Ferritin and LDH predict thrombotic risk beyond conventional scoring systems in DLBCL

Abstract

Thrombotic events are clinically significant complications in diffuse large B-cell lymphoma (DLBCL) that adversely affect treatment continuity and outcomes. Despite their importance, no standardized predictive approach exists, and traditional cancer-associated thrombosis models perform suboptimally in hematological malignancies. We retrospectively analyzed 133 adult patients with DLBCL treated at a single center between 2009 and 2021. Thrombotic events required objective radiological confirmation. Baseline demographic, clinical, and laboratory parameters were compared between patients with and without thrombosis. ROC analysis determined the optimal ferritin cutoff, and Cox proportional hazards and Fine–Gray competing-risk models (with death as a competing event) identified independent predictors. Thrombosis occurred in 16 patients (12.0%), all venous. Baseline clinical characteristics and the Khorana score did not differ between groups. Patients with thrombosis had significantly higher ferritin (median 331.5 vs. 113 ng/mL; p = 0.026) and LDH (median 459.5 vs. 256 U/L; p = 0.022). ROC analysis identified a ferritin cutoff of ≥ 227 ng/mL (AUC 0.686, 95% CI 0.532–0.840), associated with a 6.1-fold increased risk (OR 6.10, 95% CI 1.75–21.28). In multivariate Cox regression, ferritin ≥ 227 ng/mL (HR 3.77, 95% CI 1.13–12.60, p = 0.031) was an independent predictor and remained significant in competing-risk analysis (sHR 4.68); LDH was significant in Cox analysis (HR 1.002 per unit increase, p = 0.034) but was attenuated in the competing-risk model. In DLBCL, where no standardized thrombosis risk-stratification approach exists, elevated ferritin at diagnosis independently predicted thrombotic events, while LDH was significant in univariate analysis. These readily available parameters may contribute to lymphoma-specific thrombosis scoring systems and warrant prospective validation.

Introduction

Thrombotic events are a major cause of morbidity and mortality in patients with cancer and are among the primary non-cancer-related complications affecting treatment continuity and survival [1–6].

In recent years, it has been demonstrated that the incidence of thromboembolic events in hematological malignancies, particularly lymphoma, is comparable to and even higher than that in solid tumors [7]. Among lymphomas, the incidence of non-Hodgkin lymphomas, and particularly their aggressive subtypes, is much higher. This increased risk is recently reported to be associated with tumor-related factors, systemic inflammation, endothelial dysfunction, chemotherapy, and supportive therapies (growth factors, central venous catheter use, etc.) [8–14].

Various clinical models have been developed to identify patients at high risk of developing cancer-related thrombosis, including the Khorana score [15]. However, these models are primarily derived from solid tumor populations and have shown limited predictive performance in hematologic malignancies. This highlights the need for alternative biomarkers that better reflect the disease biology and thrombotic inflammatory burden in this population group.

This study aimed to investigate the risk factors of thrombotic events in patients with lymphoma and to determine whether any factor provides prognostic information beyond established clinical risk scores and laboratory markers.

Materials and methods

This retrospective, single-center cohort study includes 133 adult patients diagnosed with diffuse large B-cell lymphoma between 2009 and 2021 and treated at Ankara Dışkapı Yıldırım Beyazıt Training and Research Hospital. Clinical, laboratory, and radiological data were obtained from institutional medical records.

All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki Declaration and its later amendments or comparable ethical standards. Approval for this retrospective cohort study was granted by the Clinical Research Ethics Committee of Ankara Dışkapı Yıldırım Beyazıt Training and Research Hospital (approval number 107/30; date 22 March 2021). Informed consent was waived due to the retrospective design.

Patients who developed thrombosis during the study period were recorded. Thrombosis was diagnosed based on patient-reported symptoms suggestive of thromboembolic disease (including dyspnea, extremity pain, swelling, or acute ischemic symptoms), followed by radiological confirmation of the diagnosis. Only radiologically confirmed thrombotic events were included in this study.

