Von Willebrand, factor VIII, and D-dimer levels as predictors of risk and clinical outcomes in acute myeloid leukemia

Abstract

Coagulopathies, infection, and CNS infiltration are common complications during induction therapy of acute myeloid leukemia (AML). Early identification of adverse-risk patients may improve outcomes. This study evaluated the predictive value of coagulation markers, von Willebrand Factor antigen (vWF-ag), von Willebrand Factor–ristocetin cofactor (vWF-RCof), antihemophilic factor (FVIII), and D-dimer on risk stratification, severity and outcomes of AML patients. Fifty AML patients treated at Oncology Center Mansoura University hospital, from February 2023 to February 2024 were recruited and stratified into three risk groups favourable, intermediate, and adverse risk groups respectively according to the 2022 European leukemianet (ELN) risk stratification for AML. vWF-ag, vWF: RCof, FVIII, and D-dimer were measured at diagnosis and at remission. Adverse-risk group had the highest median levels of vWF-ag, vWF-RCof, FVIII and D-dimer at, both diagnosis and remission (p < 0.05). All markers significantly declined after remission (p < 0.05). vWF-ag, vWF-RCof, and D-dimer differed significantly among risk groups at both diagnosis and remission. ROC analysis showed that in the adverse-risk group, D-dimer had an AUC of 0.813 (cutoff > 1.6), vWF: Ag 0.780 (> 295), and vWF: RCo 0.761 (> 223). FVIII showed lower predictive value (AUC 0.625). Infection was the most frequent complication, followed by bleeding and thrombosis. Infection correlated with higher vWF: Ag after remission; thrombosis correlated with elevated vWF: Ag at diagnosis and remission and bleeding with lower vWF-ag. CNS infiltration showed no association. vWF-ag, vWF–RCof, and D-dimers may be associated with risk and complications in AML, but these findings require external validation in larger cohorts.

Supplementary Information

The online version contains supplementary material available at 10.1007/s10238-026-02198-0.

Introduction

Acute myeloid leukemia (AML) is the most prevalent acute leukemia in adults, with a median diagnosis age of 68 years. It is a heterogeneous hematologic malignancy that results from the unchecked expansion of clonal myeloid precursors [1]. Although adults’ 5-year overall survival is predicted to be around 30%, results differ by age; for example, those over 65 had an overall survival of less than 10% to 15% [2]. The US Food and Drug Administration has authorized more than 15 new AML agents or combinations since 2017. Advanced molecular understanding of etiology, clonal evolution, progression, and resistance mechanisms, in addition to targeted advancements, have revolutionized AML diagnosis, risk assessment, and treatment management [3].

Significant morbidity and mortality are caused by thromboembolic events in AML, particularly during induction treatment. Predicting the onset or recurrence of thrombosis is generally difficult in hematological malignancies [4]. Myeloid progenitor proliferation results in a hypercoagulable condition and endothelial dysfunction, which in turn alters the complete blood count [5]. The risk assessment methods now in use are not optimal for evaluating consequences in hematological malignancies because they are primarily utilized for solid tumors [6].

Despite the fact that all chemotherapeutics depress the immune system, hematological malignancies have an additional risk of infection, bleeding, or thrombosis because of poor hematopoiesis [7]. Despite being rare, CNS invasion is nonetheless a recognized AML consequence that lacks early warning signs [8]. Routine diagnostics like blood cell counts and coagulation markers only represent current status and are not very useful in predicting future difficulties, making it difficult to predict complications in patients with AML.

By supporting targeted treatment at the time of diagnosis, the identification of biological markers that would offer early insight into AML risk stratification and risk of consequences would improve prognosis. It may seem reasonable to think that von Willebrand factor (vWF), which is produced by both megakaryocytes and vascular endothelial cells, indicates hypercoagulability or injury to these cells. However, vWF may be dysfunctional, which could result in bleeding. An accurate sense of problems may be obtained by analyzing both markers, as vWF-ag determines the amount of vWF released while vWF-RCof determines its functionality. According to reports, vWF-ag may serve as a useful biomarker for stratifying the risk of thrombotic and bleeding complications in patients with AML [9].

When the pretest likelihood is unclear, D-dimer is utilized clinically to diagnose thromboembolic events because it reflects coagulation and fibrinolysis [10]. When AML is diagnosed, elevated D-dimer levels are a robust predictor of symptomatic thromboembolic events after chemotherapy [11]. However, its measurement at remission has not been studied, which could help define predictive cutoffs, at diagnosis, to potentially identify high risk groups and anticipate susceptibility to AML-associated complications.

Due to vWF’s function as a protective carrier molecule that delays the early degradation of procoagulant factor VIII (FVIII), epidemiological studies show that cancer patients have elevated FVIII, which is also independently linked to an increased risk of venous thromboembolism (VTE). It is still unclear what heightened FVIII is based on [12–14].

