Fluid dynamic and in vitro blood study to understand catheter-related thrombosis

Abstract

Purpose:

Central venous catheters (CVCs) provide a direct route to the venous circulation but are prone to catheter-related thrombosis (CRT). A known CRT risk factor is a high catheter-to-vein ratio (CVR), or a large catheter diameter with respect to the indwelling vein size. In this study, the CVR’s effect on CVC hemodynamics and its impact on CRT is investigated with in vitro and in silico experiments.

Methods:

An in vitro flow loop is used to characterize the hemodynamics around CVCs using particle image velocimetry. In addition, CRT is investigated using an in vitro flow loop with human blood and clinical catheters. The wall shear rate of flow around the CVC is computed numerically. CVRs of 0.20, 0.33, and 0.49 and Reynolds numbers of 200, 800, and 1300 are evaluated. No flow is used through CVC lumens to model chronic indwelling catheters.

Results:

Results show CVR ≥ 0.33 promotes platelet-rich clot growth at the device tip and at an increased rate compared to lower CVR cases. A high wall shear rate gradient on the CVC tip and an extended wake distal to the tip exists for higher CVR cases, promoting the aggregation of platelets and subsequent stagnation for clot formation. Further, the combination of the CVR and Reynolds number are crucial to CRT potential, not the CVR alone. Specifically, thrombosis risk is increased with low (stasis driven) and/or high (platelet activation driven) flow conditions, with the CVR and CVC’s geometry playing an additional role in promoting fluid mechanic driven thrombus development. A high CVR (≥ 0.33) and high flow condition (≥ 1300) results in the highest risk for clot growth at the tip of the device; other locations of the device are at risk for thrombus development in lower flow conditions, regardless of the CVR. The importance of the device geometry and flow in promoting thrombus and fibrin sheath formation is also shown for the device investigated.

Conclusions:

This work demonstrates that the CVR, flow, and device geometry affect CRT. For clinical cases with CVR ≥ 0.33 and/or Re ≥ 1300, the device tip may be monitored more consistently for clot formation. Thrombosis risks remain on the entire catheter, regardless of the flow condition, for a CVR = 0.49. Device placement should be chosen carefully with respect to the combination of the Reynolds number and CVR. Further study is needed on the effect of catheterization to confirm these findings.

Introduction

Central venous catheters (CVCs) are essential vascular access devices used in applications such as hemodialysis, oncology, intensive care, and post-surgical care. CVCs provide a direct route to the venous circulation and reduce the need for frequent vascular punctures. The clinical benefit of CVCs, especially over the past few decades, is undeniable; with more than 5 million placed in the United States alone annually [1]. However, CVCs cause complications, such as catheter-related thrombosis (CRT), which constitutes 70% of all hospital cases of upper extremity deep vein thrombosis (DVT) [1]. CRT can present as a DVT, a venous thromboembolism, a clot within or attached to a CVC, and/or a fibrin sheath laminating a CVC surface [2]. CVC complication rates vary based on study type and patient risk factors. The incidence of CRT, specifically in the form of upper extremity DVT, is reported between 2.7–37% [2–5].

CRT consequences can include pulmonary embolism, loss of venous access, post-thrombotic syndrome, recurrent DVT, and death [2]. Multiple risk factors have been associated with CRT, including age, cancer, endothelial damage during insertion, longer indwelling catheter duration, and a catheter-to-vein ratio (CVR) > 0.45 [1, 6]. The CVR is defined as the outer diameter of a catheter divided by the inner diameter of the vein where the CVC is placed (Equation 1).

$$CVR=\frac{\mathit{CVC}\mathit{Outer}\mathit{Diameter}}{\mathit{Vein}\mathit{Inner}\mathit{Diameter}}$$ (1)

Previous clinical work has investigated these and other risk factors to understand the significance of each with respect to CRT formation [2, 7]. However, this body of work often presents conflicting data: for example, there are inconsistent reports on the optimal insertion vessel and side of the body to mitigate thrombosis [2, 8, 9]. Further, although the CVR has been found to affect patient outcomes significantly, clinical studies have not established a standard practice or vessel measurement location for the CVR: in some cases a value of 0.33 is recommended, while others suggest 0.45 [2, 10, 11]. The lack of CVR standardization in clinical settings is due, in part, to the fact that CVCs are manufactured in a wide range of sizes for multiple applications. Some patients need a CVC for short-term use (<1 week) and may not require a large diameter. However, clinical center standard sizes for short-term venous access in adult patients can range anywhere between 5–11.5 French, which may inadvertently result in an increased CVR (> 0.45) in a smaller patient [6, 11–13]. Other CVCs are large caliber to meet flow requirements for dialysis applications and a higher CVR may be unavoidable. Further, a multitude of CVC designs are also available on the market, all having different tip and side hole geometries. CVC tip and side hole designs have been found to significantly affect thrombosis risk from a fluid mechanics perspective, and clinical reports show that a decreased number of lumens (side holes) can minimize thrombosis risk [6, 14–17]. Some side hole and tip geometries have also been presented as more “optimal,” but there is not a consensus on the best design(s) to mitigate thrombosis [14, 18, 19]. Therefore, while a smaller CVC size with fewer side holes would be ideal to lower CRT risk, the wide clinical size range along with varying designs make subsequent thrombosis risk, especially as a result of a higher CVR, extremely challenging to avoid.

In addition to these clinical challenges, there is also a lack in mechanistic understanding of CRT development from a fluid mechanics and hematologic perspective [20]. Previous CVC numerical and flow visualization work investigated how flow develops around CVCs, but limited work has focused on the effect of the CVR [14, 21]. Lucas et al. paired a numerical analysis of blood flow around hemodialysis catheters with an investigation of adhered fibrin on catheters that had been removed from patients, observing that fibrin adhesion was greatest at the side hole rims, where the shear rate was high (> 5000 s⁻¹) [22]. Peng et al. and Park et al. found that CVC insertion leads to higher velocities and time averaged wall shear stress in the central vein and shortened blood residence times along the CVC [23, 24]. However, while Peng et al. showed a dramatically increased time averaged wall shear stress and flow velocity with CVC insertion, native venous flow has the potential to decrease after CVC insertion due to an increased hydraulic resistance in the vessel. Nifong and McDevitt previously supported this notion, finding that an increased CVR in vitro can reduce venous flow rates significantly [25]. These reports demonstrate a need in the field to experimentally investigate the effect of the CVR on CRT risk, while also completing experiments over multiple flow conditions to help highlight the potential scenarios of decreased flow.

Additionally, apart from Lucas et al., numerical or flow visualization experiments rarely incorporate blood studies in CRT work and have subsequently reported disagreements with clinical thrombosis data [24]. This demonstrates a need for more canonical CRT studies using both fluid mechanics and in vitro blood work to directly link the fluid mechanic changes upon CVC insertion to thrombosis risk [24]. Previous work focused on CRT blood experiments has assessed the thrombus potential of CVCs in venous-mimicking flow loops [26–29]. While such studies can build a framework for understanding CRT, it is important to note that the CVC location was not controlled in experimental flow loops in these studies. From a fluid mechanics perspective, this is problematic, because the local flow will be altered significantly if the catheter is in constant motion or its location is altered across experiments, resulting in erroneous interpretation, and/or inaccurately estimating clotting in vitro. Additionally, the CVR is generally not an experimental condition that is assessed in these studies, leaving open questions about its effect on thrombosis potential in vitro.

As such, no studies have investigated CRT in vitro across multiple flow conditions and CVRs using both fluid mechanics and blood studies with controlled CVC placement. In this study, CVC hemodynamics with altered CVRs (0.20, 0.33, and 0.49) and flow conditions (Reynolds number = 200, 800, and 1300) were investigated using particle image velocimetry (PIV), and in vitro blood experiments were conducted to analyze the effect of an altered CVR on CRT with a fixed CVC in a rigid flow tube. A dual lumen CVC design was used in this study to evaluate the impact of a side hole on the flow field and simultaneous thrombosis risk due to the clinical prevalence of multi-lumen device use. Numerical simulations were also performed as a supplemental analysis to compute the wall shear rate on the CVC surface to further understand the CRT risk in each condition. These in vitro and in silico experiments aim to decrease the knowledge gaps that exist for CRT by investigating fluid mechanics and thrombosis on commercially available medical catheters. The insights gained through this study can help guide clinical standardization for catheter size recommendations and device management, as well as suggestions for device design criteria to mitigate thrombosis.