Only objectively confirmed deep vein thrombosis (DVT) and/or pulmonary embolism (PE) were counted as qualifying events; catheter-related thrombosis and superficial thrombophlebitis were not included. Cause of death was ascertained from medical records and classified as lymphoma-related (progressive, relapsed, or refractory disease), treatment-unrelated (an independent cause in a patient in sustained complete remission), or indeterminate (death within six months of treatment completion without a documented cause); a VTE-related death was defined as death with pulmonary embolism documented as the proximate cause, or death within 30 days of a thrombotic event without an alternative explanation. Deaths were not formally adjudicated.

In time-to-event analyses, time zero was defined as the date of lymphoma diagnosis. The time to thrombosis was calculated from the date of diagnosis to the first objectively confirmed thrombotic event or the last follow-up, whichever occurred first.

The baseline variables included age, sex, comorbidities, ECOG performance status, disease stage, IPI score, Khorana score, treatment-related variables, catheter use, treatment response, and laboratory parameters, including complete blood count indices, ferritin, LDH, C-reactive protein (CRP), erythrocyte sedimentation rate (ESR), albumin, vitamin B12, β2-microglobulin.

All laboratory parameters, including ferritin, CRP, ESR, vitamin B12, and β2-microglobulin, were obtained at the time of diagnosis, prior to the initiation of lymphoma-directed therapy, as part of the routine institutional diagnostic and staging work-up. Owing to the retrospective design and the long study period, a subset of tests was not uniformly available for every patient; ferritin, in particular, was available in 100 of 133 patients (75.2%).

First-line treatment followed contemporary international guidelines (NCCN/ESMO); most patients received R-CHOP, with dose reductions in elderly or frail patients and in those with cardiac comorbidity. Thrombotic events were managed with low-molecular-weight heparin (LMWH) as standard initial anticoagulation, preferred for its shorter duration of action in the context of active chemotherapy and treatment-related cytopenias. No patient received erythropoiesis-stimulating agents; symptomatic or severe anemia was managed with red blood cell transfusion. Granulocyte colony-stimulating factor was not administered as routine primary prophylaxis but was given as short-course therapy to patients who developed neutropenia.

The lymphoma-specific Thrombosis Lymphoma (ThroLy) score was calculated for all patients. Where reduced mobility was not explicitly documented, an ECOG performance status ≥ 2 was used as a surrogate, and body mass index was applied where available and otherwise scored as absent. Patients were classified as low- (0–1), intermediate- (2–3), or high-risk (> 3). The Khorana score was likewise categorized; because lymphoma contributes to the site component, no patient scored 0, and patients were grouped as intermediate- (1–2) or high-risk (≥ 3).

Statistical analyses were performed using the Number Cruncher Statistical System (Utah, USA). Normality was assessed using the Shapiro–Wilk test and graphical analysis. Continuous variables were compared using the Student’s t-test or the Mann–Whitney U test, as appropriate. Categorical variables were analyzed using Pearson chi-square, Fisher’s exact, or Fisher–Freeman–Halton tests.

ROC curve analysis was used to determine the optimal ferritin cutoff value for predicting thrombosis. Time-to-thrombosis analysis was conducted using Cox proportional hazards regression. Variables with p < 0.20 in the univariate analyses were included in the multivariate models using backward elimination method. The results are reported as hazard ratios (HRs) with 95% confidence interval (CIs). A two-sided p-value < 0.05 was considered statistically significant.

To account for the competing risk of death, cumulative incidence functions were estimated using the Aalen–Johansen method. Subdistribution hazard ratios (sHR) for the Fine–Gray model were obtained within the survival package by constructing inverse-probability-of-censoring-weighted risk sets (the finegray function) followed by a weighted Cox regression with robust standard errors. Cause-specific Cox models were fitted as a sensitivity analysis. Analyses were performed in R version 4.3.3 (survival package, version 3.5.8), and discrimination comparisons (DeLong test) in Python (statsmodels and scikit-learn packages).

Results

A total of 133 patients were included, with a mean age of 63.13 ± 14.85 years. Thrombotic events occurred in 16 patients (12.0%), and these events were venous thrombosis. Of these patients, 10 (62.5%) had unilateral deep vein thrombosis, 2 (12.5%) had bilateral deep vein thrombosis, 2 (12.5%) had pulmonary embolism, and the remaining 2 (12.5%) had both pulmonary embolism and deep vein thrombosis. No catheter-associated thrombotic event occurred. The median time from diagnosis to thrombosis was 2.18 months (interquartile range 1.05–5.82; range 0.43–11.79), with 50% of events occurring within the first two months and 75% within six months of diagnosis.

