Clinical Significance of FLT3‐ITD Variation Allele Frequency in Acute Myeloid Leukemia: A Multicenter Study

ABSTRACT

The specific prognostic impact of FLT3‐ITD variant allele frequency (VAF) in acute myeloid leukemia (AML) remains unclear. In this multi‐center study of 190 AML patients harboring FLT3‐ITD mutations, a VAF cut‐off of 40% was applied. Patients with high VAF (≥ 40%) exhibited significantly higher white blood cell counts, lactate dehydrogenase levels, and bone marrow blast percentages, alongside markedly shorter median overall survival (OS). Although high VAF was independently associated with inferior outcomes in the overall multivariate Cox regression analysis, its effect was profoundly amplified in the presence of co‐mutations in NPM1 or DNMT3A. This suggests that high VAF may function not as an absolute, independent determinant, but primarily as a potent “risk amplifier.” Notably, allogeneic hematopoietic stem cell transplantation (allo‐HSCT) effectively mitigated this survival disadvantage. Furthermore, we characterized a non‐linear relationship between VAF and mortality risk, with a sharp increase observed once the VAF burden exceeded 34%. In conclusion, a quantitative VAF ≥ 40% should be considered a context‐dependent prognostic marker for risk stratification. Allo‐HSCT represents a critical intervention capable of overcoming the exceedingly poor prognosis associated with high VAF.

1. Introduction

Internal tandem duplication (ITD) mutations of the FMS‐like tyrosine kinase 3 (FLT3) gene are the most common genetic mutations in acute myeloid leukemia, occurring in approximately 15%–30% of AML patients [1, 2, 3]. FLT3‐ITD is closely associated with a high relapse rate and poor prognosis in AML [4, 5, 6, 7, 8].

The FLT3‐ITD high allelic ratio (AR) is significantly associated with the prognosis of AML patients [9, 10]. Multiple studies have shown that in both adult and pediatric AML patients, high AR is significantly associated with higher tumor burden, chemotherapy resistance, and poorer overall survival [11, 12, 13]. The European LeukemiaNet 2022 guidelines have included FLT3‐ITD AR in the risk stratification system, emphasizing its importance in AML prognosis assessment [11]. Although AR is closely related to VAF, there are certain differences between them in clinical application [14]. At present, the clinical significance of VAF for common mutations in myeloid tumors such as TP53, DNMT3A, and NPM1 has become relatively clear [15, 16, 17, 18, 19]. However, the prognostic value of FLT3‐ITD VAF still needs to be further validated in large‐scale clinical cohorts.

Therefore, we aim to explore the relationship between FLT3‐ITD VAF and survival outcomes in AML patients through a multicenter retrospective study, and to assess whether HSCT can overcome the poor prognosis caused by high VAF.

2. Materials and Methods

2.1. Study Design and Participants

This retrospective study included 1017 newly diagnosed non‐M3 acute myeloid leukemia patients who presented at the First Affiliated Hospital of Soochow University, The Third Affiliated Hospital of Soochow University, and Yancheng First People's Hospital between 2016 and 2025. The study strictly adhered to the principles of the Declaration of Helsinki and was approved by the ethics committees of all three centers. Given the retrospective nature of the research, informed consent was waived following ethics committee review. After screening, we excluded 50 patients aged < 18 years, 17 patients with non‐FLT3 mutations/non‐classical FLT3 mutations/FLT3 double mutations, 6 patients with missing FLT3‐ITD VAF data, and 15 patients lost to follow‐up. Ultimately, 190 patients with FLT3‐ITD mutations (18.7% of the initial cohort) and 75 patients with FLT3‐TKD mutations (7.4% of the initial cohort) were included. The detailed screening process is illustrated in Figure 1A.

FIGURE 1.