Methodology

1. In Vitro Flow Visualization with Particle Image Velocimetry:

1.1. Particle Image Velocimetry Experiments:

Teleflex Arrow^® polyurethane dual lumen 5 French (Fr) catheters (Arrow International, Reading, PA; AK #14502) were rendered into SOLIDWORKS (Dessualt Systèmes, Wartham, MA) drawings to create acrylic models. A 5-Fr sized device (outer diameter ~1.67 mm) was chosen for its relevance in central access device use. The geometry of the CVCs was replicated by measuring each part of the CVC with calipers and implementing each feature into SOLIDWORKS (Dessualt Systèmes, Wartham, MA). The CVCs contained one side hole and had a tapered tip design. Three optically clear CVC acrylic models and an outer channel representing the superior vena cava (SVC) were fabricated (Atlantic Industrial Models, Essex, MA). The SVC channel (inner diameter 1.6 cm) and three CVCs (3.20, 5.28, and 7.85 mm diameter) were fabricated to evaluate the impact on the flow with altered CVRs (0.20, 0.33, and 0.49, respectively) (Fig. 1). This scaling approach for the acrylic CVCs was used due to the limitations of the resolvable dimensions in traditional PIV techniques; a smaller tube and CVC were not feasible. This approach also allowed comparisons across multiple side hole and tip sizes (Fig. 1). The acrylic CVC was solid and did not have hollow internal lumens (the side hole and tip therefore did not have flow through them to represent a clinically relevant period of no infusion) [30].

Fig. 1.

Cross sectional view of acrylic channels and CVCs. This cross-sectional centerline plane of the channel was visualized with PIV. Five regions of interest were imaged along the CVC from 16 – 36 cm downstream from the incoming flow. The five flow visualization regions are segmented with dashed black lines. The flow direction is shown with a thick black arrow. The flow straightener which stabilized the incoming flow is shown in gray. Acrylic channel inner diameter (d) = 1.6 cm. Each acrylic catheter has one side hole and a tapered tip. The streamwise length of each side hole is 4.25, 7.90, and 10.50 mm for the 0.20, 0.33, and 0.49 CVR cases, respectively. The length of the tapered section of the tip is 11 mm, 17 mm, and 25 mm for the 0.20, 0.33, and 0.49 CVR cases, respectively. a: 0.20 CVR, b: 0.33 CVR, c: 0.49 CVR.

A non-Newtonian blood analog fluid was created with water, glycerin (McMaster-Carr, Elmhurst, IL), sodium iodide (Chem-Impex International, Wood Dale, IL), and Xanthan gum (MP Biomedicals, LLC, Santa Ana, CA) (Table 1). Sodium iodide was used to match the refractive index (RI) of the acrylic channel and CVC (1.49), and the RI was measured with an Abbe Mark III Refractometer (Reichert Technologies, Depew, NY). Xanthan gum was used to attain similar viscoelastic properties to blood [31]. Sodium thiosulfate (LabChem Inc., Zelienople, PA) was also added to the fluid over time to maintain optical transparency. A Vilastic-3 Viscoelasticity Analyzer (Vilastic Scientific Inc., Austin, TX) was used to confirm the viscoelastic properties of the blood analog fluid, and data were matched to whole blood samples reported in Long et al [32]. The asymptotic dynamic viscosity of the blood analog fluid was 4.0 ± 0.17 cP. The blood analog fluid was seeded with 10 μm glass beads (Potters Industries, LLC, Malvern, PA) for flow visualization.

Table 1.

Non-Newtonian blood analog fluid formulation (wt. %).

Fluid Constituent	Sodium Iodide	Glycerin	Water	Xanthan Gum	Sodium Thiosulfate
Weight %	51.47	15.50	33.00	0.03	< 1.00

A two-dimensional PIV system (Fig. 2a) was used to capture flow around the acrylic CVCs using a continuous flow pump (Heartmate II, Abbott Cardiovascular, Chicago, IL), and measurements were completed as previously described [19, 33–37]. Briefly, a high-speed Terra PIV laser (Continuum, San Jose, CA) was coupled with a laser light arm (TSI Inc., Shoreview, MN), −25 mm cylindrical lens, and a 500 mm spherical lens to illuminate a single plane in the model. Two laser pulses, controlled by a LaserPulse Synchronizer (TSI Inc., Shoreview, MN) and Insight 4G^™ software (TSI, Inc., Shoreview, MN), excited the glass beads using a frame straddling technique. The laser sheet was 1 mm thick and illuminated the center plane of the flow field. Experimental conditions and corresponding condition names are shown in Table 2. All experiments were completed at room temperature. The Reynolds numbers were chosen to replicate flow conditions that CVCs can experience: the SVC flow peaks between Re 1300–1800 in systole and has an average value of around Re 800. If the CVC is placed in a smaller vein, the Re can range from 100–500; Re 200 is therefore used as the low flow condition [19, 38, 39].

Fig. 2: Technical schematics of experimental flow loops for PIV and in vitro blood studies (a, b). CVC segmentation and imaging techniques (c-f).

a: Technical schematic of the PIV system and experimental flow loop designed for flow visualization studies. Flow is driven by a continuous flow pump through the acrylic model and is captured using a 2-D PIV system. The flow direction is counterclockwise and noted with black arrows. The flow loop fluid volume was 750 mL. b: Flow loop for in vitro blood studies. Blood flow is driven through a closed loop and the rigid tube (gray) held the CVC in place. The flow direction is counterclockwise and noted with the thick black arrow. The blood flow loop fluid volume was ~150 mL. c: Schematic of the catheter inside the rigid tube and segmentation of the catheter into zones. Zone 1 represents the part of the catheter partially angled upon entering the channel; Zones 2 and 3 are straight sections of the catheter; Zone 4 includes the device’s side hole and tip regions. The side hole is traced with a black curved line for visualization purposes. The catheter’s top and bottom face approximations for confocal imaging purposes are shown segmented with a dashed red line. d: Silicone isolator in-between two coverslips for confocal imaging. e: Schematic of the three sequential images taken in each of Zones 1–3. Total image area acquired is 4.67 mm². f: Imaging regions of interest for Zone 4, highlighting the device side hole and tip with red boxes.

Table 2.

Catheter dimensions, flow conditions, and corresponding dT values for PIV experiments. Flow visualization data were collected for each CVR at the three flow conditions (9 Cases total).

Condition Name	CVR	Reynolds Number	Flow Rate (L/min)	PIV dT (μs)
CVR0.20 Re200	0.20	200	0.57	1500
CVR0.33 Re200	0.33
CVR0.49 Re200	0.49
CVR0.20 Re800	0.20	800	2.40	300
CVR0.33 Re800	0.33
CVR0.49 Re800	0.49
CVR0.20 Re1300	0.20	1300	3.60	150
CVR0.33 Re1300	0.33
CVR0.49 Re1300	0.49

A highspeed Phantom charge couple device camera (resolution ~33 μm/pixel) captured 200 image pairs at each location shown in Fig. 1 for each condition. A flow straightener was used to stabilize the incoming flow, and it comprised of a bundle of 14 1.15-mm diameter cylindrical tubes oriented parallel to the flow direction. Data were collected in the centerline plane of the channel at five locations between 16 and 36 cm downstream from the incoming flow (Fig. 1). The minimum velocity resolvable by the PIV system was 0.004, 0.020, and 0.040 m/s for the Re = 200, 800, and 1300 flow conditions, respectively. The continuous pump operated at 5470, 6900, and 8178 RPM for the Re = 200, 800 and 1300 flow conditions, respectively, and flow rates were confirmed during testing with a Transonic 6PXL Clamp-on Tubing Flowsensor (Transonic Systems, Ithaca, NY). Continuous flow is used in these experiments to obtain a foundational understanding of flow in catheterized vessels without the added complexity of pulsatility. A clamp was also used downstream of the pump to achieve the desired flow rates.

1.2. Flow Visualization Data Post Processing:

Images were masked manually by tracing along the inner wall of the flow channel and around the CVC surface in Insight 4G^™ software (TSI, Inc., Shoreview, MN). All processing was completed in Insight 4G^™ software: image pairs were cross-correlated using a direct correlator and processed with a bilinear peak, recursive Nyquist grid, and a final interrogation window size of 16×16 pixels with a 50% overlap. Vector conditioning and local validation were also applied during processing. Velocity fields were averaged across the 200 image pairs for each flow condition. Velocity data were postprocessed using TECPLOT 360 to create contour plots of the flow (TECPLOT, Inc., Bellevue, WA). These methods have been used in previous studies [19, 33, 36].