Each of these markers has limits on its own. Nonspecific, D-dimer is upregulated in cancer, trauma, inflammation, infection, and surgery [15]. Being acute phase reactants, FVIII and vWF may be indicative of underlying disease, with inter-patient release variations impacted by comorbidities, hepatic function, and genetics. Although they might indicate infection, disease severity, endothelial damage, or platelet dysfunction, these are not specific indicators on their own [16].

To improve early prediction of AML risk grade and associated complications, this study examined whether vWF-ag, vWF–RCof activity, FVIII, and D-dimer levels, at diagnosis and remission, can predict outcomes, such as infections, CNS infiltration, thromboembolic events or bleeding, in patients of differing AML risk grades and if so, determine the significant cut off values of these markers at diagnosis and remission.

Patients and methods

Fifty newly diagnosed AML patients were enrolled in the study before the initiation of chemotherapy between February 2023 and February 2024 are the subjects of the current prospective cohort study. Participants were chosen from among the inpatients and outpatients of the Oncology Center, Mansoura University (OCMU). According to the 2022 European LeukemiaNet (ELN) risk stratification system, AML patients are classified into favorable, intermediate, and adverse risk groups which is based on cytogenetic and molecular abnormalities at diagnosis [17]. Prior to research enrollment, all patients provided written consent, and the Institutional Review Board (IRB) approved the study protocol.

Newly diagnosed AML patients who had their diagnosis verified by cytogenetic, flow cytometry, and bone marrow tests met the inclusion requirements. Patients who were on anticoagulant medication or thromboprophylaxis, had incomplete or missing data, or had begun chemotherapy were not included. Age, sex, history of prior VTE, treatment history, related comorbidities, known thrombophilia, and whether bedridden or immobilized were among the clinical and demographic information gathered. In order to evaluate the illness status and identify any associated comorbidities, comprehensive clinical and laboratory evaluations were also carried out. To confirm the diagnosis, all patients had regular peripheral blood film, bone marrow aspiration, flow cytometry, cytogenetic investigations, and complete blood counts. Additional cytogenetic and molecular research evaluated AML risk categorization. The French-American-British (FAB) classification method, which divides AML into subtypes based on the appearance and differentiation of leukemic cells, was also used to classify the patients. These categories range from M0 to M7 [18].

To examine hematological parameters, eight milliliters of whole blood specimens were withdrawn into 0.109 mol/L (3.2% weight/volume [w/v]) buffered sodium citrate collection tubes. Plasma was rapidly separated from whole blood samples at normal temperature and was used for the following laboratory assays: vWF-ag, vWF-RCof, FVIII activity, and D-dimer. Sysmex CS-2500, Kobe, Japan was used to assess FVIII activity via clot-based method, while turbidimetric immune assay provided levels of vWF-ag, vWF-RCof, and d-dimer. These hematological parameters were examined both at diagnosis and after remission in all participants.

Remission was defined as complete blood cell counts reaching reference values and the inability to identify leukemia cells in the bone marrow and blood. Following a conventional chemotherapy course, all patients were monitored for a median of six months to evaluate any problems related to bleeding, infection, thrombosis, or CNS infiltration. They were then reassessed after full remission. Deep vein thrombosis (DVT) was diagnosed by duplex ultrasonography or venography if the patient had symptoms and any other thrombotic event was noted. Depending on the infection site, bacterial, viral, and fungal cultures were used to diagnosis the infections. Imaging was used to diagnose CNS infiltration in patients exhibiting new neurological impairments. Based on the patient’s history and follow-up, bleeding symptoms were documented.

Statistical analysis

Data was analyzed using IBM SPSS Statistics for Windows, Version 25.0 from IBM Corp [19]. Categorical data comparisons were conducted using the chi-square test to assess the association between variables. The Mann-Whitney U test was applied to compare two independent groups when the data was not normally distributed. The Kruskal-Wallis test, was utilized to compare continuous variables among multiple groups, when the assumptions of normality were not met. Analysis of Variance (ANOVA) was employed to compare means among three or more groups when assumptions of normality criterion were met. The Wilcoxon Signed-Rank test was used to compare paired samples in which normal distribution assumptions could not be applied. The Spearman correlation coefficient was calculated to assess the strength and direction of the relationship between two continuous variables. ROC curve analysis was used to evaluate the ability of hemostatic parameters to predict adverse-risk AML, with AUC and 95% CIs reported and cut-offs determined by the Youden index. To evaluate predictors of ELN adverse-risk AML, univariable logistic regression analysis was initially performed. To account for the limited number of events and reduce small-sample bias, a penalized logistic regression model using Firth’s correction was subsequently applied. Model variables were selected based on clinical relevance and univariable significance, with restriction of model size to avoid overfitting. Internal validation was conducted using bootstrap resampling (1000 iterations), and bias-corrected and accelerated (BCa) 95% confidence intervals were reported. Due to the small sample size and multiple comparisons, analyses were considered exploratory without adjustment for multiple testing.