Cohort construction, mutation profiles, and survival analysis in patients with FLT3‐mutated acute myeloid leukemia. (A) Flowchart illustrating patient selection criteria and study design. (B) Oncoplot comparing somatic co‐mutation profiles (e.g., NPM1, DNMT3A, IDH1/2) between FLT3‐ITD and FLT3‐TKD cohorts. Clinical characteristics are annotated at the top. (C) Oncoplot displaying mutation profiles across distinct FLT3‐ITD molecular subgroups. (D) Kaplan‐Meier curve comparing overall survival (OS) between FLT3‐ITD and FLT3‐TKD groups (p = 0.0079).

2.2. Data Collection and Definition

Baseline clinical characteristics and laboratory data were collected through a systematic review of electronic medical records. Data dimensions included demographic information (age, gender), disease‐specific indicators (peripheral blood white blood cell count, hemoglobin concentration, platelet count, and serum lactate dehydrogenase level at initial diagnosis), and the status of allogeneic hematopoietic stem cell transplantation during the first complete remission. The primary endpoint of this study was overall survival, defined as the time from the date of pathological diagnosis to death from any cause. For patients alive at the time of the last follow‐up, data were censored at the date of last confirmed survival.

2.3. Molecular Analysis

All analyses were based on newly diagnosed bone marrow samples. Genomic DNA was extracted using the QIAamp DNA Blood Mini Kit and sequenced using an NGS panel covering 100 hematologic malignancy‐related genes, including FLT3, NPM1, DNMT3A, IDH1, IDH2, and TP53. Sequencing was performed on Illumina or BGI Genomics platforms, with quality control standards of: total DNA ≥ 100 ng, purity OD260/280 = 1.6–1.8; sequencing data required an average depth ≥ 500×, alignment rate ≥ 90%, Q30 ≥ 85%, and coverage ≥ 90%. Bioinformatics analysis was based on the hg19 reference genome and annotated using databases such as dbSNP, COSMIC, ClinVar, ExAC, and 1000 Genomes. The FLT3‐ITD VAF calculation formula is: mutant reads/(wild‐type reads + mutant reads) × 100% [9].

2.4. Statistical Analysis

Categorical variables were described as frequencies and percentages. Intergroup comparisons were performed using the chi‐square test or Fisher's exact test. Continuous variables were expressed as medians and ranges, with intergroup comparisons conducted using the Mann‐Whitney U test or Kruskal‐Wallis test. Survival analysis was performed using the Kaplan‐Meier method, and differences in survival between groups were assessed using the log‐rank test. Cox proportional hazards regression models were constructed to evaluate the independent prognostic value of FLT3‐ITD VAF. Univariate analysis was conducted, followed by the inclusion of variables with p < 0.1 or clear clinical significance into a multivariate Cox model for adjustment. Restricted cubic splines were employed to model the relationship between continuous VAF and hazard ratios, exploring the nonlinear association between VAF and mortality risk. All statistical analyses were performed using R software (version 4.4.0). Bilateral p values < 0.05 were considered statistically significant.

3. Results

3.1. Clinical Characteristics and Prognosis in FLT3‐Positive Acute Myeloid Leukemia

Our study included 265 patients with FLT3‐mutated AML. We found FLT3‐ITD mutations in 190 patients (71.7%). The remaining 75 patients (28.3%) had FLT3‐TKD mutations. The median age of the entire group was 49.5 years, ranging from 18 to 89. A normal karyotype appeared in 68% of the patients. Additionally, medical teams performed a hematopoietic stem cell transplant (HSCT) on 126 patients (47.5%). Our genomic analysis revealed several co‐mutations. NPM1 was the most common co‐mutation at 34%. DNMT3A followed at 26%, and TET2 at 14% (Figure 1B, Supporting Information S1: Table S1). NPM1 and DNMT3A remained the top co‐mutations in both the ITD and TKD subgroups. However, their exact frequencies varied slightly between the two groups (Figure 1C). We then compared the two patient subgroups. Patients with FLT3‐ITD mutations were significantly older. Their median age was 52.5 years, compared to 44.0 years for the TKD group (p = 0.002). The FLT3‐ITD group also presented with higher white blood cell counts. Their median count was 43.6 × 10⁹/L, while the TKD group was 23.9 × 10⁹/L (p = 0.044). The genetic profiles differed as well. The FLT3‐ITD cohort had a higher rate of normal karyotypes (72% vs. 56%, p = 0.017). These patients also showed more co‐occurring NPM1 and CEBPA mutations. Conversely, the FLT3‐TKD group had different associations. These patients more frequently exhibited KIT and ASXL1 mutations. They also had more core‐binding factor abnormalities, such as t(8;21) and inv(16) (Supporting Information S1: Table S2). Finally, our survival analysis revealed worse outcomes for the FLT3‐ITD group. These patients experienced a significantly shorter median overall survival (OS) of 630 days. In contrast, the FLT3‐TKD group had not yet reached its median OS (p = 0.0079) (Figure 1D).