2. In Vitro Blood Experiments:

2.1. Human Blood Flow Loop Experiments:

An in vitro blood flow loop was designed to mimic a CVC inserted in venous flow conditions (Fig. 2b). A fresh Teleflex Arrow^® polyurethane dual lumen 5-Fr catheter (Arrow International, Reading, PA; AK #14502) was used in the flow loop for each experiment. A dual lumen design was chosen in this study to evaluate the impact of a side hole on thrombosis risk. A 5-Fr size device was chosen for its relevance in central access device use. The flow loop comprised of a rigid tube to hold the CVC in place, Tygon PVC tubing (Soft Plastic Tubing for Air and Water, McMaster-Carr, Elmhurst, IL), and a peristaltic flow pump (Masterflex L/S Easy-Load II, Model 77200–60, Cole Parmer, Vernon Hills, IL) (Fig. 2b). The rigid tube that held the CVC in place was fabricated with a stereolithography 3-D printer (FormLabs, Sommerville, MA) with clear resin and was used to hold the CVC in place during the tests for repeatability. Three tube sizes were fabricated to alter the CVR: 8 mm, 5 mm, and 3.4 mm inner diameter for 0.20, 0.33, and 0.49 CVRs, respectively. The CVC was also fixed with a wire and <0.5 mL of cured clear resin (FormLabs, Sommerville, MA) in its central lumen to provide it with structural rigidity. The CVC was inserted at a 30° angle and held in place by the inlet of the rigid tube (Fig. 2c). The rigid tube and the internal fixation technique of the CVC lumen ensured that the device remained in place along the tube’s centerline during each test for experimental control and reproducibility. Eight centimeters of CVC length were inside the channel for all experiments, and the side hole orientation was kept the same for each experiment, with its long axis oriented parallel to the flow direction. The rigid tube and flow loop were soaked in 3% by weight bovine serum albumin (Sigma Aldrich, Burlington, MA) overnight to minimize non-specific binding of adhesive blood proteins and cells during experiments.

Following a Penn State institutional review board-approved protocol, fresh human blood (250 mL) was collected for each experiment in a 3.2% sodium citrate solution (JT Baker, Phillipsburg, NJ). Blood was collected via venipuncture at a 1:9 volume ratio for a final sodium citrate wt. % of 0.32%, and blood donors were not fasted at the time of the donation. All blood donors were healthy, adult volunteer males and females and had a complete blood count test performed on their blood prior to donation. All donors abstained from taking steroids, nonsteroidal anti-inflammatory drugs, and allergy medications for 5 days before the blood draw. Donor whole blood was also tested before experiments with a rotational thromboelastometry machine (Werfen, Bedford, MA) to ensure coagulation and fibrinolytic activity were within published reference ranges (data shown in Supplemental Table S1) [40]. The blood was separated via density centrifugation at 300g for 30 minutes, followed by 2700g for 30 minutes to remove platelet-rich plasma and platelet-poor plasma, respectively. The blood was reconstituted to a fixed platelet count (223±14 platelets/mL) and hematocrit (40%/vol) for all experiments. Experiments were performed within 5 hours of the blood draw to reduce loss in platelet activity.

Blood was recalcified to 20 mM CaCl₂ (Sigma Aldrich, Burlington, MA) and then run through the closed flow loop (Fig. 2b) for 1 hour; flow rates were verified with 5PXL and 6PXL Clamp-on Tubing Flowsensors for the different CVR tubing sizes (Transonic Systems, Ithaca, NY). This 1 hour time duration for in vitro device thrombogenicity testing has been used previously [26–28]. Before the tests, the blood flow loop was primed and degassed using 1X phosphate-buffered saline (PBS); the blood then was pumped into the loop. The one empty CVC lumen was filled with 1X PBS and clamped to prevent backflow of blood during the experiment. There was no continuous flow through the CVC lumens during the 1-hour experiment; an extended period of no infusion through a CVC is common clinically, even with a multi-lumen device [30]. Experimental conditions are shown in Table 3 (n=6). Conditions in Table 3 are assigned condition names corresponding to those in the PIV studies, but with an additional “B”, denoting that they are from the blood studies.

Table 3.

Dimensional and flow conditions for in vitro blood experiments. The “B“ following the condition name is used to denote that the experiment is from the blood studies.

Case Name	CVR	Reynolds Number	Blood Flow Rate (L/min)	Channel Diameter (cm)	Pump RPM
CVR0.33 Re200 B	0.33	200	0.110	0.50	17
CVR0.49 Re200 B	0.49	200	0.115	0.34	21
CVR0.20 Re800 B	0.20	800	1.10	0.80	40
CVR0.33 Re800 B	0.33	800	0.70	0.50	50
CVR0.33 Re1300 B	0.33	1300	1.10	0.50	85
CVR0.49 Re1300 B	0.49	1300	0.78	0.34	100

The blood experiment CVR conditions were chosen by performing 0.33 CVR at each flow condition, along with an additional CVR at each flow condition; this expands our understanding across multiple CVRs from 0.20 – 0.49 and flows from Re = 200 to 1300. The pump operational conditions and channel size used in each experiment are also shown in Table 3. The tubing size used in the experiments matched that of the rigid tube diameter (+/− 5%). A peristaltic pump was used for all blood studies to avoid additional mechanical trauma and contact activation to the blood cells from a blood-wetted continuous pump head.

After each experiment, the loop was drained, and the CVC was gently rinsed with 1X PBS and assessed for macroscopic thrombus deposition. Macroscopic clot data collected included clot location on the device and clot surface area (length and width). The CVC was segmented into four “Zones” for data presentation, corresponding to 2-cm sections along the CVC (Fig. 2c). Clots that formed at the side hole and tip on the device were preserved in 4% paraformaldehyde (PFA) (Sigma Aldrich, Burlington, MA) for histological processing. After clot removal, the CVC surface was rinsed again with 1X PBS, cut into the four Zones with wire cutters, and fixed in 1% PFA (Sigma Aldrich). CVC samples were subsequently blocked and permeabilized with 0.5% Triton X-100 (Sigma Aldrich, Burlington, MA) in 5% goat serum (Sigma Aldrich, Burlington, MA). An indirect antibody labeling protocol was then used to identify surface adherent platelets and fibrin. In brief, CD41a and fibrinogen gamma chain primary antibodies (Thermo Fisher, Waltham, MA, Catalog #’s 14–0419-82 and MA534761, respectively) were used, followed by Alexa Fluor 488 and 647 conjugated secondary antibodies (Thermo Fisher, Waltham, MA, Catalog #’s A11001 and A21244, respectively) in 5% goat serum, respectively.

2.2. Confocal Imaging and Image Processing:

After immunostaining was complete, the CVC samples were immersed in 1X PBS in-between two coverslips (VWR International, Radnor, PA) separated by a silicone isolator (Electron Microscopy Sciences, Hatfield, PA) for imaging (Fig. 2d). Samples were imaged on an Olympus FLUOVIEW FV 3000 Confocal Laser Scanning Microscope with 488 and 640 nm lasers and a 10X dry objective (Olympus America Inc, Center Valley, PA). Image acquisition settings were kept constant for all experiments.

For Zones 1–3, three sequential z-stack images were acquired along the catheter length, for a total image acquisition area of 3.24 × 1.44 mm per zone (Fig. 2e). Z-stacks captured the catheter surface’s apex to its maximum visible width. Approximate image locations for each of Zones 1–3 were at ~7 cm (Zone 1), ~5 cm (Zone 2), and ~3 cm (Zone 3). This imaging process was repeated for the two “faces” of the device, and adhesion data averaged over the six total images for each Zone (Fig. 2c, e). For Zone 4, both the side hole and the tip were imaged in their entirety, as these regions were of interest due to known disruptions in flow that the geometrical changes can induce (Fig. 2f).

All confocal images were processed with a set of in-house ImageJ processing macros (National Institutes of Health). Briefly, z-stacks were projected to a single image as a maximum intensity projection and quantified for platelet and fibrin surface area coverage using a local thresholding technique developed by Phansalkar [41].

