Thrombin generation testing in routine clinical practice: are we there yet?


Dr J. J. van Veen, Department of Haematology, Level 2, Sandringham building, Leicester Royal Infirmary, Infirmary Square, Leicester LE1 5WW, UK. E-mail:


Thrombin is the central enzyme in the coagulation cascade. Estimation of an individual’s potential to generate thrombin may correlate more closely with a hyper- or hypo-coagulable phenotype, compared to traditional coagulation tests. The possible correlation and recent technical advances in thrombin generation measurement has caused a significant interest in the method and the development of commercial assays. Several variations of the assay exist depending on the defect to be investigated. Fluorogenic thrombin generation assays have acceptable intra-laboratory variation but a higher inter-laboratory variation. Variation in preanalytical variables makes comparisons between studies difficult. Thrombin generation is highly variable between individuals and there are suggestions that this may allow individualized treatment based on global haemostatic response in patients with bleeding disorders or on anticoagulant therapy. In patients with thrombotic disorders it may be possible to identify those at higher risk of recurrent thrombosis. For both scenarios, however, data from large prospective studies are lacking or inconclusive and a good relationship between thrombin generation and phenotype remains to be established. Further standardization of the assay is needed before large multicentre studies can be conducted and until then thrombin generation in routine clinical practice is not yet a reality.

Thrombin generation measurement

Thrombin generation, converting fibrinogen to fibrin and leading to clot formation, is the endpoint of a complex series of proteolytic reactions started by the formation of the tissue factor (TF) – activated factor VII (FVIIa) complex following vessel wall damage. This leads to the generation of small amounts of thrombin through activation of the prothrombinase complex, insufficient to cause full scale conversion of fibrinogen to fibrin but causing feedback activation of factor V (FV), factor VIII (FVIII) and factor XI (FXI) leading to a thrombin burst (Gailani & Broze, 1991; Hemker & Beguin, 1995) and converting fibrinogen to fibrin. Thrombin generation through the TF – FVIIa pathway is inhibited by the action of tissue factor pathway inhibitor (TFPI) (Rapaport, 1991), thought to be activated by protein S in a protein C-independent manner (Hackeng et al, 2006). Cell surfaces containing phosphatidyl serine have traditionally been seen as a template for these reactions (cascade model of coagulation) but are now thought to actively coordinate them (cell-based model of coagulation) (Hoffman & Monroe, 2001). Thrombin not only accelerates its own generation by positive feedback systems but also inhibits it through its interaction with thrombomodulin. By binding to this endothelial receptor, thrombin loses its procoagulant function and can then activate protein C. Activated protein C (APC) inhibits activated FVIII (FVIIIa) and activated FV (FVa) with protein S acting as a vital cofactor (Esmon, 1989; Walker & Fay, 1992). A further major natural anticoagulant inhibiting thrombin generation is antithrombin (Lawson et al, 1993).

Traditional coagulation tests, such as the prothrombin time (PT) and activated partial thromboplastin time (APTT), do not assess the whole coagulation system. These tests use clot formation as their endpoint, which occurs when only around 5% of all physiologically relevant thrombin is formed (Hemker & Beguin, 1995; Rand et al, 1996) and also are insensitive to prothrombotic states. Coagulation factor assays can identify specific deficiencies but these do not always closely correlate with the clinical phenotype. The limitations of the traditional tests are also demonstrated by the observation from Butenas et al (1999) who showed that thrombin generation varies up to 40-fold when measurements are done with individual coagulation factors at the extremes of the normal ranges in a synthetic plasma system.

Measurement of an individual’s capacity to generate thrombin, however, captures the end result of the interaction between proteases and their inhibitors and is therefore potentially more useful as a reflection of a thrombotic (high thrombin generation) or haemorrhagic (low thrombin generation) phenotype compared to conventional coagulation tests. The in vitro capacity of plasma to generate thrombin over time has been termed the endogenous thrombin potential (ETP) (Hemker et al, 1986) and must be distinguished from in vivo markers of thrombin generation such as prothrombin fragments 1 and 2 (F1+2), thrombin anti-thrombin complexes and fibrinopeptide A. Measurement of thrombin generation was initially described in 1953 (Macfarlane & Biggs, 1953; Pitney & Dacie, 1953). This method however was difficult to perform and subject to high variability. Over the last 30 years Hemker’s group developed an automated method capable of measuring thrombin generation in multiple samples simultaneously (Hemker et al, 1986, 1993, 2000, 2003), thus facilitating measurement. These techniques initiate thrombin generation by the addition of a trigger to recalcified plasma in the presence of phospholipids causing the cleavage of a chromogenic or fluorogenic substrate by thrombin. The chromogenic method has to be performed in defribrinated plasma, whereas this is not necessary with fluorogenic methods that can be performed in platelet rich (PRP) as well as platelet poor (PPP) plasma (Hemker et al, 2000; Ramjee, 2000). Recently, a fluorogenic assay in whole blood (WB) has been described (Tappenden et al, 2007). For a detailed discussion on the technical aspects of thrombin generation assays the reader is referred to a recent review (Baglin, 2005).