3.2. Prognostic Value of the 40% Variant Allele Frequency Threshold in FLT3‐ITD Positive AML

Based on a cutoff value of 40% variant allele frequency (VAF) [19], the study cohort was divided into two distinct subgroups: the low VAF group (VAF < 40%) and the high VAF group (VAF ≥ 40%). Patients with a high VAF (40% or more) had more aggressive disease. They showed significantly higher white blood cell counts, bone marrow blasts, and lactate dehydrogenase levels (Table 1, Supporting Information S1: Figure S1A). In fact, the VAF level directly correlated with these three disease markers (Supporting Information S1: Figure S1B). However, the different VAF groups showed no significant differences in age, gender, coexisting mutations, or risk stratification. Crucially, patients with a high VAF faced a much shorter median overall survival. They survived for 375 days compared to 2024 days for the lower VAF group (Figure 2A). In the 99 patients with isolated FLT3‐ITD mutation (excluding NPM1 and DNMT3A mutations), the difference in overall survival between the high‐VAF and low‐VAF groups did not reach statistical significance (p = 0.1, Supporting Information S1: Figure S2); nevertheless, the survival curve of the VAF ≥ 40% group remained consistently below that of the low‐VAF group. We also looked at the impact of stem cell transplants. Without a transplant, the high VAF group survived significantly shorter than the low VAF group. Yet, receiving a transplant erased this survival gap entirely (Figure 2C–D). Finally, we divided the patients into four VAF tiers. These tiers were under 10%, 10%–20%, 20%–40%, and 40% or higher. The highest VAF group was the most common. As expected, this group also had the worst overall survival among all four tiers (Figure 2B).

TABLE 1.

Characteristics of patients with de novo AML with FLT3‐ITDmut (N = 190).