2.3. Histological Processing of Macroscopic Blood Clots:

Side hole and tip clots that formed on the catheters were preserved in 4% PFA (Sigma Aldrich, Burlington, MA) for 48 hours before being transferred to a Leica TP1020 automatic tissue paraffin processor (Leica, Wetzlar, Germany) for histological processing. The blood clots were then embedded in wax and sliced into 5 μm thick sections with a Leica Shandon finesse paraffin microtome (Leica, Wetzlar, Germany) and transferred to glass slides (VWR International, Radnor, PA). The slides were dewaxed with a Leica autostainer ST5010 XL (Leica, Wetzlar, Germany) and stained with Carstairs’ method (all stains purchased from Electron Microscopy Sciences, Hatfield, PA) before being mounted and cover-slipped (Sigma Aldrich, Burlington, MA) with xylene mounting media (Epredia, Kalamazoo, MI) [42]. Samples were imaged on a light microscope and analyzed via an in-house MATLAB R2023a (The MathWorks, Inc., Natick, MA) code to quantify the percentage of red blood cells (RBCs), fibrin, and platelets. In brief, a thresholding technique was applied to each image based on pre-determined hue, saturation, and value (H, S, V, respectively) ranges for blue (128<H<230; 28<S<254; 0<V<255), red (0<H<10; 14<S<255; 0<V<255), and yellow (12<H<50; 28<S<254; 0<V<255) which corresponded to platelets, fibrin, and RBCs, respectively. The area occupied by each color was normalized to the size of the image to quantify the percentage of the clot occupied by platelets, fibrin, and/or RBCs.

2.4. Statistical Methods:

Macroscopic clot data and confocal imaging data were analyzed via two-sample, unpaired t-tests across experimental conditions within the Re = 200 cases (CVR_0.33 Re₂₀₀ B, CVR_0.49 Re₂₀₀ B), the Re = 800 cases (CVR_0.20 Re₈₀₀ B, CVR_0.33 Re₈₀₀ B), and the Re = 1300 cases (CVR_0.33 Re₁₃₀₀ B, CVR_0.49 Re₁₃₀₀ B). T-tests were also used to evaluate any differences between the 0.49 cases (CVR_0.49 Re₂₀₀ B, CVR_0.49 Re₁₃₀₀ B). One way analysis of variance (ANOVA) was performed on all data in the 0.33 CVR experiments (CVR_0.33 Re₂₀₀ B, CVR_0.33 Re₈₀₀ B, CVR_0.33 Re₁₃₀₀). Statistical significance is assessed at the value of 0.05. Post-hoc Tukey tests were performed after ANOVA when tests resulted in significant differences. All statistical analysis was completed in MATLAB R2023 (The MathWorks, Inc., Natick, MA).

3. Computational Methods:

The Navier-Stokes equations (Equations 2–3) were solved for blood flow around the CVC in computational fluid dynamics (CFD) simulations to compute the wall shear rate (WSR) on the surface of the catheter. Computational investigation is included to supplement the PIV experiments by extracting WSR data. Numerical simulations were completed in OpenFOAM^® (Version 7, CFD Direct, England, United Kingdom). As the size of the rigid flow tube and CVCs was smaller in the blood experiments than in the PIV experiments, the WSR was computed numerically for both systems separately. The CVC and outer tube geometries in the numerical experiments matched those of the PIV experiments and the blood experiments, respectively. The SOLIDWORKS (Dessualt Systèmes, Wartham, MA) model of the acrylic CVCs was used for the numerical simulations that mimicked the PIV experiments. For the blood experiment simulations, the Teleflex Arrow^® 5 French catheters were scanned with a micro-computed tomography scanner (General Electric v|tome|x L300 Multi-Scale Nano/Micro CT System, Boston, MA, Penn State Center for Quantitative Imaging) to create a 3-dimensional model. Fifteen simulations were completed in total: 9 conditions from the PIV experiments and 6 conditions from the blood experiments (Table 2, Table 3).

$$\nabla \cdot \mathit{u}=0$$ (2)

$$\frac{\partial \mathit{u}}{\partial t}+\mathit{u}\cdot \nabla \mathit{u}=\frac{1}{\rho }\left(𢈒\nabla \mathrm{p}+\mathrm{\mu }{\nabla }^2\mathit{u}\right)$$ (3)

Blood was assumed to be an incompressible non-Newtonian fluid with a density of 1,050 kg/m³, and the kinematic viscosity was modeled using the Cross-Power equation (Equation 4) [43]:

$$\nu ={\nu }_{\mathrm{\infty }}+\frac{\left({\nu }_0𢈒{\nu }_{\mathrm{\infty }}\right)}{1+{\left(m\dot{\gamma }\right)}^j}$$ (4)

where $v_0$ is 754 cSt at low shear rates (≤ 0.1 s⁻¹); $v_0$ is 3.5 cSt at high shear rates (≥ 100 s⁻¹); m is 2.433 s; $j$ is 1.229; and $\dot{\gamma }$ is the shear rate in s⁻¹.

Mesh generations were performed using cfMesh (Creative Fields Holding LTD, London, United Kingdom) and implemented in OpenFOAM. A mesh independence study was performed using three mesh densities (0.64, 1.12, and 2.11 million cells for the coarse, medium, and fine meshes, respectively). A boundary layer mesh is used in each mesh with 5 layers to adequately capture near-wall phenomena for the WSR quantification. The flow simulations were performed by merging the controls of the hybrid pressure-implicit with the splitting of operators (PISO) and the semi-implicit method for pressure-linked equations (SIMPLE), pimpleFoam solver in OpenFOAM. A no-slip boundary condition was applied to surfaces such as the outer walls of the CVC and the inner tube walls. The inflow boundary conditions applied match those presented in Table 2 and Table 3 for the PIV and blood experiments, respectively, and a steady flow was used for all simulations. The pressure at the outlet of the flow tube was set to 0 Pa and a zero gradient pressure condition was applied to all walls and the inlet. The wall shear rate was calculated as the wall shear stress divided by the fluid’s viscosity.

The PIV and CFD flow fields were compared for model validation and to ensure the WSR data were an accurate representation of the in vitro tests. The error in the CFD result compared to the PIV data is quantified by $\mathit{E}$, as shown in Equation 5, where ${\left|u_{2D}\right|}_{PIV,i}$ represents the velocity data at each point as measured through PIV tests, and ${\left|u_{2D}\right|}_{CFD,i}$ is the velocity data predicted by CFD at each corresponding location, and ${\stackrel{-}{\mid u_{2D}\mid }}_{PIV}$ is the average in plane PIV velocity measurement. This error metric has been used for comparisons between CFD and PIV data in our previous work[19].

$$E(\mathrm{\%})=\frac{1}{n}\sum _{i=1}^n \left|\frac{{\left|u_{2D}\right|}_{PIV,i}𢈒{\left|u_{2D}\right|}_{CFD,i}}{{\stackrel{-}{\mid u_{2D}\mid }}_{PIV}}\right|\times 100$$ (5)

Results

1. Particle Image Velocimetry

Figure 3 shows the flow around the acrylic CVCs in each condition. The bulk flow direction is left to right in all flow visualization figures. The flow increases with each respective condition, and some velocity magnitude increases are observed with an increased CVR in each flow condition. There is stagnant flow in the side hole and in the wake distal to the tip in each case. Fig. 4 shows a closer view of the side hole of the acrylic CVCs. With a CVR = 0.20 at any flow condition, there is stagnant flow, but no prominent flow reversal in the side hole. However, flow reversal is observed with a larger side hole size as for CVRs = 0.33 and 0.49 in any flow condition. The presence of reversed flow also increases as the Re increases for both CVR = 0.33 and 0.49.

Fig. 3.

Flow field around CVCs at Re = 200, 800, and 1300. Solid black lines indicate the data collection locations as shown in Fig. 1.

Fig. 4.

The flow field in the side hole of the acrylic CVCs at each flow condition and CVR. The streamwise length of each side hole is 4.25, 7.90, and 10.50 mm for the 0.20, 0.33, and 0.49 CVR cases, respectively.

PIV data were sampled along a 10 mm long line parallel to the bulk flow direction distal to the CVC tip to assess the recovery of the flow through the wake in each case (Fig. 5a). Fig. 5b shows the relative velocity in the wake past the CVC tip. The relative velocity is equivalent to the sampled flow velocity divided by the peak incoming flow velocity for each Re ($U_{\text{Peak,Incoming}}=0.09\mathrm{m}/\mathrm{s}$, 0.20 m/s, and 0.60 m/s for Re = 200, 800, and 1300, respectively). The peak incoming velocity is calculated through Equation 5, where $Q$ is the incoming flow rate and $r$ is the acrylic tube radius (0.8 cm).

Fig. 5.

