In vitro methodology for medical device material thrombogenicity assessments: A use condition and bioanalytical proof‐of‐concept approach

Abstract Device manufacturers and regulatory agencies currently utilize expensive and often inconclusive in vivo vascular implant models to assess implant material thrombogenicity. We report an in vitro thrombogenicity assessment methodology where test materials (polyethylene, Elasthane™ 80A polyurethane, Pebax®), alongside positive (borosilicate glass) and negative (no material) controls, were exposed to fresh human blood, with attention to common blood‐contact use conditions and the variables: material (M), material surface modification (SM) with heparin, model (Mo), time (T), blood donor (D), exposure ratio (ER; cm2 material/ml blood), heparin anticoagulation (H), and blood draw/fill technique (DT). Two models were used: (1) a gentle‐agitation test tube model and (2) a pulsatile flow closed‐loop model. Thrombogenicity measurements included thrombin generation (thrombin‐antithrombin complex [TAT] and human prothrombin fragment F1.2), platelet activation (β‐thromboglobulin), and platelet counts. We report that: (a) thrombogenicity was strongly dependent (p < .0001) on M, H, and T, and variably dependent (p < .0001 – > .05) on Mo, SM, and D (b) differences between positive control, test, and negative control materials became less pronounced as H increased from 0.6 to 2.0 U/ml, and (c) in vitro‐to‐in vivo case comparisons showed consistency in thrombogenicity rankings on materials classified to be of low, moderate, and high concern. In vitro methods using fresh human blood are therefore scientifically sound and cost effective compared to in vivo methods for screening intravascular materials and devices for thrombogenicity.


| INTRODUCTION
When an intravascular medical device is placed in contact with a patient's circulatory system a variety of reactions are recognized to take place in blood and on the device material surfaces. [1][2][3] Factors such as surface area of exposure, supplementary anticoagulants, the type of medical device material, and device form can influence the nature and extent of these reactions and potentially influence the level of safety or risk to the patient. Fortunately, the materials used in medical devices come from a select list of high-performance polymers, metals, ceramics, and biological tissues and have an established history of safe use in humans. Novel devices, which introduce new or unproven materials are increasingly less common, given the costs associated with material qualification processes. As a result, device and material evolution can be protracted and is often achieved through simple or slight changes in materials, material geometry, contact surface chemistry, source (vendor), composition, or manufacturing processes as they are introduced in next generation devices.
Within the context of the highly diverse and regulated environment of medical device/material applications, sound material science, toxicology science, and biocompatibility science are required to define device safety and risk.
One of the greatest limitations in the risk assessment process involving intravascular biomaterials is in methods available to assess risk of thrombosis. Even with today's technologies, few standardized in vitro material "thrombogenicity" tests exist, 4 and existing tests rely on test conditions that are moderately removed from device use conditions (uc). Often initial testing resorts on simple clotting time and platelet counts to measure material thrombogenicity. 5,6 These metrics represent a single data point in a complex coagulation process and often do not offer a distinction between individual devices and iterative device improvements. Another reason for this discrepancy is the multifactorial nature of the process of thrombosis and the lack of systematic and controlled studies on key variables involved in material thrombogenicity assessment. Such variables include, for example: material, material surface roughness, surface chemistry, overall material/device geometry, species of blood, freshness of blood, subject/ donor genetics, anticoagulant type and amount, blood exposure time, and material blood exposure ratio (surface area [cm 2 ] of test material to volume of blood [ml] in the test model). This list goes on to include factors such as hematocrit, temperature, presence of an air interface, and an assortment of test model particularities, such as model complexity, blood-contact surface area of the model itself, gentle mixing versus physiological flow, and so forth. In addition, modern bioanalytical techniques are often antibody-based and species-specific and require careful scrutiny for sensitivity to detect statistically and clinically meaningful differences between positive controls, negative controls, and test materials.
In this proof-of-concept report we attempt to address actual use conditions and bioanalytical limitations associated with in vitro thrombogenicity assessment of medical device materials. Our report begins with a series of simple exploratory screening experiments that take a first principles look at some of the key variables that can affect the outcome of an in vitro material thrombogenicity test. To control cost and complexity of these material thrombogenicity exploratory studies (MTESs), investigations were limited to: examination of five different materials; testing all conditions at n = 2; limited exploration on the effect of exposure time and test material exposure ratio (cm 2 /ml blood); blood type and freshness restricted to human blood that was directly-drawn into the models with resultant instant exposure to test and control materials; experiments conducted using blood from a small pool of healthy male-only donors (4; age range 35-60 years old); heparin anticoagulant ranging from 0.6-2.0 U/ml; material heparin surface modification without rigorous process optimization; and testing using two simple in vitro models. Measurements for assessing material thrombogenicity included assays for thrombin generation (thrombin-antithrombin complex (TAT) and prothrombin fragment F1.2 (F1.2) ELISAs) and platelet activation (betathromboglobulin (ßTG) ELISA and platelet counts), as recommended in Reference 7. From the outcome of these screening studies, a simplified use condition approach and scoring scheme for an in vitro assessment of material (M) thrombogenicity is proposed based upon thrombin (T) and platelet (P) activation (A) assays. This "ucMTPA" method was subsequently applied to six medical device/material case studies with the in vitro results compared to in vivo thrombogenicity evaluations on the same or similar devices/materials using common animal models. The latter consisted primarily of the nonanticoagulated venous implant (NAVI) model. 3,7 2 | MATERIALS AND METHODS

