Introduction

Peripherally inserted central venous catheters (PICCs) are commonly used in cancer patients to administer chemotherapy and supportive treatment1. PICCs are easy to insert, safe, and cost-effective2, and carry a lower risk of mechanical injury (e.g., pneumothorax) than centrally inserted venous catheters (CVCs)3. However, PICCs are associated with a higher risk of PICC-related venous thrombosis (PICC-RVT), one of the most common and harmful catheter-related complications. The incidence of symptomatic PICC-RVT varies from 6.7 to 10.6% in cancer patients4,5. PICC-RVT could cause patient discomfort, unplanned catheter removal, prolonged hospital stays, and increased financial burden6. In severe cases, PICC-RVT could lead to lethal pulmonary embolism and post-thrombotic syndrome7,8. Therefore, identifying patients at greater risk of PICC-RVT earlier is clinically significant, which helps guide subsequent prophylaxis.

Researchers have constructed some risk prediction models for PICC-RVT, such as Seeley9 and Michigan10, Peng et al.11, and Hu et al.12. Nevertheless, some prediction models were built based on retrospective studies with selection bias9,10,11, and some studies failed to consider catheter-related risk factors as candidate predictors9,12. The current guideline13 does not recommend a specific predictive PICC-RVT model and states that the Caprini model may be an alternative but with the risk of over-prophylaxis. To summarize, the predictive model for PICC-RVT needs to be further explored. Thus, we conduct a prospective cohort study to construct a new prediction model for PICC-RVT in patients with cancer. The result of our study would add evidence and guide clinicians in the risk assessment of PICC-RVT.

Methods

Study setting and population

Between April 2023 and January 2024, a prospective cohort study of patients with cancer receiving PICC insertions was conducted at a cancer center in Guangzhou, China. The study passed the review of the hospital Ethics Committee. All patients signed the informed consent before the study. Inclusion criteria are as follows: (1) received PICC insertion (4 F/5F, Bard Access System); (2) diagnosed with cancer; (3) age ≥18 years old; (4) received subsequent catheter maintenance in our hospital until catheter removal. Exclusion criteria include patients known or suspected to be allergic to the components contained in PICC catheters and PICC inserted in the lower limbs. This study was performed in accordance with the Declaration of Helsinki. This study was approved by the Ethics Committee of Sun Yat-sen University. The purpose and significance of the study have been explained to all participants before the investigation. The research followed the principle of confidentiality. All patients have signed an informed consent form.

PICC placement and maintenance

All catheters in the study are Bard brand, including 4 F Groshong PICC (Bard Access Systems Inc., Salt Lake City, UT, USA) and 4 F/5F Bard PowerPicc Solo2(Bard Access Systems Inc., Salt Lake City, UT, USA). All PICC catheterization had been inserted by an experienced (≥10 years) and qualified professional nurse (“PICC nurse”) under ultrasound guidance. PICC operation procedures were performed under the guidelines13. The catheter tip position was located with an intra-cavitary electrocardiogram (IC-ECG) and confirmed via X-ray. The PICCs used in this study were single-lumen (4-French) or double-lumens (5-French). Catheter maintenance was implemented by specialist nurses under the standard protocol and guidelines in our intravenous catheter clinic.

Data collection

Data extracted were (1) patient-related: gender, age, body mass index (BMI), smoking, cancer type, comorbidities (e.g. hypertension, diabetes, hyperlipidemia, coronary heart disease, severe lung disease, superior vena cava syndrome), history of thrombosis/anticoagulants/radiotherapy/chemotherapy/trauma/major surgery/CVC placement, the activity of the catheterized limb (evaluated with Muscle Strength Testing14); (2) laboratory test results before PICC insertion: D-dimer level, white blood cell count (WBC), and Platelet (PLT); (3) PICC insertion data: catheter type, number of catheter lumens, catheter dwell time, vein/arm of insertion, catheter-to-vein ratio, catheter tip position, puncture times, number of catheter malposition times. The researcher and the PICC nurse independently extracted all data from electronic health records and via face-to-face communication with patients on the day of catheterization. Following catheter insertion, all patients were followed up within 48 h and weekly until PICC-RVT or catheter removal occurred. Follow-up consisted of catheter maintenance and evaluation of signs and symptoms of PICC-RVT (e.g., pain, swelling). The PICC-RVT outcome was diagnosed by ultrasonography in the presence of signs and symptoms. On ultrasonic examination, PICC-related thrombosis appeared as an anechoic or hypoechoic image, partially or fully occluding the vessel lumen. Specialist nurses recorded data on the PICC-RVT occurrence during catheter maintenance. The PICC nurse, specialist nurses for catheter maintenance, and the radiologist did not participate in the study design. The researchers checked missing data weekly to ensure data quality.

