Preoperative neutrophil-to-lymphocyte ratio and postoperative deep vein thrombosis after tibial plateau fracture: a retrospective analysis of a prospectively collected cohort Authors: Nuoman Han, Guangzhao Hou, Mengke Fan, Rui Chen, Zhenbang Yang, Haoran Wang, Haoyu Wang, Wei Chen, Hongzhi Lv Published: 2026 ABSTRACT AbstractBackgroundDeep vein thrombosis (DVT) is a common and serious postoperative complication of tibial plateau fracture that significantly impacts patient prognosis, but its accurate risk prediction remains challenging. The utility of neutrophil-to-lymphocyte ratio (NLR) as a simple marker of inflammatory-immune status in thrombosis risk assessment is not clear, especially in the tibial plateau fracture population.MethodsThe study enrolled 1582 adult patients with tibial plateau fracture who underwent surgery from October 2019 to December 2023. Demographic characteristics, medical co-morbidities, surgical factors, and laboratory indicators were collected, with DVT within 14 days of surgery as the primary outcome. LASSO regression was used to screen key variables, and the dose-effect relationship between NLR and DVT was analyzed by 4-node restricted cubic spline (RCS) and the optimal cut-off value was determined, combined with multivariable logistic regression to assess the strength of the association, and sensitivity analyses to verify the robustness of the results.ResultsBaseline data showed that 54.4% were male with a mean age of 48.2 ± 12.5 years. Postoperative DVT occurred in 548 patients (34.64%). LASSO regression screened 16 variables suitable for inclusion in the model for adjustment. RCS revealed a significant nonlinear association (P < 0.001), and the optimal cut-off value was determined to be 3.265. After PSM, high NLR remained an independent risk factor for DVT. Sensitivity analyses confirmed robustness.ConclusionNLR > 3.265 independently predicts tibial plateau fracture postoperative DVT. Integrating NLR into risk models may aid personalized antithrombotic strategies, improving patient prognosis.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12891-026-09994-5.Keywords: Tibial plateau fracture, Deep vein thrombosis, Neutrophil-lymphocyte ratio, Restricted cubic spline, Nonlinear association Introduction Tibial plateau fractures, a complex knee trauma, account for 1–2% of systemic fractures and 90% of lateral tibial plateau fractures [1]. Deep vein thrombosis (DVT), a serious postoperative orthopaedic complication [2], prolongs hospital stays, increases medical burden, and may cause fatal pulmonary embolism or long-term lower limb dysfunction [3, 4]. Despite the current guidelines recommending conventional drug prevention, the incidence of DVT remains high [5–7], which suggests that we need to deepen our knowledge of its epidemiological characteristics and risk factors. Recent evidence has also highlighted the substantial burden of postoperative venous thromboembolism across a wide spectrum of orthopaedic trauma populations, including patients with pathologic fractures, emphasizing that thrombosis prevention remains a critical and unmet clinical need in orthopaedic surgery [8–10]. Therefore, exploring more accurate risk prediction indicators is essential to optimise prevention and management strategies. Existing studies have confirmed that a variety of factors, including age and underlying disease, are associated with the development of DVT after tibial plateau fracture [11, 12], but most of these studies have focused on external factors, and the exploration of the internal pathophysiological mechanisms is particularly insufficient. Indeed, the interaction between the inflammatory response and the coagulation system is a central mechanism of thrombosis, and the key role of systemic inflammation in venous thrombosis has been widely appreciated. In the field of risk prediction, although D-dimer as a traditional biomarker has been used to identify patients at high risk of DVT [13], its limitations in early and accurate prediction have gradually emerged, such as insufficient specificity, which has prompted researchers to explore more sensitive, convenient, and cost-effective biological indicators. Neutrophil-to-lymphocyte ratio (NLR), as a simple indicator reflecting the systemic inflammatory state and immune balance of the organism [14], has demonstrated significant predictive value in a