Early risk stratification of acute postoperative deep vein thrombosis with inflammatory and hemostatic molecular markers in Chinese patients undergoing unicompartmental knee arthroplasty

Abstract

Background

Deep vein thrombosis (DVT) is a feared postoperative complication following knee arthroplasty. However, the incidence of acute DVT, associated early risk stratification models, and optimal thromboprophylaxis after unicompartmental knee arthroplasty (UKA) remain unclear. This study aimed to comprehensively evaluate the perioperative dynamics of systemic-inflammatory and hemostatic molecular markers in patients with knee osteoarthritis undergoing UKA and to analyze their association with the occurrence of acute DVT.

Methods

The first multi-center prospective study enrolled consecutive patients undergoing unilateral UKA without routine postoperative pharmacological thromboprophylaxis from three institutions (two for derivation and one for external validation). Systemic-inflammatory and hemostatic molecular markers were measured at admission, 2-hour, 1-day, and 3-day post-UKA. Lower-extremity ultrasound was routinely conducted prior to and at 4–5 days following UKA to detect asymptomatic and symptomatic DVT. Temporal trends in these biomarkers and their relationship with acute DVT were analyzed. A nomogram was then developed to visualize individualized near-term risk stratification of acute postoperative DVT. Finally, a retrospective cohort of UKA-treated patients who received routine thromboprophylaxis was enrolled to evaluate the risk stratification performance of the identified indicators in an anticoagulated population.

Results

The derivation cohort included 87 patients (mean age 67.8 ± 6.5 years, 78.2% female), of whom 21 (24.1%) developed acute postoperative DVT (20 asymptomatic and 1 symptomatic). Multivariate Firth’s penalized logistic regression identified that higher body mass index (BMI), and elevated postoperative day 3 levels of interleukin (IL)-6, thrombin-antithrombin complex (TAT), and plasmin-α2 plasmin inhibitor complex (PIC) independently associated with an increased risk of acute DVT detected at 4–5 days post-UKA. A nomogram incorporating these four indicators demonstrated good discriminative ability, calibration, and exploratory net benefit. These findings were further validated in an independent external cohort (n = 42; mean age 67.2 ± 7.7 years; 69% female; DVT incidence, 23.8%) and in a separate cohort of 135 patients (mean age 66.4 ± 6.8 years; 77% female; DVT incidence, 20.7%) who received routine thromboprophylaxis.

Conclusion

A preliminary nomogram incorporating BMI and postoperative day 3 levels of IL-6, TAT, and PIC serves as a promising auxiliary tool for near-term risk stratification of acute DVT detected at 4–5 days following UKA.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12893-026-03964-3.

Background

Deep vein thrombosis (DVT) and its sequelae, including pulmonary embolism (PE) and post-thrombotic syndrome, are among the most frequent and severe complications following knee arthroplasty, contributing substantially to postoperative morbidity and mortality [1–4]. In recent years, unicompartmental knee arthroplasty (UKA) has been increasingly adopted in clinical practice due to its minimally invasive nature, shorter hospital stays, and faster recovery compared with total knee arthroplasty (TKA) [5–7]. While routine pharmacological thromboprophylaxis is standard after TKA given the well-recognized high risk of postoperative DVT, this strategy raises concerns regarding bleeding risk and wound-related complications [8–10]. In contrast, the incidence of symptomatic DVT appears to be lower in patients undergoing UKA; however, evidence to guide optimal thromboprophylaxis in this population remains limited [4, 11–14]. Specifically, several studies have reported a lower overall incidence of venous thromboembolism (VTE) in Asian populations after UKA, raising questions about the necessity of routine pharmacological thromboprophylaxis, particularly in patients at low risk of DVT [15–17]. For example, a Korean study of patients undergoing UKA without pharmacological thromboprophylaxis reported newly detected VTE (including DVT and PE) in 18 of 70 (25.7%) cases. Notably, all thromboembolic lesions remained asymptomatic and completely resolved regardless of their location during the six‑month follow‑up period [18]. Consequently, key clinical questions persist: Should routine pharmacological thromboprophylaxis be applied to low-risk UKA patients as in TKA? Can early markers be identified to guide individualized anticoagulation strategies? Currently, however, reliable laboratory measures for the early risk stratification of acute postoperative DVT in patients undergoing UKA are still lacking.

Orthopedic surgeries provoke venous thrombosis through Virchow’s triad of endothelial damage, venous stasis, and hypercoagulability. Although traditionally considered separately, emerging evidence has revealed a complex interplay between inflammation and thrombosis, known as immunothrombosis [19–21]. Aberrant activation of immunothrombosis following knee arthroplasty may significantly contribute to the development of acute postoperative DVT [19–21]. Recent studies have explored the utility of complete blood count (CBC)-derived systemic-inflammatory markers, such as neutrophil-to-lymphocyte ratio (NLR) and systemic immune inflammation index (SII; platelet × NLR), in predicting the risk of arterial thrombotic events (e.g., ischemic stroke and myocardial infarction) and VTE, including DVT and PE [22–27]. Furthermore, novel hemostatic molecular markers, including thrombin-antithrombin complex (TAT), plasmin-α2 plasmin inhibitor complex (PIC), tissue plasminogen activator-plasminogen activator inhibitor complex (t-PAIC), and soluble thrombomodulin (sTM), are increasingly investigated as early indicators of coagulation and fibrinolysis activation as well as endothelial injury, and thus as potential predictors of arterial or venous thrombosis [28–34]. While some of these markers have been studied in the setting of TKA [25–29], their perioperative dynamics and utility for early risk stratification for acute postoperative DVT in patients undergoing UKA remain unexplored.

To address this gap, we first conducted a multi-center prospective study involving patients with knee osteoarthritis who underwent unilateral UKA and did not receive routine postoperative pharmacological thromboprophylaxis. Our primary aims were (1) to comprehensively characterize the temporal profiles of systemic-inflammatory and hemostatic molecular markers in the perioperative period (at admission, 2-hour, 1-day, and 3-day post-UKA), (2) to evaluate the near-term risk stratification value of these prespecified biomarkers for acute DVT detected at 4–5 days following UKA, and (3) to perform an exploratory analysis to preliminarily develop a nomogram for visualizing individualized near-term risk stratification for acute postoperative DVT. Additionally, we retrospectively enrolled a separate cohort of UKA-treated patients who received routine pharmacological thromboprophylaxis, with their blood samples collected prospectively. The secondary aims were: (1) to compare the incidence of acute DVT between low-risk patients with and without routine pharmacological thromboprophylaxis following UKA, and (2) to evaluate the near-term risk stratification performance of the identified indicators and the nomogram in the anticoagulated population.