Statistical analysis

Data analysis was performed with SPSS 29.015 for model development and R package (version 4.4.0)16 for model internal validation. Categorical data were quantified with numerical values and percentages, while continuous data were presented as mean±standard deviation. Statistical differences among categorical data were assessed using the Chi-square or Fisher’s exact tests. Those with a Pvalue < 0.2 in the univariate logistic regression analysis were entered into the multivariate analysis. Using stepwise regression analysis, the independent risk factors of PICC-RVT were determined. A nomogram prediction model was constructed based on the multivariable logistic regression analysis. The area under the receiver-operating characteristic curve (AUC) was reported for discrimination evaluation. AUC ranged from 0 to 1, with a higher AUC representing a higher discriminatory ability. An AUC greater than 0.9 indicates excellent discriminatory ability, 0.7 to 0.9 indicates moderate discrimination, and 0.5 to 0.7 means low discrimination17. The model calibration was evaluated using the Hosmer–Lemeshow test. For internal validation, a bootstrap with 1000 bootstrap resamples was used to quantify the overfitting of the developed model and optimism in its predictive performance. A two-sided P value of less than 0.05 was considered statistically significant.

Results

Study design and patients

From April 2023 to January 2024, 281 patients received PICC insertions at the cancer center, accounting for 30,799 catheter days. The sample size for the final analysis was 275 after the exclusion of 3 cases for transferring to other hospitals and 3 cases discontinuing the scheme for catheter maintenance scheme. 203(73.8%) patients were men, and 72 (26.2%) were women, with a mean age of 53.21 ± 12.67 years (range 18–87 years). Head and neck cancer (58.9%) are the most common cancer types, followed by digestive system cancer (12.0%) and reproductive system cancer (10.9%). Most patients (98.5%) were inserted with single-lumen catheters. 63.3% of the catheters inserted were polyurethane, and 36.7% were silicone. The most common punctured vein was the basilic one (76.3%), and the other 23.7% was the brachial one. The mean catheter dwelling time was 114.6 ± 45.6 days (range 3–274 days). The demographic-, laboratory-, and catheter-related data of both groups are shown in Tables 1 and 2.

Characteristics of PICCRVT in cancer patients

18 patients (6.5%) developed symptomatic PICC-RVT. The mean duration from PICC insertion to the occurrence of PICC-RVT was 83.4 ± 34.6 days (range 28–143 days). The most common sites of PICC-RVT were basilic veins (n = 14), followed by axillary veins (n = 11), and subclavian veins (n = 5).

Univariate and multivariate analyses for PICCRVT

Univariate analysis showed that diabetes requiring insulin (P = 0.036), trauma (within 1 month) (P = 0.174), major surgery (within 1 month and operation time > 45 min) (P = 0.017), reduced limb activities of the PICC arm (P < 0.001), catheter material (P = 0.196), number of catheter malposition during insertion (≥3 times) (P = 0.102) were correlated with PICC-RVT (P < 0.2) (Tables 1 and 2).

Table 1 Demographic characteristics (N = 275).
Full size table
Table 2 Laboratory test and catheterization characteristics (N = 275).
Full size table

Multivariate analysis revealed that diabetes requiring insulin (P = 0.035, OR:8.016, 95% CI 1.157–55.536), major surgery (within 1 month and operation time > 45 min) (P = 0.023, OR:0.023,95% CI 1.296–30.77), reduced limb activities of the PICC arm (P < 0.001, OR:6.687, 95% CI 2.024–22.09) were found to be independently associated with PICC-RVT (Table 3). Although catheter material was not statistically significant in the multivariate analysis with a Pvalue of 0.062 (OR:3.319, 95% CI 0.940–11.723), it was considered clinically important and supported by evidence18. Thus, it was included in the final model.

Table 3 PICC-RVT Risk factors in a stepwise entry process multivariate analysis.
Full size table

Nomogram for PICC-RVT and validation

Four risk factors including diabetes requiring insulin, major surgery, reduced limb activities of the PICC arm, and catheter material, were retained in the final model (Fig. 1). The AUC of the model was 0.796 (95% CI 0.695–0.897) (Fig. 2), indicating good discriminatory ability17. The results of the Hosmer–Lemeshow goodness-of-fit test (1.685, P = 0.194) indicated that the model had a satisfactory prediction effect, and no further calibration is required.

Fig. 1
Fig. 1The alternative text for this image may have been generated using AI.
Full size image

Nomogram for predicting PICC-related venous thrombosis.

Fig. 2
Fig. 2The alternative text for this image may have been generated using AI.
Full size image

The area under the ROC curve of the risk prediction model.

Discussion

This study observed an incidence of 6.5% for symptomatic PICC-RVT in cancer patients, comparable to results from relevant studies of 3.6%19 and 6.7%5. Nevertheless, the incidence of PICC-RVT was assumed to be higher as this study did not include asymptomatic cases. A meta-analysis20 found that asymptomatic cases (46.7%, 77/165) nearly accounted for half of the total PICC-RVT events. Thus, the thrombotic complications of PICC insertions remain a significant source of morbidity and mortality in patients with cancer.