| Control and test materials
Each in vitro MTES utilized the following control and test materials: Medtronic Santa Rosa, CA), and (6) Figure 1. The order of test material exposure to blood in each blood draw was randomized, with each draw including additional separate initial (Bi) and final (Bf) blood samples to monitor draw quality, that is, the extent of blood activation during the draw procedure. The randomization was applied to eliminate any bias during the fill process, such as a slow increase or decrease in blood activation. After blood exposure, blood was withdrawn from each tube into syringes prefilled with CTAD solution (citrate, theophylline, adenosine, and dipyridamole; BD reference no. 367947; 1:10 by volume) and put on ice to arrest any further coagulation and platelet activation, post experiment. 0.5-1.0 ml samples were taken for CBC analysis and remaining blood was centrifuged (2,500 x g, 20 mins) and plasma samples (100-300 μl) were aliquoted into cryotubes and stored at −70 C for subsequent ELISA analyses.
Test materials were also gently rinsed with Plasma-Lyte A to remove nonadherent blood and photographed for visible thrombus on the material surfaces. Consult the Appendix for additional details.

| Dynamic closed-loop model
This more advanced in vitro model employed closed circular loops of PVC tubing as the "test tube" for insertion of test and control materials.
These torus-shaped loops have specific features that allow priming with heparinized saline, no-air-exposure blood filling via saline displacement draw technique, and pulsatile flow created by the combination of an integral check valve and applied pulsatile rotational motion ( Figure 2). See References 8-12 and the Appendix for additional details.

| Thrombogenicity assessment using in vitro assays for thrombin generation and platelet activation
Thrombogenicity measurements consisted of ELISA assays for the key coagulation proteins TAT 13 and F1.2, 14 as well as β-thromboglobulin (βTG), an indicator of platelet activation. 15  2.4 | MTESs MTES1, MTES2, and MTES3 on essential blood interaction variables Exploratory "screening" study variables and the study designs are described in Table 1. Completing each study required a series of separate blood donations from each donor (four donors used, designated A through D), allocated by distinct heparin level, and other variables, such as time, model, and so forth. All conditions were tested in duplicate in tubes or loops of test materials exposed to 3.0, 3.2, or 5.0 ml blood per tube/loop (depending on model). All blood donations involved informed consent and required donors to be healthy and drug and aspirin refraining.

| Case studies on devices or device materials tested for in vitro thrombogenicity
Six medical device or medical device material cases were evaluated for in vitro thrombogenicity using variable and measurement selection based upon MTES findings. Table 2 provides details on each case study and Figure 3 shows the geometric representation of the generic in vitro study design. The later consisted of using either the tube or loop model along with: blood from two donors (selected from the same pool of MTES donors, and an additional donor E), an exposure ratio of 6.0 cm 2 /ml, heparin anticoagulation at one or two levels (within range of 0.6-2.0 U/ml), blood exposure restricted to 60 minutes, and replication n = 2 (1 run) or 4 (2 runs) per condition.