Results

Patient demographics, stratified risk levels and complications

This study assessed 50 AML patients, 48% males and 52% females, with a mean age of 45.82 years, ranging from 7 to 81 years. The most common FAB classification was M4 (28%), followed by M2 and M3, both 20%. Risk stratification according to the 2022 ELN for AML was distributed, nearly evenly, between favourable, intermediate, and adverse-risk groups, 36%, 36%, and 28%, respectively. Infection was the most common complication at 50%, followed by bleeding 14%, thrombosis 12%, and CNS infiltration was found in 2%. The demographic, clinical and laboratory characteristics of AML patients are shown in (Table 1).

Table 1.

Patients’ Clinical and Laboratory Characteristics

		AML N = 50
Male		24(48%)
Female		26(52%)
Age (years)		45.82 ± 18.11 (7–81)
Hemoglobin (g/dL)		8.1 ± 1.69 (4.9–11)
WBCs (X10^9/L)		25.5 (0.8–439)
Platelet (X10^9/L)		32.5(8.4–251)
BM blasts (%)		76.5(15–95)
Peripheral blasts (%)		36(0–95)
vWF-ag (%) at diagnosis		293.5(63–583)
vWF-RCof (%) at diagnosis		213.5(57–425)
FVIII (%) at diagnosis		138.5 (72–275)
D-dimer (mg/L) at diagnosis		1.4(0.5–15.6)
FAB classification	M1	6(12%)
	M2	10(20%)
	M3	10(20%)
	M4	14(28%)
	M5	9(18%)
	M6	1(2%)
Risk	Favourable risk	18(36%)
	Intermediate risk	18(36%)
	Adverse risk	14(28%)
Infection		25(50%)
Thrombosis		6(12%)
Bleeding		7(14%)
CNS infiltration		1(2%)

Categorical data are expressed as count (%); parametric data are expressed as mean ± SD (minimum-maximum); non parametric data are expressed as median (minimum-maximum)

Individual parameters, age and thromboembolic events, remained statistically significant when comparing risk-stratified groups. Specifically, patients in the intermediate-risk group had a significantly higher mean age, 52.5 years, compared to the favourable-risk group at 36.94 years, with the adverse-risk group mean at 48.64 years (p = 0.025). Thrombosis was significantly more frequent among the adverse-risk group at 35.7%, compared to the intermediate and favourable risk groups, 5.6% and 0%, respectively (p = 0.007). Other parameters, such as sex, hemoglobin, white blood cells (WBCs), platelet counts, bone marrow (BM) blasts, peripheral blasts, FAB classification, infection, bleeding, and CNS infiltration, did not show a statistically significant association with risk groups (p > 0.05). Association of risk with other studied parameters among AML patients are shown in (Table 2).

Table 2.

Association of clinical and hematological parameters with risk level in all studied AML patients

		Risk			p
		Favourable risk	Intermediate risk	Adverse risk
Male		9(50%)	10(55.6%)	5(35.7%)	0.525
Female		9(50%)	8(44.4%)	9(64.3%)
Age (years)		36.94 ± 17.18	52.5 ± 17.19	48.64 ± 16.87	0.025*
Hemoglobin (g/dL)		7.86 ± 1.8	8.34 ± 1.65	8.1 ± 1.66	0.700
WBCs (X10^9/L)		25.65(0.8–171)	21.5(1.3–439)	27(1.7–238)	0.535
Platelet (X10^9/L)		27.5(8.4–184)	37.5(10–251)	32.5(9–57)	0.393
BM blasts (%)		80(26–95)	63(15–93)	76(22–95)	0.194
Peripheral blasts (%)		65(0–95)	11(0–95)	54(0–95)	0.178
FAB classification	M1	2(11.1%)	3(16.7%)	1(7.1%)	0.891
	M2	3(16.7%)	5(27.8%)	2(14.3%)
	M3	5(27.8%)	3(16.7%)	2(14.3%)
	M4	4(22.2%)	5(27.8%)	5(35.7%)
	M5	4(22.2%)	2(11.1%)	3(21.4%)
	M6	0(0%)	0(0%)	1(7.1%)
Infection		8(44.4%)	11(61.1%)	6(42.9%)	0.629
Thrombosis		0(0%)	1(5.6%)	5(35.7%)	0.007*
Bleeding		1(5.6%)	2(11.1%)	4(28.6%)	0.255
CNS infiltration		0(0%)	1(5.6%)	0(0%)	1

Categorical data are expressed as count (%); parametric data are expressed as mean ± SD; non parametric data are expressed as median (minimum-maximum). *, p < 0.05 is considered significant

AML risk levels and associations with hemostasis parameters

Significant differences were observed across risk groups for vWF-ag, vWF-RCof, and D-dimer at diagnosis and after remission, while FVIII exhibited significant differences among risk groups, only after remission. The adverse-risk group had the highest median vWF-ag, vWF-RCof, FVIII and D-dimer at diagnosis as well as after remission. All hemostatic parameters demonstrated significant reduction following remission when compared to those at diagnosis. Association of risk level with hemostatic parameters among AML patients is detailed in (Table 3; Fig. 1& Fig. 2).