Baseline demographic characteristics, comorbidities, ECOG performance status, disease stage, IPI score, Khorana score, and catheter use did not differ significantly between patients with and without thrombosis (Table 1).

Table 1.

Clinical findings at the time of diagnosis based on the presence of thrombosis

		Thromboembolic event
		No (n = 117) (%)	Yes (n = 16) (%)	Total	p
Age	Mean ± SD	63.50 ± 14.46	60.38 ± 17.77	63.13 ± 14.85	0.673
	Median (Min-Max)	65 (20–86)	66.5 (24–85)	65 (20–86)
Gender	Female	59 (50.4)	8 (50.0)	67 (50.4)	0.974
	Male	58 (49.6)	8 (50.0)	66 (49.6)
Comorbidity	None	43 (36.8)	6 (37.5)	49 (36.8)	0.954
	Present	74 (63.2)	10 (62.5)	84 (63.2)
Number of Comorbidity	Median (Min-Max)	1 (0–5)	1 (0–3)	1 (0–5)	0.681
ECOG	Median (Min-Max)	1 (0–4)	1 (0–3)	1 (0–4)	0.607
Khorana score	Median (Min-Max)	2 (1–3)	2 (1–4)	2 (1–4)	0.188
Stage	Median (Min-Max)	3 (1–4)	3 (1–4)	3 (1–4)	0.290
IPI score	Median (Min-Max)	3 (0–5)	2 (1–4)	3 (0–5)	0.732
Bone Marrow	Negative	91 (82.0)	13 (86.7)	104 (82.5)	1.000
Involvement	Positive	20 (18.0)	2 (13.3)	22 (17.5)
Bulky	Negative	85 (80.2)	11 (73.3)	96 (79.3)	0.510
	Positive	21 (19.8)	4 (26.7)	25 (20.7)
Extranodal	Negative	54 (53.5)	10 (62.5)	64 (54.7)	0.500
Involvement	Positive	47 (46.5)	6 (37.5)	53 (45.3)
CVC	None	81 (69.2)	12 (75.0)	93 (69.9)	0.776
	Present	36 (30.8)	4 (25.0)	40 (30.1)

ECOG Eastern Cooperative Oncology Group, IPI International Prognostic Index, CVC central venous catheter

p < 0.05

Laboratory analysis demonstrated significantly higher ferritin levels in patients with thrombosis than in those without (median, 331.5 vs. 113 ng/mL; p = 0.026). LDH levels were significantly elevated in the thrombosis group (median, 459.5 vs. 256 U/L; p = 0.022). Other hematologic and inflammatory parameters did not show any significant differences (Table 2).

Table 2.

Laboratory findings at the time of diagnosis based on the presence of thrombosis

		Thromboembolic event
		No (n = 117)	Yes (n = 16)	Total	p
HB	Mean ± SD	11.98 ± 2.21	11.2 ± 2.07	11.88 ± 2.2	0.187
	Median (Min-Max)	12 (6.6–17)	11.6 (7.5–14.3)	11.9 (6.6–17)
HCT	Mean ± SD	36.01 ± 6.8	34.06 ± 6.14	35.78 ± 6.73	0.278
	Median (Min-Max)	36.7 (15–49.1)	34.1 (22.9–44.6)	36.5 (15–49.1)
PLT	Mean ± SD	282178.63 ± 136413.43	301312.5 ± 114750.58	284480.45 ± 133747.53	0.422
	Median (Min-Max)	267000 (10000–1089000)	291500 (95000–537000)	268000 (10000–1089000)
WBC	Mean ± SD	8181.71 ± 3743.06	8775.63 ± 3051.11	8253.16 ± 3661.66	0.359
	Median (Min-Max)	7400 (1740–21900)	8045 (5080–16500)	7500 (1740–21900)
ESR	n	103	13	116	0.806
	Mean ± SD	39.42 ± 26.67	37.38 ± 27.24	39.19 ± 26.62
	Median (Min-Max)	32 (5–120)	30 (2–89)	32 (2–120)
CRP	n	103	15	118	0.529
	Mean ± SD	46.63 ± 60.77	51.73 ± 48.39	47.28 ± 59.19
	Median (Min-Max)	23.6 (1.4–374)	30 (2.8–161)	24 (1.4–374)
Ferritin	n	86	14	100	0.026*
	Mean ± SD	253.07 ± 427.6	357.42 ± 293.19	267.68 ± 411.83
	Median (Min-Max)	113 (2.6–2831)	331.5 (0.9–980)	129 (0.9–2831)
LDH	Mean ± SD	392.56 ± 350.6	544.69 ± 350.82	410.86 ± 352.81	0.022*
	Median (Min-Max)	256 (120–2314)	459.5 (167–1340)	268 (120–2314)
B2M	n	93	9	102	0.128
	Mean ± SD	3.84 ± 3.14	15.46 ± 35.1	4.86 ± 10.84
	Median (Min-Max)	2.9 (1.3–25.6)	3.9 (2.1–109)	2.9 (1.3–109)
Albumin	n	96	14	110	0.844
	Mean ± SD	3.59 ± 0.71	3.62 ± 0.56	3.6 ± 0.69
	Median (Min-Max)	3.6 (1.6–5.3)	3.7 (2.4–4.3)	3.7 (1.6–5.3)