Evaluation of flow distal the CVC tip in PIV studies. a: Schematic of the sampling line to extract velocity values from PIV data in the wake for each case. b: Relative velocity distal the CVC tip for each flow condition and CVR. The inset shows a zoomed-in plot of X = 0 to 2 mm along the 1 cm long sampling line.

$$U_{\mathit{Peak,Incoming}}=\frac{2Q}{\pi r^2}$$ (5)

There is a lower relative velocity (U_Rel) in the 0.49 CVR cases compared to the 0.20 and 0.33 CVR cases across the 10 mm sampling distance for all three flow conditions (Fig. 5b). There is also an extended streamwise distance where U_Rel ≤ 0.05 for all cases with CVR = 0.49 compared to those at CVR = 0.20 and 0.33. Specifically, with CVR = 0.49, U_Rel ≤ 0.05 up to 2 mm distal the CVC tip in all flow conditions, where this is only observed in the lower CVRs (0.20, 0.33) at Re = 1300 (Fig. 5b, Inset). There is a slight change in U_Rel values observed from the 0.20 to 0.33 CVRs, specifically in the ability of the Re = 800 condition to recover flow past the wake more adequately in the 0.20 CVR. However, the CVR’s effect in flow recovery through the wake just distal to the CVC tip (X ≤ 2 mm) is most pronounced with either Re = 1300 or a CVR = 0.49.

2. In Vitro Blood Experiments

2.1. Macroscopic Clot Data:

Clot surface area data (length x width) were collected along with clot location for each experiment. Clot locations and clot sizes across each of the flow conditions and CVRs in the blood experiments are detailed in Table 4 and Fig. 6a, respectively. As clots did not form at the same frequency for each experiment, the number of each clot for each condition is notated in Table 4 and Fig. 6a. The clots formed at CVR_0.20 Re₈₀₀ B were all located at the side hole, however, for all cases with CVR > 0.20, clots that formed in Zone 4 were located at both the side hole and the tip (Table 4). With an increased flow condition and CVR (CVR_0.33 Re₁₃₀₀ B and CVR_0.49 Re₁₃₀₀ B), the Zone 4 clots developed more frequently at the tip than at the side hole.

Table 4.

Clot locations for each experimental condition (n=6). Zone 4 is divided into clot formation at the side hole and the tip.

Condition	Zone 1	Zone 2	Zone 3	Zone 4			Total Number of Clots
				# At Side Hole	# at Tip	Total Number of Zone 4 Clots
CVR0.33 Re200 B	1	3	5	2	2	4	13
CVR0.49 Re200 B	4	3	2	4	4	8	17
CVR0.20 Re800 B	0	0	0	5	0	5	5
CVR0.33 Re800 B	0	0	0	4	3	7	7
CVR0.33 Re1300 B	0	2	0	2	4	6	8
CVR0.49 Re1300 B	1	6	3	1	5	6	16

Fig. 6. Clot sizes (a) and histological clot characterizations (b-g) in in vitro blood experiments.

A: Box and whisker plot showing the surface area (mm²) of clots on the device for all blood experiments (n=6). (* = p < 0.05). b-g: Percent of platelets, fibrin, and red blood cells in the side hole and tip clots. Each case has the number of clots noted on it (“n= #”). b, c Show the percentage of platelets in the side hole and tip clots, respectively; d, e show the percentage of fibrin in the side hole and tip clots, respectively; f, g show the percentage of RBCs in the side hole and tip clots, respectively. The black dot corresponds the data mean; the red line is the median. Any outliers are shown with a red hash mark.

Clots in the CVR_0.20 Re₈₀₀ B condition were larger than those in the CVR_0.33 Re₈₀₀ B condition (Fig 6a, p<0.05). Further, clots were larger in size in the CVR_0.33 Re₂₀₀ B condition as compared to the CVR_0.33 Re₈₀₀ B condition (Fig 6a, p<0.05) and also formed along all Zones of the catheter, rather than in just Zone 4. In the CVR_0.49 Re₂₀₀ B, CVR_0.33 Re₁₃₀₀ B, and CVR_0.49 Re₁₃₀₀ B cases, clots also formed in multiple zones (Table 4). There was no significant difference between clot sizes in the CVR_0.33 Re₁₃₀₀ B condition and the CVR_0.33 Re₂₀₀ B or CVR_0.33 Re₈₀₀ B conditions. There was also no significant difference between clot sizes in the CVR_0.33 Re₁₃₀₀ B condition and the CVR_0.49 Re₁₃₀₀ B condition or between clots in either 0.49 CVR condition.

Side hole and tip clots that formed in experiments were processed in histological sections and data are shown in Fig. 6b–g. As clots did not form at the same frequency at the side hole and tip for each experiment, the number of each is notated on Fig. 6. Additionally, due to difficulties in histological processing, a small fraction of clots formed in experiments did not make it to the end of the histological process (4 out of 36 total clots). Therefore, the exact number of clots that were characterized is noted on Fig. 6b–g for each case. The percentage of RBCs was lowest for all clots across the five experimental conditions. In the CVR_0.49 Re₂₀₀ B, CVR_0.33 Re₁₃₀₀ B, and CVR_0.49 Re₁₃₀₀ B conditions, the percentage of platelets in the side hole clots was higher compared to the other conditions. A similar trend is observed for tip clots, where the CVR_0.49 Re₂₀₀ B, CVR_0.33 Re₁₃₀₀ B clots have a higher average percentage of platelets than other conditions. Interestingly, the mean platelet percentage in the tip clots in all conditions increased at least slightly as compared to side hole clots, reflecting a platelet-rich behavior in all tip clots. Additionally, the mean percentage of platelets in tip clots increases with a higher CVR or Re (Fig. 6c). The fibrin percentages follow the opposite trend for each case.

Fibrin adhesion to the CVC surface as quantified by confocal microscopy is shown in Fig. 7. Platelet adhesion was <1.5 surface area coverage (%) in all zones and is therefore not included. The side hole generally has a higher average adhesion compared with other Zones and the tip for all conditions (Fig. 7a). Zone 1 and the tip show slight increases in overall adhesion compared to Zones 2 and 3 in some cases. Fibrin adhesion is higher at the tip in the CVR_0.33 Re₈₀₀ B condition compared to both CVR_0.33 Re₂₀₀ B and CVR_0.20 Re₈₀₀ B (p < 0.05) (Fig. 7a). There is a slight increase in adhesion in some cases for CVR_0.33 Re₁₃₀₀ B, reflected by outliers in the data sets, though this was not significant. Further, there is a significant increase in adhesion of fibrin in Zones 1 and 3 for CVR_0.49 Re₂₀₀ B (p < 0.05) compared to the CVR_0.33 Re₂₀₀ B condition (Fig. 7a). There is also a significant increase in fibrin adhesion at the side hole and tip for the CVR_0.33 Re₁₃₀₀ B case compared to that in the CVR_0.49 Re₁₃₀₀ B case (p<0.05) (Fig. 7a). Fibrin adhesion was also locally increased within the side hole pocket/corners (Fig. 7b, white arrows) and on the side hole rims (Fig. 7b, dashed white lines) across all cases. Representative images are shown for select cases to highlight the fibrin adhesion changes across conditions and are shown in Fig 7c.

Fig. 7. Fibrin adhesion to the CVC in in vitro blood experiments quantified as surface area coverage (%) (a), with representative adhesion images (b, c). Quantified data are sorted in box and whisker plots for each Zone (Zone 1, Zone 2, and Zone3) and the side hole and tip.

a: Box and whisker plots of fibrin adhesion (surface area coverage, %) to the CVC in each Zone and at the side hole and the tip (Zone 4). Red line denotes the median, black dot denotes the mean, and the red hash denotes an outlier. The * denotes statistical significance (* = p < 0.05). b: Representative confocal images of fibrin and platelet adhesion to the side hole. Adhesion is increased near the side hole corners (white arrows), and on the side hole rims (white dashed lines). c: Representative confocal images of Zone 1 (CVR_0.33 Re₂₀₀ B and CVR_0.49 Re₂₀₀ B), the Side Hole (CVR_0.33 Re₁₃₀₀ B and CVR_0.49 Re₁₃₀₀ B), and the Tip (CVR_0.20 Re₈₀₀ B and CVR_0.33 Re₈₀₀ B). Increases in fibrin adhesion are observed for CVR_0.49 Re₂₀₀ B in Zone 1 and for CVR_0.33 Re₁₃₀₀ B at the Side Hole. Dense aggregates are present on the tip at CVR_0.33 Re₈₀₀ B. (Fn = fibrin; P = platelets; Scale bar = 400 um).