F I G U R E 3
The generic in vitro case study experimental design. This design involves testing a positive control (pos cntl), a No Material negative control (neg cntl), a biomaterial control (PE), and a legally marketed comparator device (LMCD) material alongside the test material (Test) at two heparin levels (where use conditions indicate test devices/materials may be used with and without anticoagulation). It specifies testing with blood from two donors, and potential repeat runs (allowing n = 2 or n = 4) to increase robustness and confidence in results. This gives a Materialx2x2x2 full factorial design that involves 4-8 separate small volume blood draws (bd; i.e., bd1 to bd8) 2.6 | In vivo methods for device/material testing for thrombogenicity All animals utilized in this research were cared for according to the policies and principles established by the Animal Welfare Act and the NIH Guide for Care and Use of Laboratory Animals. The test devices/ materials listed in Table 2 Case Studies 1, 4, 5, and 6 were tested for in vivo thrombogenicity using the NAVI model. 3,7 Briefly, the NAVI tests involved inserting a 10-15 cm portion of each device or material, in catheter form, into a vein of a large animal. In these investigations, either a canine or ovine femoral or jugular vein model was used.
In all cases, the test material/device was positioned in one vein and the control LMCD material was positioned similarly in the contralateral site. The level of replication differed between these studies, as did the in situ blood exposure duration (1-4 hours). Immediately following euthanasia, the implants were carefully exposed in situ and gently rinsed with buffer to remove nonadherent blood. They were then photographed and assessed for apparent surface thrombus using the scoring scheme shown in Table 3

| MTES1
This study examined responses in blood to the test materials over time using the test tube model (SDF blood filling) with nutation mixing, blood from three donors, exposure ratio = 9.0 cm 2 /ml whole blood, and heparin anticoagulation at two levels (0.6 and 1.0 U/ml).
Results across all test variables are shown graphically in the Appendix.

| MTES2 and MTES3
These two exploratory studies had the same experimental design, and each examined the responses in blood to test materials using: two dif-  (Table 7). As would be expected, there was a general tendency for TAT, F1.2, and βTG responses to be lower with less material  Here too, ANOVA-analysis (Table 6)

| In vitro thrombogenicity results on devices or device materials tested under a select set of test variables and measurements
The results for the coagulation and platelet activation thrombogenicity measurements (TAT and/or F1.2; βTG and/or platelet loss) for each of the six in vitro case studies are shown graphically and by heat map tables. The results were scored based on the criteria described in Table 5, to identify results with classifications of low, moderate, and high concern. For illustrative purpose, Case 4 shows a material of lowthrombogenicity concern (see Figure 5 and heat map Table 8) and Case 6 shows a test material of moderate thrombogenicity concern (see Figure 6 and heat map Table 9). The figures and heat map tables for the other cases are presented in the Appendix. All scores are summarized in Table 10. The test devices/materials in Cases 1 through 4 consistently revealed thrombogenicity measurements below the 2.0 threshold of concern. The test devices/materials in Case studies 5 and 6 revealed a number of conditions that gave rise to thrombogenicity scores above the 2.0 threshold of concern. Figure 7 shows representative gross images of some of the in vitro study samples after gentle rinsing with buffer to remove nonadherent blood.

