For decades, warfarin therapy has served as the backbone of chronic anticoagulation therapy to prevent thrombotic morbidity secondary to hypercoagulable states (e.g., antithrombin deficiency) or the presence of a blood-biomaterial interface (e.g., artificial heart valve, ventricular assist device). The well-characterized mechanism of action of warfarin is the inhibition of the synthesis of vitamin K-dependent coagulation enzymes Factors II (FII), VII (FVII), IX (FIX), and X (FX). Administration of warfarin-based anticoagulation is monitored by prolongation of prothrombin time (PT), expressed as an index such as international normalized ratio (INR, typically the patient PT value divided by normal laboratory PT value). Management of patients implanted with devices with artificial surfaces typically have INR values maintained between 2 and 4.1–4 The rationale for this range is that INR values below 2 enhance the probability of thrombosis, whereas INR values above 4 increase the chance of hemorrhagic morbidity.1–4 Unfortunately, as recently reviewed,5 despite maintaining INR values within the therapeutic range, major bleeding involving the gastrointestinal or urinary tracts occurs in up to 6.5% of patients per year, and fatal intracranial hemorrhage occurs with an incidence of 1% per year. Further, a persistent incidence of thrombosis and thromboembolism is observed despite maintenance of therapeutic INR values.1–4 Thus, warfarin therapy improves patient outcomes but is far from being a panacea for thromboembolism from blood-biomaterial interfaces.
The mechanism responsible for the inadequacies of warfarin-based therapy in the setting of blood-biomaterial interfaces is likely the lack of enzyme-specific inhibition of key proteins known to be involved with contact activation. As recently reviewed,6 contact protein activation of coagulation results in thrombi that grow faster and are more resistant to lysis than clots initiated by tissue factor (TF). The enhanced resistance to clot lysis by tissue-type plasminogen activator (tPA) mediated by contact protein activation is secondary to greater thrombin generation and subsequence activation of Factor XIII (FXIII) and thrombin activatable fibrinolysis inhibitor (TAFI).7 In sum, a critical gap in knowledge existed concerning the effects of warfarin treatment on thrombus growth and fibrinolytic disintegration following either TF or contact pathway protein- mediated coagulation initiation.
The purpose of the present investigation was to define the effects of warfarin therapy on thrombus growth and disintegration following either contact pathway protein or TF initiation to provide insight into why patients may simultaneously suffer from hemorrhagic and thromboembolic complications. This goal was achieved using the clot lifespan model (CLSM) of coagulation,6–8 a thrombelastography-based method used to determine the effects of clinical/biochemical interventions on the kinetics of clot growth and fibrinolysis.
Materials and Methods
Plasma, unlike whole blood from volunteers, is devoid of individual hemostatic variation mediated by platelets. Further, plasma sets obtained from screened subjects are commercially available, are noncellular, and cannot be linked to individual donors. Thus, institutional ethical approval is not required as per the guidelines of the National Institutes of Health.
Experimentation used a normal subject plasma set (George King Bio-Medical, Overland Park, KS) anticoagulated with sodium citrate. The gender ratio was 1:1, and there were six subjects in the set. All subjects were nonsmokers and not exposed to either prescription or over-the-counter medications. Hemostatic normalcy was defined as an INR of 0.9 (n = 3) or 1.0 (n = 3), a partial thromboplastin time of 23.4–37.3 seconds and a Factor VIII activity of 50%–150% of normal. For simplicity, all normal subjects were considered to have an INR of 1.0. Lastly, all subjects were demonstrated to have no serological evidence of hepatitis, human immunodeficiency virus or syphilis.
Plasma Obtained from Patients Exposed to Warfarin
A total of 17 samples were obtained from 10 individuals exposed to warfarin (George King Bio-Medical). The specific INR values were 1.8, 1.9, 2.0, 2.1, 2.4, 2.8, 3.0, 3.4, 3.5, 3.6, 4.0, 4.7, 5.3, 5.4, 5.7, 8.2, and 9.6. The rationale for only using one sample for each specific INR value is that there is significant variability in CLSM variables between specific lots of pooled plasma or individual plasma based on differences in individual coagulation factor activity that are not dependent on vitamin K, such as fibrinogen concentration and antifibrinolytic protein activities (e.g., TAFI) as has recently been published.9,10 This approach isolates the difference between individual samples based on INR values with a wide range of intensity of anticoagulation without the confounding effects of individual procoagulant/antifibrinolytic enzymes not affected by warfarin.