variety of diseases due to its advantages of being accessible and inexpensive based on routine blood tests, e.g., in the fields of malignant tumours, cardiovascular diseases, etc., NLR has proven to be effective in the assessment of the severity and prognosis of diseases [15, 16]. In the field of orthopaedics, preliminary studies have also explored the association between NLR and the risk of postoperative DVT in hip fracture patients, suggesting that the inflammatory response triggered by fracture trauma and surgical stress may indirectly reflect the risk of thrombosis through the indicator of NLR [17]. However, the association between NLR and postoperative DVT has not been systematically elucidated for the specific population of tibial plateau fracture, especially the key questions of whether there is a nonlinear association between the two, whether there is a clinically meaningful cut-off value, and whether NLR can enhance risk stratification independently of traditional risk factors such as D-dimer, etc. There is still a lack of targeted studies. Tibial plateau fractures are intra-articular injuries with characteristics of high-energy trauma, severe soft tissue damage, and long postoperative immobilization. The inflammatory and coagulation activation patterns differ from those of hip fractures and joint arthroplasty. Current studies on thromboembolism are mostly concentrated on hip fractures, polytrauma, and joint replacement, while large-sample studies focusing on the specific population of tibial plateau fractures are insufficient. Therefore, this study selected tibial plateau fracture patients to explore the predictive value of preoperative NLR for postoperative DVT, so as to fill the evidence gap in this field. Based on the above questions, this study established a prospectively maintained tibial plateau fracture database (2019–2023), which included 1582 operated patients, with the aim of addressing two key questions: (1) Is there a nonlinear association between preoperative NLR and the occurrence of postoperative DVT in patients with tibial plateau fractures? (2) If a nonlinear association exists, what is the optimal cut-off value for NLR and its predictive efficacy? By combining restricted cubic spline (RCS), propensity score matching (PSM) and multivariable logistic regression, we systematically assessed for the first time the threshold effect and clinical utility value of NLR in postoperative DVT of tibial plateau fracture, providing an evidence-based basis for individualised anticoagulation prevention. Method Data source Data were from the Tibial Plateau Fracture Database, prospectively maintained since October 2014 at our central China university-affiliated tertiary referral teaching hospital (serving a 74.61-million-plus population area). A 15-member expert team (orthopaedic surgeons, radiologists, etc.) established the database 60 interns, trained for a month and grouped, collected data: (1)EMR/clinical data (Beijing Kaihua); (2)imaging data (iMedical PACS, DHC); (3)lab/bacterial results (RMLIS, Rimage). Follow-up, at 1, 6, 12 months post-discharge, involved clinical nurses (≥ 3-year training) and specialists, via phone, WeChat, or outpatient visits. Unresponsive patients, re-contacted after 24 h and still unresponsive, were lost. EpiData (3.1) enabled dual data entry/validation to reduce errors. Study cohort and study design The study population consisted of 2,854 patients enrolled in the database between October 2019 and December 2023, and 1,582 adult patients who underwent surgical fixation for tibial plateau fractures were screened (Fig. 1). Patients who met the following criteria were excluded: (1) < 18 years old; (2)pathological fractures, tumour-related fractures, or multiple fractures; (3)not treated with reduction surgery; (4)recent use of anticoagulant medication; (5)comorbid severe organ dysfunction of the heart, lungs, liver, and kidneys; and (6)those who were lost to visit or had missing data. The study protocol was reviewed and approved by the ethics committee of our hospital, in accordance with the principles of the Declaration of Helsinki and following the STROBE criteria [18].Since the clinical data involved in this article are secondary uses, the local ethics committee exempted the requirement of informed consent for the use of patient data. Flowchart of patient selection in the study cohort