Materials and methods

Study design and participants

In the first multi-center prospective study, consecutive patients with knee osteoarthritis scheduled for unilateral UKA were enrolled from three institutions (two for derivation and one for external validation) (Fig. 1). The derivation cohort comprised patients recruited from Xuanwu Hospital, Capital Medical University (Beijing, China) and Beijing Jishuitan Hospital, Capital Medical University (Beijing, China) between March 2024 and December 2024. This cohort was used to analyze the perioperative profiles of systemic-inflammatory and hemostatic molecular markers and to identify potential biomarkers for acute postoperative asymptomatic and symptomatic DVT. An independent external validation cohort was subsequently established from patients screened at Tianjin Hospital (Tianjin, China) between February 2025 and June 2025 to verify the near-term risk stratification performance of the identified indicators. Patients in both the derivation and external validation cohorts were excluded if they had high-risk factors for DVT [35, 36]. These patients did not receive routine postoperative pharmacological thromboprophylaxis. Therapeutic anticoagulation was initiated only when symptomatic or asymptomatic DVT was detected on postoperative ultrasonography examination, regardless of its location.

Fig. 1.

Flowchart of patient selection for the derivation cohort and external validation cohort. Abbreviations: DVT, deep vein thrombosis; UKA, unicompartmental knee arthroplasty; VTE, venous thromboembolism

In the second retrospective cohort, consecutive patients who underwent unilateral UKA and received routine thromboprophylaxis were retrospectively screened from Xuanwu Hospital, Capital Medical University (Beijing, China) between March 2023 and November 2023. Their blood samples had been prospectively collected and maintained in our institutional biobanks.

The inclusion criteria for all enrolled patients were: (1) aged ≥ 18 years; (2) knee osteoarthritis eligible for unilateral UKA. Exclusion criteria were: (1) a history of lower extremity surgery; (2) a personal or family history of VTE or thrombophilia; (3) severe cardiac, hepatic or renal failure; (4) malignant tumors; (5) rheumatoid arthritis, gout, or concurrent/recent infection; and (6) unavailable blood samples or incomplete clinical data. In the absence of explicit guideline recommendations, the decision to withhold routine postoperative pharmacological thromboprophylaxis did not deviate from the local standard of care for low-risk patients undergoing UKA. The study protocol was approved by the Ethics Committee of Xuanwu Hospital, Capital Medical University (ID-CRB-062/2021) and was conducted according to the principles of Declaration of Helsinki. Written consent was obtained from patients or their legal representatives.

Surgical procedure and postoperative management

All patients underwent the standard Oxford UKA procedure under intraspinal anesthesia, following uniform perioperative protocols as previously described [37, 38]. Briefly, antiplatelet agents were discontinued 7 days before surgery, and anticoagulant medications were withheld for 24 h. Tranexamic acid (1 g) was administered intravenously 30 min before surgery initiation to minimize bleeding. A tourniquet inflated to systolic blood pressure plus 100 mmHg was applied from skin incision until prosthesis cementation. The procedure was performed through a medial parapatellar approach using the cemented Oxford Phase III medial UKA prosthesis (two-peg design; MP instrumentation; Biomet UK Ltd). Before wound closure, a single intra-articular injection of long-acting betamethasone (DIPROSPAN^®; 1 mL, containing 5 mg betamethasone dipropionate and 2 mg betamethasone sodium phosphate) was administered. No surgical drain was placed postoperatively. Patients were encouraged to begin ankle flexion and quadriceps contraction exercises as early as possible. Full weight-bearing ambulation with crutches was initiated four hours after surgery. A lower-limb intermittent pneumatic compression device was applied on the operative night and removed the following morning. Intravenous analgesia was administered for the first three postoperative days. In the first multi-center prospective study, no routine pharmacological thromboprophylaxis was administered. Therapeutic anticoagulation with low‑molecular‑weight heparin (LMWH; enoxaparin sodium injection, 0.4 mL: 4000 AxaIU, twice daily) was administered only when asymptomatic or symptomatic DVT was confirmed by lower‑extremity vascular ultrasonography on postoperative day (POD) 4–5. In the second retrospective cohort, all patients received routine prophylactic anticoagulant therapy with enoxaparin sodium injection (0.4 mL: 4000 AxaIU, once daily). For patients in whom asymptomatic or symptomatic DVT was confirmed, the LMWH dose was escalated to the therapeutic regimen (0.4 mL: 4000 AxaIU, twice daily).

Clinical and outcome data collection

We collected the following clinical data: (1) age, sex, and body mass index (BMI); (2) medical history, including hypertension, diabetes mellitus, dyslipidemia, coronary heart disease, prior ischemic stroke, varicose veins; (3) smoking and drinking habits; (4) surgery-related variables, including operation time, tourniquet time, and intraoperative blood loss; (5) postoperative adverse events, including bleeding‑related events (anemia due to blood loss or need for transfusion), wound‑related complications (dehiscence or exudate suspicious for infection), and infectious events [superficial/deep incisional surgical‑site infection and periprosthetic joint infection (PJI)] [12, 39]; and (6) lower limb DVT. To diagnose both asymptomatic and symptomatic DVT, color duplex ultrasonography examinations of the lower limb veins, including the iliac, femoral, popliteal, and calf muscular veins, were conducted upon hospital admission and on POD4-5 by experienced radiologists who were blinded to the clinical data, in accordance with established diagnostic criteria [40]. DVT was classified as either proximal (involving the popliteal vein or any vein proximal to it) or distal (involving any vein distal to the popliteal vein).

Systemic-inflammatory markers

Peripheral venous blood samples were routinely collected upon admission, and at 2-hour, 1-day, and 3-day post-UKA, to measure CBC, C-reactive protein (CRP), and interleukin-6 (IL-6) in the clinical laboratory. Absolute counts of neutrophils, lymphocytes, monocytes, and platelets were subsequently used to calculate NLR (neutrophil/lymphocyte), SIRI (systemic inflammation response index; monocyte × NLR), SII (platelet × NLR), and AISI [aggregate inflammation systemic index; (neutrophil × monocyte × platelet)/lymphocyte].

Hemostatic molecular markers

At each tested time point, peripheral venous blood samples were collected into sodium citrate tubes (final concentration 0.32%) and centrifuged at 1500g for 15 min, followed by a second centrifugation of the supernatant at 13000g for 3 min to obtain cell-free plasma. TAT, PIC, t-PAIC, and sTM were measured using a microparticle chemiluminescent immunoassay performed on an automatic chemiluminescence analyzer (MI600; Pushkang Biotech, Shaoxing, China), utilizing specific detection kits (Pushkang Biotech). The international normalized ratio (INR) and D-dimer levels, measured as part of routine clinical care, were obtained from medical records.