Plasma Obtained from Subjects with Congenital Deficiency of FXII or FVII
Plasma deficient in either FXII (<1% normal activity, INR = 1.0) or FVII (<1% normal activity, INR = 3.4) was obtained from George King Bio-medical and exposed to either celite or TF (n = 1 per condition) as subsequently described to illustrate the impact of specific protein deficiencies compared with the effect of INR value alone on CLSM parameter values.
Clot Lifespan Model of Coagulation
The final volume for all subsequently described plasma sample mixtures was 360 μL. Sample composition consisted of 320 μL of plasma; 10 μL of tPA (580 U/μg, Genentech, Inc., San Francisco, CA) diluted with 10 mM potassium phosphate buffer (pH 7.4) for a final activity of 100 IU/mL; 10 μL of 1% celite (final concentration 0.28 mg/mL; Sigma Aldrich, Saint Louis, MO) or TF (final concentration 0.1%; Diagnostica Stago, Asnieres, France) in 0.9% NaCl, and 20 μL of 200 mM CaCl2.
Plasma sample mixtures were placed in a disposable cup in a computer-controlled thrombelastograph (TEG) hemostasis system (Model 5000, Hemoscope Corp., Niles, IL), with addition of CaCl2 as the last step to initiate clotting. Data were collected until clot lysis time (CLT) occurred. The following variables were determined at 37°C: clot growth time (CGT, time from clot amplitude of 2 mm (102 dynes/cm2) until maximum strength occurred in sec), CLT (time from when maximum strength was observed to 2 mm amplitude in sec) and clot lifespan (CLS, the sum of CGT and CLT). Additional elastic modulus-based parameters previously described6–8 were determined and are displayed in Figure 1. The nomenclature used to describe these phenomena is as follows: time to maximum rate of thrombus generation (TMRTG), this is the time interval (s) observed before the maximum speed of the clot growth; maximum rate of thrombus generation (MRTG), this is the maximum velocity of clot growth observed (dynes/cm2/s); total thrombus generation (TTG), this is the total area under the velocity curve during clot growth (dynes/cm2), representing the amount of clot strength generated during clot growth; time to maximum rate of lysis (TMRL), this is the time interval (s) measured from the time of maximum strength to the time when the velocity of clot disintegration is maximal; maximum rate of lysis (MRL), this is the maximum velocity of clot disintegration observed (−dynes/cm2/s); and area of clot lysis (ACL), this is the total area under the velocity curve during clot disintegration (−dynes/cm2), representing the amount of clot strength lost during clot disintegration.
Normal subject data are presented as mean ± standard deviation (SD). Normal subject data were analyzed to assess the effects of celite activation compared with TF activation using a paired Student’s t-test (SigmaStat 3.0, SPSS Inc., Chicago, IL). Data obtained from subjects exposed to warfarin were simply displayed with celite-activated and TF-activated sample data presented as parallel line graphics. However, the linear relationships of the difference between the CLSM parameter values obtained from paired celite and TF-activated samples and corresponding INR value were determined and reported as coefficients of determination (r2). Graphical representation of the data was generated with commercially available software (Origin 7.5, OriginLab Corp., Northampton, MA). A p value <0.05 was considered significant.
Effects of Warfarin on CGT, CLT, and CLS
As depicted in Figure 2, contact protein activation (solid line) resulted in clots with significantly larger CGT, CLT, and CLS values than thrombi initiated with TF (dash line) in samples from normal subjects with an INR of 1.0. Between INR values of 1.0 and 2.8, there was no activator-specific pattern observed with CGT, whereas in general both CLT and CLS tended to be greater in samples activated with celite compared with matched samples activated with TF. CGT tended to be greater in samples with contact protein activation compared with plasma activated with TF when the INR values varied from 3.0 to 9.6. With regard to CLT and CLS, essentially all samples with an INR of 3.0 or greater had significantly larger values when activated with celite when compared with the matching sample activated with TF. Interestingly, three samples with INR values between 3.0 and 9.6 activated with TF had no discernable thrombus formation by 1,800 sec and were considered to be devoid of measureable coagulation. The overall trend for contact activated samples from patients exposed to warfarin was a modest decrease in CLT and CLS (within a SD of TF-activated normal values) as INR increased, with no important decrease in CGT noted. In sharp contrast, TF-activated samples with INR values of 3.0 or greater had a marked decrease in CLT and CLS compared with samples with INR values of 1.0–2.8. In sum, progressive increases in INR value resulted in a rapid decrease in time required for thrombi to dissolve and decrease in CLS in samples activated with TF compared with those with contact protein activation of coagulation.