Exposure variable The primary exposure variable was the NLR. Haematological samples from all patients were collected within 24 h prior to surgery to ensure that they reflected the baseline inflammatory state prior to surgery. The testing procedure strictly followed the International Committee for Standardisation in Haematology (ICSH) guidelines [19], where neutrophil counts and lymphocyte counts were determined during haematological investigations and the ratio of the two was calculated to obtain the NLR. Covariates Based on the factors influencing postoperative deep vein thrombosis in patients with tibial plateau fractures identified previously in the literature and our experience, we considered four types of covariates: demographic information, medical co-morbidities, surgical factors, and laboratory indicators. Demographic information included sex, age, marriage, season, residence, blood type, length of stay(LOS), and cause of injury (high-energy injury or low-energy injury); medical co-morbidities included diabetes, hypertension, combined ligament injury, combined meniscus injury, and hypoproteinemia; surgical factors included anaesthesia, reset mode, and types of internal fixations; and laboratory indicators included platelet (Plt), hypersensitive C-reactive protein (hs-CRP), albumin (Alb), haemoglobin (Hb), creatinine (Cr). This study also included D-dimer as a key covariate; plasma levels of D-dimer reflect increased fibrinolytic activity and hypercoagulability secondary to thrombotic events, which is a recognised sensitive marker of thrombotic events [17–20]. Fracture severity indicators (e.g., Schatzker classification) were not included in the primary analysis due to incomplete documentation in the database; however, cause of injury was used as a proxy. Outcomes The primary outcome was the occurrence of DVT within 14 days after surgery, including both symptomatic and asymptomatic DVT, which was further categorized into proximal DVT and distal DVT according to ultrasound location. All patients underwent routine bilateral lower extremity venous color Doppler ultrasonography for systematic screening on postoperative day 1, 7, and 14, regardless of the presence or absence of clinical symptoms suggestive of DVT. The diagnosis of DVT was established based on color Doppler ultrasonography, strictly in accordance with the Guidelines for the Diagnosis and Treatment of Deep Vein Thrombosis (3rd edition) formulated by the Chinese Medical Association [21, 22]. Statistical analysis Categorical variables were expressed as numbers (%) and analysed using the χ2 test. For continuous variables, normality was assessed in this study using the Kolmogorov-Smirnov test. When continuous variables conformed to normal distribution, they were expressed as mean ± standard deviation and analysed using the t-test, and if they did not conform to normal distribution, they were expressed as median[Q1-Q3] and analysed using the the Mann-Whitney test. Multiple interpolation (MI) was used to deal with missing values [23] (Supplementary Table S1). Prior to any regression analysis and modelling, we tested for multicollinearity between all variables by checking the variance inflation factor (VIF), and a VIF greater than 3 was considered multicollinearity (Supplementary Table S2). To reduce the potential overfitting problem in the model, the least absolute shrinkage and selection operator (LASSO) regression is used to screen the variables suitable for inclusion in the model, combined with 10-fold cross-validation to optimise the regularisation parameter (λ), and variable selection is achieved through the introduction of the L1 regularisation term. Restricted cubic spline (RCS) was used to assess the dose-effect relationship between preoperative NLR and postoperative DVT. The nonlinear association between NLR and the risk of developing postoperative DVT was visually assessed using an RCS model with 4 nodes adjusted for all covariates with P < 0.1. Based on the results of the RCS analyses, the optimal cut-off value for NLR was determined, and patients were divided into a “low NLR group” (≤ cut-off value) and a “high NLR group” (> cut-off value). To reduce potential confounding effects from baseline characteristics, propensity score matching (PSM) was performed using a 1:1 nearest-neighbour matching algorithm with a caliper value set at 0.01. Matching