3. Computational Fluid Dynamics Results

3.1. Mesh Independence Study and Comparison to in vitro PIV Data:

A cross-sectional view of the coarse, medium, and fine meshes is shown in Fig. 8a. A boundary layer mesh with five layers was used to capture the near wall phenomena on the CVC for each mesh. The flow field at each mesh density is shown in Fig. 8b (condition used: CVR_0.33 Re₈₀₀); minimal differences are observed across the three meshes. As shown in Table 5, the maximum velocity values did not change significantly across any of the mesh types (<0.01%). There is a 6.9% difference in the maximum wall shear stress between the medium and coarse meshes, and there is < 1% difference between the prediction in the medium versus the fine mesh (Fig 8c., Table 5). As such, the medium mesh was used for all CFD predictions in this study.

Figure 8.

Computational mesh independence study and comparison between CFD and PIV study flow fields. Axes are provided for reference in each panel, and flow is left to right. a: Three meshes used for the mesh independence study. The meshes contained 0.64, 1.12, and 2.11 million cells for the coarse, medium, and fine mesh densities, respectively. A five-layer boundary layer mesh is shown to capture the near wall flow phenomena for WSR calculation. b: Flow field in the coarse, medium, and fine meshes for the CFD studies. c: Wall shear stress around the CVC side hole in the coarse, medium, and fine meshes, as predicted from the CFD simulations. d: Comparison between CFD and PIV data for two cases. The CVC is outlined in blue. The green line represents the CFD prediction, and the red line represents the PIV data.

Table 5.

Number of cells and maximum velocity and wall shear stress magnitudes in the coarse, medium, and fine meshes used for the mesh independence study for CFD experiments.

Mesh	Number of Cells (millions)	Maximum Velocity Magnitude (m/s)	Maximum Wall Shear Stress (Pa)
Coarse	0.64	0.33	4.45
Medium	1.12	0.33	4.16
Fine	2.11	0.33	4.19

Further, the flow field is compared in the xy cross sectional plane to the PIV data at the CVR_0.20 Re₁₃₀₀ and CVR_0.33 Re₈₀₀ cases (Fig 8d). There are very minor differences observed in the flow field between the CFD and PIV data for both cases. The error metric, E, was calculated for each case, and was equivalent to 4.6% and 6.2% for the CVR_0.20 Re₁₃₀₀ and CVR_0.33 Re₈₀₀ cases, respectively. Based on these results, the CFD method using the medium mesh was deemed suitable for the WSR analysis for this study.

3.2. CVC Wall Shear Rate Contours:

The computed WSR on the CVC surface in the PIV studies is shown for the side hole and tip regions in Fig. 9 for each flow condition and CVR. The trends observed in Fig. 9 follow an increased flow rate and CVR leading to an overall increased WSR. The side hole rims experience a locally increased WSR, while a decreased WSR is observed inside the side hole and on the tip surface where the device geometry begins to taper. The maximum WSR on the CVC in the CVR_0.20 Re₁₃₀₀, CVR_0.33 Re₁₃₀₀, and CVR_0.49 Re₁₃₀₀ conditions reaches 2200 s⁻¹, whereas WSRs are within the range of 100–300 s⁻¹ in the CVR_0.20 Re₂₀₀, CVR_0.33 Re₂₀₀, and CVR_0.49 Re₂₀₀ conditions. The WSR in the side hole increases with an increased CVR and flow condition. The slight asymmetries in the flow seen in Fig. 3 are captured in these data, though resulting WSR magnitudes are not significantly different between the top and bottom surface of the CVC.

Fig. 9.

WSR contours from numerical simulations for each condition in the PIV studies. Top (left column) and bottom (right column) views of the CVC are provided. The bulk flow direction is left to right.

Due to spatial restrictions in PIV experiments, the PIV setup was scaled up as compared to the blood experimental tube and CVC. As such, the wall shear rate data were computed for the blood study set up as well. The computational WSR contours on the CVC surface in the blood experimental conditions are shown in Fig. 10. The inside of the side hole has a lower WSR in each case as compared to the rest of the catheter, though this value increases with an increased flow condition (i.e. CVR_0.49 Re₁₃₀₀ B). The WSR increases on the side hole rims, reaching values of 10,000–15,000 s⁻¹ in the CVR_0.33 Re₁₃₀₀ B and CVR_0.49 Re₁₃₀₀ B cases. The WSR is greater in CVR_0.33 Re₈₀₀ B as compared to CVR_0.20 Re₈₀₀ B and CVR_0.33 Re₂₀₀ B. The WSR increased significantly with a high CVR even with a decreased flow condition (CVR_0.49 Re₂₀₀ B). Similar to the computational WSR contours shown in Fig. 9, a decreased WSR is observed on the tip surface where the device geometry begins to taper. The wall shear rate gradient (WSRG) on the tip was computed from the data shown in Fig. 10 and is shown in Table 6. The WSRG is calculated as the spatial change in WSR magnitude on the tip from the point it begins to taper to its distal end point. The overall WSRs and WSRG are the greatest in CVR_0.49 Re₁₃₀₀ B. The WSRG increases with an increasing CVR and increasing flow condition.

Fig. 10.

WSR contours from numerical simulations in the six blood experimental conditions. An expanded view of the side hole is provided for each case to highlight local changes.

Table 6.

Numerical spatial wall shear rate gradient along the device’s tip in the six blood experimental conditions.

Condition	Wall Shear Rate Gradient (cm−1*s−1)
CVR0.33 Re200 B	949
CVR0.49 Re200 B	5,993
CVR0.20 Re800 B	2,194
CVR0.33 Re800 B	7,578
CVR0.33 Re1300 B	13,421
CVR0.49 Re1300 B	18,848

Discussion

This study provides an initial framework to study CRT and the CVR through fluid mechanics and in vitro blood studies. An acrylic CVC model was used to analyze the flow field around CVCs using PIV and a rigid tube with a fixed CVC was used to control blood experiments and investigate thrombosis on commercially available medical catheters. The catheter was kept at the center line of the channel for all experiments for experimental control and reproducibility. CFD simulations were performed in OpenFOAM to analyze the wall shear rate on the CVC surface in PIV and blood experiments. Both flow visualization and blood experiments were completed over multiple CVRs and Reynolds numbers to understand thrombosis risks across a comprehensive set of physiologic conditions.

Results first show that with an increasing CVR, the tip is a critical area for CRT under any flow condition: with CVR = 0.20, tip clots did not form in any experiment, but with CVR ≥ 0.33, tip clots formed across all conditions (Table 4). The presence of tip clots at CVR ≥ 0.33 and the variation in tip clot frequency in different conditions as the CVR increases can be explained through the effective hemodynamic changes in each case. More specifically, PIV data show that the flow in the wake distal the tip in CVR_0.33 Re₈₀₀ takes a longer streamwise distance to recover as compared to CVR_0.20 Re₈₀₀ (Fig. 5b, Inset). Additionally, the WSRG along the tip surface in CVR_0.33 Re₈₀₀ B is 3.45 times higher than that in CVR_0.20 Re₈₀₀ B (Table 6). A shear deceleration zone in a cardiovascular device such as the WSRG present on the tip surface has been found to preferentially promote platelet aggregation which can increase clot formation as compared to the cases with lower WSRGs [44]. Further, the extended wake zone just distal to the tip in the higher CVR case can stagnate the aggregated platelets and other agonists to form a clot. This also extends to the CVR_0.49 Re_200, CVR_0.33 Re₁₃₀₀, and CVR_0.49 Re₁₃₀₀ cases where U_Rel < 0.05 up to 2 mm distal of the CVC tip and the WSRGs on the tip are supraphysiologic (> 5000 cm⁻¹s⁻¹; Fig. 5b, Table 6). Subsequently, the number of tip clots in these three conditions was the highest of all blood experiments (Table 4), and tip clots become the dominant clot type in Zone 4 at Re = 1300 for either CVR. These data show that the CVR and flow condition play a role in the tip clot development, with conditions at CVR ≥ 0.33 and Re ≥ 1300 promoting the most dominant tip clot formation.