Data are presented as median (range) or mean ± SD. Units ferritin, ng/mL LDH, U/L hemoglobin, g/dL hematocrit, % platelets and leukocytes, /µL CRP, mg/L ESR, mm/h albumin, g/dL, β2-microglobulin (B2M), mg/L. SD standard deviation, LDH lactate dehydrogenase, CRP C-reactive protein, ESR erythrocyte sedimentation rate

*p < 0.05 **p < 0.01

ROC analysis identified ferritin ≥ 227 ng/mL as the optimal cut-off for predicting thrombosis, with a sensitivity of 71.4%, specificity of 70.9%, and an AUC of 0.686 (95% CI 0.532–0.840, p = 0.026). Ferritin ≥ 227 ng/mL was associated with a 6.1-fold increased risk of thrombosis (OR 6.100, 95% CI 1.749–21.280) (Fig. 1).

Fig. 1.

Receiver operating characteristic (ROC) curve for serum ferritin in predicting thrombosis

In the multivariable Cox regression analysis, ferritin ≥ 227 ng/mL remained independently associated with thrombotic events (HR 3.77, 95% CI 1.13–12.60, p = 0.031), along with LDH (HR 1.002 per unit increase, p = 0.034). The Khorana score was not significant in the multivariate analysis (Table 3).

Table 3.

Univariate and Multivariate Assessments of Risk Factors Associated with Thrombosis

	Thrombosis		Univariable		Multivariable
	No	Yes	HR (95% CI)	p	HR (95% CI)	p
Khorana score
Mean ± SD	1.64 ± 0.65	1.94 ± 0.85	1.875 (0.939–3.743)	0.075
Median (Min-Max)	2 (1–3)	2 (1–4)
HB
Mean ± SD	11.98 ± 2.21	11.20 ± 2.07	0.845 (0.679–1.052)	0.133
Median (Min-Max)	12 (6.6–17)	11.6 (7.5–14.3)
LDH
Mean ± SD	392.56 ± 350.6	544.69 ± 350.82	1.002 (1.000–1.003)	0.010*	1.002 (1.001–1.003)	0.034*
Median (Min-Max)	256 (120–2314)	459.5 (167–1340)
Ferritin
< 227	61 (70.9)	4 (28.6)	Reference		Reference
≥ 227	25 (29.1)	10 (71.4)	5.229 (1.638–16.686)	0.005**	3.769 (1.127–12.602)	0.031*
First-line treatment response
CR	80 (68.4)	8 (50.0)	Reference	0.058
PR	4 (3.4)	2 (12.5)	3.571 (0.758–16.822)	0.108
RR	33 (28.2)	6 (37.5)	3.315 (1.110–9.902)	0.032*

