Development and validation of a predictive nomogram for venous thromboembolism in adult patients undergoing orthotopic liver transplantation

To the Editor: Venous thromboembolism (VTE) is a common complication following orthotopic liver transplantation (OLT), with an incidence of 2.8–8.6%,1–3 which affects the quality of life of post-transplant patients. Current VTE risk assessment tools, such as the Caprini score, have limitations when applied to this population, underscoring the necessity of developing an early VTE risk assessment model tailored for OLT recipients. This single-center retrospective study aimed to identify the risk factors of VTE and to develop a predictive nomogram model. This study was approved by the Ethics Committee of Beijing Tsinghua Changgung Hospital (No. 19242-401) and was registered at ClinicalTrials.gov (NCT05209048). Informed consent was waived by Institutional Review Board because of the retrospective nature of our study. The study included 280 adult OLT recipients in the training cohort (August 2018–December 2020) and 76 in the validation cohort (January–December 2021). Supplementary Figure 1, https://links.lww.com/CM9/C126 shows the participant selection criteria and the study design process. The cohort was divided into case and control groups based on the occurrence of VTE within 30 days post-OLT. Post-OLT VTE events primarily include deep vein thrombosis (DVT) and pulmonary embolism (PE). VTE screening involved daily ultrasound or CT venography within seven days post-surgery, with PE clinical suspicion confirmed using CT pulmonary angiography. Patients received prophylactic, mechanical, or pharmacological treatment. In this study, the overall incidence of VTE after OLT was 11.5% (41/356). The incidence of DVT was 11.0% (39/356), with 38 cases of peripheral DVT and 1 case of mixed-type DVT, whereas the incidence of PE was 0.6% (2/356). Among DVT cases, 74.4% (29/39) were asymptomatic and incidentally diagnosed during routine postoperative ultrasonographic screening, all involving isolated calf muscle vein thrombosis, whereas the rest presented with limb swelling. The onset of VTE ranged from postoperative day 1 to 20, with a median onset on day 9. Demographic and perioperative clinical data, such as surgical information and laboratory test results, were collected, and logistic regression was used to identify the posttransplant VTE risk factors. Before univariate logistic regression analysis, the variables were transformed and processed to enhance the applicability of the predictive model. Optimal cutoff values for continuous variables were determined using receiver operating characteristic (ROC) curves: age (60 years), body mass index (BMI) (25 kg/m2), Model for End-Stage Liver Disease (MELD) score (25), central venous pressure (CVP) ≤5 cmH2O time (1 h), and surgical duration (12 h). This finding led to the conversion of these variables into dichotomous ones. Multicategory variables, such as the American Society of Anesthesiologists (ASA) Physical Status and Child–Pugh classifications, were converted into dichotomous variables using optimal scaling regression. Referring to the diagnostic and grading criteria for postoperative acute kidney injury (AKI) proposed by the Kidney Disease: Improving Global Outcomes (KDIGO) organization4 in 2012, and considering the availability of observational indicators in retrospective data, this study defined postoperative AKI Grade 1 as an increase in serum creatinine by ≥26.5 μmol/L or 1.5–1.9 times above baseline within 48 h postoperatively; AKI Grade 2 as an increase of 2.0–2.9 times above baseline within seven days postoperatively; and AKI Grade 3 as an increase in serum creatinine by ≥353.6 μmol/L or ≥3.0 times above baseline within seven days postoperatively. The baseline creatinine level was defined as the lowest recorded serum creatinine level in the week before surgery. The occurrence of postoperative AKI was transformed into a dichotomous variable using the optimal scaling regression, categorizing it as "Not occurred or Grade 1" and "Grade 2 or higher". The D-dimer and fibrinogen ratio (DFR) has a higher sensitivity and specificity in predicting thrombosis than relying solely on D-dimer or fibrinogen (FIB). A higher DFR indicates a greater likelihood of thrombosis. In this study, the average DFR value in the week before surgery and the DFR value on the first postoperative day were compared, and the increase in DFR on the first postoperative day compared with the preoperative levels was used as a dichotomous variable for analysis. Supplementary Table 1, https://links.lww.com/CM9/C126 shows the univariate analysis of perioperative clinical data of the patients in the training cohort. We found significant correlations (P 0.05, indicating that the predicted probabilities of the model were statistically consistent with the observed outcomes. This finding suggests good overall agreement between predicted and actual probabilities, affirming the model's reliability and performance. Decision curve analysis (DCA) for the VTE nomogram model and Caprini score are presented in Supplementary Figure 4, https://links.lww.com/CM9/C126. The diagonal line represents the net benefit of the intervention, where all individuals predicted to be positive received anticoagulant therapy. In contrast, the horizontal line indicates the net benefit of no intervention for all individuals predicted to be negative. The slope of the diagonal line reflects the net benefit gained from the intervention, with a negative slope indicating a decrease in benefit. The farther the curve was from these reference lines, the higher the clinical utility of the model. The VTE prediction model showed a higher net benefit within a range of risk thresholds of above 5.0%. Conversely, the decision curve for the Caprini score closely resembles the lines representing "all receive or adjust anticoagulant therapy" and "none receive or adjust anticoagulant therapy," suggesting a minimal net benefit across the range of risk thresholds. The inclusion of comprehensive predictive factors in the model reflects various aspects of the patient's preoperative disease status, postoperative coagulation function, and organ function recovery, thereby improving predictive accuracy. Patients with preoperative comorbidities, such as hypertension, hyperlipidemia, diabetes, and coronary heart disease, are at a significantly increased risk of postoperative VTE due to associated vascular damage, alterations in coagulation status, and hemodynamic changes. The MELD score serves as a standard for assessing the severity of end-stage liver disease and is often associated with poor prognosis. The complex changes in coagulation function during the perioperative period of LT make it unreliable to rely solely on single coagulation function indicators, such as FIB or D-dimer, to predict thrombotic risk. Instead, the D-dimer/FIB ratio can provide a more comprehensive reflection of the balance between fibrinolysis and coagulation processes. After surgery, the status of important organs, such as the liver and kidneys, is a decisive factor in systemic coagulation. The model's internal and external validations demonstrated good discrimination and calibration, outperforming the Caprini score, making it applicable in post-liver transplant clinical practice for identifying high-risk patients. The study's limitations include a single center's clinical data and the majority of VTE events being mild DVT, requiring caution when applying the model to different populations. The model should be used as a supplementary tool in liver transplant patients, not as a direct guide to clinical decisions due to the varying severity of VTE. Further research is needed to investigate the relationship between post-liver transplant immunotherapy regimens and postoperative VTE, as the use of immunosuppressive medications may increase the risk of platelet aggregation and thrombosis.5 Prospective observational studies with multicenter data are required to confirm the model's clinical applicability. Acknowledgment We would like to thank Editage (www.editage.cn) for English language editing. Funding None. Conflicts of interest None.