variables were based on 16 key covariates screened by LASSO regression to ensure that the two groups were well-balanced in terms of demographic characteristics, medical co-morbidities, surgical factors, and laboratory indicators. A good balance of covariates between groups was indicated when the standardised mean difference (SMD) was less than 0.1. Subsequently, a multivariable logistic regression model was constructed to incorporate the above key covariates to assess the significance and magnitude of the association between the “high NLR group” and the risk of postoperative DVT. To verify the robustness of the main findings, firstly, to address the potential clinical variability of the NLR cut-off value, we redefined the exposure grouping based on the threshold commonly used in clinical practice (NLR = 3.0): NLR > 3.0 was defined as the “high NLR group”, and NLR ≤ 3.0 was defined as the “low NLR group”. The strength of the association between high NLR and postoperative DVT was repeatedly assessed using a multivariable logistic regression model (including the same set of covariates) that was consistent with the main analysis. Secondly, population stratification analysis was performed by key variables. Based on clinical experience, previous research results and biological mechanisms [24], we pre-determined stratification variables with clear clinical significance and theoretical basis, including patients’ age (< 45 years, 45–64 years, ≥ 65 years), cause of injury (high-energy injury, low-energy injury), LOS (0–7 days, 8–14 days, ≥ 15 days) and D-dimer levels (< 1 mg/L, 1–2 mg/L, > 2 mg/L). These variables were all measured at baseline and were not influenced by exposure factors versus outcome variables, minimising confounding bias. multivariable logistic regression models were constructed separately for each subgroup, and the stratified results were presented by forest plots to assess the interaction effect of stratified variables with NLR in order to explore the synergistic effect of inflammatory markers and coagulation indicators. All analyses were performed using R software 4.4.2 and SPSS 26.0, with P < 0.05 indicating a statistically significant difference. Results Baseline characteristics A total of 1582 patients were included, 54.4% were male and 45.6% were female; the mean age was 48.2 ± 12.5 years; and high-energy injuries accounted for 65.7%. Postoperative DVT occurred in 548 cases (34.64%), of which 266 (48.5%) were male and 282 (51.5%) were female (Table 1). Table 1Baseline characteristics of participantsVariablesOverall(n = 1582)DVT(n = 548)Without DVT(n = 1034)P valueSex0.001 Males860(54.4)266(48.5)594(57.5) Females722(45.6)282(51.5)440(42.6)Age49.0[37.0–58.0]53.0[42.3–62.0]46.0[35.0–56.0]< 0.001Marriage< 0.001 Unmarried78(4.9)15(2.7)63(6.1) Married1449(91.6)502(91.6)947(91.6) Divorced/Widowed55(3.5)31(5.7)24(2.3)Season0.459 Spring441(27.9)144(26.3)297(28.7) Summer374(23.6)135(24.6)239(23.1) Autumn460(29.1)169(30.8)291(28.1) Winter307(19.4)100(18.3)207(20.0)Residence0.004 Rural938(59.3)352(64.2)586(56.7) Urban644(40.7)196(35.8)448(43.3)Blood type0.427 A410(25.9)153(27.9)257(24.9) B531(33.6)181(33.0)350(33.9) O485(30.7)167(30.4)318(30.8) AB156(9.9)47(8.6)109(10.5)LOS< 0.001 0–7 days257(16.2)59(10.8)198(19.2) 8–14 days739(46.7)236(43.1)503(48.7) ≥ 15 days586(37.0)253(46.2)333(32.2)Cause of injury0.037 High-energy damage1040(65.7)379(69.2)661(63.9) Low-energy damage542(34.3)171(31.2)371(35.9)Diabetes0.127 Yes291(18.4)112(20.4)179(17.3) No1291(81.6)436(79.6)855(82.7)Hypertension0.088 Yes315(19.9)122(22.3)193(18.7) No1267(80.1)426(77.7)841(81.3)Ligament injury0.305 Yes520(32.9)171(31.2)349(33.8) No1062(67.1)377(68.8)685(66.3)Meniscus injury0.049 Yes398(25.2)154(28.1)244(23.6) No1184(74.8)394(71.9)790(76.4)Reset mode0.117 Yes216(13.7)85(15.5)131(12.7) No1366(86.3)463(84.5)903(87.3)Types of internal fixations0.039 Plate1395(88.2)500(91.2)895(86.6) Screw129(8.2)31(5.7)98(9.5) Intramedullary nail44(2.8)12(2.2)32(3.1) Pin14(0.9)5(1.0)9(0.9)Anaesthesia0.023 General418(26.4)126(23.0)292(28.2) Regional306(19.3)104(19.0)202(19.5) Local11(0.7)7(1.3)4(0.4) Combined847(53.5)311(56.8)536(51.8)Hypoproteinemia0.006 Yes1344(85.0)447(81.6)897(86.8) No238(15.0)101(18.4)137(13.3)Plt229.1[194.0-272.0]229.1[195.6-280.7]229.1[193.9-269.1]0.478Hs-CRP20.5[9.0-36.3]20.6[9.2–37.0]20.5[8.9–35.8]0.535Alb40.7[38.2–43.1]40.2[37.7–42.1]40.9[38.6–43.1]< 