Table 3.

Association of hemostatic parameters with risk levels in AML patients at diagnosis and following remissions

		Total	Risk			P1
			Favourable risk	Intermediate risk	Adverse risk
vWF-ag (%)	at diagnosis	293.5(63–583)	205.5(87–424)	304.5(65–563)	468(63–583)	0.002*
	after remission	176.5(15–482)	112.5(23–315)	183(15–387)	311.5(15–482)	0.006*
	P2	< 0.001*	< 0.001*	< 0.001*	0.001*
vWF-RCof (%)	at diagnosis	213.5(57–425)	174.5(65–298)	217(57–298)	277(67–425)	0.007*
	after remission	122.5(22–284)	102.5(28–187)	161(28–193)	159.5(22–284)	0.013*
	P2	< 0.001*	< 0.001*	< 0.001*	0.001*
FVIII (%)	at diagnosis	138.5(72–275)	121.5(87–189)	147.5(72–234)	157.5(73–275)	0.131
	after remission	99.5(32–189)	98(32–111)	98.5(47–167)	122(44–189)	0.011*
	P2	< 0.001*	< 0.001*	< 0.001*	0.001*
D-dimer (mg/L)	at diagnosis	1.4(0.5–15.6)	0.9(0.5–5.2)	1.5(0.5–11.2)	6.05(0.6–15.6)	< 0.001*
	after remission	0.8(0.1–12.3)	0.35(0.2–3.7)	0.9(0.2–9.5)	4.3(0.1–12.3)	< 0.001*
	P2	< 0.001*	< 0.001*	< 0.001*	< 0.001*

Non parametric data are expressed as median (minimum-maximum). *, p < 0.05 is considered significant. P1, comparison between favourable, intermediate and adverse risk groups; p2, comparison between at diagnosis and after remission

Fig. 1.

Hemostatic parameters levels change at diagnosis and after remission (A) vWF-ag, (B) vWF-RCof, (C) FVIII and (D) D-dimer

Fig. 2.

Hemostatic parameters levels at diagnosis and after remission stratified by AML risk regarding (A) vWF-ag, (B) vWF-RCof, (C) FVIII and (D) D-dimer

AML complications in association with hemostasis parameters

Thrombosis was significantly associated with a high median vWF-ag at diagnosis (535%) and after remission (392%), compared to patients without thrombosis. Bleeding was associated with the lowest median vWF-ag at diagnosis (85%) and after remission (23%), compared to patients without bleeding. Similar trends were noticed for vWF-RCof, D-dimer and FVIII, with thrombosis consistently associated with higher median values and bleeding with lower values. Infection was associated with a higher median vWF-ag after remission (214%) compared to patients without infection (p = 0.046). CNS infiltration did not show significant associations with all hemostatic parameters. All hemostatic parameters improved significantly after remission when compared to levels at diagnosis in both AML cases with and without studied complications. This data is detailed in (Table 4).

Table 4.

Association of hemostatic parameters with complications in patients at diagnosis and following remissions.

		Complications
		Infection			Thrombosis			Bleeding			CNS infiltration
		Negative	Positive	P1	Negative	Positive	P1	Negative	Positive	P1	Negative	Positive	P1
vWF-ag (%)	at diagnosis	286(65–583)	314(63–563)	0.299	286(63–583)	535(482–563)	< 0.001*	312(124–563)	85(63–583)	0.056	292(63–583)	428(428–428)	0.400
	after remission	129(15–375)	214(15–482)	0.046*	144(15–375)	392(318–482)	< 0.001*	182(79–482)	23(15–312)	0.019*	175(15–482)	345(345–345)	0.280
	P2	< 0.001*	< 0.001*		< 0.001*	0.028*		< 0.001*	0.018*		< 0.001*	#
vWF-RCof (%)	at diagnosis	212(57–298)	222(65–425)	0.299	204(57–298)	292.5(239–425)	< 0.001*	220(112–425)	71(57–278)	0.011*	213(57–425)	298(298–298)	0.160
	after remission	119(22–193)	145(25–284)	0.111	119.5(22–193)	183.5(127–284)	0.004*	127(72–284)	28(22–167)	0.003*	122(22–284)	178(178–178)	0.440
	P2	< 0.001*	< 0.001*		< 0.001*	0.028*		< 0.001*	0.018*		< 0.001*	#
FVIII (%)	at diagnosis	124(73–189)	160(72–275)	0.074	124.5(72–225)	224.5(118–275)	0.003*	146(105–275)	77(72–160)	0.003*	132(72–275)	225(225–225)	0.160
	after remission	99(44–132)	110(32–189)	0.336	98(32–165)	149.5(100–189)	0.001*	100(78–189)	52(32–124)	0.019*	99(32–189)	165(165–165)	0.160
	P2	< 0.001*	< 0.001*		< 0.001*	0.027*		< 0.001*	0.018*		< 0.001*	#
D-dimer (mg/L)	at diagnosis	1.2(0.5–12.3)	1.4(0.5–15.6)	0.326	1.2(0.5–12.3)	10.15(5.8–15.6)	< 0.001*	1.4(0.5–15.6)	0.9(0.5–12.3)	0.546	1.4(0.5–15.6)	1.9(1.9–1.9)	0.600
	after remission	0.8(0.1–10.9)	0.9(0.2–12.3)	0.307	0.8(0.1–10.9)	8.95(3.2–12.3)	< 0.001*	0.9(0.2–12.3)	0.3(0.1–10.9)	0.322	0.8(0.1–12.3)	1.5(1.5–1.5)	0.520
	P2	< 0.001*	< 0.001*		< 0.001*	0.028*		< 0.001*	0.018*		< 0.001*	#