A comparison can also be made between the CVR_0.33 Re₂₀₀ and CVR_0.49 Re₂₀₀ cases, where the relative velocity past the tip is markedly decreased in the 0.49 CVR case compared to that in 0.33 CVR. Further, the WSRG is 6.3 times higher in the CVR_0.49 Re₂₀₀ B case and resulting number of clots in the CVR_0.49 Re₂₀₀ B case is then double that of that in the CVR_0.33 Re₂₀₀ B case. In CVR_0.33 Re₂₀₀ B tip clots formed in lower frequency than other high CVR conditions, which can be explained primarily by a decreased impact on flow recovery out of the wake distal the tip and a lower surface WSRG compared to other cases (Fig. 5b, Table 6). Additional comparisons can be made with respect to flow between the CVR_0.33 Re_200, CVR_0.33 Re_800, and CVR_0.33 R₁₃₀₀ cases, where the relative velocity in the wake consistently decreases and WSRGs increase from Re = 200 up to Re = 1300. Subsequently, the clot number at the tip increases for each condition in CVR 0.33 from Re = 200 to Re = 1300. These results show that there is an increased risk on tip clot potential at a high CVR (≥ 0.33) based on the flow condition chosen: with a relative higher CVR the tip clot formation is at higher risk, and this risk generally increases with an increasing Reynolds number.

The proposed fluid mechanic and CVR mechanism of tip clot formation is also supported by histology data showing a slight increase in the mean percentage of platelets in tip clots compared to side hole clots as shown in the CVR_0.33 Re₂₀₀ B, CVR_0.33 Re₈₀₀ B, CVR_0.33 Re₁₃₀₀ B, and CVR_0.49 Re₁₃₀₀ B cases (Fig. 6b–g). Further, the highest platelet percentage in tip clots are in CVR_0.49 Re₂₀₀ B, CVR_0.33 Re₁₃₀₀ B, and CVR_0.49 Re₁₃₀₀ B cases showing the potential for high WSRGs to aggregate platelets and subsequently stagnate them in the flow to form a platelet rich clot. There is also an increased RBC percentage found in the tip clots at Re = 800–1300 in the blood experiments, showing the impact of the stasis in the large wake in these cases on tip clot formation and constituents (Fig. 6b–g). Although the number of clots analyzed is small (n=2–3) for each case, these findings are encouraging towards a potential mechanism of CVR- and flow- induced CRT.

These data show the impact that the flow and CVR have on tip clot formation and can explain one potential mechanism impacting CRT (platelet driven aggregation in a high WSRG zone and subsequent stagnation in a spatially extended wake). Previous clinical work has also reported that a CVR > 0.45 measured specifically at the tip level of the device leads to more catheter failures (CRT or catheter-related blood stream infections) for mid-line and peripherally inserted catheters, supporting the finding that the tip may be the region of concern for CRT in high CVR cases [45]. This previous work and our study therefore suggest that if a CVC is placed at a CVR ≥ 0.33, the device tip should be monitored consistently with imaging or other clinical intervention for clot formation to reduce catheter occlusion and/or embolus potential [45]. Further, our data show that depending on the vessel where the CVC tip resides and there is a higher Re (~1300), the tip clot risk may be even higher if the CVR ≥ 0.33. These insights on the tip clot growth potential are significant in the field and can guide clinical recommendations for device placement and monitoring if the CVR must be ≥ 0.33 for the clinical application or infusion requirement(s).

The side hole is also a major source of thrombus deposition as these clots formed in all cases (Table 4), which is consistent with previous studies [16, 46]. Wall shear rate data demonstrate that the side hole causes both locally low WSRs inside the side hole promoting protein and platelet adhesion, as well as high WSRs on the side hole rims promoting increased fibrin adhesion and potential for platelet activation and/or aggregation (Fig. 10). High shear rates and increased fibrin adhesion on side hole rims have also been reported in previous in vivo studies [47]. The stagnant flow within the side hole promoted clot formation as well as fibrin deposition in the distal corners of the hole pocket (Fig. 7b, white arrows). The frequency of side hole clot formation in each case is related to the shear rate magnitude within the side hole. More specifically, the WSR is highest in the side hole at CVR_0.49 Re₁₃₀₀ B, resulting in the fewest side hole clots formed (Table 4, Fig. 10). Additionally, the platelet percentage in the side hole clots in CVR_0.49 Re₂₀₀ B, CVR_0.33 Re₁₃₀₀ B and CVR_0.49 Re₁₃₀₀ B were greater than other cases, which correlated to CVC surface WSRs > 3,000⁻¹ s (Fig. 6, 10), which presumably facilitated platelet activation in the bulk flow [48]. Over the 1-hour experiment, activated platelets likely adhered to the device in the low flow zone within the side hole; this further shows the potential of platelet-rich clots to form in high CVR cases.

Further, PIV data show that flow recirculation within the side hole increases with increasing Re and side hole size (Fig. 4). Enhanced recirculation could promote washing of the side hole surface to break up microaggregates and adhered protein layers. However, in the blood experiments, the side hole size was kept consistent in all cases, and the side hole WSR is low for most cases (< 400 s⁻¹) (Fig. 9). Therefore, this flow may have been insufficient to wash the side hole to prevent clot formation.

The mentioned PIV data also suggest that a larger side hole size in vessels with increased flow (Re > 800) may help with side hole washing to reduce thrombosis risk in clinical applications. The CVC geometry, such as that of the side hole and/or tip, is of major interest in decreasing thrombosis potential. Mareels et al. completed PIV and CFD studies on several different CVC tip designs, showing significant changes in hemodynamics with slight geometrical alterations [49]. The influence of fluid mechanics and device geometry on adhesion in the side hole and side hole corners has also been shown in this study (Fig. 4, 8) and in our previous in silico study using Teleflex Arrow catheters [19]. Our previous work showed that removing the side hole corners to reduce stagnation zones as well as reducing the tip cross sectional area greatly lowered thrombus development after 15 minutes in silico [19]. Here we have further shown the differences in the sizes of the side holes and tips with our PIV experiments: with a greater side hole size, there is ability for greater vortex formation and flow recirculation, which may increase washing of the device, and with a smaller tip size, the wake flow recovery is less impacted (Fig. 4, 5). Previous work has also shown that an increase in the side hole size, to a point, can help mitigate thrombosis risks [50]. In total, these data together highlight thrombosis risks due to device geometry and suggest that a device redesign could consider removal of the side hole corners as well as a larger side hole area/size to increase the surface washing, along with a lower cross sectional area tip to mitigate thrombosis in these regions. Further, clinical incidence of thrombosis at the side holes can be reduced if a larger side hole size is used in high flow applications due to potential influence of vortical, washing flow.

This work further demonstrates that the combination of both the flow condition and the CVR are critical to thrombus formation and not necessarily the CVR alone. Comparing an increased CVR in the same flow condition (CVR_0.20 Re₈₀₀ B vs CVR_0.33 Re₈₀₀ B), the flow velocities and WSRs increase, resulting in a significantly decreased thrombus size in the higher CVR case (Fig. 3, Fig. 6, Fig. 10). However, in this case, increased fibrin adhesion in CVR_0.33 Re₈₀₀ B is observed in Zone 1 and the tip, where the local fluid mechanics are altered due to bending and geometrical changes in these regions obstructing flow. Further, thrombus risk increased at the tip with the higher CVR as discussed previously. With a decreased Reynolds number within the same CVR (CVR_0.33 Re₂₀₀ B vs. CVR_0.33 Re₈₀₀ B), clot formation drastically increased in frequency, location, and size. At lower Reynolds numbers, all surfaces of the device are susceptible to thrombus deposition: flow visualization data show the lowest surface velocities along the device, formed clots have the greatest frequency and sizes across all Zones in CVR_0.33 Re₂₀₀ B and CVR_0.49 Re₂₀₀ B (Fig. 3, Table 4). At high Reynolds numbers (CVR_0.33 Re₁₃₀₀ B), increased adhesion and clotting outside of Zone 4 can be caused by high shear rate phenomena, as demonstrated by supraphysiologic shear rates on the CVC surface (> 5000 s⁻¹) (Fig. 10). The fluid shear rate may be sufficiently high to promote mechanical platelet activation and von Willebrand factor unfurling causing thrombosis at an increased rate and on smooth surfaces of the catheter outside of Zone 4[48, 51]. Importantly, SVC flow can reach Re = 1400 in the cardiac cycle, bearing the potential for the high shear phenomena to promote CRT [52]. In total, both the CVR and flow condition should be considered upon CVC insertion, not just the CVC size. Risks remain on the entire CVC (regardless of the CVR) at low flow conditions (Re = 200), whereas some risks may be mitigated with a smaller CVC at higher Reynolds numbers (800–1300) to decrease overall WSRs on the CVC surface and WSRGs on the tip. The tip remains a problem area for high CVRs, and this risk is increased with an increased flow condition. A CVR of 0.49 is not recommended due to the thrombosis frequencies regardless of the flow condition.