| In vivo thrombogenicity results on the medical devices or medical device material
The in vivo thrombogenicity scores on the same or similar materials/ devices evaluated in the in vitro thrombogenicity case studies are shown in Table 11. The test devices/materials in Cases 1 through  to these two models given the use condition simulation, the advantage of small volume, ease of use with fresh human blood, and the many practical drawbacks of large-volume animal-blood models. Unmistakably, the combination of small volume models with advanced multifactor experimental designs allowed a more robust study of the factors influencing blood-material interaction.

| Effect of the test model
The Model factor, examined in MTES2 and MTES3, was found to be significant across all thrombogenicity measurements, with only one exception-TAT in MTES2 (Table 6). However, while material-specific MTES3 was the exposure ratio of 9.0 versus 6.0 cm 2 /ml, respectively.
The less dense material packing and resultant more even blood mixing in MTES3 may have been an influential factor. It is noteworthy that the exposure ratio of 6.0 cm 2 /ml was arbitrarily chosen yet is recommended as a reasonable target in other types of studies. 25 It is also a reasonable worse-case exposure ratio as it is representative of the high-exposure ratio seen in common extra corporeal membrane oxygenation (ECMO) procedures. In the latter case, an ECMO blood oxygenator with a surface area of 25,000 cm 2 may come into contact with an average human adult blood volume of 5,000 ml, to create an exposure ratio of 5 cm 2 /ml. Some methods suggest blood interaction studies use higher exposure ratios (as high as 12 cm 2 /ml, 6 ) to increase measurement signal-to-noise ratio. However, such high-ratios depart greatly from typical clinical use conditions. Moreover, at such high-exposure ratios the important ability of blood to evenly distribute and mix over the material surface can be substantially diminished, as we observed at even 9 cm 2 /ml in MTES2.

| Effect of device use conditions variables
As each medical device application involves diverse patients, unique  to be relatively safe to avoid excessive coagulation. 8,11,12 Conversely, levels of heparin greater than 2.0 U/ml were anticipated to generate minimal/background responses across all conditions and materials due to the known therapeutic effectiveness of heparin to quench thrombotic processes at or above this level. In general, heparin levels of 0.6, 1.0, and 2.0 U/ml whole blood were sufficient to consistently avoid excessive coagulation yet still see significant differences between positive (Glass) and negative (No Material) controls. The 0.6 and 1.0 U/ml whole blood heparin concentrations used in MTES1 worked well to simulate a "challenging" (low) and This suggests that such testing using only a small number of donors will generally give consistent results and a fair estimate of material thrombogenicity. It is important to note, however, that baseline

| Correlation of in vitro to in vivo thrombogenicity
The second half of this investigation examined the ability of simple multiparameter in vitro thrombogenicity tests to give results comparable to in vivo models. A common and somewhat controversial in vivo approach to material/device thrombogenicity assessment is the NAVI model (and its anticoagulated model counterpart-AVI). 1 A critical factor that guides interpretation of device/material thrombogenicity in these models is inclusion of a LMCD (a clinically approved device used in the same application). To assess device risk, such studies are evaluated according to the criteria in Table 3 and the test device/material is examined for score/response similarity to the LMCD. Table 11 Table 5 interpretation. The in vivo thrombogenicity scores in these same four cases were also low