HR hazard ratio, CI confidence interval, LDH lactate dehydrogenase

*p < 0.05 **p < 0.01

In the competing-risk analysis, the cumulative incidence of thrombosis was 6.8% at 3 months, 9.0% at 6 months, and 12.1% at 12 months, after which it plateaued. In Fine–Gray models, ferritin ≥ 227 ng/mL remained significantly associated with thrombosis (subdistribution hazard ratio [sHR] 5.19, 95% CI 1.64–16.40, p = 0.005), as did LDH (sHR 1.001 per unit, 95% CI 1.000–1.002, p = 0.039), whereas the Khorana (sHR 1.87, 95% CI 0.85–4.09, p = 0.117) and ThroLy (sHR 1.16, 95% CI 0.87–1.55, p = 0.312) scores were not. In the multivariable model (ferritin-available subset, n = 100), ferritin remained independently predictive (sHR 4.68, 95% CI 1.32–16.59, p = 0.017), while LDH was attenuated. Cause-specific Cox models yielded concordant estimates (ferritin hazard ratio 5.23, p = 0.005; LDH 1.001, p = 0.012). The cumulative incidence of thrombosis stratified by ferritin is shown in Fig. 2 (≥ 227 ng/mL: 20.0% and 29.0% at 6 and 12 months, versus 4.6% and 6.2% for lower values; Fine–Gray p = 0.005) (Table 4).

Fig. 2.

Cumulative incidence of thrombosis according to serum ferritin (≥ 227 vs. <227ng/mL), with death treated as a competing event (Fine–Gray p = 0.005)

Table 4.

Competing-risk (Fine–Gray) models and comparative performance of the Khorana and ThroLy scores

Variable	sHR/value (95% CI)	p	n
Fine–Gray subdistribution hazard models
Ferritin ≥ 227 ng/mL (univariable)	5.19 (1.64–16.40)	0.005	100
LDH, per unit (univariable)	1.001 (1.000–1.002)	0.039	133
Khorana score (univariable)	1.87 (0.85–4.09)	0.117	133
ThroLy score (univariable)	1.16 (0.87–1.55)	0.312	133
Ferritin ≥ 227 ng/mL (multivariable)	4.68 (1.32–16.59)	0.017	100
LDH, per unit (multivariable)	1.000 (0.999–1.001)	0.443	100
Thrombosis rate by risk-score category
Khorana intermediate (1–2)	13/119 (10.9%)
Khorana high (≥ 3)	3/14 (21.4%)
ThroLy low (0–1)	7/69 (10.1%)
ThroLy intermediate (2–3)	6/47 (12.8%)
ThroLy high (> 3)	3/17 (17.6%)
Discrimination (AUC) and incremental value (DeLong)
Khorana alone	0.561
Khorana + ferritin ≥ 227	0.745 (p = 0.029)
Khorana + ferritin + LDH	0.759 (p = 0.026)
ThroLy alone	0.575
Ferritin ≥ 227 ng/mL	0.712

sHR subdistribution hazard ratio, CI confidence interval, AUC area under the receiver operating characteristic curve, LDH lactate dehydrogenase. Death without thrombosis was treated as the competing event. The multivariable model was restricted to patients with available ferritin (n = 100)

When categorized, thrombosis occurred in 13 of 119 intermediate-risk (10.9%) and 3 of 14 high-risk (21.4%) patients by the Khorana score, and in 10.1%, 12.8%, and 17.6% of low-, intermediate-, and high-risk patients by the ThroLy score, respectively; notably, 7 of the 16 events (44%) occurred in the ThroLy low-risk group. Adding ferritin ≥ 227 ng/mL to the Khorana score significantly improved discrimination, increasing the area under the ROC curve from 0.561 to 0.745 (DeLong p = 0.029), and to 0.759 with the further addition of LDH (p = 0.026). The ThroLy score did not outperform ferritin alone (AUC 0.575 vs. 0.712) (Table 4).

During follow-up, 64 deaths were recorded. Cause of death was lymphoma-related in 45 patients (70.3%), unrelated in 13 (20.3%), and indeterminate in 6 (9.4%); no death was attributable to a thrombotic event. Of the 16 patients with thrombosis, 9 died (6 lymphoma-related, 2 unrelated, 1 indeterminate), with an interval between thrombosis and death ranging from 0.2 to 59.1 months, and in no case was a thrombotic event the proximate cause of death. Overall survival did not differ significantly between patients with and without thrombosis (log-rank p = 0.565) (Table 5); because mortality was driven almost entirely by lymphoma progression, death represented a strong competing event, and the incidence of thrombosis was therefore analyzed within a competing-risk framework.

Table 5.