Statistical analysis

Statistical analysis was performed using R software (version 4.3.1; The R Foundation) and SPSS (version 26.0; IBM Corporation, Armonk, NY, USA). Quantitative variables were reported as mean (standard deviation, SD) or median (inter quartile range, IQR) based on the Kolmogorov-Smirnov test. Intergroup differences were analyzed using Student’s t-test or Mann-Whitney U test, as appropriate. Categorical data were presented as count (percentage), and intergroup differences were analyzed by Chi-square test or Fisher’s exact test, as appropriate. To analyze the perioperative profiles of systemic‑inflammatory and hemostatic molecular markers, and to compare their profiles between patients with and without DVT, we employed generalized estimating equations (GEE) with a Gaussian distribution, an identity link function, and an exchangeable working correlation structure. The statistical significances of the main effects of time and group, and their interaction, were assessed using the Wald χ² test. For multiple comparisons of biomarker levels at different time points between groups, p‑values were adjusted using the Bonferroni method. Associations between indicators and acute postoperative DVT were analyzed using multivariate Firth’s penalized logistic regression analysis adjusting for potential confounders (p < 0.05 in univariate Firth’s penalized logistic regression analysis). Based on the logistic regression results, a nomogram was subsequently constructed to visualize individualized near-term risk stratification for acute postoperative DVT. The discriminative ability of the independent indicators and the nomogram was evaluated by generating receiver operating characteristic (ROC) curves and calculating the area under the curve (AUC). Their optimal cutoff values were determined using the Youden index, which maximized the sum of sensitivity and specificity for discriminating acute postoperative DVT. Model calibration was assessed using calibration curves. Calibration intercept and slope were derived from standard logistic recalibration (i.e., logistic regression of observed outcomes on predicted logits), with 95% confidence intervals (CIs) estimated by percentile bootstrap using 1000 resamples. Furthermore, clinical net benefit was examined via decision curve analysis (DCA). The nomogram developed in the derivation cohort was externally validated in the external validation and the retrospective anticoagulation cohorts. In both cohorts, predicted probabilities were calculated directly using the original regression coefficients, without recalibration, coefficient updating, or model refitting. All statistical tests were two-tailed, and a p-value of < 0.05 was considered statistically significant.

Results

Baseline characteristics of the derivation cohort

The derivation cohort included 87 patients (mean age 67.8 ± 6.5 years; 78.2% female) with knee osteoarthritis who underwent unilateral UKA (Fig. 1). These patients received only postoperative mechanical thromboprophylaxis and did not receive routine pharmacological thromboprophylaxis. Among the postoperative adverse events assessed within 4–5 days after UKA, only anemia due to blood loss (6.9%; none required transfusion) and wound exudate (5.7%) occurred. Of these patients, 21 (24.1%) developed acute DVT following UKA, including 20 asymptomatic cases (2 proximal and 18 distal) and one symptomatic distal case. All 21 patients subsequently received therapeutic anticoagulation. During the 3‑week follow‑up period, 2 patients with asymptomatic distal DVT developed increased but self-limited wound exudate; no patient required blood transfusion or reoperation for hemostasis (Supplementary Table 1). Compared with patients without DVT, those who developed DVT had a significantly higher BMI (p = 0.022) and a higher prevalence of diabetes mellitus (p = 0.045), whereas no significant differences were observed in other variables. Detailed demographic and clinical characteristics of the cohort are summarized in Table 1.

Table 1.

Baseline characteristics of the derivation cohort (n = 87)

Characteristics	Total (n = 87)	DVT
		Yes (n = 21)	No (n = 66)	P
Age, year, mean (SD)	67.8 (6.5)	67.1 (4.0)	68 (7.1)	0.492a
Gender, male/female, n	19/68	5/16	14/52	0.770b
BMI, kg/m2, mean (SD)	24.2 (3.0)	25.3 (2.3)	23.8 (3.1)	0.022a
Risk factors, n (%)
Hypertension	60 (69.0)	15 (71.4)	45 (68.2)	0.993c
Diabetes mellitus	28 (32.2)	11 (52.4)	17 (25.8)	0.045c
Dyslipidemia	46 (52.9)	9 (42.9)	37 (56.1)	0.421c
Coronary artery disease	14 (16.1)	3 (14.3)	11 (16.7)	1.000b
Stroke history	6 (6.9)	2 (9.5)	4 (6.1)	0.628b
Varicose veins	4 (4.6)	2 (9.5)	2 (3.0)	0.244b
Smoking	17 (19.5)	5 (23.8)	12 (18.2)	0.544b
Drinking	10 (11.5)	2 (9.5)	8 (12.1)	1.000b
Procedure-related variables, median (IQR)
Operation time, min,	74 (68-99.5)	74 (70–88)	73 (68-104.5)	0.666d
Tourniquet time, min	63 (57–70)	67 (57–71)	62 (57–68)	0.140d
Intraoperative blood loss, mL	44 (36.5–50)	40 (36–48)	44 (37–50)	0.937d
Postoperative adverse events, n (%)
Anemia due to blood loss	6 (6.9)	3 (14.3)	3 (4.6)	0.149b
Wound exudate	5 (5.7)	3 (14.3)	2 (3.0)	0.088b

^aAnalysed by Student’s t test

^bAnalysed by Fisher’s exact test

^cAnalysed by Chi-square test

^dAnalysed by Mann-Whitney U test

Abbreviation: BMI Body mass index, DVT Deep venous thrombosis, IQR Inter quartile range, SD Standard deviation

Perioperative profiles of systemic-inflammatory markers

At admission, none of the systemic-inflammatory markers differed significantly between the two groups (all p > 0.05; Fig. 2A-C). Postoperatively, levels of neutrophil count, SIRI, AISI, and IL-6 showed a sustained increase at 2-hour and on POD1, then decreased by POD3 (all p_time <0.001; Fig. 2A-C). NLR and SII levels increased at 2-hour, remained elevated on POD1, and declined by POD3 (both p_time <0.001; Fig. 2B). CRP levels exhibited a gradual increase over the first 3 postoperative days (p_time <0.001; Fig. 2C). Monocyte and lymphocyte counts dropped rapidly at 2-hour after surgery and recovered within 3 days (both p_time <0.001), whereas platelet count (p_time =0.084) remained stable throughout the first 3 perioperative days (Fig. 2A). Among all markers examined, only the time courses of CRP and IL-6 differed significantly between patients with and without DVT (interaction effect: CRP, p_interaction =0.032; IL-6, p_interaction =0.001; Fig. 2C). Subsequent analysis at individual time points revealed that patients who developed DVT exhibited higher CRP and IL‑6 levels than those without DVT on POD3 (CRP, p = 0.013; IL-6, p = 0.001). Detailed data for these results are provided in Supplementary Table 2.

Fig. 2.

Dynamic changes in levels of CBC counts (A), CBC-derived systemic-inflammatory markers (B), CRP and IL-6 (C) in patients with DVT compared to those without DVT during the perioperative period of UKA. P_time indicated the significance of the individual effect of time, while P_interaction denoted the significance of the interaction effect between time and outcome, as determined by the generalized estimating equations. Abbreviations: AISI, aggregate inflammation systemic index; CBC, complete blood count; CRP, C-reactive protein; DVT, deep vein thrombosis; IL-6, interleukin-6; NLR, neutrophil-to-lymphocyte ratio; SII, systemic immune inflammation index; SIRI, systemic inflammation response index; UKA, unicompartmental knee arthroplasty. *p < 0.05 and **p < 0.01

Perioperative profiles of hemostatic molecular markers

At admission, there were no significant intergroup differences in hemostatic molecular markers (Fig. 3A-B). After UKA, levels of D-dimer (p_time <0.001) and t-PAIC (p_time =0.019) increased persistently at 2-hour and on POD1, then declined by POD3 (Fig. 3A-B). INR, TAT, and PIC levels rose rapidly at 2-hour post-UKA, and returned toward baseline within 3 days (all p_time <0.001; Fig. 3A-B). Level of sTM remained stable throughout the perioperative period (p_time =0.093; Fig. 3B). The overall time courses of D‑dimer, TAT, and PIC differed significantly between the two patient groups (interaction effect: D-dimer, p_interaction =0.043; TAT, p_interaction =0.046; PIC, p_interaction =0.013; Fig. 3A-B). Analysis at each time point revealed that, compared to patients without DVT, those who developed DVT had significantly higher levels of D-dimer (p = 0.004), TAT (p < 0.001), and PIC (p < 0.001) on POD3 (Fig. 3A-B). Detailed data are shown in Supplementary Table 3.