Non parametric data are expressed as median (minimum-maximum). *, p < 0.05 is considered significant. P1, comparison between negative and positive groups; p2, comparison between at diagnosis and after remission. #, excluded from analysis due to small sample size

Hemostatic parameters correlations with risk stratification and other patients’ data

At diagnosis and after remission, risk stratification showed the strongest positive correlation with all four hemostatic parameters (p < 0.05 for each). Age also showed a significant positive correlation with D-dimer at diagnosis and after remission (rs = 0.305 and 0.292, respectively, both p < 0.05). Bone marrow blasts demonstrated a significant positive correlation with FVIII after remission (rs = 0.283, p = 0.046). Other parameters like white blood cells, hemoglobin, platelets, peripheral blasts did not show significant correlations with the hemostatic parameters. Correlation of hemostatic parameters with other data among AML patients was shown in (Table 5).

Table 5.

Correlation of hemostatic parameters with other data among AML patients

		vWF-ag		vWF-RCof		FVIII		D-dimer
		at diagnosis	after remission	at diagnosis	after remission	at diagnosis	after remission	at diagnosis	after remission
Age	rs	0.216	0.276	0.100	0.183	0.116	0.062		0.292
	p	0.132	0.053	0.490	0.204	0.421	0.667		0.039*
WBCs	rs	0.140	0.205	-0.025	0.114	0.087	0.093	0.162	0.220
	p	0.333	0.152	0.866	0.432	0.549	0.520	0.261	0.125
Hemoglobin	rs	-0.033	0.061	0.040	0.022	-0.016	-0.038	0.135	0.143
	p	0.822	0.671	0.783	0.881	0.914	0.791	0.348	0.323
Platelet	rs	-0.014	0.044	-0.043	0.138	0.069	-0.029	-0.205	-0.152
	p	0.920	0.760	0.765	0.339	0.635	0.841	0.153	0.294
BM blasts	rs	0.239	0.223	0.216	0.160	0.226	0.283	0.232	0.246
	p	0.095	0.120	0.132	0.267	0.114	0.046*	0.106	0.086
Peripheral blasts	rs	0.028	0.049	-0.150	-0.027	-0.073	-0.051	0.119	0.146
	p	0.850	0.735	0.298	0.851	0.615	0.723	0.411	0.311
Risk	rs	0.502	0.454	0.448	0.399	0.280	0.418	0.592	0.571
	p	< 0.001*	0.001*	0.001*	0.004*	0.049*	0.003*	< 0.001*	< 0.001*

Categorical data are expressed as count (%); parametric data are expressed as mean ± SD; non parametric data are expressed as median (minimum-maximum). *, p < 0.05 is considered significant

Predictive value of hemostatic parameters at AML diagnosis for adverse-risk disease

The D-dimer stands out with an AUC of 0.813, a highly significant p-value (< 0.001), and a cutoff of > 1.6. It demonstrates good sensitivity 78.6% and specificity 75%, indicating its potential as a reliable predictor. VWF-ag demonstrates an AUC of 0.780 at a cutoff value of > 295, sensitivity of 85.71% and specificity of 66.67%, with a significant p-value of 0.003, suggesting good discriminative ability. The vWF-RCof exhibits a notable AUC of 0.761at a cutoff of > 223, sensitivity 85.7% and specificity 75%, with a significant p-value of 0.004. FVIII revealed an AUC of 0.625 with sensitivity of 71.43% and a specificity of 58.33%, and a non-significant p-value of 0.183, suggesting limited discriminatory power for adverse-risk AML. These results underscore the importance of considering both sensitivity and specificity when assessing the predictive value of hemostatic parameters in diagnosing adverse-risk AML, with D-dimer showing particular promise in this context. Validity of hemostatic parameters at diagnosis for prediction of adverse-risk AML is shown in (Table 6& Fig. 3).

Table 6.