Variable	VAF < 40% N = 102	VAF ≥ 40% N = 88	p value
Age (years)			0.56
Median (min, max)	50.5 (18.0, 89.0)	54.5 (18.0, 83.0)
Gender			0.39
Female	39 (38%)	40 (45%)
Male	63 (62%)	48 (55%)
White blood cell count (× 109/L)			< 0.001
Median (min, max)	25.6 (0.8, 285.9)	71.7 (0.8, 449.6)
Hemoglobin (g/L)			0.211
Median (min, max)	78.5 (45.0, 136.0)	86.5 (29.5, 156.0)
Platelet (× 109/L)			0.547
Median (min, max)	51.0 (28.0, 89.0)	53.5 (9.0, 498.0)
Lactate dehydrogenase (U/L)			0.003
Median (min, max)	422.5 (101.0, 2232.0)	589.5 (122.0, 3908.0)
Bone marrow blast percentage (%)			< 0.001
Median (min, max)	63.0 (15.0, 94.0)	77.0 (20.0, 97.0)
Cytogenetic category, n (%)
Normal karyotype	69 (68%)	68 (77%)	0.189
Abnormal karyotype	33 (32%)	20 (23%)	0.189
‐Of which complex karyotype	2 (2.0%)	2 (2.3%)	> 0.999
Specific aberrations, n (%)
t(8;21)	10 (9.8%)	3 (3.4%)	0.146
inv(16) or t(16;16)	1 (1.0%)	0 (0%)	> 0.999
t(6;9)	0 (0%)	1 (1.1%)	0.941
ELN 2022 risk			> 0.999
Favorable	0 (0%)	0 (0%)
Intermediate	78 (76%)	68 (77%)
Adverse	24 (24%)	20 (23%)
HSCT	54 (53%)	30 (34%)	0.014
Mutation
NPM1	36 (35%)	36 (41%)	0.519
DNMT3A	31 (30%)	24 (27%)	0.755
IDH1	5 (4.9%)	8 (9.1%)	0.394
IDH2	10 (9.8%)	11 (13%)	0.72
TP53	2 (2.0%)	2 (2.3%)	> 0.999
ASXL1	4 (3.9%)	2 (2.3%)	0.816
RUNX1	13 (13%)	7 (8.0%)	0.403
TET2	13 (13%)	14 (16%)	0.679
CEBPA	15 (15%)	7 (8.0%)	0.221
WT1	15 (15%)	8 (9.1%)	0.337
BCOR	4 (3.9%)	4 (4.5%)	> 0.999
CSF3R	4 (3.9%)	0 (0%)	0.17
EZH2	1 (1.0%)	3 (3.4%)	0.512
GATA2	10 (9.8%)	3 (3.4%)	0.146
JAK2	1 (1.0%)	2 (2.3%)	0.897
KIT	4 (3.9%)	2 (2.3%)	0.816
KRAS	5 (4.9%)	0 (0%)	0.099
MPL	0 (0%)	1 (1.1%)	0.941
NRAS	16 (16%)	4 (4.5%)	0.024
PTPN11	1 (1.0%)	2 (2.3%)	0.897
SRSF2	1 (1.0%)	3 (3.4%)	0.512
STAG2	3 (2.9%)	2 (2.3%)	> 0.999
U2AF1	2 (2.0%)	2 (2.3%)	> 0.999

FIGURE 2.

Prognostic impact of FLT3‐ITD variant allele frequency (VAF) burden. (A) Kaplan‐Meier curves compare survival outcomes between low VAF (< 40%) and high VAF (≥ 40%) patients. Log‐rank tests were used to assess differences in survival distributions (p = 0.00049). (B) Distribution of patients stratified by four VAF levels (< 10%, 10%–20%, 20%–40%, and ≥ 40%),Kaplan‐Meier curves comparing overall survival (OS) across the four VAF groups in the entire cohort (p = 0.006). (C) Significantly inferior survival in non‐transplant patients with VAF ≥ 40% (p = 0.04); (D) No significant survival difference observed in patients undergoing HSCT (p = 0.29).

3.3. Clinical Characteristics and Prognosis of NPM1 and DNMT3A Co‐Mutations in FLT3‐ITD‐Positive Acute Myeloid Leukemia

We analyzed co‐mutation patterns in FLT3‐ITD‐positive AML patients. NPM1 was the most common co‐mutation (∼45%), followed closely by DNMT3A (∼28%) (Figure 1C, Supporting Information S1: Figure S1C). We first compared patients with FLT3‐ITD/NPM1 co‐mutations (n = 72) to those with FLT3‐ITD alone (n = 118). The co‐mutated group was older (median age 56.5 vs. 49.0, p = 0.008) and exhibited more normal karyotypes (85% vs. 64%, p = 0.004) (Supporting Information S1: Table S3). Next, we examined the FLT3‐ITD/DNMT3A subgroup (n = 55). These patients were also significantly older (59.0 vs. 49.0, p < 0.001) with higher LDH levels (604.0 vs. 419.0 U/L, p = 0.017) and lower transplant rates (29% vs. 50%, p = 0.012) (Supporting Information S1: Table S4).

We then assessed the FLT3‐ITD/NPM1/DNMT3A triple‐mutation cohort (n = 36). Compared to single‐mutation patients (n = 99), this triple‐mutated group was older (58.0 vs. 48.0, p = 0.001) and had more myeloid blasts (78.5% vs. 65.3%, p = 0.013) (Supporting Information S1: Table S5). Interestingly, WT1 mutations showed strong exclusivity. WT1 appeared in 15% of single‐mutation patients but in zero triple‐mutation patients (p = 0.007). Ultimately, these co‐mutations negatively impacted survival. The FLT3‐ITD/NPM1 group had a drastically shorter median overall survival compared to the single‐mutation group (375.0 vs. 2024.0 days, p = 0.044) (Figure 3A). Patients with DNMT3A and triple mutations faced similar survival disadvantages (Figure 3C, E).