Fig. 3.

Dynamic changes in levels of hemostatic molecular markers in patients with DVT compared to those without DVT during the perioperative period of UKA. P_time indicated the significance of the individual effect of time, while P_interaction denoted the significance of the interaction effect between time and outcome, as determined by the generalized estimating equations Abbreviations: DVT, deep vein thrombosis; IL-6, interleukin-6; INR, international normalized ratio; PIC, plasmin-α2 plasmin inhibitor complex; sTM, soluble thrombomodulin; TAT, thrombin-antithrombin complex; t-PAIC, tissue plasminogen activator-plasminogen activator inhibitor complex; UKA, unicompartmental knee arthroplasty. **p < 0.01 and ***p < 0.001

Associations between systemic-inflammatory and hemostatic molecular markers and acute DVT detected at 4–5 days post-UKA

Logistic regression analysis was performed to evaluate the potential of BMI, history of diabetes mellitus, and levels of POD3 CRP, IL‑6, D-dimer, TAT, and PIC—all of which had showed significant differences in prior analyses (Figs. 2 and 3, and Table 1)—for near-term risk stratification of acute DVT detected at 4–5 days post-UKA. Univariate Firth’s penalized logistic regression analysis showed that patients who developed acute DVT had significantly higher BMI [odds ratio (OR), 1.18; 95%CI, 1.00-1.41; p = 0.047], a greater prevalence of diabetes mellitus (OR, 3.10; 95%CI, 1.14–8.57; p = 0.026), and elevated levels of POD3 CRP (OR, 1.03; 95%CI, 1.01–1.06; p = 0.009), IL-6 (OR, 1.12; 95%CI, 1.05–1.20; p < 0.001), D-dimer (OR, 5.19; 95%CI, 1.69–17.77; p = 0.004), TAT (OR, 1.24; 95% CI, 1.12–1.41; p < 0.001), and PIC (OR, 16.17; 95%CI, 3.72–86.85; p < 0.001) (Table 2). After adjustment, multivariate logistic regression revealed that BMI [adjusted OR (aOR) per 1-kg/m² increase, 1.38; 95%CI, 1.08–1.86; p = 0.008], POD3 IL-6 (aOR per 1-pg/mL increase, 1.11; 95%CI, 1.01–1.24; p = 0.024), TAT (aOR per 1-ng/mL increase, 1.21; 95%CI, 1.02–1.47; p = 0.025), and PIC (aOR per 1-µg/mL increase, 8.21; 95%CI, 1.02–86.58; p = 0.048) remained independently associated with an increased risk of acute postoperative DVT. We further performed sensitivity analysis using standardized Z‑scores for biomarkers in both univariate and multivariate Firth’s penalized logistic regression models. As expected for linear transformations in generalized linear models, the p-values were identical to those in the primary analysis, whereas the ORs for biomarkers were rescaled to reflect associations per 1-SD increase. In the multivariate model, the standardized aORs for POD3 IL-6, TAT, and PIC were 2.67 (95%CI, 1.13–8.31; p = 0.024), 2.64 (95%CI, 1.13–7.23; p = 0.025), and 2.09 (95%CI, 1.01–4.77; p = 0.048) per 1-SD increase, respectively (Supplementary Table 4). The AUCs for POD3 IL-6, TAT, and PIC in discriminating acute DVT were 0.756 (95%CI, 0.627–0.886; sensitivity, 57.1%; specificity, 87.9%; cutoff, 25.44 pg/mL), 0.791 (95%CI, 0.688–0.895; sensitivity, 85.7%; specificity,60.6%; cutoff, 9.5 ng/mL), and 0.764 (95%CI, 0.652–0.877; sensitivity, 81%; specificity,68.2%; cutoff, 0.78 µg/mL), respectively (Fig. 4A).

Table 2.

Associations between systemic-inflammatory and hemostatic molecular markers and DVT occurrence

	Univariate		Multivariate
	OR (95%CI)	P	OR (95%CI)	P
BMI, kg/m2	1.18 (1.00-1.41)	0.047	1.38 (1.08–1.86)	0.009
Diabetes mellitus, yes/no	3.10 (1.14–8.57)	0.026	2.18 (0.49–10.62)	0.301
CRP (day 3 post-UKA), mg/L	1.03 (1.01–1.06)	0.009	1.01 (0.97–1.05)	0.687
IL-6 (day 3 post-UKA), pg/mL	1.12 (1.05–1.20)	< 0.001	1.11 (1.01–1.24)	0.024
D-dimer (day 3 post-UKA), µg/mL	5.19 (1.69–17.77)	0.004	2.51 (0.49–14.83)	0.273
TAT (day 3 post-UKA), ng/mL	1.24 (1.12–1.41)	< 0.001	1.21 (1.02–1.47)	0.025
PIC (day 3 post-UKA), µg/mL	16.17 (3.72–86.85)	< 0.001	8.21 (1.02–86.58)	0.048

Variables with p < 0.05 in univariate Firth’s penalized logistic regression were included in multivariate Firth’s penalized logistic regression

Abbreviation: BMI Body mass index, CI Confidence interval, CRP C-reactive protein, DVT Deep vein thrombosis, IL-6 Interleukin-6, OR Odds ratio, PIC Plasmin-α2 plasmin inhibitor complex, TAT Thrombin-antithrombin complex, UKA Unicompartmental knee arthroplasty

Fig. 4.