As the formation of fibrin sheaths is known to be a driver of CVC failure (catheter occlusions, thrombosis, and infections) in vivo, this study is valuable in understanding where risks remain for fibrin sheath development[53]. The CVCs used in this study are made from medical-grade polyurethane (PU). Surface chemistry, hydrophilicity, and blood flow conditions have been found to strongly influence fibrinogen adhesion to PU [54, 55]. Additionally, the microstructure of the PU surface can influence fibrinogen adhesion: the developed fibrin strands seen in Fig. 7c (CVR_0.49 Re₂₀₀ B) are typical of smooth PUs, while some micropatterning has been found to decrease overall fibrin adhesion and strand development in vitro [56]. As mentioned, the tip in CVR_0.33 Re₈₀₀ B, Zones 1 and 3 in CVR_0.49 Re₂₀₀ B, and the side hole corners and rims in all cases were potent drivers of fibrin adhesion, demonstrating a fluid mechanic and geometrical influence of the CVCs on fibrin adhesion. The increase at the tip in CVR_0.33 Re₈₀₀ B is likely related to increased tip clot formation, while the increases seen in the side hole are likely both low shear-rate and geometrically driven. The increases in CVR_0.49 Re₂₀₀ B at Zones 1 and 3 are likely due to increased WSRs allowing platelet activation and adhesion leading to fibrin formation. Further, this may suggest that the major cause of fibrin sheath formation on smooth regions of the CVC (non-geometrically irregular regions) is due to locally increased shear rates. These insights also suggest that CVC and its surface can be redesigned with reduced sharp corners and/or with surface micropatterning to lower potential for fibrin sheath development. Further, if the local shear rates can be reduced by insertion of a smaller CVC, this may be beneficial for a lowered overall fibrin sheath formation in higher flow cases.

Overall, the PIV experiments support our understanding of the CVC clotting mechanisms that occur at different regions. The blood experiments closely supported the data generated using PIV, specifically the side hole and tip data, thus demonstrating that this model is an accurate depiction of the fluid mechanics in catheterized vessels. This model can be used in future studies (along with another supplementary CFD study to compute WSRs) to evaluate additional catheter designs, placements, and/or locations. Although the CVC size was altered for each CVR in the PIV tests, the results are still valuable and represent the relative changes with geometry and obstruction due to a higher CVR on CRT risk. As the CVC size was kept constant for the blood experiments, the influence of the increased CVR on tip clot formation may have been related more to increased WSRGs on the device surface at the tip. However, the combination of the PIV and blood experiments provide more clarity as to how the CVR and geometrical changes at the side hole and tip may cause thrombosis risks in clinical cases, and these results can be used to inform clinical standards for catheter placement.

Through this study we have shown that the tip is a critical area of concern for thrombosis development when the CVR ≥ 0.33, and this is heightened with an increasing Reynolds number. A CVR = 0.49 will result in thrombus formation in most cases, and this CVR should be avoided. Further, we have shown that there are areas of the CVC which can be redesigned to mitigate thrombosis and fibrin sheath development with suggestions based on both previous studies and the data presented here. Last, data in this study show that the combination of the CVR and the flow condition will impact the thrombosis risk, not just the CVR alone, suggesting that different vessels may require different CVR standards. Clinical guidelines should leverage in vitro and in vivo data to inform CVR and CVC placement standards, optimizing the combination of the two, and monitoring the device in locations that may be prone to thrombosis.

Limitations:

A primary limitation is that the in vitro blood studies are run for one hour. Though this is a widely accepted time duration for in vitro studies, CVCs are placed in vivo for weeks to months [26–28, 57]. Regardless, the insights from this study are valuable and lay a foundation for CRT understanding across multiple flow rates and CVRs. The recirculatory blood flow loop without the catheter inserted can also cause some basal cell damage during the 1-hour experiment. However, we mitigate this issue by coating the materials with bovine serum albumin overnight before completing any experiments, decreasing contact activation potential and adhesion of blood products in the loop. Clinical grade tubing was not used for this study, however, this is consistent with many other in vitro thrombogenicity testing evaluations [28]. Further, prior research has shown that there is a basal level of platelet activation and hemolysis in similar flow loops, but the background effects are significantly different from the test conditions using the device of interest [27, 34, 58]. Therefore, the use of in vitro flow loops does provide some limitation in the form of basal cell activation, but this is a small effect on overall results and the insights are still deemed valuable

This study has also only considered an ideal centerline catheter placement. However, CVCs are flexible and respond to patient movement and may rest against vessel walls, initiating the endothelium to cause mural thrombosis. In this study, the CVC is kept in one place for experimental control and reproducibility. Future experimental studies should consider CVC positioning and its inherent flexibility. A further limitation is that there were slight asymmetries in the flow field due to manufacturing issues with the acrylic CVC models causing minor bending, the effect of which was most prominent at CVR = 0.49 (Fig. 4). Still, these data captured an adequate understanding of the flow for the purposes of this study. This study also used a rigid flow tube for the venous model with the CVC inserted, while more compliant venous structures are the physical reality. This will have some effect on the results. Notably, the resistance within the tube will not change over time, whereas in vivo, the vein would slightly expand and relax over the cardiac cycle. If the vein does partially expand over a cardiac cycle, the flow speed will decrease and subsequent hydrodynamic drag on surface adherent clots may decrease, thus reducing embolization potential in vivo, a phenomenon which is not present in our current study. This is left to future studies for a more detailed analysis on the impact of a compliant tube rather than a rigid structure for the veinous model. We also did not have flow through the CVC lumens to mimic the clinically relevant case of chronic indwelling catheters with no flow. We acknowledge that results can differ with flow through the CVC lumens, but the scenario presented here was chosen to mimic one potential clinical situation that occurs often. Therefore, future studies can include flow through the CVC lumen(s) to investigate thrombosis risk, and the results of this study are still relevant to the field.

A continuous condition was used in the PIV experiments for simplicity. A peristaltic pump was used for all blood studies to avoid a blood-wetted pump head and additional mechanical trauma and contact activation to the blood cells during the experiments. Previous work in our lab using the same PIV flow loop but with a pulsatile flow condition showed very similar results to the steady flow conditions we present here, and as such, the continuous flow method is deemed acceptable in this study [19]. The PIV experiments were completed at room temperature. During experiments, the temperature change in the PIV fluid was <1° C, resulting in <5% differences in viscosity. Further, for our 1-hour blood experiments, the maximum temperature change associated with the peristaltic pump was < 1°C. It is therefore reasonable to assume that temperature changes were small and the effect on viscosity was negligible in our system. Last, for the in vitro blood experiments, the CVC was inserted at a 30° angle, but for the PIV experiments, the CVC was inserted as a straight pipe. This can affect the results as there will be some flow stagnation around the CVC insertion point in the in vitro blood tests. However, this minimally influenced the results. Future studies will investigate a more realistic insertion angle with novel materials for the CVC model, as it is difficult to manufacture acrylic transparent models in this way.

Conclusions

This study investigates the effects of flow and CVR on CRT through both in vitro blood and fluid mechanics studies. The CVC tip has a high potential for thrombus formation in high CVR cases, and the CVC side hole is quite susceptible to thrombus formation. Clinical guidelines can be extracted from this study, including CVR ≥ 0.33 should be monitored closely at the device tip for thrombus formation, especially at a high flow scenario (Re ≥ 800). At lower flow scenarios, the entire catheter is susceptible to clot formation, regardless of the CVR and should be monitored closely. Further, CVR > 0.49 will result in thrombus formation in most cases. Potential redesigns of a CVC could consider PU micropatterning, increased side hole size or surface area without sharp corners or stagnation zones, and a tip with a decreased cross-sectional area that does not result in supraphysiologic shear rate gradients and spatially extended wake zones. Additionally, this study showed the importance of both the CVR and flow condition on thrombus and fibrin sheath formation on medical catheters. The CVC size chosen for a patient should be considered carefully based on the target insertion vein, acknowledging that the CVR and flow condition will have an impact on thrombosis, not just the CVR alone. The results provide valuable information for CVC placement.

Data availability:

Data can be made available to interested parties via reasonable written request to the authors.