The parameters measured in the fluorogenic method include the lag time (LT, defined as the moment that the signal deviates by more than 2 standard deviations from the horizontal baseline), peak thrombin generation, the time to peak thrombin generation (ttpeak) and the ETP. A typical trace is shown in Fig 1. Thrombin generation assays must be distinguished from assays that are dependent on thrombin generation. Instead of quantifying its generation, similar assays measure its effect on clot formation. The latter include wave form analysis, overall haemostasis potential and thromboelastography/thromboelastometry. These methods have been reviewed (Barrowcliffe et al, 2006). Thromboelastometry has recently been modified to use the first derivative of the generated curve, which produces graphs similar to those of thrombin generation tests (Sorensen et al, 2003). This new method uses a low concentration of TF to trigger the assay, making it sensitive to feedback activation of thrombin by the intrinsic pathway (Sorensen et al, 2003; Sorensen & Ingerslev, 2004). Although the first derivative curve mirrors the thrombin generation curve in appearance, it is not equivalent. It represents the velocity profile of whole blood clot formation and is also dependent on fibrinogen content and integrity, fibrin polymerisation, cross linking and fibrinolysis. Thromboelastography and thromboelastometry have been comprehensively reviewed (Luddington, 2005).

Figure 1.

 Thrombin generation curve using the calibrated automated thrombogram showing the four main parameters of lag time, time to peak (ttpeak), peak thrombin and endogenous thrombin potential (ETP).

The simplification and improved reproducibility of thrombin generation measurement has sparked a significant increase in laboratories performing thrombin generation tests and the development of three commercially available assays (two fluorogenic and one chromogenic assay). The increased use of these assays raises the question about the feasibility and practical use of thrombin generation testing in routine clinical practice. This review focuses on the technical and clinical aspects of the fluorogenic method as the majority of the recent published literature in this field has been obtained using this method.

Current technical issues

For a method to be accepted in routine laboratory practice, it needs to be reproducible, standardized and relatively easy to perform. There are currently two commercially available fluorogenic thrombin generation assays, the calibrated automated thrombin generation assay (CAT) (Thrombinoscope B.V., Maastricht, The Netherlands) and the Technothrombin® TGA assay (Technoclone, Vienna, Austria). Published studies have shown acceptable reproducibility with intra- and inter-assay coefficient of variation (CV) of <10% with both PPP and PRP (Hemker et al, 2003; Vanschoonbeek et al, 2004; Gerotziafas et al, 2005).

The variation within a laboratory is however dependent on standardization of reagents and methods, particularly if a laboratory uses self prepared reagents. The introduction of new reagents or batches of reagents can introduce significant variation, requiring the assay to be recalibrated and the introduction of new local normal ranges. Internal controls should be used to assess the stability of the system over time. The preparation of plasma, storage and different blood collection systems are also potential sources of variation (Chantarangkul et al, 2003, 2004; Hemker et al, 2003; Regnault et al, 2004) and should be kept constant during a study. Lowering the concentrations of phospholipids and TF used have also been shown to increase variation. Increased variation with decreasing concentrations of phospholipids was shown at very low concentrations ranging between 0·0 and 1·5 μmol/l (Chantarangkul et al, 2003). Most laboratories, however, use concentrations of 4 μmol/l, as discussed below. Low TF concentrations (1 pmol/l) increase assay variability but this has not been a universal finding. One study showed intra-assay CVs ranging from 1·4–20·4% whilst examining three pathological samples in five different laboratories using standardized reagents (TF 1·5 pmol/l and phospholipids 4 μmol/l) and methods (Dargaud et al, 2007). Another study showed intra-assay variation of up to 12·5% for the ETP and 25% for peak thrombin generation in a sample with INR 2·1 and variation for peak thrombin between 5·8% and 15·8% in the plasma of four normal individuals (measured at 1·0 pmol/l TF and 4 μmol/l phospholipids) (van Veen et al, 2008a). In both studies the degree of variation depended on the sample and in the first study also on the centre. In contrast, Chantarangkul et al reported between run variations of less than 5·6% measured at 1 pmol/l TF and 1·5 μmol/l phospholipids (Chantarangkul et al, 2003). In the above studies the intra-assay variation could be reduced to acceptable (<9·1%) levels when contact pathway activation was inhibited with corn trypsin inhibitor (CTI), suggesting that activation of this pathway is associated with increased variability. The influence of contact pathway activation was strongest at TF below 2 pmol/l and CTI did not make a difference at 5 pmol/l (van Veen et al, 2008a). Contact pathway activation may therefore explain some of the differences reported in intra-assay variation at low TF concentrations. If CTI is used, blood samples need to be taken into pre-filled syringes (Luddington & Baglin, 2004).