Effects of Warfarin on the Kinetics of Clot Growth and Strength
As seen in Figure 3, there was no significant activator-specific effect on TMRTG, whereas contact protein activation resulted in significantly greater MRTG and TTG compared with TF activation. With regard to clot initiation, progressive increases in INR values beyond 2.0 prolonged TMRTG values in TF-activated samples to a greater extent than samples activated with celite. With regard to the rate of clot growth, MRTG values were well below normal values in samples activated with TF with INR values of 1.8 and greater, with <10% normal rates of clot growth observed when INR values were >5.0. Samples activated with celite also had decreases in MRTG as INR values increased, a relative plateau within a SD of normal TF-activated sample values was observed when INR was 3.0 or greater. The relationship of clot strength, TTG, and increasing INR values followed a similar pattern, with TTG values of TF-activated samples markedly diminished compared with normal samples when INR values were 3.0 or greater. Remarkably, a few samples with INR values between 2.0 and 4.0 had MRTG and TTG values far greater than samples with an INR of 1.0 when activated with celite. In nearly every case, contact protein activation resulted in a faster growing, stronger thrombus compared with samples activated with TF throughout the INR range.
Effects of Warfarin on the Kinetics of Clot Fibrinolysis
As depicted in Figure 4, contact protein activation resulted in significantly greater TMRL and ACL values than that observed following TF activation in normal subject plasma. There were no significant, activator-specific differences in MRL. However, in samples with 2.0 or greater INR values, as a rule the onset of maximum fibrinolysis, maximum rate of fibrinolysis and strength lost during fibrinolysis were greater in samples activated with celite compared with matched samples that were TF activated. Further, TMRL and MRL values in both celite and TF-activated samples tended to variably decrease as INR values increased. However, ACL values were markedly decreased as INR values increased in samples activated with TF. Remarkably, a few samples with INR values between 2.0 and 4.0 had TMRL, MRL, and ACL values far greater than samples with an INR of 1.0 when activated with celite.
A comparison of matched samples activated by celite (black trace) or TF (gray trace) with INR values of 1.0, 2.0, and 4.0 are depicted in Figure 5. Although warfarin treatment initially prolongs the time to onset of coagulation at the therapeutic threshold of INR 2.0 regardless of activation, the speed of clot formation and clot strength are not markedly changed, with fibrinolytic kinetics still maintaining activator specific patterns. However, when an INR value of 4.0 is reached, contact-activated thrombi behave kinetically in a fashion similar to that associated with normal subject plasma activated by TF. Importantly, TF activation of plasma with an INR value of 4.0 results in a slow growing, weak and short-lived thrombus.
Coefficients of Determination of the Relationships of INR and Celite-TF Differences of CLSM Parameters
As displayed in Table 1, with the exception of the time to maximum rate of lysis (Figure 6), variation of INR value had no significant relationship with the difference in CLSM parameter observed between individual paired celite and TF-activated samples. Given that r2 is indicative of the percentage change of the dependent variable that can be explained by the change in the independent variable, only 36% of the changes in TMRL values were explainable by changes in INR. The remainder of the relationships were not significant and were all <15% explainable by the changes in INR.
Effects of FXII and FVII Deficiency on Clot Growth and Fibrinolysis Following Contact Protein or TF-mediated Activation
As displayed in Figure 7, TF activation (gray trace) resulted in a normal CLS, growth, and fibrinolytic profile in FXII-deficient plasma with a normal INR value. However, contact activation (black trace) resulted in a weak, short-lived thrombus in FXII-deficient plasma that forms and dissolves after its matched TF-activated sample had already completed a normal CLS.
In sharp contrast, Figure 8 demonstrates that celite- activated (black trace) FVII-deficient plasma with an INR value of 3.4 had a more than normal speed of clot growth and strength, and a prolonged CLS. However, when activated with TF (gray trace), FVII-deficient plasma demonstrated a marked delay in the onset of coagulation, slow growth, poor strength, and slow rate of fibrinolysis resulting in a relatively prolonged CLS.