Predictive Value of Hemostatic Parameters at AML Diagnosis for adverse-Risk Disease

Variable	AUC (95% CI)	p-value	Cut-off	Sensitivity % (95% CI)	Specificity % (95% CI)	PPV % (95% CI)	NPV % (95% CI)	Accuracy % (95% CI)
vWF-ag	0.780 (0.640–0.885)	0.003*	> 295	85.7 (57.2–98.2)	66.7 (49.0–81.4)	50.0 (28.2–71.8)	92.3 (74.9–99.1)	72.0 (57.5–83.8)
vWF-RCof	0.761 (0.619–0.870)	0.004*	> 223	85.7 (57.2–98.2)	75.0 (58.8–87.3)	57.1 (34.0–78.2)	93.1 (77.2–99.2)	78.0 (64.0–88.5)
FVIII	0.625 (0.477–0.758)	0.183	> 132	71.4 (41.9–91.6)	58.3 (41.0–74.5)	40.0 (21.1–61.3)	84.0 (63.9–95.5)	62.0 (47.2–75.3)
D-dimer	0.813 (0.678–0.910)	< 0.001*	> 1.6	78.6 (49.2–95.3)	75.0 (58.8–87.3)	55.0 (31.5–76.9)	90.0 (73.5–97.9)	76.0 (61.8–86.9)

AUC, area under ROC curve; CI, confidence interval; PPV, positive predictive value; NPV, negative predictive value

Fig. 3.

ROC analysis of vWF-ag, vWF-RCof, FVIII and D-dimer at diagnosis for prediction of adverse-risk AML

Predictors of ELN adverse-risk AML: univariable, penalized, and bootstrap analyses

In univariable analysis, both vWF-ag and D-dimer were significantly associated with ELN adverse-risk AML. In the penalized (Firth) regression model, these associations remained statistically significant, indicating robustness after correction for small-sample bias. Bootstrap validation further confirmed the stability of these findings, with consistent effect estimates across resampled datasets. In contrast, age was not significantly associated with adverse-risk disease in any model (Table 7).

Table 7.

Predictors of ELN adverse-risk AML: univariable, penalized, and bootstrap analyses

Variable	Univariate logistic regression		Penalized (Firth) regression		Bootstrap validation of logistic regression
	OR(95% CI)	p-value	OR(95% CI)	p-value	OR(95% CI)	p-value
Age (per year)	1.008(0.987–1.029)	0.484	1.02(0.99–1.06)	0.158	1.01(0.98–1.04)	0.412
WBC	1.002(0.996–1.007)	0.599
Platelet	0.989(0.970–1.008)	0.253
Infection/Inflammation	0.788(0.375–1.655)	0.529
vWF: Ag	1.903(1.202–3.101)	0.004*	1.58(1.05–2.63)	0.032*	1.52(1.08–2.41)	0.021*
D-dimer	1.444(1.142–1.826)	0.002*	1.39(1.08–1.82)	0.014*	1.34(1.06–1.74)	0.018*

OR, odds ratio; CI, confidence interval. Univariable logistic regression was initially performed to identify candidate predictors. A penalized logistic regression model using Firth’s correction was subsequently applied to reduce small-sample bias and account for sparse data. Internal validation was performed using bootstrap resampling (1000 samples), with bias-corrected and accelerated 95% confidence intervals reported. Continuous variables were scaled to clinically meaningful units. All analyses are exploratory and should be interpreted with caution given the limited sample size

Discussion

AML is characterized by a complex spectrum of coagulopathies ranging from hemorrhagic to thrombotic symptoms [20]. In this study we investigated the pattern of haemostatic markers, vWF-ag level, vWF–RCof activity, FVIII, and D-dimer, in newly diagnosed AML patients and the change in these parameters after remission.

The predictive value of therapy type (intensive vs. supportive) for DVT in patients with AML was not assessed. In Farge et al. [21] study, patients treated with intensive therapy developed thrombosis more frequently. Intensive chemotherapy can increase the risk of thrombosis via direct endothelial damage, destruction of leukemic cells releasing thrombogenic substances, and reduced synthesis of natural anticoagulants due to liver damage [22–25]. According to the studies conducted in the general population, in patients with AML, CVL insertion is a proven risk factor for thrombosis development [26–30]. Considering that intensive treatment requires CVL insertion and that chemotherapy can be thrombogenic, we decided to include the therapy type in our final model.

Our results show significant differences in vWF-ag, vWF-RCof and D-Dimer across risk groups. In the adverse-risk group, vWF-ag demonstrated an AUC of 0.780, at a cutoff value of > 295, sensitivity of 85.71% and specificity of 66.67% at a p-value of 0. 003.Similarily, vWF-RCof had an AUC of 0.761 at a cutoff of > 223 in the adverse-risk group, with sensitivity of 85.7% and sensitivity of 75% with a p-value of 0.004. D-dimer demonstrated the highest predictive value with an AUC of 0.813 at a cutoff of > 1.6 in the adverse-risk group, with sensitivity of 78.6% and specificity of 75% and a p-value of < 0.001. FVIII was not associated with significant differences amount the risk stratified groups.

