Outcomes and Patterns of Relapse of NPM1 mutated acute myeloid leukemia treated with venetoclax based therapies

Graphical Abstract

To the editor:

Mutations in nucleophosmin 1 (NPM1-mut) are detected in approximately 30% of newly diagnosed (ND) acute myeloid leukemia (AML), and are considered founder events due to their persistence at relapse.¹ While treatment with both intensive chemotherapy (IC) and low-intensity therapy (LIT) are associated with favorable responses in NPM1-mutated AML, relapses are common, occurring in approximately 50% of IC-treated, and more frequently with LIT-treated patients.^1,2

Recent data suggests that NPM1 wildtype (NPM1-wt) relapses occur in up to 5–10% of NPM1-mut AML treated with 7+3.^3–5 This, coupled with the observation that NPM1-mut is a defining event in leukemogenesis, which rarely occurs in isolation,⁶ has raised the question of whether the NPM1-wt relapses represent clonal evolution of a common pre-leukemic stem cell, or a de novo AML arising from a pre-existing, selected clone.

We sought to determine the outcomes and patterns of relapse among patients with NPM1-mut AML treated with venetoclax-based therapies and investigate whether these were impacted by treatment intensity.

ND patients with NPM1-mut AML treated in one of the following clinical trials were included: fludarabine/cladribine, cytarabine and idarubicin with venetoclax (FLAG-IDA+VEN / CLIA+VEN [IC+VEN], NCT03214562^7,8 and NCT02115295⁹), cladribine, low dose cytarabine with venetoclax (CLAD+LDAC+VEN, NCT03586609¹⁰), and hypomethylating agents with venetoclax (HMA+VEN, NCT03404193¹¹ and NCT02203773¹²). NCT02115295 initially allowed concurrent FLT3 inhibition, but this was subsequently stopped due to myelosuppression. All other studies did not allow FLT3 inhibitors for patients with FLT3 mutated AML.

All patients had bone marrow samples sent for morphology, flow cytometry, cytogenetics and next-generation sequencing (NGS) with either a 28- or 81-gene panel at diagnosis and relapse. Measurable residual disease (MRD) was assessed via flow, sensitivity 10⁻⁴.

Response was assessed using the European LeukemiaNet (ELN) 2022 criteria. Overall survival (OS) was defined as the time from treatment initiation to death from any cause. Non-response, relapse, and deaths were considered events for event-free survival (EFS). Landmark analysis used the median time to allogeneic stem cell transplant (SCT) as the landmark. Cumulative incidence of relapse (CIR) was measured in responders from time of best response to relapse with death as a competing risk and compared with the Fine-Gray test. Statistical analysis was done with R version 4.4.2 (R Foundation, Vienna, Austria).

The trials enrolled 516 patients (192 IC+VEN, 190 CLAD+LDAC+VEN, 134 HMA+VEN), of whom 103 (20%) had NPM1-mut (22% of IC+VEN, 18% of CLAD+LDAC+VEN, 19% of HMA+VEN). Baseline and treatment characteristics of the NPM1-mut cohort are shown in Table 1 and depicted in Figure S1. Patients treated with IC+VEN were younger (median 48 years [range, 20–67]) than those treated with CLAD+LDAC+VEN (68 years) or HMA+VEN (72 years). There were no differences between ELN risk stratification and karyotypes between the treatment cohorts. Cytogenetics were diploid in 87%; only 1 patient had a complex karyotype. Rates of co-mutations were similar between treatment groups.

Table 1:

Baseline & Treatment Characteristics of NPM1-mut AML

	Overall (N=103)	IC+VEN (N=42)	CLAD+LDAC+VEN (N=35)	HMA+VEN (N=26)	P-value
Age
Median [Min, Max]	66 [20, 89]	48 [20, 67]	68 [61, 78]	72 [66, 89]	<0.01
≥60	68 (66)	7 (17)	35 (100)	26 (100)	<0.01
ELN 2022 Risk
Favorable	85 (83)	31 (74)	30 (86)	24 (92)	0.58
Intermediate	14 (14)	9 (21)	4 (11)	1 (4)
Adverse	4 (4)	2 (5)	1 (3)	1 (4)
Cytogenetics
Diploid	90 (87)	36 (86)	34 (97)	20 (77)	0.34
Other Intermediate	11 (11)	6 (14)	1 (3)	4 (15)
Complex	1 (1)	0 (0)	0 (0)	1 (4)
Missing	1 (1)	0 (0)	0 (0)	1 (4)
Mutations
NPM1	103 (100)	42 (100)	35 (100)	26 (100)	N/A
ASXL1	3 (3)	0 (0)	2 (6)	1 (4)	0.51
CEBPAbZIP	2 (2)	0 (0)	2 (6)	0 (0)	0.26
DNMT3A	45 (44)	22 (52)	15 (43)	8 (31)	0.38
FLT3-TKD	20 (19)	11 (26)	8 (23)	1 (4)	0.14
FLT3-ITD	16 (16)	11 (26)	4 (11)	1 (4)	0.08
IDH1	13 (13)	4 (10)	4 (11)	5 (19)	0.70
IDH2	25 (24)	9 (21)	9 (26)	7 (27)	0.96
KRAS	6 (6)	4 (10)	1 (3)	1 (4)	0.62
NRAS	24 (23)	11 (26)	7 (20)	6 (23)	0.94
PTPN11	14 (14)	6 (14)	7 (20)	1 (4)	0.34
RUNX1	1 (1)	1 (2)	0 (0)	0 (0)	0.69
SF3B1	6 (6)	2 (5)	2 (6)	2 (8)	0.86
SRSF2	17 (17)	4 (10)	7 (20)	6 (23)	0.21
TET2	19 (18)	6 (14)	7 (20)	6 (23)	0.82
TP53	2 (2)	0 (0)	1 (3)	1 (4)	0.69
WT1	5 (5)	2 (5)	1 (3)	2 (8)	0.86
Best Response
CR	92 (89)	39 (93)	32 (91)	21 (81)	0.69
CRi	7 (7)	2 (5)	2 (6)	3 (12)
MLFS	3 (3)	1 (2)	0 (0)	2 (8)
Died	1 (1)	0 (0)	1 (3)	0 (0)
Relapsed	21 (20)	3 (7)	5 (14)	13 (50)	<0.01
SCT	47 (46)	27 (64)	18 (51)	2 (8)	<0.01
FLT3 inhibitor	5 (5)	5 (12)	0 (0)	0 (0)	0.05

Note: Bold values indicate statistical significance p < 0.05.

FLT3-ITD was present in 26% of NPM1-mut patients treated with IC+VEN, 11% of CLAD+LDAC+VEN and 4% of HMA+VEN, with a median allelic ratio of 0.09 (range, <0.01–0.65). Five patients (median allele ratio 0.4 [range, 0.24–0.65]) in the IC+VEN group received gilteritinib. FLT3-TKD was detected in 19% of the cohort.

SCT in first complete remission (CR1) was performed in 64% of IC+VEN, 51% of CLAD+LDAC+VEN and 8% of HMA+VEN. Indications for SCT included persistent MRD positivity and co-occuring mutations or cytogenetic abnormalities, at the discretion of the treating physician.

The overall response rate (ORR) was 99%, with a composite complete remission (CRc) rate of 96%, and MRD-negative CRc in 84%. There was no significant difference in response rates between treatments (Table S1).

At a median follow-up of 49 months (95% confidence interval [CI], 44–58), the three-year OS was 78% (95%CI, 65–93) for IC+VEN, 78% (95%CI, 64–94) for CLAD+LDAC+VEN and 58% (95%CI, 42–80) for HMA+VEN (Figure S2). OS was not significantly different between IC+VEN and CLAD+LDAC+VEN (p=0.98), but both were better than HMA+VEN. Similar results were observed for EFS (Figure S3).

Landmark analysis showed no benefit for SCT in CR1 among patients treated with IC+VEN (Figure S4). However, if treated with CLAD+LDAC+VEN, 3-year OS was 94% (95%CI, 84–100) with SCT versus 69% without (95% CI, 48–99, p=0.06, Figure S5). Landmark analysis was not performed for HMA+VEN as only 2 patients underwent SCT.

The 3-year CIR of NPM1-mut AML was 9% (95%CI, 2–22), 10% (95%CI, 3–24), and 35% (95%CI, 17–53) when treated with IC+VEN, CLAD+LDAC+VEN and HMA+VEN, respectively (p<0.01, Figure S6). Of the 21 relapses, paired diagnosis and relapse samples were available for 20 patients. Nine (45%) relapses were NPM1-wt. Two (10%) of the nine relapses occurred after SCT.

Although the CIR for NPM1-mut AML was low for IC+VEN and CLAD+LDAC+VEN, we observed NPM1 wildtype (NPM1-wt) relapses in 2/3 (66%) of patients treated with IC+VEN and 3/5 (60%) of patients treated with CLAD+LDAC+VEN. With HMA+VEN, 4/12 (30%) relapses were NPM1-wt. NPM1-wt relapses appeared to be characterised by the persistence or emergence of myelodysplasia-related mutations (Figure 1, 78% of NPM1-wt relapses), especially splicing mutations. The median time to relapse was 33 months (95%CI, 9-not estimable [NE]) for NPM1-wt relapses, compared with 9 months (95% CI, 9-NE) for NPM1-mut relapses (p=0.05, Figure S7). NPM1-wt relapses were immunophenotypically distinct from diagnostic samples in 8/8 cases with paired immunophenotypic samples (Table S2). One patient had an isolated extramedullary relapse; flow cytometry was not performed at relapse.

