# Predictive model for CRT risk in cancer patients with central venous access devices: a systematic review and meta-analysis

## TL;DR

This study reviews and evaluates existing models for predicting catheter-related thrombosis in cancer patients with central venous access devices.

## Contribution

The paper provides a systematic review and meta-analysis of CRT risk prediction models in cancer patients.

## Key findings

- Nineteen papers with 29 predictive models were included, showing AUC values of 0.470–1.000.
- The combined AUC of six validated models was 0.81, indicating good discrimination.
- All studies had a high risk of bias, mainly due to poor reporting of analyzed areas.

## Abstract

With the high incidence of central venous access device catheter-related thrombosis (CRT) in patients with cancer, its early onset, and the characteristics of clinically insignificant symptoms, risk assessment is essential for the targeted application of thromboprophylaxis. The aim of this paper was to review the risk prediction models developed for central venous access device CRT in patients with cancer and to evaluate their performance.

PubMed, Embase, Web of Science, Cochrane Library, CNKI, SinoMed, Wanfang Data, and VIP databases were searched, and the search timeframes ranged from the establishment of the database to May 22, 2024. Two researchers independently performed literature screenings, data extractions, and quality assessments. The risk of bias and applicability of the included studies were assessed using the Predictive Model Risk of Bias Assessment Tool. A meta-analysis of the areas under the curve (AUC) values for model validation was performed using Stata 17.0 software.

Nineteen papers with 29 predictive models were included in this systematic review, reporting AUC values of 0.470–1.000. The incidence of central venous access device CRT in cancer patients ranges from 2.02 to 39.4%. The most commonly used predictors are D-dimer levels, BMI, and diabetes. All studies were judged to have a high risk of bias, mainly due to poor reporting of the areas analyzed. The combined AUC value of the six validated models was 0.81 (95% confidence interval: 0.76–0.86), indicating good model discrimination.

Most available CRT prediction models exhibited moderate-to-good predictive performance. However, all the studies were rated as having a high risk of bias according to the PROBAST scale. Future studies should adhere to methodological and reporting guidelines for large-sample, multi-center external validation of models, focusing on studies that report rigorous design and optimization or on the development of new models.

PROSPERO, identifier: CRD42024516563.