A D-dimer-, platelet count-, and albumin-based model for predicting 30-day central venous catheter-related thrombosis in patients with malignancy: a single-center retrospective cohort study

Abstract

Catheter-related thrombosis (CRT) is a clinically relevant complication in patients with malignancy receiving central venous access devices, but early individualized risk assessment remains limited. This study developed and internally validated a prediction model incorporating D-dimer, platelet count, and albumin for estimating 30-day CRT risk after catheter placement. This single-center retrospective cohort study included hospitalized adult patients with malignancy who underwent peripherally inserted central catheter (PICC), central venous catheter (CVC), or implantable venous access port (PORT) placement between January 2018 and December 2025. The primary outcome was imaging-confirmed CRT within 30 days after catheter insertion, including both symptomatic events and incidentally detected asymptomatic thrombosis. Core laboratory predictors included log-transformed D-dimer, platelet count, and albumin. Non-core clinical, tumor-related, and catheter-related predictors were selected using least absolute shrinkage and selection operator regression. A multivariable logistic regression model was developed, internally validated using 1,000 bootstrap resamples, and evaluated by discrimination, calibration, decision curve analysis, risk stratification, and sensitivity analyses. Among 1,876 eligible patients, 145 developed 30-day CRT, corresponding to an incidence of 7.73%. Median time to CRT diagnosis was 11.0 days. The final model incorporated log-transformed D-dimer, platelet count, albumin, body mass index, adenocarcinoma histology, distant metastasis, recent chemotherapy, prior venous thromboembolism, baseline infection, device type, basilic vein targeting, and repeated puncture attempts. The laboratory-integrated model showed good discrimination, with an apparent area under the curve of 0.816 and an optimism-corrected area under the curve of 0.798. Calibration was acceptable, with a calibration intercept of 0.015, calibration slope of 0.962, and Brier score of 0.052. Risk stratification identified low-, intermediate-, and high-risk groups with observed CRT incidences of 2.0%, 8.0%, and 25.5%, respectively. A prediction model integrating D-dimer, platelet count, albumin, and selected clinical and catheter-related variables showed acceptable performance for estimating 30-day CRT risk in patients with malignancy. External validation is required before broader clinical application.

The authors' abstract, as published at the source. Thrombosis Journal, 2026 · DOI ↗