FIGURE 3.

Prognostic impact of co‐mutation patterns and FLT3‐ITD allele burden. Kaplan‐Meier curves analyzing overall survival (OS). (A, C, E) Impact of co‐mutations: Comparison of OS between patients with FLT3‐ITD alone versus those with (A) NPM1 co‐mutations (p = 0.044), (C) DNMT3A co‐mutations (p = 0.0053), and (E) FLT3‐ITD/NPM1/DNMT3A triple mutations (p = 0.0081). (B, D, F) Impact of VAF burden (cutoff ≥ 40%) within molecular subgroups: Survival stratified by VAF (< 40% vs. ≥ 40%) in patients with (B) NPM1 co‐mutations (p = 0.0012), (D) DNMT3A co‐mutations (p < 0.0001), and (F) Triple mutations (p < 0.0001). All p‐values were calculated using the log‐rank test.

To eliminate age bias, we performed a stratified analysis. In younger patients (< 60 years), high VAF remained significantly associated with poor prognosis (p < 0.001); however, the difference was not significant in patients aged ≥ 60 years (p = 0.28) (Supporting Information S1: Figure S3). To further eliminate the potential confounding effect of treatment selection bias, we specifically analyzed the subgroup of younger patients (< 60 years) who did not receive HSCT. In this specific setting, patients with VAF ≥ 40% still exhibited significantly shorter overall survival compared to those with VAF < 40% (p = 0.012) (Supporting Information S1: Figure S4).

3.4. Comparative Analysis of Survival Outcomes in Patients With FLT3‐ITD VAF Stratification and Companion Co‐Mutations in AML

To evaluate the prognostic role of FLT3‐ITD variant allele frequency (VAF) in specific molecular contexts, patient subgroups defined by coexisting mutations were analyzed. Kaplan‐Meier analysis revealed significantly shorter median overall survival (326 days) in patients with FLT3‐ITD/NPM1 double mutations and VAF ≥ 40%. Compared to relatively favorable outcomes in the FLT3‐ITD‐mutant‐only group (1163 days) and the FLT3‐ITD/NPM1 double‐mutant group with VAF < 40% (1515 days) (p = 0.0012) (Figure 3B).

Among patients with FLT3‐ITD/DNMT3A dual mutations, those with VAF ≥ 40% (n = 24) exhibited a significantly shorter median OS of 143.5 days, compared to a median OS of 1515.0 days in patients with VAF < 40% (n = 31) (p < 0.0001) (Figure 3D). This trend was also observed in the triple‐mutation subgroup (FLT3‐ITD, NPM1, and DNMT3A). Specifically, patients with triple mutations and VAF ≥ 40% (n = 18) had a median OS of 176.5 days, which was significantly shorter than the median OS of 1515.0 days in the VAF < 40% subgroup (n = 18) (p < 0.0001) (Figure 3F).

3.5. Predictive Value of FLT3‐ITD Allele Frequency for Overall Survival: Validation of Nonlinear Dose‐Response Relationship and Independent Prognostic Significance

Univariate analysis identified VAF ≥ 40%, advanced age (≥ 60 years), and DNMT3A co‐mutations as significant risk factors (Table 2). In the multivariate Cox regression model, a VAF ≥ 40% was a significant adverse prognostic factor (HR = 1.78, 95% CI: 1.19–2.66, p = 0.005). HSCT provided the strongest survival benefit (HR = 0.07, p < 0.001). Furthermore, TP53 mutation was identified as a potent independent risk factor (HR = 5.66, p = 0.001), whereas the impact of DNMT3A was attenuated in the multivariate analysis (p = 0.063).

TABLE 2.