Survival Analysis According to the Presence of Thrombosis

	N	Deaths	Survivor	Survival Rate	Average Survival Time
Thrombosis (-)	117	55	62	53.0	68.18 ± 5.98
Thrombosis (+)	16	9	7	43.8	46.07 ± 10.71
Overall	133	64	69	51.9	66.96 ± 5.57

Discussion

Globally, particularly in developing countries, the increase in the number of patients diagnosed with lymphoma, along with improvements in treatment options and outcomes, has increased overall survival rates for patients [16, 17]. These developments have made drug side effects experienced during treatment, treatment and disease-related morbidities, and their management more important.

Venous thromboembolism (VTE) is a major cause of morbidity and mortality [18].Although various treatment options are available, VTE may cause delays in the treatment of existing malignancies and prevent the application of optimal treatments, requiring more inpatient and outpatient care. Therefore, preventing VTE and identifying high-risk patients have become increasingly important [9]. The prevention of thrombotic complications in patients with hematologic malignancies remains a topic of ongoing debate. Current international guidelines, including those from the American Society of Clinical Oncology (ASCO) and the International Thrombosis and Cancer Association (ITAC/IACT), do not recommend routine primary thromboprophylaxis for all patients with cancer receiving systemic therapy. Instead, selective thromboprophylaxis is recommended for patients identified as being at high risk for venous thromboembolism (VTE), and the necessity of accurate risk classification strategies has been emphasized [11, 19]. The increased risk of thrombocytopenia in hematological malignancies is one of the biggest challenges in implementing thromboembolism prophylaxis. Although various scoring systems exist for this purpose, they have not yet reached a sufficient or effective level.

To recommend thromboembolism prophylaxis, the relationship between the disease and thrombosis must first be established, and risk factors must be identified. When examining the relationship between cancer and VTE, increased inflammatory activity and consequently cytokines (IL-6, TNF-α) and tissue factors released from tumor cells are implicated [8, 10, 20–24]. Although the development of VTE has been reported at different times in various studies, it is most commonly observed within the first year of diagnosis and during treatment [2, 25–28]. Although the prevalence of VTE in patients with lymphoma has been reported at widely varying rates in different studies, its prevalence in the general population is quite high and has increased significantly in DLBCL compared to other subtypes [29–31]. Central venous catheter (CVC), advanced disease, high grade, high IPI score, mediastinal mass, CNS involvement, bulky disease, high lactate dehydrogenase (LDH), high beta-2 microglobulin (B2MG) level, use of erythropoiesis-stimulating agents in treatment, and active chemotherapy process have been identified as risk factors [2, 11–13, 15, 32–35].

Various risk factors have been established to predict VTE using the defined risk factors. The most well-known of these is the Khorana score [15]. The Khorana score is the most widely used scoring system. It was derived from a multicenter study on chemotherapy-related VTE in patients with cancer and included various types of cancer, with lymphoma constituting only a portion of this cohort. In our study, only three patients who developed thrombosis were defined as high-risk. The ThroLy score incorporates eight clinical and laboratory variables including performance status, BMI, comorbid vascular events, extranodal and mediastinal involvement, neutropenia, and anemia [36]. When the ThroLy score was applied to our cohort, only 17 patients were classified as high-risk and 7 of the 16 thrombotic events (44%) occurred in the low-risk group; the score did not significantly predict thrombosis (subdistribution hazard ratio 1.16, p = 0.312) and did not outperform ferritin. This limited discrimination, consistent with several external validations of ThroLy, highlights the shortcomings of existing scoring systems in predicting thrombosis and indicates that new approaches are still needed.

Simultaneously, when we evaluated the risk factors previously identified in various studies and scoring systems, the absence of any significant difference in terms of white blood cell count, neutrophil count, platelet count, sedimentation rate, CRP, beta-2 microglobulin, albumin, LDH levels, mass disease, high IPI score, and CVC presence also supports our claim.

Ferritin is traditionally considered a marker of iron storage; however, its biological role extends far beyond iron homeostasis. It functions as an acute-phase reactant closely associated with systemic inflammation, macrophage activation, and immune dysregulation [37, 38]. In hematological malignancies, including DLBCL, serum ferritin levels have been shown to correlate with tumor mass and disease activity, and elevated pretreatment ferritin has been identified as an independent adverse prognostic factor for both progression-free survival and overall survival [39–41]. These observations suggest that ferritin in DLBCL does not merely reflect iron stores but serves as a composite marker of tumor burden and systemic inflammatory response.