| Limitations
Given the exploratory nature of this work, there were some notable drawbacks. To start, blood exposure to air, the degree of blood dilution, blood mixing motion, the amount of material exposed to blood (exposure ratio), and the type of anticoagulant (including combinations of anticoagulants and antiplatelet drugs) all clearly have some degree of influence on blood responses. However, these experimental variables were not implemented systematically into our exploratory study designs, impeding thorough assessment of their impact. Also, the small number of healthy blood donors used in the exploratory studies (4) and in the case studies (5) is likely only moderately representative of individuals in the human population and not typical of the patient population that receives medical device therapies. From our studies and those of others 26,27 it is expected that differences in the human population and random selection of blood donors will lead to in vitro blood interaction studies conducted on blood that comes from a spectrum of individuals, for example, from low to high-level responders. Nonetheless, treating blood donor as a variable in blooddevice/material interaction studies offers an understanding of blood interaction response variability with inter-and intra-donor comparison of test material to LMCD, positive and negative controls. Conversely, blood interaction studies that use single source or pooled blood lose this information and can give a potential false sense of consistency of patient blood responses.
To better simulate use conditions, our studies utilized a unique blood draw procedure that involved blood drawn directly into the models and resultant immediate blood exposure to test and control materials. However, no comparison of this procedure was made to other more common approaches of test material exposure to blood, for example, approaches that involve a time delay between the blood draw and blood use, and the use blood with prior contact with one or more nontest/nonmodel materials. The importance of using fresh blood within 4 hr of blood collection (and preferably within 2 hr) has been reported. 29 Interestingly, examination of our platelet count data expressed relative to the No Material control according to Reference 6 showed the HDPE reference biomaterial to be consistently outside (60-80%) the assay validation condition of 80-120% (see Appendix). This low level of HDPE platelet reactivity has been reported by others using alternative blood preparation/exposure methods. 30,31 This difference in HDPE reactivity, which is supported by extensive additional unpublished work in our lab using a commercially available HDPE reference biomaterial suggests that blood "freshness" and exposure conditions may influence thrombogenicity measurements. Conversely, exposure of test materials to aged blood, citrated blood bank blood, recalcified and heparinized blood, or fractionated blood (e.g., blood plasma, platelet rich plasma, fresh or pooled/frozen serum) deviates significantly from most use conditions and may give unreliable results.
On the material side, our screening study use of a heparin-coated material that excluded heparin coating process optimization steps (that increase immobilized heparin bioactivity and uniformity) led to some observations of nonstatistically significant trends between coated and uncoated materials. Despite this drawback, the expected trend was apparent.
Regarding models, we chose to examine two well-established simple in vitro models, yet each model presents its own drawbacks. One T A B L E 9 Heat map of comparisons of interest in in vitro thrombogenicity Case Study 6. Red indicates a statistically significant difference as determined from Tukey-Kramer HSD analysis of dataacross the two models and with data from both donors. Comparisons with marginal statistical significance (0.05 < p ≤ 0.08) are shown

| SUMMARY
In the 19th century the famous physician Rudolf Virchow first alluded to thrombosis being influenced by a small number of critical factors. 32 Clearly the mechanisms of thrombosis are much better understood today, and in association with cardiovascular devices thrombosis can be viewed to be influenced by at least six important factors: (1) disturbed/nonphysiological flow through/around devices (2) blood "hypercoagulability", a term describing a perturbation in the mechanisms of hemostasis and/or differing coagulation potentials recognized to exist between and within individuals (3) vascular/ endothelial injury, a highly influential factor that can impact thrombosis associated with medical devices in terms of tissue trauma due to the device or implant procedure, as well as in vitro blood studies in terms of trauma associated with the blood draw technique and general blood handling (4) anticoagulants, antiplatelet drugs, and antithrombotic therapies-powerful drugs that can remarkably impact thrombus formation associated with medical devices (5) medical device materials and the growing list of surface-modified and drug-eluting materials that impact molecular and cellular interactions at the blood-contacting interface, and (6) general device application specifics, which addresses unique device use conditions such as blood-contact duration and surface area of exposure, macroscopic and microscopic material geometry, tissue engineering factors for example, living endothelial cells applied to devices, [33][34][35][36] and target vasculature (venous vs. arterial, heart vs. brain, etc.). This report shares a first principles-based approach to designing and executing studies that evaluate blood-material interactions with respect to this hexad of important factors. We describe applying a use condition (uc) approach to thrombogenicity testing that uses modern bioanalytical tools to assess medical device materials (M) based upon their capacity to induce thrombin (T) and platelet (P) activation (A) (a "ucMTPA" test). Based on these studies, in vitro models using human blood with minimal trauma and immediate exposure to test materials within specified ranges of exposure ratio, exposure time, and heparin anticoagulation, gave bioanalytical measurements of coagulation and platelet activation that were consistent with acute in vivo device/material thrombogenicity assessments.