Univariate and multivariate cox regression analyses for overall survival in patients with FLT3‐ITD–mutated acute myeloid leukemia.

Clinical features	Univariate analysis		Multivariate analysis
	HR (95% CI)	p‐value	HR (95% CI)	p‐value
VAF ≥ 40%	1.95 (1.33–2.85)	< 0.001	1.78 (1.19–2.66)	0.005
Gender	0.98 (0.67–1.43)	0.906
Age (≥ 60 years)	4.13 (2.79–6.11)	< 0.001	0.94 (0.61–1.45)	0.773
WBC (≥ 50 × 109/L)	1.35 (0.93–1.97)	0.113
HSCT	0.08 (0.05–0.13)	< 0.001	0.07 (0.04–0.14)	< 0.001
ELN 2022 risk	1.19 (0.77–1.83)	0.442
ASXL1Mutated	2.29 (0.93–5.63)	0.071
RUNX1Mutated	1.20 (0.67–2.15)	0.533
TP53Mutated	2.26 (0.83–6.16)	0.109	5.66 (2.00–16.01)	0.001
U2AF1Mutated	2.25 (0.83–6.11)	0.112
DNMT3AMutated	1.75 (1.18–2.61)	0.006	1.48 (0.98–2.23)	0.063
TET2Mutated	1.41 (0.85–2.34)	0.181
IDH1Mutated	1.27 (0.64–2.51)	0.497
IDH2Mutated	1.59 (0.93–2.70)	0.088		\

Time‐dependent ROC analysis demonstrated consistent predictive utility of VAF, with optimal performance observed during the mid‐stage follow‐up (2–3 years; peak AUC = 0.69 at 3 years, Figure 4A). Restricted cubic spline analysis revealed a non‐linear, U‐shaped dose‐response relationship between VAF and mortality. The mortality risk was lowest near the median VAF (approx. 34.4%) but escalated sharply beyond the 50% threshold. Compared to the baseline, the hazard ratio rose to 2.22 at 70% VAF and 4.17 at ≥ 97% VAF (Figure 4B). Notably, patients with extremely low VAF (< 10%) also exhibited distinct risk profiles, suggesting complex biological underpinnings beyond a simple linear correlation.

FIGURE 4.

Predictive value and dose‐response relationship of FLT3‐ITD VAF. (A) Time‐dependent ROC analysis predicting overall survival (OS) at 1, 2, 3, and 5 years. AUC values are annotated for each time point. (B) Restricted cubic spline (RCS) analysis modeling the non‐linear association between continuous VAF levels and mortality risk. The solid line represents the estimated hazard ratio (HR) relative to the median VAF (reference HR = 1), and the shaded area indicates the 95% confidence interval (CI).

4. Discussion

We conducted an in‐depth analysis of 265 AML patients with FLT3 mutations, elucidating the clinical significance of FLT3‐ITD VAF and the associated molecular genetic background in these patients. Our study further confirms that FLT3‐ITD mutations are important drivers of AML aggressiveness, in line with previous studies that identified them as key adverse prognostic markers [20, 21, 22, 23].

Currently, the European LeukemiaNet has set the AR of FLT3‐ITD mutations as a key indicator for assessing the prognosis of AML [24]. Research indicates that AR measurements themselves are not absolutely reliable, as their values are highly dependent on the genetic material used for detection [25]. Is it possible that a quantitative indicator exists that complements AR and is more readily standardized for clinical laboratory application? To this end, we systematically investigated the prognostic significance of FLT3‐ITD variant allele frequency (VAF).

An important and clinically relevant finding from our study is the validation of the 40% VAF threshold as a significant prognostic factor in adults with newly diagnosed acute myeloid leukemia. This cutoff effectively delineates FLT3‐ITD VAF ≥ 40% as an abnormally aggressive disease subtype characterized by significantly higher tumor burden—manifested by elevated white blood cell counts, lactate dehydrogenase levels, and bone marrow blast percentages—and an approximately 80% increased risk of death after adjusting for key confounders such as age and transplantation. This finding is consistent with previous quantification of FLT3‐ITD mutation burden [24, 26]. However, after excluding patients with NPM1 and DNMT3A mutations, no significant difference in overall survival was observed between the two groups of isolated FLT3‐ITD mutation patients (p = 0.1). This suggests that high VAF may not function as an absolute, independent determinant, but rather acts primarily as a potent “risk amplifier” when combined with other adverse mutations.

