Prognostic value of NGS-MRD before and after allogeneic hematopoietic cell transplantation in patients with AML: A longitudinal study

Next-generation sequencing (NGS) has shown reliability in the detection of specific mutations at both time of diagnosis and complete remission (CR) in patients with acute myeloid leukemia (AML). However, NGS-based measurable residual disease (MRD) detection, has its limitations related to sensitivity and specificity, and its inability to differentiate between residual and clonal hematopoiesis. There is limited evidence on the role of NGS-MRD in patients with AML receiving allogeneic hematopoietic cell transplantation (allo-HSCT).

The AML Hub has previously reported a collection of video interviews on MRD assessment. Kim et al.¹ recently published a study in the Blood Cancer Journal, investigating the prognostic impact of NGS-MRD in patients with AML at several timepoints, and the key findings are summarized below.

Study design

A longitudinal study with two independent prospective cohorts of patients with AML who had underwent allo-HSCT at CR. Eligible patients (N = 132) were aged ≥19 years, had bone marrow (BM) DNA available both at diagnosis and in CR before allo-HSCT. Patients in Cohort 1 (n = 63) and Cohort 2 (n = 69) received transplants from matched unrelated donor (MUD) and haploidentical familial donors. In addition, patients in Cohort 2 also received transplants from matched sibling donors.

Patient samples were obtained at three time points:

• At time of diagnosis, pre-HSCT
• Post-HSCT
• At relapse

NGS analysis was performed using a NGS panel for acute leukemia. Cumulative incidence was used to estimate the probability of cumulative incidence of relapse (CIR) and nonrelapse mortality.

Baseline characteristics

• A total of 132 pre-HSCT and 114 post-HSCT at 1-month (post-HSCT-1m) patients underwent NGS-MRD. Persistent mutations were detected in 43% and 20% of pre-HSCT and post-HSCT-1m patients, respectively, with persistent pre-HSCT mutations being more frequent in Cohort 2 (Table1).
• No significant differences in the pretransplant characteristics were observed in both cohorts, except patients in Cohort 1 had a higher incidence of second complete remission, and Cohort 2 included more elderly patients (≥60 years) than Cohort 1 (26% vs 11%, respectively) (Table1).

Table 1. Comparison of characteristics between Cohort 1 and Cohort 2*

Results

Somatic mutation landscape

• The genetic landscape of 132 patients demonstrated 389 somatic mutations in 47 genes, with a median of three mutations (interquartile range [IQR]: 2−4) per patient.
• CEBPA was the most common somatic mutation followed by DNMT3A, NPM1, and NRAS with a median variant allele fraction (VAF) of 34.39% (IQR: 11−46) in all patients. The most common mutations post-HSCT-1m included DNMT3A and TET2.
• Median VAF was 2.69% (IQR, 0−16) and 0.19% (IQR, 0−1) in pre-HSCT and post-HSCT-1m samples, respectively, showing a significant reduction in VAF from diagnosis, pre-HSCT to post-HSCT-1m.
• Allo-HSCT demonstrated a significant impact on the remaining pre-HSCT mutations, clearing an additional 84% of DNMT3A, 71% of TET2, and 100% of ASXL1.

Prognostic value of NGS-MRD detection

• At a median follow-up duration of 33 months for survivors, the overall CIR was 18.7% (95% CI, 12−26).
• Patients with persistent mutations had a significantly greater risk of relapse compared with those without persistent mutations: pre-HSCT, 34.8% vs 6.7%; p <0.001, and post-HSCT-1m, 43.5% vs 12.3%; p <0.001, respectively.
• The optimal VAF threshold to detect the most effective positive and negative predictive values was 0%, and therefore, NGS-MRD positivity (defined by a failure of complete clearance of mutations) was independently associated with increased CIR and inferior survival (Table 2).

Table 2. Multivariate analysis for survival outcomes*

Prognostic impact of persistent DTA or CHIP mutations

• A significantly higher CIR was observed in patients with measurable DTA mutations at 31% (95% CI, 16−48; p = 0.031) for pre-HSCT and 42% (95% CI, 14−68; p = 0.029) for post-HSCT-1m, compared with those without DTA mutations.
• Similarly, patients with measurable CHIP mutations showed a higher CIR at 29% (95% CI, 16−44; p = 0.009) for pre-HSCT and 42% (95% CI, 14−68; p = 0.003) for post-HSCT-1m, compared with those without CHIP mutations.
• Measurable DTA or CHIP mutations and higher CIR at post-HSCT-1m translated into inferior survival.

Conditioning intensity and prognostic value of NGS-MRD detection

• In patients with myeloablative conditioning (MAC), pre-HSCT NGS-MRD detection was significantly associated with posttransplant relapse with a CIR of 45% (95% CI, 21−67; p < 0.001). However, post-HSCT-1m NGS-MRD status showed no difference in relapse.
• On the contrary, in patients receiving reduced intensity conditioning (RIC), post-HSCT-1m NGS-MRD detection was significantly associated with posttransplant relapse with a CIR of 53% (95% CI, 27−74; p <0.001), while showed no difference in the pre-HSCT NGS-MRD status.
• No differences were observed in nonrelapse mortality according to NGS-MRD status at each time point in the RIC group.

Clonal dynamics of mutations

• Twenty-three among the 38 NGS-MRD-positive patients at pre-HSCT, were converted to NGS-MRD-negative at post-HSCT-1m. Also, 10 of the 12 patients with persistent mutations at post-HSCT-1m became NGS-MRD-negative.
• Longitudinal NGS-MRD monitoring revealed that many residual clones at both pre-HSCT and post-HSCT-1m in patients who did not relapse, disappeared after allo-HSCT. NGS-MRD monitoring also revealed the reappearance of clonal mutations 2 or 3 months before relapse.

Conclusion

The study showed that the prognostic impact of detectable mutations at each pre-specified time point was correlated to the conditioning intensity. Persistent mutations were associated with higher risks of relapse and mortality both at pre-HSCT and post-HSCT-1m, and there was an advantage of serial NGS-MRD monitoring after allo-HSCT. The practicality of NGS-MRD monitoring will enable future studies investigating the feasibility of MRD-based approaches to reduce relapse in patients with AML.