These findings are consistent with previously published reports. Elevated vWF-ag and vWF-RCof have previously been documented in hematological malignancies. In AML and ALL patients Liu et al. [9, 31] demonstrated high vWF-ag levels correlated with high-risk phenotypes; another study linked high vWF levels with unfavorable outcomes [32]. Similarly Gavriatopoulou et al. [33] showed that in Waldenstrom’s macroglobulinemia, higher vWF levels corelated to high International Prognostic Scoring System (IPSS) score indicating poor prognosis. A pilot study in myelodysplastic syndrome also reported high vWF-ag levels to be associated with high IPSS scores, reduced survival, and poorer response to therapy [34].

The association between D-dimer levels and disease severity in our study parallels prior reports in solid tumors including breast [35, 36], colorectal [37, 38], lung [39, 40] and brain cancers [41]. A metanalysis [42] demonstrated elevated D-dimer reflects enhanced activation of coagulation and fibrinolysis, and independently predicts adverse prognosis, often unrelated to overt thromboembolic events. Previous studies highlighted the prognostic and predictive value of D-dimer in chemotherapy response. According to Inanc et al. [43], lower D-dimer was linked to partial response in colorectal cancer patients while those with progressive disease had significantly higher levels. Moreover, Komurcuoglu et al. [44], reported high levels in lung cancer chemotherapy non-responders compared to responders. Antoniou et al. [45], reported high Dimer levels predicted poor response and tumor progression in advanced lung cancer patients. Alkhoder et al. [46], observed decreased D-dimer after chemotherapy in breast cancer patients. These reports are congruent with the results of this study, which is the first to suggest the clinical significance of D-dimer (and other hemostatic markers) as predictor for therapeutic response in AML patients.

Data regarding the predictive value of the DIC score according to the ISTH criteria are conflicting.Libourel et al. showed that DIC at the time of AML diagnosis was associated with an increased VTE rate. Moreover, D-dimer was marked as individual predictor [11]. Higher D-dimer levels are a widely used biomarker for DVT prediction in patients with cancer [47, 48]. In contrast, Martella et al. failed to prove the predictive value of D-dimer levels [49].

Although the use of prophylactic anticoagulation is not uncommon in patients with AML, the evident bleeding risk could outweigh its clinical benefit [50]. Therefore, developing a model for adequate risk assessment would be useful. The new International Initiative on Thrombosis and Cancer guidelines recommend thromboprophylaxis for patients with cancer at high risk for VTE development and low risk for bleeding [21].

A unique finding in our study is all hemostatic markers significantly decreased after remission, supporting their potential as biomarkers for treatment monitoring. Although some studies have shown evaluated vWF as a biomarker for response, this was mainly in other cancers and few have examined the specific panel we studied. For example, in hepatocellular carcinoma, vWF and ADAMTS13 levels have been proposed to predict response to angiogenesis inhibitors [51], hepatic artery infusion chemotherapy [52] and liver resection [53]. In metastatic colorectal cancer, baseline vWF and angiogenic markers correlated with progression-free intervals and overall survival during treatment [54].

We also observed that thrombosis was more frequent in the adverse-risk group 35.7% compared to the intermediate (5.6%) and low risk group (0%) (p = 0.007). Similar findings have been reported in other studies [55, 56], and [57]. Lee et al. [58], demonstrated VTE incidence increased with advanced molecular and cytogenetic risk. Possible explanations include the use of intensive re-induction chemotherapy in high-risk cases leading to vascular endothelial damage and activation of platelet and coagulation cascade [59–61], and [62]. In addition, disease severity with myeloid precursor proliferation can trigger a pronounced hypercoagulable state.

In our cohort, thrombosis was correlated with elevated vWF-ag, vWF-RCof and FVIII and D- dimer at both diagnosis and after remission compared to patients without thrombosis. This was shown in studies in other solid cancers. For example, high preoperative D-dimer levels predicted postoperative DVT in colorectal surgery [63]. Indeed, existing VTE risk prediction scores for ambulatory solid cancer patients such as Vienna Cancer and Thrombosis Study (CATS) score incorporate D-dimer as a major component [64] and a threshold of ≥ 8% has helped improved efficacy of drug thromboprophylaxis [65].

Beyond D-dimer, both vWF and FVIII correlate with VTE risk within the general population [66]. Interestingly, studies indicate elevated vWF levels as an independent VTE predictor in cancer patients [67–70]. In an assessment of vWF pro-peptide, a marker of vWF secretion, Nossent et al. concluded high vWF correlated with thrombosis via increased endothelial vWF secretion [71]. In Bauer et al. [72], , cultured tumor cell lines stimulate angiogenesis and vWF secretion. Melanoma tumor biopsies revealed ultra-large von Willebrand factor (UL-vWF) multimers within micro vessels. Tumor‐mediated inhibition of ADAMTS13 supported aggregation of UL-vWF multimers. Collectively, outcomes imply tumor cells induce UL‐VWF release, simultaneously, downregulating ADAMTS13. Directly, this rendered to prothrombotic states, with vWF‐platelet aggregates and fibrin deposition in tumor vasculature.