Previous studies have shown that the high mutation frequency of genes such as DNMT3A and TP53 in myeloid tumors indicates a poor prognosis [18, 19]. Another study on MDS also demonstrated the prognostic value of TP53 VAF in MDS [27].

Using a restricted cubic spline (RCS) model, we further revealed a nonlinear U‐shaped relationship between VAF and mortality risk. A slight increase in risk was observed at VAF < 10%, suggesting that these low‐burden clones may not be harmless subclones and may represent more complex clonal dynamics [28, 29]. When VAF > 50%, the hazard ratio shows a steep, dose‐dependent increase, confirming that high‐burden FLT3‐ITD clones are the dominant oncogenic driver [12].

Patients with both FLT3‐ITD and NPM1 co‐mutations showed a worse prognosis compared to patients with isolated FLT3‐ITD mutations, a finding that differs from previous studies [30]. The reason for this may be related to the fact that the co‐mutant patients in our study cohort were older.

In the FLT3‐ITD/NPM1 co‐mutant cohort, patients with FLT3‐ITD VAF < 40% had median overall survival (OS) comparable to those in the FLT3‐ITD‐only mutation group; however, when VAF ≥ 40%, the median OS was significantly shorter in the FLT3‐ITD/NPM1 cohort. These results are consistent with the conclusion that high AR predicts poor prognosis [31, 32]. Similarly, co‐mutations of FLT3‐ITD and DNMT3A also constitute a clearly defined high‐risk group. This study found that patients with both FLT3‐ITD and DNMT3A mutations had significantly shorter overall survival than those with only FLT3‐ITD mutations, a finding consistent with recent research conclusions [33]. Patients with FLT3‐ITD/DNMT3A VAF ≥ 40% also had worse survival outcomes than those with FLT3‐ITD single mutation and FLT3‐ITD/DNMT3A VAF ≤ 40%. This conclusion also holds true for the FLT3‐ITD/NPM1/DNMT3A triple mutation subtype. It should be specifically noted that patients in the co‐mutation groups were significantly older than those in the isolated mutation group (p < 0.01), suggesting that the poorer prognosis of patients with co‐mutations may have been confounded by age. After performing age‐stratified analysis, we found that among patients < 60 years, high VAF (≥ 40%) remained strongly associated with significantly shortened survival (p < 0.001), whereas this effect disappeared in patients ≥ 60 years (p = 0.28). This indicates that the genetic aggressiveness driven by high VAF is fully manifested in younger patients but is overshadowed by other factors in the elderly. In the multivariate Cox model, we found that age lost its independent prognostic significance, which may be primarily attributable to the strong association between advanced age and failure to receive HSCT (p < 0.001), together with an enrichment of high‐risk mutations (such as DNMT3A, ASXL1, and SRSF2) in the older group (Supporting Information S1: Table S6). When we removed HSCT from the multivariate model, the adverse prognostic effect of age re‐emerged prominently (HR = 3.66, 95% CI 2.44–5.50, p < 0.001) (Supporting Information S1: Table S7). This indicates that the survival disadvantage of older patients is largely mediated by the lack of curative transplantation.

The poorer survival rate in patients with high VAF may indicate the presence of persistently activated FLT3 receptors, leading to the continued activation of downstream signaling pathways such as STAT5, MAPK, and PI3K/AKT, which promotes cancer cell proliferation [4, 5, 6]. When this strong FLT3 signaling combines with the epigenetic dysregulation caused by DNMT3A mutations and the interference of NPM1 mutations on hematopoietic differentiation, a multi‐pathway synergistic pathogenic network is formed. In particular, in the case of triple mutations (FLT3‐ITD, NPM1, DNMT3A), it can lead to poor prognosis and ultimately promote a highly aggressive leukemia phenotype [34, 35, 36].