Discriminative ability of POD3 IL-6, TAT, and PIC levels and a nomogram model for assessing the risk of acute DVT detected on POD4-5 following UKA in derivation cohort. A ROC curve analysis for IL-6, TAT, and PIC levels measured on POD3 in relation to acute DVT identified on POD4-5. B Development of a nomogram model incorporating BMI and POD3 IL-6, TAT, and PIC levels. C ROC curve analysis of the nomogram for its discriminative ability regarding acute DVT risk. D Calibration curve analysis of the nomogram model. E DCA analysis comparing the near‑term risk stratification performance of POD3 IL-6, TAT, and PIC levels, and the nomogram model for acute DVT detected on POD4-5. Abbreviations: AUC, area under the curve; BMI, body mass index; CI, confidence interval; DCA, decision curve analysis; DVT, deep vein thrombosis; IL-6, interleukin-6; PIC, plasmin-α2 plasmin inhibitor complex; ROC, receiver operating characteristic; TAT, thrombin-antithrombin complex; UKA, unicompartmental knee arthroplasty

Development of a preliminary nomogram for near-term risk stratification of acute DVT detected at 4–5 days post-UKA

A multivariable logistic regression model was developed to discriminate acute DVT detected at 4–5 days post-UKA based on the four indicators, incorporating BMI, and POD3 IL-6, TAT, and PIC. The final model equation was formulated as follows: Logit(P) = − 20.0352 + 0.410×BMI + 0.151×IL-6 + 3.144×PIC + 0.218×TAT, where P denoted the probability of acute DVT occurrence, calculated as P = 1/[1 + exp(–Logit(P))]. To facilitate near-term clinical risk stratification, a nomogram based on this equation was further constructed (Fig. 4B). Higher total scores on the nomogram corresponded to an increased risk of DVT, while lower scores indicated a decreased risk. ROC analysis demonstrated that the nomogram discriminated acute DVT with an AUC of 0.921 (95%CI, 0.854–0.988), a sensitivity of 76.2%, and a specificity of 93.9% (Fig. 4C). Calibration curve analysis showed close agreement between predicted and observed probabilities (Hosmer-Lemeshow test; χ² = 10.7, p = 0.221; calibration intercept = 0.000, 95%CI, -0.742-0.703; calibration slope = 1.000, 95%CI, 0.612–1.544; Brier score = 0.084; Fig. 4D), indicating good model calibration. DCA curves further revealed that, across a DVT probability threshold range of 0.10 to 0.99, the nomogram yielded better net benefits compared with “all (assuming DVT occurred in all patients)”, “none (assuming none of the patients developed DVT)”, and any single indicator alone, suggesting an exploratory net benefit for discriminating acute DVT in patients with knee osteoarthritis who underwent unilateral UKA (Fig. 4E).

Validation of the individual indicators and the nomogram for DVT risk stratification in an independent external cohort

The near-term risk stratification performance of the individual indicators and the nomogram was further evaluated in an independent external validation cohort comprising 42 patients (mean age 67.2 ± 7.7 years; 69% female) (Fig. 1). Among them, 10 patients (23.8%) developed acute asymptomatic DVT (1 proximal and 9 distal) after UKA; no symptomatic DVT occurred. These patients subsequently received therapeutic anticoagulation, and no bleeding or wound events occurred during the 3‑week follow‑up period (Supplementary Table 1). Baseline characteristics did not differ significantly between groups, except that patients who developed acute DVT had a significantly higher BMI (p = 0.036). Detailed demographic and clinical characteristics of the cohort are summarized in Table 3.

Table 3.

Baseline characteristics of the external validation cohort (n = 42)

Characteristics	Total (n = 42)	DVT
		Yes (n = 10)	No (n = 32)	P
Age, year, mean (SD)	67.2 (7.7)	68.8 (5.1)	66.7 (8.4)	0.344a
Gender, male/female, n	13/29	3/7	10/22	1.000b
BMI, kg/m2, mean (SD)	23.4 (2.7)	24.9 (2.3)	22.9 (2.6)	0.036a
Risk factors, n (%)
Hypertension	28 (66.7)	7 (70.0)	21 (65.6)	1.000b
Diabetes mellitus	16 (38.1)	6 (60.0)	10 (31.3)	0.142b
Dyslipidemia	23 (54.8)	4 (40.0)	19 (59.4)	0.468b
Coronary artery disease	5 (11.9)	1 (10.0)	4 (12.5)	1.000b
Stroke history	3 (7.1)	1 (10.0)	2 (6.3)	1.000b
Smoking	10 (23.8)	3 (30.0)	7 (21.9)	0.678b
Drinking	9 (21.4)	2 (20.0)	7 (21.9)	1.000b
Procedure-related variables, median (IQR)
Operation time, min	77.5 (66.5–92)	80 (72-89.5)	75 (66–93)	0.515c
Tourniquet time, min	65 (57.25–71.75)	66 (58–68)	64 (57.75–72.5)	0.668c
Intraoperative blood loss, mL	42 (37–54)	40.5 (38.5–55)	42 (36.75-54)	0.836c
Postoperative adverse events, n (%)
Anemia due to blood loss	3 (7.1)	1 (10.0)	2 (6.3)	1.000b
Wound exudate	5 (11.9)	2 (20.0)	3 (9.4)	0.557b
Biomarkers, median (IQR)
IL-6 (day 3 post-UKA), pg/mL	15.52 (12.86–21.41)	21.96 (15.57–28.8)	15.2 (12.67–19.5)	0.042c
TAT (day 3 post-UKA), ng/mL	10.12 (7.30-14.81)	14.7 (11.11–16.38)	9.05 (6.6-13.04)	0.020c
PIC (day 3 post-UKA), µg/mL	0.76 (0.50–0.98)	0.97 (0.77–1.09)	0.71 (0.47–0.95)	0.035c

^aAnalysed by Student’s t test

^bAnalysed by Fisher’s exact test

^cAnalysed by Mann-Whitney U test

Abbreviation: BMI Body mass index, DVT Deep venous thrombosis, IL-6 Interleukin-6, IQR Inter quartile range, PIC Plasmin-α2 plasmin inhibitor complex, SD Standard deviation, TAT Thrombin-antithrombin complex, UKA Unicompartmental knee arthroplasty

In this cohort, the AUCs for individual biomarkers POD3 IL-6, TAT, and PIC in discriminating acute DVT were as follows: 0.717 (95%CI, 0.518–0.915; sensitivity, 50%; specificity, 90.6%; cutoff value, 23.56 pg/mL), 0.746 (95%CI, 0.582–0.910; sensitivity, 80%; specificity, 62.5%; cutoff value, 10.8 ng/mL), and 0.725 (95%CI, 0.552–0.897; sensitivity, 80%; specificity, 59.4%; cutoff value, 0.76 µg/mL), respectively (Fig. 5A). The nomogram maintained good near-term risk stratification performance, with an AUC of 0.887 (95%CI, 0.773-1.000; sensitivity, 70.0%; specificity, 93.8%; Fig. 5B). In calibration curve analysis, the calibration slope was close to unity (slope = 1.007, 95%CI, 0.427–1.983), suggesting no major evidence of overfitting or underfitting, although the confidence interval was wide (Hosmer-Lemeshow test: χ² =11, p = 0.199; Brier score = 0.116; Fig. 5C). However, the calibration intercept was 0.711 (95%CI, -0.238-1.59), suggesting a potential tendency toward systematic underestimation of risk, although the confidence interval included zero. Visual inspection of the calibration curve (Fig. 5C) also showed some deviation from the ideal 45° line, consistent with this potential miscalibration. In DCA, the nomogram still yielded a good exploratory net benefit (Fig. 5D).

Fig. 5.