Conversely, patients with bleeding manifestations had the lowest vWF-ag, vWF-RCof, and FVIII levels at diagnosis and remission. This is in keeping with Shen [73], describing nonfunctional VWF multimers in AML [73]. and Federici and D’Amico [74] reporting proteolytic degradation of vWF released from leukemic cells, which reduces vWF–RCof activity and promoting bleeding despite vWF release. Similar findings were shown by others (Abaza et al. [75], and Lethagen et al. [76]).

This study highlights the critical burden of severe bleeding complications in patients with AML, revealing a complex interplay of disease biology, treatment-related factors, and clinical vulnerability. Abdulgayoom et al. [77] reported that among 214 newly diagnosed AML patients treated at our center over a ten-year period, 16.3% experienced Grade ≥ 3 bleeding, with intracranial hemorrhage being the predominant and most devastating manifestation. This aligns with previous studies that have consistently identified central nervous system bleeding as a major source of morbidity and early mortality in AML [78] and is comparable to the 5%-20% incidence reported in other international series [79–81]. Although current guidelines recommend maintaining platelet counts above 10 × 10⁹/L in stable AML and 20–50 × 10⁹/L in high-risk settings, most ICH events occurred below these thresholds (median 11 × 10⁹/L), suggesting that existing practices may not always prevent catastrophic bleeding. Beyond thrombocytopenia, qualitative platelet dysfunction, coagulopathy, and disseminated intravascular coagulation (DIC) may further amplify bleeding risk [82].

The study findings are exploratory and hypothesis-generating due to small sample size and multiple testing. ROC-derived cut-offs are data-driven and require external validation. The present study demonstrates that vWF: Ag and D-dimer are consistently associated with adverse-risk AML across multiple analytical approaches, including conventional logistic regression, penalized regression, and bootstrap validation. The concordance of these findings supports the robustness of the observed associations despite the limited sample size. Given the exploratory nature of the analysis and the potential for residual confounding, these results should be interpreted cautiously.Limitations include small cohort size, low event rates, and no internal validation. Larger prospective studies are needed to confirm these findings and validate predictive thresholds.

Overall, and despite some limitations related to the relatively sample size and single center research, the study concludes vWF-ag, vWF-RCof, and D-dimer levels are significantly associated with AML risk, thrombotic and bleeding complications, and shows a dynamic pattern with remission, indicating their potential as biomarkers for prognosis, risk stratification, and treatment monitoring in AML. We provide evidence in support of conducting larger studies to validate these observations and to explore integration of these biomarkers into risk stratification models in hematological malignancies.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

ADAMTS13

A distntegrin and metalloproteinase with a thrombospondin type 1 motif, member 13

AML

Acute myeloid leukemia

ANOVA

Analysis of Variance

AUC

Area under the ROC curve

BM

Bone marrow

CATS

Cancer and Thrombosis Study

CNS

Central Nervous System

EC

Endothelial cells

FAB

French-American-British

FVIII

Factor VIII

IRB

Institutional Review Board

OCMU

Oncology center Mansoura university

RCO

Ristocetin cofactor

ROC

Receiver Operating Characteristic

VTE

Venous thromboembolism

VWF

Von Willebrand factor

WBCs

White blood cells

WPB

Weibel-Palade bodies

Author contributions

Conception and design, analysis or interpretation of data, reviewing article: Hanaa Ali EL-Sayed, Doaa H. Sakr, Reham Alghandour, Eman Eid, May Denewer, Mohamed Awad Ebrahim, and Hanan Azzam; Data (or sample) collection: Hanaa Ali EL-Sayed and Doaa H. Sakr; Performing experiments: Hanaa Ali EL-Sayed and Hanan Azzam; Writing the article: Hanaa Ali EL-Sayed, Balvinder Saarsalu, Maha, Othman, and Hanan Azzam. Intellectual contribution, data and manuscript review and critique: Maha Othman. All authors read and approved the final manuscript.

Funding

This study was primarily personally funded. Mansoura University facilitated patient recruitment and provided research lab facilities.

Data availability

all data is available by the corresponding author on a reasonable request.

Declarations

Competing interests

The authors declare no competing interests.

Ethics approval and consent to participate

The research adhered to the Code of Ethics of the World Medical Association (Declaration of Helsinki) for studies involving human subjects. The project was approved by Mansoura University board of ethics with code R.24.10.2858. All the cases involved in the study provided the written consent for participation and publication of scientific data.

Consent for publication

The manuscript does not contain information or images that could lead to identification of a study participant.