0.001Hb122.0[111.4–133.0]122.0[108.7–128.0]122.5[114.0-135.0]< 0.001Cr62.0[53.4–71.2]61.1[52.54–70.53]62.8[54.0-71.6]0.051D-dimer2.1[1.0-4.4]2.4[1.44–5.28]1.8[0.9-4.0]< 0.001NLR3.6[2.5–4.7]3.7[2.6–5.2]3.6[2.3-5.0]0.028Values are median [Q1–Q3] for continuous variables and n (%) for categorical variables. Differences in characteristic were analyzed using Mann–Whitney tests for continuous variables and χ2 -tests for categorical variablesLOS Length of stay, NLR Neutrophil-to-Lymphocyte Ratio, Plt Platele, Hs-CRP hypersensitive C-reactive protein, Alb albumin, Hb Haemoglobin, Cr Creatinine Baseline characteristics of participants Values are median [Q1–Q3] for continuous variables and n (%) for categorical variables. Differences in characteristic were analyzed using Mann–Whitney tests for continuous variables and χ2 -tests for categorical variables LOS Length of stay, NLR Neutrophil-to-Lymphocyte Ratio, Plt Platele, Hs-CRP hypersensitive C-reactive protein, Alb albumin, Hb Haemoglobin, Cr Creatinine Variable screening and modeling Twenty-three variables were initially included. Although lambda.1se contributed to model simplicity, we preferred lambda.min to retain 16 variables (Fig. 2). The optimal lambda value was 0.00931695. as this method maximises predictive efficiency while maintaining clinically relevant factors (Supplementary Table S3). Screening and filtering of variables using the Least Absolute Contraction and Selection Operator (LASSO) regression. A Misclassification error of the LASSO coefficient distribution. B In the LASSO coefficient distribution of the 23 relevant covariates, different colors correspond to different covariates The results of the RCS suggested a significant non-linear association between the two. That is, when the NLR exceeds a certain threshold, the risk of DVT rises sharply. The optimal cut-off value of NLR was determined to be 3.265, and the patients were divided into the “low NLR group” (≤ 3.265) and the “high NLR group” (> 3.265) accordingly. The risk of DVT (HR) increased significantly with increasing NLR, and the confidence intervals of the curves did not overlap with the HR = 1 line, suggesting that high NLR is an independent risk factor for DVT (Fig. 3). This result provides a basis for the early identification of DVT risk groups and the development of preventive intervention strategies by monitoring NLR in the clinic. The dose-effect relationship between preoperative NLR and DVT in patients with tibial plateau fractures. It was assessed using a four-node restricted cubic spline model adjusted for all preoperative covariates (P < 0.1). Relative risk is shown as a solid red line, and 95% confidence intervals are marked with a red shaded area, where the reference point for NLR is 3.265. Blue shaded areas show the frequency distribution of NLR. NLR=Neutrophil-to-Lymphocyte Ratio, DVT=Deep Vein Thrombosis PSM successfully matched 940 patients, of which 470 were in each of the high and low NLR groups (Table 2). The SMDs of all covariates after matching were < 0.1, suggesting a good balance of preoperative baseline characteristics between the two groups (Fig. 4). multivariable logistic regression analysis showed that high NLR remained an independent risk factor for DVT after controlling for confounding by PSM, and the risk estimate was slightly enhanced (OR = 1.050, 95% CI, 1.009–1.093, P < 0.05) (Table 3). Table 2Comparison of baseline characteristics of the “low NLR” and “high NLR” subgroupsVariablesUnmatchedP valueSMDAfter PSMP valueSMDlow NLR(n = 722)high NLR(n = 860)low NLR(n = 470)high NLR(n = 470)Sex0.0020.1620.004 Males361 (50.0)499 (58.0)252 (53.6)251 (53.4) Females361 (50.0)361 (42.0)218 (46.4)219 (46.6)Age47.0[37.0–57.0]49.5[38.0–59.0]0.0070.13748.0[37.0–57.0]48.0[37.0–58.0]0.0210.007Marriage0.7020.0420.6820.081 Unmarried39 ( 5.4)39 ( 4.5)25 ( 5.3)28 ( 6.0) Married657 (91.0)792 (92.1)434 (92.3)425 (90.4) Divorced/Widowed26 ( 3.6)29 ( 3.4)11 ( 2.3)17 ( 3.6)Season0.0520.1410.2620.065 Spring184 (25.5)257 (29.9)131 (27.9)134 (28.5) Summer161 (22.3)213 (24.8)111 (23.6)99 (21.1) Autumn229 (31.7)231 (26.9)137 (29.1)139 (29.6) Winter148 (20.5)159 (18.5)91 (19.4)98 (20.9)Residence0.0200.1200.1180.009 Rural405 (56.1)533 (62.0)285 (60.6)283 (60.2) Urban317 (43.9)327 (38.0)185 (39.4)187 (39.8)Blood type0.6710.0630.8480.074 A179 (24.8)231 (26.9)120 (25.5)116 (24.7) B240 (33.2)291 (33.8)157 (33.4)145 (30.9) O227 (31.4)258 (30.0)147 (31.3)162 (34.5) AB76 (10.5)80 ( 9.3)46 ( 9.8)47 (10.0)LOS< 0.0010.2030.0140.073 0–7 days111 (15.4)146 (17.0)80 (17.0)70 (14.9) 8–14 days376 (52.1)363 (42.2)210 (44.7)225 (47.9) ≥ 15 days235 (32.5)351 (40.8)180 (38.3)175 (37.2)Cause of injury0.1970.0680.8670.005 High-energy damage462 (64.0)578 (67.2)314 (66.8)313 (66.6) Low-energy damage260 (36.0)282 (32.8)156 (33.2)157 (33.4)Diabetes0.0120.1310.1620.022 Yes609 (84.3)682 (79.3)387 (82.3)383 (81.5) No113 (15.7)178 (20.7)83 (17.7)87 (18.5)Hypertension0.0940.0880.8720.027 Yes592 (82.0)675 (78.5)380 (80.9)385 (81.9) No130 (18.0)185 (21.5)90 (19.1)85 (18.1)Ligament injury0.0020.1610.4180.014 Yes455 (63.0)607 (70.6)315 (67.0)312 (66.4) No267 (37.0)253 (29.4)155 (33.0)158 (33.6)Meniscus injury0.0020.1610.0230.015 Yes513 (71.1)671 (78.0)350 (74.5)347 (73.8) No209 (28.9)189 (22.0)120 (25.5)123 (26.2)Reset mode0.7780.0180.7150.055 Yes621 (86.0)745 (86.6)400 (85.1)409 (87.0) No101 (14.0)115 (13.4)70 (14.9)61 (13.0)Types of internal fixations< 0.0010.2580.0350.069 Plate612 (84.8)783 (91.0)429 (91.3)424 (90.2) Screw86 (11.9)43 ( 5.0)25 ( 5.3)32 ( 6.8) Intramedullary nail20 ( 2.8)24 ( 2.8)13 ( 2.8)12 ( 2.6) Pin4 ( 0.6)10 ( 1.2)3 ( 0.6)2 ( 0.4)Anaesthesia0.6590.0640.8060.040 General184 (25.5)234 (27.2)125 (26.6)129 (27.4) Regional134 (18.6)172 (20.0)94 (20.0)96 (20.4) Local5 ( 0.7)6 ( 0.7)2 ( 0.4)3 ( 0.6) Combined399 (55.3)448 (52.1)249 (53.0)242 (51.5)Hypoproteinemia0.3150.0540.5800.070 Yes621 (86.0)723 (84.1)401 (85.3)389 (82.8) No101 (14.0)137 (15.9)69 (14.7)81 (17.2)Plt229.1[195.0-281.0]229.1[193.6-266.3]0.1090.081229.1[190.0-277.1]229.1[194.4-266.3]0.6410.030Hs-CRP18.4[6.3–33.8]24.5[12.6–40.4]< 0.0010.31520.5[6.8–38.0]22.2[12.2–34.6]< 0.0010.055Alb40.5[38.2–43.1]40.7[38.3–43.1]0.5010.03440.6[38.0-43.6]40.7[38.4–43.1]0.0490.017Hb122.5[113.2-135.4]122.0[110.2–131.0]< 0.0010.184122.0[111.9-134.1]122.0[111.0-132.8]0.1030.028Cr62.0[53.2–71.6]61.8[53.6–70.9]0.2060.06461.2[52.8–70.4]62.3[50.1–71.7]0.9090.037D-dimer1.5[0.7–2.8]2.8[1.4–5.7]< 0.0010.4921.8[0.9-4.0]2.1[1.2-4.0]0.0040.053Values are median [Q1–Q3] for continuous variables and n (%) for categorical variables. Differences in characteristic were analyzed using Mann–Whitney tests for continuous variables and χ2 -tests for categorical variablesSMD < 0.1 indicated adequate between-group balance. LOS Length of stay, NLR Neutrophil-to-Lymphocyte Ratio, Plt Platelet, Hs-CRP hypersensitive C-reactive protein, Alb albumin, Hb Haemoglobin, Cr Creatinine, PSM propensity score match, SMD standardized mean difference Comparison of baseline characteristics of the “low NLR” and “high NLR” subgroups Values are median [Q1–Q3] for continuous variables and n (%) for categorical variables. Differences in characteristic were analyzed using Mann–Whitney tests for continuous variables and χ2 -tests for categorical variables SMD < 0.1 indicated adequate between-group balance. LOS Length of stay, NLR Neutrophil-to-Lymphocyte Ratio, Plt Platelet, Hs-CRP hypersensitive C-reactive protein, Alb albumin, Hb Haemoglobin, Cr Creatinine, PSM propensity score match, SMD standardized mean difference SMD distributions of the variables and their changes before and after propensity score matching (PSM). SMD < 0.1 indicated adequate between-group balance. LOS=Length of stay; NLR=Neutrophil-to-Lymphocyte Ratio; Plt=Platelet; Hs-CRP=hypersensitive C-reactive protein; Alb=albumin; Hb=Haemoglobin; Cr=Creatinine. PSM=propensity score match; SMD=standardized mean difference Table 3Risk of DVT associated with high NLR group before and after propensity score matchingVariablesUnmatchedAfter PSMAdjustedOR (95% CI)P valueAdjustedOR (95% CI)P valuehigh NLR(>3.265)1.045(1.005–1.087)0.0411.050 (1.009–1.093)0.017 Risk of DVT associated with high NLR group before and after propensity score matching Sensitivity analysis Sensitivity analyses further confirmed the robustness of the results. When NLR = 3.0 was used as the cut-off value, the risk of postoperative DVT remained significantly increased in the high NLR group (OR = 1.265, 95% CI, 1.009–1.585, P < 0.05). Subgroup analyses showed that, with the exception of D-dimer, the interactions between the other stratification variables and NLR were not statistically significant (P > 0.05 for the interaction term), suggesting that this association is stable in most populations(Fig. 