Discriminative ability of POD3 IL-6, TAT, and PIC levels and a nomogram model for assessing the risk of acute DVT detected on POD4-5 following UKA in external validation cohort. A ROC curve analysis for IL-6, TAT, and PIC levels measured on POD3 in relation to acute DVT identified on POD4-5. B ROC curve analysis of the nomogram for its discriminative ability regarding acute DVT risk. C Calibration curve analysis of the nomogram model. D DCA analysis comparing the near‑term risk stratification performance of POD3 IL-6, TAT, and PIC levels, and the nomogram model for acute DVT detected on POD4-5. Abbreviations: AUC, area under the curve; CI, confidence interval; DCA, decision curve analysis; DVT, deep vein thrombosis; IL-6, interleukin-6; PIC, plasmin-α2 plasmin inhibitor complex; ROC, receiver operating characteristic; TAT, thrombin-antithrombin complex; UKA, unicompartmental knee arthroplasty

Validation of the individual indicators and the nomogram for DVT risk stratification in UKA patients who received routine thromboprophylaxis

Finally, we retrospectively enrolled a separate cohort comprising 135 patients (mean age 66.4 ± 6.8 years; 77% female) who underwent unilateral UKA and received routine pharmacological thromboprophylaxis (Supplementary Fig. 1). Baseline characteristics of these patients were not statistically different from those of the 129 patients in the combined derivation (n = 87) and external validation (n = 42) cohorts who did not receive pharmacological thromboprophylaxis (Supplementary Table 5). Among them, 28 patients (20.7%) developed acute asymptomatic DVT (4 proximal and 24 distal) following UKA, which was not statistically different from the incidence in the non‑prophylaxis group (31 of 129, 24.0%; p = 0.622). These 28 patients underwent anticoagulation dose escalation. Among them, one patient with asymptomatic proximal DVT developed a wound hematoma that required reoperation for hemostasis; 2 patients with asymptomatic distal DVT developed increased but self-limited wound exudate (Supplementary Table 1).

We then evaluated the near-term risk stratification performance of the individual indicators and the nomogram in this cohort. Baseline characteristics did not differ significantly between the DVT and non‑DVT groups, except that patients who developed acute DVT had significantly higher age (p = 0.046) and BMI (p = 0.002) and a higher prevalence of diabetes mellitus (p = 0.044). Detailed demographic and clinical characteristics of the cohort are summarized in Supplementary Table 6.

In this cohort, the AUCs for individual biomarkers POD3 TAT and PIC in discriminating acute DVT were 0.734 (95%CI, 0.633–0.834; sensitivity, 71.4%; specificity, 72.0%; cutoff value, 11.98 ng/mL) and 0.718 (95%CI, 0.630–0.805; sensitivity, 92.9%; specificity, 55.1%; cutoff value, 0.83 µg/mL), respectively (Supplementary Fig. 2A). In contrast, POD3 IL-6 showed limited discrimination: 0.628 (95%CI, 0.498–0.759; sensitivity, 35.7%; specificity, 91.6%; cutoff value, 21.41 pg/mL). Nevertheless, the full nomogram maintained good near-term risk stratification performance, with an AUC of 0.805 (95%CI, 0.714–0.897; sensitivity, 75.0%; specificity, 79.4%; Supplementary Fig. 2B). The calibration curve indicated good agreement between predictions and observations (Hosmer-Lemeshow test: χ² =7.36, p = 0.498; calibration intercept = 0.319, 95%CI, -0.226-0.838; calibration slope = 0.853, 95%CI, 0.546–1.243; Brier score = 0.122; Supplementary Fig. 2C), and DCA showed an exploratory net benefit (Supplementary Fig. 2D).

Discussion

In this study, we comprehensively analyzed the temporal changes in systemic-inflammatory and hemostatic molecular markers during the perioperative period of UKA, and evaluated their associations with the occurrence of acute postoperative asymptomatic or symptomatic DVT. Our findings demonstrated that elevated BMI and increased levels of IL-6, TAT, and PIC on POD3 were independently associated with acute DVT detected on POD4-5 following UKA. Based on these four indicators, we developed a preliminary nomogram for near-term risk stratification of acute DVT detected on POD4-5, which exhibited good performance in discrimination, calibration, and exploratory net benefit, regardless of whether pharmacological thromboprophylaxis was administered.

A key methodological strength of our study is the routine use of preoperative Doppler ultrasonography, which unequivocally confirmed that all identified DVT events were newly developed, eliminating confounding by pre‑existing thrombi. Regardless of pharmacological thromboprophylaxis, the overall acute postoperative DVT incidence on POD4-5 in our cohorts exceeded 20%. Although this rate appears high for UKA, it is largely attributable to systematic screening, and the inclusion of asymptomatic and subclinical DVT [4, 11, 12, 15]. Notably, only one patient not receiving routine postoperative anticoagulation developed a symptomatic distal DVT, which aligns with previously reported symptomatic rates (0.28–1.6%) following UKA [15]. We included asymptomatic DVT as an endpoint for two reasons. First, the limited sample size resulted in few symptomatic events. Second, and more importantly, we propose that asymptomatic DVT may serve as a reference indicator to guide thromboprophylaxis decisions. Specifically, initiating therapeutic anticoagulation only when asymptomatic DVT is detected on POD4-5 imaging. Accordingly, in the present study, we deliberately adopted no routine pharmacological anticoagulation unless asymptomatic or symptomatic DVT was confirmed on POD4-5. This approach was based on several considerations. First, current major society guidelines provide only weak recommendations for routine anticoagulation after UKA, supported by low‑certainty evidence regardless of the agent used (LMWH, direct oral factor Xa inhibitors, warfarin, or aspirin) [13, 14]. Second, VTE incidence is notably lower in Asian populations, likely due to genetic and lifestyle factors [15–17]. A Korean study of UKA-treated patients without thromboprophylaxis reported newly detected asymptomatic VTE in 25.7% of patients with all thromboembolic lesions completely resolving during six months of follow‑up [18]. Third, we restricted enrollment to low‑risk patients by excluding individuals with a history of lower extremity surgery, personal or family history of VTE or thrombophilia, or active malignancy—conditions for which routine anticoagulation might still be justified [35, 36]. As a result, our preliminary exploratory analysis showed no significant difference in acute DVT incidence between low‑risk UKA patients with versus without routine postoperative thromboprophylaxis. This finding is consistent with a previous retrospective study from Japan, which reported no differences in the incidence (31.6% vs. 37.6%), location, or characteristics of DVT following TKA with or without pharmacological prophylaxis [41]. However, given the current study design, this non‑significant p‑value did not establish equivalence, non‑inferiority, or safety. In addition, due to the comparable baseline characteristics and small sample size, propensity score matching was not performed. Whether routine postoperative pharmacological thromboprophylaxis is necessary in carefully selected low‑risk UKA patients still requires confirmation in adequately powered multicenter randomized controlled trials.