Figure 1.

Patterns of Relapse in Patients who were NPM1-mut at diagnosis. Heatmap showing cytomolecular and treatment characteristics of the 20 patients with available paired diagnosis and relapse samples. NPM1-wildtype relapses (cleared mutations in yellow) appear to be more common in patients treated with venetoclax-based therapies, especially as treatment intensity increases. They appear to be characterized by persistent or emergent myelodysplasia-related mutations. Three patients had 28-gene panels at diagnosis but 81-gene panels at relapse. Genes mutated at relapse which were not represented on the 28-gene panel are annotated as Relapse (Dx Unknown). MDS: myelodysplastic syndrome; TumSup: tumor suppressor genes; FLT3i: FLT3 inhibitor; SCT: stem cell transplant; CR1: first complete remission; EM: extramedullary; Dx: diagnosis.

The pattern of relapses among patients with NPM1-mut AML appeared to be influenced by the type of chemotherapy backbone in combination with venetoclax. Mutations in signaling genes (FLT3, NRAS, KRAS, 42% of relapses) were overrepresented among patients who relapsed after HMA+VEN, and were notably absent among patients treated with IC+VEN or CLAD+LDAC+VEN. Persistence/emergence of tumor-suppressor gene mutations (WT1 and TP53, 30% of relapses) and clonal hematopoiesis genes involved in methylation (DNMT3A, TET2, IDH1, IDH2, 80% of relapses) was seen across the 3 therapies.

Of the 16 patients with NPM1-mut and FLT3-ITD at diagnosis, only four (25%) relapsed, of whom only one HMA+VEN-treated patient had recurrent FLT3-ITD at relapse. Of interest, no patients treated with IC+VEN or CLAD+LDAC+VEN acquired FLT3-ITD at relapse. Two patients treated with HMA+VEN had a new, emergent FLT3-ITD at relapse. Only one (5%) patient with baseline FLT3-TKD (CLAD+LDAC+VEN-treated) relapsed, without FLT3-TKD.

NPM1-mut is an AML-defining founder mutation, present at diagnosis and in most cases at relapse when treated with conventional therapies.^1–3 The introduction of venetoclax has improved outcomes in AML, but its effect on longitudinal NPM1-mut dynamics is less well understood. We present a cohort of 103 NPM1-mut AML patients treated with frontline venetoclax-based therapies of varying intensity. Outcomes of NPM1-mut AML treated with IC+VEN or CLAD+LDAC+VEN were excellent, with a 3-year OS of 78% and 3-year CIR of approximately 10%. These results compare favorably to 3-year OS of 60–70% and CIR of 30% for intensive regimens without venetoclax.^13,14 Relapses were more common with HMA+VEN, despite similar response rates, possibly due to treatment intensity and fewer number of patients consolidated with SCT.

Although the number of relapses was small, we observed a larger than expected incidence of late NPM1-wt relapses, especially in the cytarabine-based venetoclax combinations. With conventional 3+7, NPM1-wt relapses comprise only 5–10% of all relapses.⁵ Myelodysplasia-related mutations were observed in 78% of NPM1-wt relapses. This, coupled with the longer (approximately 3 year) time to relapse and immunophenotypic shift suggests effective, durable control of the original NPM1-mut clone, followed by an outgrowth and “relapse” of a separate, NPM1-wt clone, likely arising from a pre-existing preleukemic population. This is in keeping with earlier work demonstrating that NPM1-wt relapses lose homeobox (HOX) expression signatures,⁵ and is consistent with data from the CAVEAT study, where 2/5 relapses were NPM1-wt.¹⁵

Our findings are limited by the low incidence of relapse in this large patient cohort. Nevertheless, the observations are hypothesis-generating, suggesting that a proportion of NPM1-mut AML may be cured by venetoclax-based therapies, and that NPM1-wt relapses may represent clonal evolution of a pre-existing clone, or a de novo secondary AML. Single-cell sequencing of paired samples is required to map the dynamics of the leukemic clone(s) and confirm our clinical observations. In addition, the occurrence of NPM1-wt relapses with venetoclax-based therapies may have considerations for mutation-agnostic MRD assessments in addition to NPM1-specific MRD monitoring, the role of NPM1-specific salvage therapy with menin inhibitors, and indications for SCT in CR1.

Data sharing:

The study data is not publicly available to respect participant confidentiality. Requests for sharing of deidentified data should be directed to the corresponding author.