5). Association analysis between NLR and risk of deep vein thrombosis (DVT) in different subgroups of patients. Interaction P values were calculated using the likelihood ratio test. LOS=Length of stay; low NLR: NLR ≤ 3.265;high NLR: NLR > 3.265 Discussion In this study, we demonstrated for the first time a nonlinear dose-effect relationship between preoperative NLR and postoperative DVT risk in patients with tibial plateau fractures by using a prospective cohort design combining RCS and machine learning methods. We identified a critical NLR threshold of 3.265, above which the risk of DVT was significantly elevated (OR = 1.050, 95% CI, 1.009–1.093). This finding not only provides a new perspective for understanding the relationship between inflammation and thrombosis, but also provides a practical risk stratification tool for clinical practice [25]. Although existing studies have confirmed the predictive role of inflammatory markers in venous thrombosis, they have mostly focused on populations such as hip or spine fractures [15, 26],and assumed linear associations, whereas in this study, for the first time, we clarified the nonlinear dose-effect relationship between NLR and DVT by using a 4-node RCS model for a specific group of tibial plateau fractures, revealing the characteristics of a sudden increase in risk after the NLR breaks through a specific threshold. This finding fills the gap in understanding inflammatory nonlinearity in post-fracture thrombotic risk and quantifies that not all inflammation elevation equally increases thrombosis risk [27–29]. In terms of biological mechanisms, elevated NLR, a simple indicator of systemic inflammation and immune balance [30], may promote thrombosis through multiple pathways: neutrophil extracellular traps (NETs) released by neutrophil activation can damage the vascular endothelium and activate the coagulation system, whereas suppression of lymphocyte function weakens the immune regulation of thrombosis [31],and this “procoagulant-anticoagulant imbalance” is highly consistent with the significant increase in the risk of DVT after elevation of NLR in the present study [32], further supporting the classical theory that inflammation-coagulation interaction is the central mechanism of venous thrombosis [33]. The optimal cut-off value of 3.265 for NLR determined in this study has important clinical practice implications. Compared with traditional thrombotic markers such as D-dimer, NLR can be obtained based on routine blood tests without additional testing costs, which is especially suitable for primary care or resource-limited scenarios, and provides a convenient and feasible tool for early risk screening of postoperative DVT in patients with tibial plateau fractures [34]. In the sensitivity analysis, even if the clinically commonly used NLR = 3.0 was used as the cut-off value, a high NLR was still significantly associated with an elevated risk of DVT, which not only verifies the robustness of the threshold selection, but also suggests that in actual clinical practice, close thresholds can be flexibly selected for risk stratification according to the testing conditions, which enhances the operationalisation of the results of the study. The findings of subgroup analysis further deepened the understanding of the predictive value of NLR, and the strength of the association between NLR and DVT was significantly enhanced in patients with higher levels of D-dimer, which confirms the biological logic that the superimposition of inflammatory activation and hypercoagulability can exacerbate the risk of thrombosis [35], suggesting that the combination of NLR and D-dimer may further enhance the risk of thrombosis. This suggests that combined monitoring of NLR and D-dimer may further improve the accuracy of risk prediction and provide a basis for individualised anticoagulation strategies. Meanwhile, the associations between NLR and DVT in the subgroups of age, cause of injury, and days of hospitalisation were consistent and not significantly heterogeneous, indicating that the predictive value of NLR was not affected by these underlying clinical characteristics, which further supports its potential as a universal risk marker for postoperative DVT after tibial plateau fracture. The high incidence of DVT (34.64%) in this study was mainly due to routine ultrasonic screening, including asymptomatic