The mechanistic link between elevated ferritin and thrombosis is multifaceted. Ferritin-mediated iron release can generate reactive oxygen species (ROS) through the Fenton reaction, leading to oxidative endothelial injury, impaired nitric oxide bioavailability, and enhanced platelet adhesion and aggregation [42]. Experimental studies have demonstrated that chronic iron loading significantly accelerates arterial thrombosis through a pro-oxidant mechanism, an effect abrogated by reactive oxygen species scavengers [43]. Furthermore, inflammatory cytokines such as TNF-α — which are characteristically elevated in aggressive lymphomas — promote intracellular iron accumulation in endothelial cells by upregulating iron import proteins and downregulating iron export proteins, thereby amplifying local oxidative stress and contributing to a prothrombotic endothelial phenotype [44]. Elevated ferritin levels are thus associated with endothelial activation, cytokine-induced hypercoagulability, and thromboinflammatory processes [38].

The clinical relevance of hyperferritinemia in thrombotic states has been demonstrated in conditions beyond lymphoma. In COVID-19, elevated ferritin levels at hospital admission have been associated with increased rates of venous thromboembolism and arterial thrombosis, in a pattern consistent with the broader concept of “hyperferritinemic syndromes” — a spectrum that includes hemophagocytic lymphohistiocytosis, macrophage activation syndrome, and catastrophic antiphospholipid syndrome, all characterized by hyperinflammation, immune dysregulation, and a markedly elevated thrombotic risk [45, 46]. The observation that DLBCL itself can trigger macrophage activation and cytokine storm further strengthens the biological plausibility of ferritin as a thrombosis predictor in this population.

A particularly noteworthy finding in our study is that ferritin emerged as an independent predictor of thrombosis despite the absence of significant differences in other established inflammatory markers, including CRP and ESR, between the thrombosis and non-thrombosis groups. This dissociation suggests that ferritin may captures a distinct aspect of thromboinflammatory risk not reflected by conventional acute-phase reactants. Unlike CRP and ESR, which primarily reflect acute systemic inflammation, ferritin at elevated levels in DLBCL likely mirrors a more specific combination of tumor-driven iron dysregulation, macrophage activation, and oxidative endothelial stress. This finding supports the use of ferritin as a complementary, rather than redundant, biomarker in thrombotic risk stratification. In our cohort, a ferritin cutoff of ≥ 227 ng/mL was associated with a 6.1-fold increased risk of thrombosis (OR 6.10, 95% CI 1.75–21.28), and this threshold remained independently predictive in multivariate Cox regression (HR 3.77, 95% CI 1.13–12.60, p = 0.031), further underscoring its clinical utility as a simple, readily available parameter obtainable at the time of diagnosis.

Lactate dehydrogenase (LDH) is a cytoplasmic enzyme that catalyzes the interconversion of pyruvate and lactate in the anaerobic glycolytic pathway and is ubiquitously expressed in virtually all tissues [47]. Elevated serum LDH levels are a recognized marker of tissue injury and increased cellular turnover, and many cancers are associated with a general rise in LDH, making it a widely used non-specific tumor biomarker [47]. In the context of DLBCL, LDH has long been established as one of the most robust independent prognostic factors, and multivariate analyses have confirmed that it is among the strongest predictors of progression-free survival, forming a cornerstone of risk stratification systems including the IPI [48].

Beyond its prognostic value, elevated LDH reflects high tumor burden and accelerated cell turnover, conditions closely linked to the prothrombotic milieu observed in aggressive lymphomas. Lactic acid generated via LDH-mediated pathways modulates interactions between tumor cells and vascular endothelial cells, and studies indicate that lactate enhances HIF-1α stability in tumor-associated endothelial cells, subsequently inducing VEGF expression and facilitating angiogenesis [49]. Furthermore, uptake of lactic acid by vascular endothelial cells triggers the phosphorylation and degradation of IκBα, activates NF-κB, and promotes IL-8 expression, thereby accelerating both angiogenesis and tumor growth [50]. These mechanisms may collectively contribute to endothelial dysfunction and a procoagulant state. Cancer-mediated hypercoagulability involves direct activation of procoagulant pathways through aberrant tissue factor expression by tumor cells, release of tumor-derived microparticles, and indirect systemic effects on leukocytes, endothelial cells, and platelets [8, 10].

