Integrating 7-day D-dimer exposure into deep vein thrombosis risk prediction after gastrointestinal surgery

Abstract

Deep vein thrombosis (DVT) is a serious complication following gastrointestinal surgery. While D-dimer is a widely used biomarker for thrombosis, its postoperative specificity is limited due to inflammatory interference. This study introduces a novel cumulative metric-7-day D-dimer exposure (7dDDE)-to quantify perioperative coagulation burden. We retrospectively analyzed 525 patients undergoing gastrointestinal surgery, performed propensity score matching, and constructed a multivariable logistic regression model incorporating 7dDDE and other clinical variables. Model performance was evaluated using ROC curves, decision curve analysis, calibration plots, SHAP values, and a nomogram. Additionally, a linear mixed-effects model assessed D-dimer trajectories over time. The results demonstrated that 7dDDE was independently associated with postoperative DVT and was the most influential predictor in the model. The model showed good discrimination and clinical utility. Longitudinal analysis further revealed significant differences in D-dimer dynamics between DVT and non-DVT groups, even after adjustment for confounders. These findings support the use of 7dDDE as a robust biomarker for thrombotic risk stratification and highlight the importance of integrating temporal biomarker patterns into perioperative DVT prediction.

Fig. 1

Study flowchart.

Fig. 2

Forestplot of subgroup analysis for the association between 7dDDE and DVT.

Fig. 3

Model performance evaluation and interpretation in the testing set. (a) ROC curve of the logistic regression model incorporating 7dDDE. (b) Decision Curve Analysis showing the net clinical benefit of the model across a range of high-risk thresholds compared to “treat-all” and “treat-none” strategies. (c) Clinical Impact Curve displaying the number of individuals identified as high-risk and the number of true positives across different thresholds. (d) Calibration curve generated via 100 bootstrap resamples. (e) SHAP plot ranking feature importance. (f) Nomogram constructed based on the top five variables.