Anticoagulants and the propagation phase of thrombin generation

Abstract

The view that clot time-based assays do not provide a sufficient assessment of an individual's hemostatic competence, especially in the context of anticoagulant therapy, has provoked a search for new metrics, with significant focus directed at techniques that define the propagation phase of thrombin generation. Here we use our deterministic mathematical model of tissue-factor initiated thrombin generation in combination with reconstructions using purified protein components to characterize how the interplay between anticoagulant mechanisms and variable composition of the coagulation proteome result in differential regulation of the propagation phase of thrombin generation. Thrombin parameters were extracted from computationally derived thrombin generation profiles generated using coagulation proteome factor data from warfarin-treated individuals (N = 54) and matching groups of control individuals (N = 37). A computational clot time prolongation value (cINR) was devised that correlated with their actual International Normalized Ratio (INR) values, with differences between individual INR and cINR values shown to derive from the insensitivity of the INR to tissue factor pathway inhibitor (TFPI). The analysis suggests that normal range variation in TFPI levels could be an important contributor to the failure of the INR to adequately reflect the anticoagulated state in some individuals. Warfarin-induced changes in thrombin propagation phase parameters were then compared to those induced by unfractionated heparin, fondaparinux, rivaroxaban, and a reversible thrombin inhibitor. Anticoagulants were assessed at concentrations yielding equivalent cINR values, with each anticoagulant evaluated using 32 unique coagulation proteome compositions. The analyses showed that no anticoagulant recapitulated all features of warfarin propagation phase dynamics; differences in propagation phase effects suggest that anticoagulants that selectively target fXa or thrombin may provoke fewer bleeding episodes. More generally, the study shows that computational modeling of the response of core elements of the coagulation proteome to a physiologically relevant tissue factor stimulus may improve the monitoring of a broad range of anticoagulants.

Figure 1. Model representation of individuals stablely anticoagulated with warfarin.

Computationally derived thrombin generation profiles reflecting each individual's actual ensemble of factor levels are presented with each anticoagulated individual's profile identified by his clinically determined INR value. A thrombin generation profile characterizing a hypothetical individual with all factors at mean physiologic levels is also shown. Panel A, 7 individuals; Panel B, One warfarin treated individual assessed 8 times over 6 months and after warfarin therapy was ended (INR = 1); Panel C, 7 individuals from study 2 with INR values of 2.3 or 2.7. Note the difference in Y-axis scale.

Figure 2. Dependence of differences between INR and cINR values on TFPI levels.

The difference between each individuals (N = 47) INR and cINR was expressed as a percent of the INR [(cINR-INR)/INR] and then plotted against each individual's TFPI level expressed as a percent of mean physiologic. Individuals with high normal TFPI levels display positive ratio values while those with low normal values display negative ratios. Linear regression analysis yielded the following expression y = 101.65x + 105.4 with an r² = 0.51. The predicted y intercept, representing the TFPI concentration where cINR and INR values do not diverge, correlates well with the mean TFPI value for the population (see Table 1). The three individuals that had a thrombotic event subsequent to the blood draw for compositional analysis are indicated in red.

Figure 3. Empirical and computational studies of coagulation factor-based variability in thrombin generation in 2 control individuals in the presence or absence of rivaroxaban.

cTGPs for 2 individuals (A,E) (open symbols) based on their plasma factor composition with corresponding empirical thrombin generation (closed symbols) measured in synthetic coagulation proteome reconstructions. Panel A, no rivaroxaban; Panel B, 4 nM rivaroxaban; Panel C, 10 nM rivaroxaban; Panel D, 21 nM rivaroxaban. Predicted thrombin concentrations are given at time points to match empirical sampling, which is performed at 1 minute intervals.

Figure 4. Model representation of hypothetical warfarin anticoagulation of control individuals.

cTGPs are presented for individuals before (e.g. A) and after hypothetical warfarin therapy (e.g. A') achieved by setting all VKD proteins at 33% their mean physiologic values with other factors retaining their individual specific values. Panel A: cTGPs of 5 control individuals with an additional cTGP representing an individual with all factors at 100% mean physiologic level. Panel B: 2 individuals with the most disparate thrombin generation profiles prior to anticoagulation. Panel C: 2 individuals with the most similar thrombin generation profiles prior to anticoagulation.

Figure 5. Computational thrombin generation profiles reporting the response of 5 control individuals to anticoagulation.

Anticoagulant concentrations are used that yield a group averaged cINR of ∼2.1 for each anticoagulant. Panel A, 1.25 nM UFH. Panel B, 100 nM Fpx. Panel C, 6 nM rivaroxaban. Panel D, 0.5 µM DAPA. Note for these 5 individuals the average sensitivity to Fpx and DAPA was different than the 32 individuals in Table 3.