The study identified four risk factors independently correlated with PICC-RVT: diabetes requiring insulin, reduced limb activities of the PICC arm, major surgery, and catheter material. A systematic review21 concluded that diabetes was a significant predictor of PICC-RVT. According to guidelines22, diabetes requiring insulin treatment indicates more severe hyperglycemia (≥ 16.7 mmol/L), leading to vascular injury and predisposing to a pro-thrombotic state23. Our study differentiates the impact of diabetes and diabetes requiring insulin treatment on the incidence of PICC-RVT. Thus, for diabetic patients receiving insulin therapy, caution should be exercised regarding the risk of thrombotic complications, and regular assessment and careful monitoring are required24.

Another patient-related risk factor identified was reduced limb activities of the PICC arm. This finding was in accord with many former studies9,25,26,27,28. The reason might be that reduced activity causes blood stasis and thrombosis due to prolonged bedridden and immobility29. However, the definition of reduced activity amount varied across studies such as “continuous bed rest > 72 h”27, “daily activity amount at the insertion site compared to pre-tubing activity26”, “activity with catheter side arm28”, use of the Eastern Cooperative Oncology Group (ECOG)30/Karnofsky Performance Status (KPS)31 score, “impaired activities of daily living25”, “recently bedridden9”. The reduced limb activities of the PICC arm in our study were evaluated with muscle strength testing (1–5). Patients with a score ≤ of 4 for the catheter side were considered to have reduced limb activities. In future studies, this prompts a uniform definition of reduced activity amount, including the extent (duration, intensity) and parts (catheter arm vs. whole body). Patients with reduced activity on the PICC arm should be educated about the importance of strengthening arm exercises. The 2024 Infusion Therapy Standards of Practice13 encouraged upper extremity exercise to reduce venous stasis. It further indicated that handgrip exercise with an elastic ball 3 or 6 times every day for 3 weeks could decrease the incidence of PICC-RVT in patients with cancer13. For patients unable to engage in active exercise, passive exercises are encouraged and future studies focusing on exercise regimens are required in this population.

Major surgery is a well-established risk factor for the development of thrombosis. Caprini32 first introduced the predictor of “major surgery > 45 minutes” (2 points) into their model for the risk of venous thromboembolism (VTE) in surgical patients. Our study builds on the work of Caprini et al. and further advances by focusing specifically on patients with PICCs. Previous literature has powerfully demonstrated independent correlations between surgery and PICC-RVT, such as “breast surgery” in Peng et al.11, and surgery lasting > 1 h33. The mechanisms behind surgery for thrombosis formation were the collective impact of decreased blood flow and stasis, hypercoagulable state, and endothelial injury34. Thus, those receiving PICCs with a history of major surgery lasting > 45 min should be assessed earlier and closely monitored for the risk of PICC-RVT.

A novel finding of our study was the catheter material, which has not yet been included in any models. Evidence18 has revealed that polyurethane catheters are more susceptible to PICC-RVT than silicone catheters. Nevertheless, thrombosis-induced polyurethane material allows a high-flow injection due to high stiffness35. Current guidelines do not recommend a specific catheter material. A choice of catheter material should be made based on the patient’s clinical requirement before catheter insertion. For those who require examinations of high-pressure injection (e.g., enhanced CT), the choice of polyurethane catheters should also consider the risk of PICC-RVT. For those who do not, silicone catheters are preferred.

The AUC of our RAM is 0.796, indicating good discriminatory ability. The Hosmer–Lemeshow goodness-of-ft test showed good calibration with a result of 1.685 (P = 0.194). The model incorporated four risk factors i.e., diabetes requiring insulin, major surgery (within 1 month and lasting > 45 min), reduced limb activities of the PICC arm, and catheter material, which are easy to access for risk assessment before PICC insertion. Our results showed that the model has acceptable predictive performance and could be a potentially helpful tool for predicting the risk of PICC-RVT. Compared with Caprini and Seeley, our model includes catheter-related risk factors, which makes it highly relevant to PICC-RT. Second, data on the risk factors of this model is more easily accessible (through inquiry and observation) than in Michigan (in need of laboratory tests). External validation of this model on a large sample size is needed.

Strengths and limitations

The study has some important strengths. First, our study is a prospective design. Second, a comprehensive list of patient-, laboratory-and catheter-related risk factors were considered as candidate predictors during model development. Nevertheless, our sample size was small from a single center, which could have diminished the ability to detect potentially significant risk factors. Besides, as it was mainly derived from head and neck cancers, this model should be used with caution when applied to other cancer types.

Conclusions

We developed and internally validated a nomogram prediction model, including diabetes requiring insulin, major surgery, reduced limb activities of the PICC arm, and catheter material. This model has a good predictive ability, which could help evaluate the risk of PICC-RVT in patients with cancer.