Abstract
OBJECTIVE: To identify key risk factors for peripherally inserted central catheter -related deep venous thrombosis (PICC-RVT) in patients with cancer through a systematic review and meta-analysis, develop risk prediction model and validate its performance. METHODS: A systematic literature search was performed in Pubmed, EMBASE, Web of Science (core collection), Scopus, CINAHL and Ovid databases from the time of their inception to June 2025. Study quality was assessed using the Newcastle-Ottawa Scale (NOS) checklist. Meta-analysis was performed using RevMan 5.4 statistical software to identify independent risk factors for PICC-RVT in adult patients with cancer, then the effect of each independent risk factor was determined using β - formation conversion for developing the risk predictive model. Finally, we collected the clinical datal of 338 adult patients with cancer who underwent PICC catheterization from June 2024 to May 2025 to evaluate the predictive performance of the risk predictive model by drawing the receiver of curve (ROC). RESULTS: A total of 32 cohort studies involving 28,813 individuals (28 from China, 2 from UK, 1 each from US and Canada) were included. Twenty-one independent risk factors were identified through meta-analysis. The risk prediction model was developed using β - coefficient transformation: pooled odds ratios (ORs) from meta-analysis were converted to β coefficients through natural logarithmic transformation (β = ln[OR]), then each β - coefficient was multiplied by 10 and rounded to one decimal place to create a point-based scoring system. The total risk score was calculated by summing individual factor scores. Preliminary external validation was conducted in 338 patients with cancer (15 thrombosis events, 4.4% incidence) from a single Chinese center, yielding an area under curve (AUC) of 0.792 (95% [confidence interval (CI)] 0.653-0.931). The model showed acceptable calibration (Hosmer-Lemeshow P = 0.082) but validation was underpowered (15 events vs. recommended 100+ events for precise performance estimation). CONCLUSIONS: This study developed a PICC-RVT risk prediction model based primarily on Chinese patients with cancer. The model demonstrated moderate discrimination in preliminary single-center validation, but requires multi-center validation with adequate event numbers before clinical implementation. The model provides a framework for PICC-RVT risk stratification in Chinese and similar healthcare settings. SYSTEMATIC REVIEW REGISTRATION: PROSPERO (CRD420250651190).