Numerous studies have identified a variety of contributing factors for postoperative DVT following TKA. These include advanced age, obesity, diabetes mellitus, current smoking status, the method of anesthesia, surgical procedure and its duration, the use of tourniquets, pain control strategies, and early rehabilitation interventions, etc [2, 3, 8, 26–28]. However, evidence regarding whether these factors apply to UKA remains limited. Cao et al. [42] reported that older age increases the likelihood of thrombosis after UKA. Li et al. [4] identified intraoperative blood loss > 102 mL, BMI > 26.8 kg/m², and D-dimer > 0.29 mg/L as independent risk factors for DVT after UKA. In our derivation cohort of 87 patients, we observed that higher BMI was independently associated with an increased risk of acute asymptomatic or symptomatic DVT, consistent with the higher VTE incidence observed in severely or morbidly obese populations after UKA/TKA [12, 15]. Obesity is generally accepted to predispose individuals to thrombosis, potentially through the exacerbation of thromboinflammatory responses [43]. In addition, diabetes mellitus is another important factor known to promote endothelial activation and thromboinflammation [44]. We previously reported that the incidence of DVT within three days after UKA was significantly higher in diabetic patients, and was proportional to glycosylated hemoglobin levels [38]. In the present study, although univariate analysis indicated a higher prevalence of diabetes mellitus in patients who developed acute DVT, diabetes was not identified as an independent risk factor. We suspect this is likely due to two reasons. First, the limited sample size may have underpowered its detection after adjusting for multiple covariates. Second, potential collinearity between diabetes and other variables, particularly BMI and IL‑6, given that diabetes is frequently associated with higher BMI and a pro‑inflammatory state characterized by elevated IL‑6 levels [44]. Larger clinical cohorts are warranted to further investigate risk factors related to patient characteristics, surgical details, and postoperative management that may contribute to DVT following UKA.

Increasing evidence highlights the significant role of immunothrombosis in DVT pathogenesis, whereby coagulation cascade activation triggers immune responses and innate immune in turn facilitate thrombosis [20, 21]. von Brühl et al. [45] revealed that venous thrombosis involves a complex interplay among innate immune cells (e.g., neutrophils), platelets, and venous endothelial cells. In the present study, we observed that systemic-inflammatory markers, including neutrophil count, NLR, SIRI, SII, AISI, CRP, and IL-6, as well as hemostatic molecular markers such as D-dimer, INR, TAT, and PIC, were significantly upregulated during the acute phase after UKA, indicating that UKA procedures acutely activate coagulation and fibrinolysis cascades and the innate immune responses. Several circulating inflammatory marker have been measured as potential biomarkers for DVT diagnosis [2, 8, 19]. CBC-derived systemic-inflammatory markers provide a comprehensive reflection of the host’s inflammatory, immune, and thrombotic status, demonstrating superior predictive ability compared to single-cell counts [22]. Early studies have indicated that certain systemic-inflammatory markers, such as NLR, SII, SIRI, and AISI, are associated with postoperative DVT or infection after arthroplasty [24–27]. However, these studies primarily focused on patients undergoing total hip and knee arthroplasty. In the present study, we investigated these systemic-inflammatory markers in patients undergoing UKA for the first time but found no significant differences between patients with and without DVT. This null finding may be partly attributable to the limited sample size. Additionally, the intra‑articular use of corticosteroids during surgery may have attenuated local inflammatory responses [46]. Furthermore, this lack of association may also be explained by milder inflammatory changes due to reduced surgical trauma in UKA [47, 48]. Nevertheless, we acknowledge that this interpretation remains speculative without direct objective measures of surgical stress (e.g., serial measurements of tissue injury‑related cytokines). Therefore, the ability of CBC-derived systemic-inflammatory markers to discriminate DVT in the UKA population warrants further investigation in larger cohorts that concurrent quantification of surgical stress biomarkers. CRP and IL-6 are among the most extensively studied inflammatory biomarkers for predicting complications following TKA [47–50]. In our cohorts, we observed significantly elevated CRP and IL-6 levels on POD3 in patients who developed DVT compared to those who did not. Multivariable logistic regression further identified that POD3 IL-6 was independently associated with an increased risk of acute DVT following UKA. In studies of PJI, Zhang et al. [39] reported that IL-6 outperforms CRP in predicting acute PJI, whereas CRP is more predictive of chronic PJI. This difference may be explained by the distinct kinetic profiles of these biomarkers: IL-6 rises rapidly during acute inflammatory responses, acting as a key upstream mediator that triggers the synthesis of acute-phase proteins such as CRP [51]. Mechanistically, IL-6 plays a pleiotropic role in linking inflammation and thrombosis during DVT pathogenesis [52]. It exerts a central procoagulant effect by upregulating tissue factor expression on monocytes and endothelial cells, shifting the venous endothelium from an antithrombotic to a prothrombotic state. This IL-6-mediated endothelial activation amplifies thrombin generation, promotes platelet-leukocyte aggregation, and enhances neutrophil extracellular trap (NET) formation, thereby fueling thrombus progression [52]. Pre-clinical studies have shown that targeting IL-6 reduces thrombus burden and vein wall inflammation [53, 54]. Although IL-6 inhibitors are not yet approved for clinical DVT management, this strategy warrants further investigation [52].

Venous endothelial disturbances in the coagulation-fibrinolysis system contribute to DVT [19]. Classic coagulation laboratory assays such as prothrombin time (PT) and activated partial thromboplastin time (APTT) primarily reflect coagulation factor consumption or impaired synthesis; thus, their alterations typically manifest later in thrombosis progression. In contrast, novel hemostatic markers—TAT, PIC, t‑PAIC, and sTM—directly reflect early activation of the coagulation‑fibrinolysis system and endothelial function [30–33]. Newly formed thrombin has a short half-life and rapidly interacts with antithrombin III to form the stable TAT, which signifies acute activation of the coagulation pathway and thrombin production. The α2-plasmin inhibitor promptly engages with plasmin to form the stable PIC, indicating activation of the fibrinolytic pathway, plasmin formation, and a potential prothrombotic state. Similarly, plasminogen activator inhibitor-1 (PAI-1) inhibits the catalytic activity of tissue plasminogen activator (t-PA) by forming an inactive t-PAIC, which suggests sustained endothelial damage and fibrinolysis activation. Additionally, sTM is produced through the proteolytic cleavage of endothelial-bound thrombomodulin following endothelial injury, serving as a marker of endothelial function [30–33]. In the present study, we demonstrated that as early as 2 h post-UKA treatment, the levels of TAT and PIC, but not t-PAIC and sTM, were higher than baseline levels measured on admission. Moreover, we identified that the levels of TAT and PIC measured on POD3 were independently associated with an increased risk of acute DVT. TAT, as a direct measure of thrombin generation, reflects not only fibrin clot formation but also thrombin‑mediated activation of platelets and inflammatory signaling via protease‑activated receptors [55, 56]. The concomitant rise in PIC indicates an activated fibrinolytic response, which paradoxically may represent an insufficient compensatory mechanism in the face of ongoing thrombosis [57]. Our results are further supported by previous studies indicating that elevated levels of TAT are associated with the occurrence of DVT in patients undergoing total joint arthroplasty, as well as in those suffering from malignancies [28–30]. Additionally, we observed a difference in POD3 D-dimer levels between patients with and without DVT, but did not prove it to be an independent risk stratification factor. It is important to note that D-dimer is the most widely used marker in clinical practice for excluding DVT due to its high sensitivity, albeit with low specificity [19, 40]. In future large‑scale studies, incorporating D‑dimer as a rule‑out indicator may improve early risk stratification efficiency.

