AUTHORS
Amanda C Winters, MD,PhD, Amanda Knarr2, Andrew T Kent, MD, PhD3, Jonathan Gutman, MD4, Mathew G. Angelos, MD, PhD5, Christine M. McMahon, MD6, Daniel Pollyea, MD7, Cailin Collins8, Maria Laura Amaya, MD PhD9, Craig Jordan, PhD10
AFFILIATIONS
1 University of Colorado, Aurora, Colorado, United States
2 University of Colorado, Aurora, Colorado, United States
3 University of Colorado, Aurora, Colorado, United States
4 University of Colorado, Aurora, Colorado, United States
5 University of Colorado, Aurora, Colorado, United States
6 University of Colorado, Aurora, Colorado, United States
7 University of Colorado, Aurora, Colorado, United States
8 University of Colorado, Aurora, Colorado, United States
9 University of Colorado, Aurora, Colorado, United States
10 University of Colorado, Aurora, Colorado, United States
ABSTRACT
Relapse is a leading cause of death in AML. Persistence of measurable residual disease (MRD) by both flow cytometry and by molecular methods (PCR, NGS) is strongly predictive of relapse in patients who initially achieve complete remission, and pre-emptive intervention at recurrence of MRD-level disease prolongs survival. Reliance on bone marrow assessments for disease monitoring limits the frequency with which patients can be evaluated for recurrent disease. Molecular MRD measurement from cell-free DNA in peripheral blood (cfDNA) is an attractive alternative. Prior literature on cfDNA assessment in MDS and AML is limited to correlation of mutational burden between bone marrow and cfDNA. The more practical question of clinical benefit, particularly predictive sensitivity and specificity of cfDNA for disease relapse, remains to be demonstrated.
We evaluated retrospective bone marrow and peripheral blood samples from AML patients in our institution’s Hematologic Malignancies Tissue Bank. We extracted cfDNA from banked peripheral blood serum. We initially compared AML mutational burden by digital PCR (dPCR) for cfDNA versus bone marrow mononuclear cells for 21 patients at diagnosis. Patients had wide variance in bone marrow blast percentage, total white blood cell count in the peripheral blood, and peripheral blood blast count. Irrespective of these differences, and similar to previous reports, cfDNA closely recapitulated variant allelic frequencies (VAF) found in the marrow for mutations such as NPM1, IDH1, IDH2, or TP53 (R2 = 0.98 for high peripheral blasts, R2 = 0.81 for low peripheral blasts). For 17 patients we also had matched dPCR MRD from marrow vs cfDNA. Contrary to prior reports, we saw higher VAF of AML mutations in cfDNA compared to bone marrow in 5/9 patients who ultimately relapsed, suggesting that cfDNA may be a more sensitive method for assessing disease burden in the MRD setting than conventional marrow sampling.
Finally, we wanted to address potential lead time to relapse with cfDNA versus traditional bone marrow sampling. For this aspect of the study, we focused on AML patients who received allogeneic stem cell transplant (SCT), as these patients had more frequent bone marrow evaluations in the post-SCT period. We identified 20 patients who were in remission prior to SCT, relapsed after SCT, and had available post-SCT/pre-relapse serum samples in our bank. Of these, 2 relapsed in CNS only and 2 relapsed >4 years out from transplant with no proximal serum samples banked. In 5 of 16 remaining patients, cfDNA MRD by dPCR was detectable prior to relapse, with a median lead time of 140 (IQR 123-179) days from first positive cfDNA to clinical relapse versus 0 (IQR 0-0) days with marrow-based standard clinical screening. In 1of these 5 patients, a new NRAS mutation at relapse was detectable in cfDNA 249 days prior. The other 11 patients had available samples at similar time points prior to relapse, yet dPCR did not detect their AML-associated mutations. As a control, we also evaluated cfDNA MRD in 18 AML patients who did not relapse after SCT despite 3-4 years’ clinical follow-up – none were cfDNA MRD positive. Therefore, in the total cohort of 34 patients (16 with relapse and 18 without), the positive predictive value of cfDNA for relapse was 100%, negative predictive value 62%, specificity was 100% and sensitivity 31%.
Limitations of our study include the retrospective nature of the samples used, including variable sample processing that could affect the stability of cfDNA and missing data points, and small cohort sizes. For the post-SCT cohort there were seldom bone marrow samples available at the same time points as cfDNA, so direct comparison between dPCR MRD between the 2 sample sources was not generally possible.
Despite these limitations, in 5 out of 16 evaluable patients with relapse the cfDNA MRD was positive well in advance of marrow relapse, which could create a window of opportunity for pre-emptive therapies to be applied. If cfDNA was detected, the patient had a 100% chance of relapsing within 3-6 months without further intervention. The finding of clonal evolution in one patient points toward the utility of implementing periodic NGS-based MRD evaluation of cfDNA to allow for mutation discovery, which is not possible with dPCR. These findings yield support for prospective studies of cfDNA as a less invasive sample source for MRD monitoring in AML patients, with clinically actionable implications.
DISCLOSURES
Amanda C Winters, MD,PhD: No relevant conflicts of interest to declare.