In our cohort, LDH was associated with thrombotic events in univariable and Cox regression analyses (HR 1.002 per unit increase, p = 0.034), consistent with its role as a marker of disease aggressiveness and systemic procoagulant burden. This effect is small in absolute terms, corresponding to an approximately 22% relative increase in risk per 100 U/L, and LDH lost statistical significance in the multivariable competing-risk (Fine–Gray) model once ferritin was included. Because LDH is a well-established surrogate of tumor burden and a component of the IPI, its univariable association with thrombosis may partly reflect overall disease aggressiveness rather than a thrombosis-specific mechanism. To our knowledge, few studies have specifically investigated the relationship between LDH and VTE risk in DLBCL. We therefore interpret LDH cautiously, as a hypothesis-generating, complementary marker rather than a robust independent predictor; ferritin was the only parameter that remained significant across Cox, competing-risk, and cause-specific analyses. Prospective studies are needed to clarify whether LDH contributes thrombosis-specific information beyond its established prognostic role.

Conclusion

Thrombotic events in DLBCL patients negatively impact treatment continuity, increase the need for additional interventions, and ultimately worsen clinical outcomes. Despite the clinical significance of this complication, there is currently no standardized diagnostic test or universally accepted risk stratification approach specific to this population. Existing scoring systems, including the widely used Khorana score and the lymphoma-specific ThroLy score, demonstrated limited ability to identify high-risk patients in our cohort, highlighting the ongoing need for more effective predictive tools.

Our study demonstrated that elevated serum ferritin — and, to a lesser extent, LDH — at diagnosis were associated with thrombotic events. Ferritin remained an independent predictor across Cox, competing-risk, and cause-specific models, whereas LDH, although significant in univariable and Cox analyses, was attenuated in the multivariable competing-risk model and is best regarded as a complementary marker of disease burden. These simple, inexpensive parameters, routinely available at diagnosis, may provide thrombotic risk information not fully captured by current clinical scoring systems.

These findings support the potential contribution of ferritin and LDH to lymphoma-specific thrombosis risk scoring frameworks. Prospective multicenter studies with larger DLBCL cohorts are needed to validate these parameters and to define their precise role within a standardized risk stratification model.

Limitations

This study has several limitations. First, the retrospective single-center design introduces inherent selection bias and limits the generalizability of our findings to broader DLBCL populations. Second, the thrombotic event group was relatively small (n = 16), which may have reduced the statistical power of the multivariate models and warrants cautious interpretation of the identified independent predictors. Third, ferritin levels were not available for all patients (100 of 133), as ferritin was not routinely measured during the earlier years of the study period; this missing data may have introduced bias in the ferritin-based analyses and potentially underestimated the true association. Fourth, data on anticoagulant prophylaxis use prior to or during follow-up were not systematically recorded, which may have influenced the observed thrombosis rates. Fifth, the lack of an external validation cohort limits the generalizability of the proposed ferritin cutoff value and regression models. These limitations notwithstanding, this study contributes novel data on the predictive value of ferritin and LDH in DLBCL-associated thrombosis.

Authors’ contributions

Material preparation and data collection were performed by B.S., M. A., A.Y., P.T., M.T., M.R.A, F.Y., and H.B.A.Ö, Statistical analysis was performed by B.S. and M.A. The first draft of the manuscript was written by B.S., and all authors commented on previous versions of the manuscript. All authors read and approved the final manuscript.

Funding

This research received no specific grants from funding agencies in the public, commercial, or not-for-profit sectors.

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethics approval

This retrospective study was approved by the Clinical Research Ethics Committee of Ankara Dışkapı Yıldırım Beyazıt Training and Research Hospital (approval number 107/30; date 22 March 2021). All procedures were performed in accordance with the 1964 Declaration of Helsinki and its later amendments.

Consent to participate

Owing to the retrospective design of the study, the requirement for informed consent was waived by the ethics committee.

Consent to publish

Not applicable. The manuscript contains no individual identifying data.

Competing interests

The authors declare no competing interests.