The finding that POD3 represents the most informative time point may be explained as follows: UKA induces an acute but transient surge in systemic-inflammatory and hemostatic molecular markers, with levels rising rapidly at 2 h and on POD1, followed by a marked decline by POD3 in patients without ongoing thrombotic events. In contrast, patients who subsequently develop DVT on POD4‑5 exhibit a sustained activation of thrombo‑inflammatory pathways, resulting in persistently elevated POD3 levels despite the general downward trend. However, given the observational and correlational nature of our study, it is difficult to definitively establish causality—whether the elevated levels of POD3 IL‑6, TAT, and PIC truly predict subsequent DVT or merely reflect biological changes from already‑present but yet‑undetected thrombosis. Additionally, this study did not sufficiently clarify which markers are most relevant or practical for everyday clinical use. Definitive resolution of the causality question, as well as identification of the key drivers of DVT, will require larger prospective studies with daily imaging from POD1 onward, combined with mechanistic investigations into the pathogenesis of post‑UKA thrombosis.

The nomogram is a visual clinical scoring model designed to assess disease risk and prognosis [58]. Lin et al. [28] recently developed a nomogram that incorporates surgical site, leg swelling, and POD1 TAT to predict the risk of DVT on POD7 following total hip and knee arthroplasty. In the present study, we generalized a nomogram integrating BMI, and POD3 IL-6, TAT, and PIC to stratify the risk of DVT detected on POD4-5 following UKA. Our preliminary nomogram showed good discrimination, calibration, and exploratory net benefit in patients not receiving routine postoperative pharmacological thromboprophylaxis. Moreover, this nomogram also demonstrated good performance for near-term risk stratification in the retrospective anticoagulation cohort. Nevertheless, confirmation in a prospective anticoagulated cohort is required. Additionally, we acknowledge three practical limitations. First, the majority of DVT events detected in this study were asymptomatic distal DVTs. This nomogram should not be interpreted as a tool for discriminating clinically severe VTE, symptomatic DVT, proximal DVT, or PE. Dedicated studies with appropriately powered endpoints are warranted to address those clinically more relevant outcomes. Second, although the nomogram showed an exploratory net benefit, the DCA analysis was performed over a broad range of risk thresholds (0.10–0.99). Defining clinically actionable thresholds that correspond to specific interventions (e.g., initiating pharmacological prophylaxis, performing additional surveillance, or delaying discharge) will require further validation in larger, multicenter prospective impact studies. Third, the limited availability and high cost of advanced hemostatic molecular markers TAT and PIC may hinder real-world implementation of the proposed model. Although these assays are increasingly used in clinical practice, their accessibility remains variable across healthcare settings, and cost-effectiveness analyses are required to support routine use. Therefore, the nomogram should be regarded as exploratory and preliminary rather than ready for clinical decision‑making.

This study has some limitations. First, although designed as a multi-center prospective study comprising both derivation and independent external validation cohorts, the overall sample size remains small. The derivation cohort had 21 DVT events, and the final nomogram includes four indicators, yielding an events‑per‑variable (EPV) ratio of 5.25:1. This borderline EPV ratio may lead to optimistic performance estimates and calls for cautious interpretation. Furthermore, the external validation cohort contained only 10 DVT events, resulting in a wide 95%CI for the AUC, which may reflect imprecision rather than definitive excellent performance. In particular, the Hosmer‑Lemeshow test is sample‑sensitive. The calibration intercept and slope estimates have wide confidence intervals, reflecting statistical uncertainty. Additionally, the wide 95%CI for PIC in the multivariable analysis suggests some instability. Second, all participants were of Chinese ethnicity, necessitating validation across diverse ethnic and geographic populations. Third, the performance of the preliminary nomogram in the context of acute clinically severe VTE, symptomatic DVT, proximal DVT, or PE requires further validation, and the present study did not address DVT beyond the early postoperative period. Fourth, temporal changes in inflammatory and hemostatic markers were not measured in UKA patients receiving routine postoperative anticoagulation. Fifth, the nomogram model was based on variables significant in univariate analyses, and multiple biomarkers were screened across several time points. This p‑value‑driven approach may increase overfitting. Therefore, the biomarker screening and model‑building process is exploratory, and the p‑values should be interpreted cautiously. In addition, we intentionally simplified the model to focus on the novel concept of whether selected systemic-inflammatory and hemostatic molecular markers could serve as early detection tools for DVT. Future larger studies should pre‑specify all candidate indicators based on clinical and biological rationale and apply shrinkage or regularization methods to obtain more stable estimates [59]. Finally, the observed associations do not establish causality, underscoring the need for more fundamental investigations.

Conclusions

Elevated BMI and increased levels of POD3 IL-6, TAT, and PIC are independently associated with an increased risk of acute asymptomatic or symptomatic DVT detected on POD4-5 following UKA. This study provides preliminary evidence supporting the role of systemic-inflammatory and hemostatic molecular markers in near-term risk stratification of acute DVT after UKA. A preliminary nomogram based on these four indicators represents a promising auxiliary tool for near-term risk stratification of acute postoperative DVT. However, due to the borderline EPV ratio and the small sample size, these findings should be interpreted with caution. The nomogram remains exploratory and preliminary, rather than robust or ready for clinical decision-making. A more comprehensive model that combines pre‑specified established clinical risk factors with the inflammatory and hemostatic biomarkers identified in this study, along with modern modeling techniques (e.g., machine learning), is needed to construct a clinically useful risk stratification tool for acute DVT after UKA. Future research should also explore the potential of these indicators to guide postoperative anticoagulation strategies in clinical practice.

Authors’ contributions

G.C., X.X., and Z.L. conceived, designed, and supervised the study. Z.L., T.L., and Z.D. developed methodology. Z.L., T.L., Z.D., M.D., and S.L. conducted blood sample testing, acquired and analyzed data, and prepared the figures and tables. B.Y. provided technical support. Z.L. and T.L. wrote the original draft of the manuscript. Z.L., T.L., X.X., and G.C. made critical revisions to the manuscript. All authors read and approved the final manuscript.

Funding

This study was supported by the Taishan Scholar Youth Expert Program of Shandong Province (tsqn202408398), the Beijing Hospitals Authority Clinical Medicine Development of Special Funding Support (grant XMLX202139), and the Wu jieping medical foundation (grant 320.6750.2024-23-23).

Data availability

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

Declarations

Ethics approval and consent to participate

This study was approved by the Ethics Committee of the Xuanwu Hospital, Capital Medical University (ID-CRB-062/2021), and was performed according to the Principles of Declaration of Helsinki. Written consent was obtained from patients or their legal representatives.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.