and distal DVT. Although the clinical significance of asymptomatic distal DVT is still debated, it is related to the long-term risk of post-thrombotic syndrome and recurrence [36–38]. Our finding that NLR > 3.265 is an independent predictor can be used for preoperative rapid risk stratification to guide individualized anticoagulation prophylaxis, optimize monitoring intensity and prevention duration, and improve the prognosis of patients with tibial plateau fractures. It should be noted that after propensity score matching, the adjusted OR value of high NLR for postoperative DVT was 1.050, which was statistically significant but relatively modest in magnitude. This suggests that NLR has a limited independent predictive effect when used alone, and its clinical value lies more in preoperative rapid screening and risk stratification relying on routine blood test results. Therefore, NLR is more suitable as a supplementary predictive marker rather than an independent decisive indicator. The combination of NLR with conventional thrombotic markers such as D-dimer may achieve better predictive performance, which also reflects the limitations of single inflammatory markers in predicting postoperative DVT in clinical practice. Of course, there are some limitations in this study; single-centre data may limit the extrapolation of the results, which needs to be further validated by multicentre studies in the future; data on the dynamic changes of NLR were not included to assess its predictive value at different postoperative time points. Furthermore, although this study has attempted to control for a variety of confounding factors, several perioperative variables, such as intraoperative blood loss, time from injury to surgery, and postoperative immobilization protocol, were not included in the analysis due to incomplete clinical documentation, which may introduce residual confounding. Future prospective studies should standardize the recording of such factors to more accurately evaluate the predictive value of NLR. Conclusion Preoperative NLR > 3.265 independently predicts tibial plateau fracture postoperative DVT. Integrating NLR into risk models may aid personalized antithrombotic strategies, improving patient prognosis. Supplementary Information Supplementary Material 1. Acknowledgements Not applicable. Abbreviations Deep vein thrombosis Neutrophil-to-lymphocyte ratio Restricted cubic spline Propensity score matching Picture Archiving and Communication System International Committee for Standardisation in Haematology Length of stay Platelet Hypersensitive C-reactive protein Albumin Haemoglobin Creatinine Multiple interpolation Variance inflation factor Least absolute shrinkage and selection operator Standardised mean difference Odds ratios 95% confidence intervals Neutrophil extracellular traps Authors’ contributions N.H.,Data curation, Software, Writing-original draft, Visualization, G.H.,Methodology, M.F.,Data curation, Formal analysis, R.C.,Investigation, Z.Y.,Methodology, H.W.,H.W.,Project administration, W.C.H.L.,Funding acquisition, Writing-review & editing. Funding This study was supported by the Beijing-Tianjin-Hebei Basic Research Cooperation project(Grant No.J230007) and 2025 government-funded clinical medicine talent cultivation project(ZF2025136). Data availability All the data will be available upon reasonable request to the corresponding author of the present paper. Declarations Ethics approval and consent to participate This study was approved by the Ethics Committee of the Third Hospital of Hebei Medical University (W2021-098-1). The study protocol was conducted in accordance with the Declaration of Helsinki and the STROBE guidelines. A waiver of informed consent was approved by the Ethics Committee of the Third Hospital of Hebei Medical University due to the retrospective and anonymized secondary use of clinical data. Consent for publication Not applicable. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Nuoman Han, Guangzhao Hou and Mengke Fan contributed equally to this work and share first authorship. Contributor Information Wei Chen, Email: surgeonchenwei@126.com. Hongzhi Lv, Email: 190099199@qq.com. References Associated Data Supplementary Materials Supplementary Material 1. Data Availability Statement All the data will be available upon reasonable request to the corresponding author of the present paper.