorphologic CR were analyzed for MRD by both FC and QPCR (mainly fusion transcripts and NPM1 mutations) in parallel applying ELN guidelines. Patients´ages at diagnosis ranged from 20 to 89 years (median 57), sex ratio (f/m) was 155/150. As anticipated due to selection for MRD assessment risk according to ELN was favorable in 191 (63%), intermediate in 92 (30%) and poor in 18 (6%). Accordingly, 5 year event-free survival (EFS) was 68% and 5 year overall survival (OS) 68% at 2.4 years median follow-up. Samples for MRD assessment were drawn between 7 days and 8.5 years after diagnosis (median 5.8 months). MRD samples were categorized into 5 intervals: I1, <2 months after diagnosis, 189 cases (20%); I2, 2 to 6 months, 286 (31%); I3, 6 to 12 months, 167 (18%); I4, 1 to 2 years, 146 (16%); I5, >2 years, 148 (16%).

Results: First, we focused on the percentages of MRD as determined by FC (%FC). %FC ranged from 0% to 5% (median, 0.003%). Overall, %FC strongly correlated with EFS (p<0.0001). Application of a cutoff of 0.1% %FC, as recommended by ELN, resulted in significant differences in 5 year EFS (76% vs. 49%, p<0.0001). Considering even lower %FC, also a cutoff of 0.01% had significant impact on 5 year EFS (79% vs. 63%, p<0.0001). Multivariate analysis considering both cutoff values revealed independent impact for both (0.1%, p=0.005; 0.01%, p<0.0001). Adding to the multivariate model the two relevant prognostic parameters determined at diagnosis, i.e. age and ELN poor risk, resulted in both cutoff values for %FC still being independently significant (0.1%, p=0.013; 0.01%, p<0.0001).

Secondly, MRD levels determined by QPCR were evaluated. The ratios between target gene expression and ABL1 expression were determined for diagnosis and follow-up and the further divided by the latter resulting in a ratio QR for each MRD assessment. QR ranged from 0 to 8 (median, 0). QR was significantly related to EFS (p=0.015). Application of a cutoff of 0.01 QR (100-fold decrease) resulted in significant differences in 5 year EFS (79% vs. 42%, p<0.0001). Considering lower QR, also a cutoff of 0.001 (1000-fold decrease) had significant impact on 5 year EFS (81% vs. 50%, p<0.0001). Multivariate analysis considering both cutoff values revealed independent impact for both (0.01, p=0.022; 0.001, p<0.0001). Adding to the multivariate model age and ELN poor risk resulted in both cutoff values for QR still being independently significant (0.01, p=0.038; 0.001, p=0.002).

Multivariate analysis for EFS considering %FC 0.1% and QR 0.01 confirmed independent significance for both methods (p<0.0001 each). Same was true for %FC 0.01% and QR 0.001 (p<0.0001 each). Multivariate analysis considering all four parameters (%FC 0.1%, %FC 0.01%, QR 0.01, QR 0.001) as well as age and ELN poor risk revealed independent significance for %FC 0.01% (p<0.0001) and QR 0.0001 (p=0.008).

Subgroup analysis revealed impact on EFS for %FC at I1, I2 and I3 and for QR at I2 to I5. In detail, independent impact was demonstrated at I1 for %FC 0.01% (p=0.018), at I2 for %FC 0.1% (p=0.002) and at I3 for %FC 0.01% (p=0.016) while independent impact was demonstrated at I2 for QR 0.001 (p=0.014), at I3 for QR 0.001 (p=0.001), at I4 for QR 0.001 (p=0.006) and at I5 for QR 0.001 (p=0.002).

MRD levels determined by both methods were also significantly related to OS. Application of %FC 0.01 resulted in significant differences in 5-year OS (87% vs. 76%, p<0.0001) as did application of QR 0.001 (89% vs. 68%, p<0.0001). Multivariate analysis for OS considering %FC 0.01% and QR 0.001 revealed independent impact for both (p=0.002 and p<0.0001, respectively). Adding age and ELN poor risk to the multivariate model still revealed independent impact for both %FC 0.01% (p=0.001) and QR 0.001 (p<0.0001) on OS.

Conclusions: Quantification of MRD by both FC and QPCR reveals independent and prognostically highly relevant information. FC is more suitable early during course of disease while QPCR is so later. MRD by FC may be considered even at lower levels than recommended by ELN.