The primary finding of this CLSM-based investigation was that contact protein-activated thrombi were faster, stronger, and lasted longer than TF-activated clots within and beyond the therapeutic range of INR values achieved with warfarin anticoagulation. This observation is consistent with clinical observations of concurrent device-associated thromboembolic events and hemorrhagic complications during warfarin therapy.1–5 Remarkably, some subjects with INR values within the therapeutic range of 2.0–4.0 demonstrated relative hypercoagulability (Figures 2–4) following contact protein activation, with abnormally fast growing, strong and long-lived thrombi compared with normal subject values. Conversely, and similarly remarkable, two subjects had essentially no measurable coagulation following TF-mediated activation within the therapeutic range of 2.0–4.0. This unpredictable, activator-specific variability in coagulation response may in part be secondary to the 50%–70% decrease in vitamin K-dependent factor activity with warfarin that is not necessarily observed equally with each factor.11 Specifically, some individuals with a plasma INR value of 3.0 may have a far greater decrease in FVII than FX, with a consequently greater disparity between celite and TF activation as FVII activity is critical for TF-mediated activation as demonstrated in Figure 7. If FX activity is decreased to a lesser extent than FVII activity by warfarin, then contact protein activation would be expected to result in superior coagulation as seen in Figure 7. Also of interest, as INR values increase, the decrease in FII, FVII, FIX, and FX activity is not linear, with values of between 5% and 40% of normal activity observed within an INR range of 5.0–10.0.12 In sum, the results of this investigation coupled with previous clinical and biochemical investigations1–7,11,12 support the concept that warfarin therapy cannot predictably provide safe anticoagulation to patients implanted with devices with a blood-biomaterial interface.
Another remarkable finding was that warfarin had little effect on the CLSM parameter values obtained from the difference between paired celite and TF-activated samples (Table 1). Changes in INR did not predictably affect the superior degree by which celite-activated thrombi formed and dissolved compared with matched TF-activated clots; rather, it seemed that warfarin treatment poorly attenuated activated FXII-derived thrombin generation and consequent FXIII and TAFI activation as previously reported.6,7 The one significant correlation between TMRL and INR may be explained by the loss of TAFI activation, which requires a minimum amount of thrombin generation to be activated and inhibit fibrinolysis.13 Thus, although there may still be differences in thrombin generation between celite and TF-activated samples with large INR values, the celite-activated sample may not have meaningful TAFI activation with consequent similar onset of fibrinolysis compared with matched TF-activated samples. In sum, it seems that the attenuation of vitamin K-dependent coagulation factor activity only slightly affects the difference between contact protein-mediated and TF-mediated clot growth and disintegration kinetics.
Given the lack of predictable biologic response to warfarin therapy, an attractive alternative for providing safe anticoagulation to patients with blood-biomaterial interfaces would include inhibition of FXII-mediated thrombin generation. As seen in Figure 6, marked FXII deficiency results in normal TF-mediated thrombus formation and fibrinolysis, whereas celite activation results in a markedly delayed formation of a weak, short-lived clot. If this situation could be translated to clinical situations involving patients with valve prostheses and ventricular assist devices with FXIIa/kallikrein inhibitors14,15 that are biocompatible, then perhaps low-dose warfarin therapy (e.g., INR 1.5–1.8) could be used with an associated decrease in hemorrhagic morbidity.
Although not central to our hypothesis, it was noted that the maximum rate of clot lysis was markedly reduced in TF-activated samples compared with celite-activated plasma. The most likely explanation for this phenomenon is a greater thrombin generation and subsequently greater FXIII activation in celite-activated samples than in TF-activated plasma. FXIII activation has recently been demonstrated to be required to maintain normal velocities of fibrinolysis, presumably by optimizing intermolecular distances and fibrin matrix structure and consequently enhancing plasmin-fibrin polymer interactions.16 Further, we have recently demonstrated that progressive reduction of thrombin generation with either argatroban or heparin results in concordantly decreased MRL values.17 Considered as a whole, it seems that reduced thrombin generation via warfarin, argatroban, or heparin results in less activation of FXIII, with a consequently weaker, more disorderly clot matrix that lyses more slowly than normal plasma.