However, the impact of high VAF is not immutable; HSCT can effectively eliminate the survival disadvantage associated with FLT3‐ITD VAF ≥ 40%. In the non‐transplant cohort, patients with high VAF have poorer outcomes. In the transplant cohort, survival curves between patients with high and low VAF tend to converge, strongly demonstrating that HSCT can overcome this high‐risk genetic trait.

This study has several limitations. First, retrospective studies are subject to selection bias [37]. Secondly, due to the long timespan of this study, FLT3 inhibitors (such as midostaurin and gilteritinib) were not yet widely available in China or were limited by economic factors during the study period; therefore, the interventional effect of targeted agents could not be uniformly and statistically evaluated. Future prospective studies are needed to further validate the prognostic value of high VAF in the era of targeted therapy. Furthermore, the application of hematopoietic stem cell transplantation (HSCT) was uneven across different genetic risk subgroups, with significantly higher transplantation rates in patients with low VAF than in those with high VAF. This treatment selection bias preliminarily suggests that the poorer survival outcomes in the high‐VAF group may be partly attributable to insufficient access to this potentially curative consolidation therapy. As revealed by our real‐world data, a substantial proportion of younger patients with high VAF did not proceed to HSCT due to extremely high rates of primary refractoriness and early mortality; this indicates that the lack of access to treatment is fundamentally driven by the exceptionally high biological aggressiveness of the disease itself. To further exclude the confounding effect of transplantation, we conducted a specific analysis in younger patients (< 60 years) who did not receive HSCT and found that high VAF still exhibited a strong adverse prognostic impact (p = 0.012, Supporting Information S1: Figure S4). Finally, the prognostic utility of the 40% VAF threshold as a universal standalone marker indeed has limitations. Our data show that the prognostic impact of high VAF is attenuated in the overall cohort of patients with isolated FLT3‐ITD mutations. In summary, the biological aggressiveness associated with high VAF is highly context‐dependent: it functions primarily as a potent “risk amplifier” in the presence of co‐mutations and reveals its strong inherent lethality in the absence of curative HSCT intervention.

In conclusion, FLT3‐ITD VAF is a key prognostic factor in AML. In the presence of concurrent NPM1 and DNMT3A co‐mutations, it is more likely to function as a significant “risk amplifier.” Our findings support the inclusion of VAF in clinical practice and are significant for identifying high‐risk patients in urgent need of intensive therapy and allogeneic HSCT. Future prospective studies should aim to validate this threshold within a unified targeted therapy context and explore the combination of novel inhibitors with optimized transplantation strategies.

Author Contributions

Laibing Guo: conceptualization, methodology, formal analysis, writing – original draft. Jingyan Song: data curation, investigation. Haoyu Zang: data curation, investigation. Yang Cao: data curation, investigation. Yan Lin: validation, visualization. Zhuojun Zheng: validation, visualization. Suning Chen: resources, data curation (contributed patient data from participating centers). Lijun Wen: resources, data curation (contributed patient data from participating centers). Yuqing Miao: resources, data curation (contributed patient data from participating centers). Weiying Gu: conceptualization, supervision, writing – review and editing, project administration.

Funding

This work was supported by the Key project of Jiangsu Province Health Committee (ZD2021043); Changzhou Medical Center, Nanjing Medical University (CMCB202311).

Ethics Statement

The study was conducted in accordance with the Declaration of Helsinki. The study protocol was approved by the Ethics Committee of The Third Affiliated Hospital of Soochow University (Approval No. [2025] 171). The study was also approved or acknowledged by the institutional review boards of the other participating centers (The First Affiliated Hospital of Soochow University and Yancheng First People's Hospital).

Consent

The requirement for informed consent was waived by the Ethics Committees due to the retrospective nature of the study.

Conflicts of Interest

The authors declare no conflicts of interest.

Permission to Reproduce Material From Other Sources

The authors have nothing to report.

Supporting information

Supporting Information S1

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.