In conclusion, warfarin-based anticoagulation did not eliminate the activator-specific effects on thrombus growth and fibrinolysis that has been previously described in normal plasma.7 There was greater thrombin generation, faster clot growth, and enhanced resistance to fibrinolysis in thrombi formed by contact activation compared with TF-mediated activation. Further, inhibition of FXIIa-mediated thrombin generation would be expected to reduce blood-biomaterial thrombus formation, given the results obtained with FXII-deficient plasma with the CLSM. Our investigation, coupled with previous clinical and laboratory data,1–6 serve as a rational basis to continue to pursue alternative molecular approaches (e.g., FXIIa/kallikrein inhibition) to warfarin-based anticoagulation to reduce thrombus formation on artificial surfaces while diminishing hemorrhagic morbidity.
This investigation was supported by the Departments of Anesthesiology and Surgery.
1.Akins CW: Results with mechanical valvular prostheses. Ann Thorac Surg
60: 1836–1844, 1995.
2.Schapkaitz E, Jacobson BF, Becker P, Conway G: Thrombo-embolic and bleeding complications in patients with mechanical valve replacements—A prospective observational study. S Afr Med J
96: 710–713, 2006.
3.Bayliss A, Faber P, Dunning J, Ronald A: What is the optimal level of anticoagulation in adult patients receiving warfarin following implantation of a mechanical prosthetic mitral valve? Interact Cardiovasc Thorac Surg
6: 390–396, 2007.
4.Koertke H, Zittermann A, Tenderich G, et al
: Low-dose oral anticoagulation in patients with mechanical heart valve prostheses: Final report from the early self-management anticoagulation trial II. Eur Heart J
28: 2479–2484, 2007.
5.Leissinger CA, Blatt PM, Hoots WK, Ewenstein B: Role of prothrombin complex concentrates in reversing warfarin anticoagulation: A review of the literature. Am J Hematol
83: 137–143, 2008.
6.Nielsen, VG, Kirklin JK, Holman WL, et al
: Mechanical circulatory device thrombosis: A new paradigm linking hypercoagulation and hypofibrinolysis. ASAIO J
54: 351–358, 2008.
7.Nielsen VG, Steenwyk BL, Gurley WQ: Contact activation prolongs clot lysis time in human plasma: Role of thrombin activatable fibrinolysis inhibitor and factor XIII. J Heart Lung Transplant
25: 1247–1252, 2006.
8.Nielsen VG, Cohen BM, Cohen E: Elastic modulus-based thrombelastographic quantification of plasma clot fibrinolysis with progressive plasminogen activation. Blood Coagul Fibrinolysis
17: 75–81, 2006.
9.Nielsen VG, Cohen BM, Cohen E: Effects of coagulation factor deficiency on plasma coagulation kinetics determined via Thrombelastography®: Critical roles of fibrinogen and factors II, VII, X and XII. Acta Anaesthesiol Scand
49: 222–231, 2005.
10.Nielsen VG: Clot lifespan model analysis of clot growth and fibrinolysis in normal subjects: Role of thrombin activatable fibrinolysis inhibitor. Blood Coagul Fibrinolysis
19: 283–287, 2008.
11.Jerkeman A, Astermark J, Hedner U, et al
: Correlation between different intensities of anti-vitamin K treatment and coagulation parameters. Thromb Res
98: 467–471, 2000.
12.Sarode R, Rawal A, Lee R, et al
: Poor correlation of supratherapeutic international normalized ratio and vitamin K-dependent procoagulant factor levels during warfarin therapy. Br J Haematol
132: 604–607, 2005.
13.Taketomi T, Szlam F, Bader SO, et al
: Effects of recombinant activated factor VII on thrombin-mediated feedback activation of coagulation. Blood Coagul Fibrinolysis
19: 135–141, 2008.
14.Ulmer JS, Lindquist RN, Dennis MS, Lazarus RA: Ecotin is a potent inhibitor of the contact system proteases factor XIIa and plasma kallikrein. FEBS Lett
365: 159–163, 1995.
15.Isawa H, Orito Y, Iwanaga S, et al
: Identification and characterization of a new kallikrein-kinin system inhibitor from the salivary glands of the malaria vector mosquito Anopheles stephensi
. Insect Biochem Mol Biol
37: 466–477, 2007.
16.Nielsen VG: Hydroxyethyl starch enhances fibrinolysis in human plasma by diminishing α2
-antiplasmin-plasmin interactions. Blood Coagul Fibrinolysis
18: 647–656, 2007.
17.Nielsen VG, Kirklin JK: Argatroban enhances fibrinolysis by differential inhibition of thrombin-mediated activation of thrombin activatable fibrinolysis inhibitor and factor XIII. Blood Coagul Fibrinolysis
19: 793–800, 2008.