Although the variation within a laboratory is acceptable in a well controlled setting, comparisons between laboratories are problematic. A collaborative study coordinated by the National Institute for Biological Standards and Control (NIBSC, Potters Bar, UK) on thrombin generation tests in PPP has shown inter-laboratory variation of 16–20% for the ETP and 15–24% for the peak thrombin generation for 41 participating centres using the CAT method and 30–50% for the peak thrombin generation for 29 centers using the Technothrombin® TGA assay (Dr E. Gray, NIBSC, personal communication). Variation to this degree makes multicentre studies difficult to perform but could be improved upon if the results are expressed as a ratio to reference plasma used by all participating laboratories. In another collaborative study involving over a 100 laboratories, using a reference plasma for normalization reduced variation from 17–21% to 5–13% for the CAT method (ETP), and from 20–24% to 10–22% for the Technothrombin® TGA assay (Tmax) (Dr E. Gray, NIBSC, personal communication). In the NIBSC study participating centres used standard reagents. These results contrast with other published observations where different laboratories often use their own reagents. This was shown in a recent multicentre study (Dargaud et al, 2007) that found inter-centre CVs of 39–109%. The incomparability of the results was mainly due to the variability of TF concentrations used in the different centres (0·5–5 pmol/l, Table I). Variation was reduced to 4·4–23·2% when standardized reagents were used and the results normalized as a ratio to a reference plasma. When CTI was used and only the centres compared with the same conditions using experienced operators, inter-centre variation was less than 7·5% (Dargaud et al, 2007). As measurements in PRP need to be done in fresh plasma, studies on inter-laboratory variation are not possible.

Table I.   Comparison of the ETP and inter-centre CV values of normal, protein C deficiency, FVIII and FIX deficient samples using locally available reagents and protocols without corn trypsin inhibitor. Adapted from Dargaud et al (2007).
SampleCentre 1Centre 2Centre 3Centre 4Centre 5Inter-centre CV%
Mean of 6 normal samples (mean ± SD)13122954 8891194 768 62
Protein C: 55%2629 (204)2963 (260)1130 (36)1635 (153)1602 (179) 39
FIX: 4 IU/dl 409 (87)2202 (251) 222 (30)  56 (17)1163 (118)109
FVIII: 7 IU/dl 521 (122)1695 (84) 202 (26) 114 (22) 668 (81) 99
Tissue factor (final concentration)Thromborel S -Dade Behring (2·5 pmol/l)Innovin – Dade Behring (5 pmol/l)Innovin – Dade Behring (1 pmol/l)Innovin – Dade Behring (0·5 pmol/l)Innovin – Dade Behring (0·56 nmol/l) 
Phospholipids (final concentration)PC 60-PS 20-PE 20 mol% (4 μmol/l)PC 20-PS 13-PE 45-PA 22 mol% (4 μmol/l)PC 60-PS 20-PE 20 mol% (4 μmol/l)PC 60-PS 20-PE 20 mol% (4 μmol/l)Pathrombin SL-Dade Behring diluted at 1:2500 
Thrombin substrate (final concentration)Z-Gly-Gly-Arg-AMC Bachem 2·5 nmol/l)Z-Gly-Gly-Arg-AMC Bachem 2·5 nmol/l)Z-Gly-Gly-Arg-AMC Bachem 2·5 nmol/l)Z-Gly-Gly-Arg-AMC Bachem 2·5 nmol/l)Z-Gly-Gly-Arg-AMC Bachem 1·96 nmol/l) 
Thrombinoscope software versionV3.0.0.25V3.0.0.25V3.0.0.`25V2.106V3.0.0.29 

Studies reporting on thrombin generation that have used the same concentration can potentially also have very different results, especially if ‘in house’ reagents are used. The manufacturers of TF do not state the concentration of TF in their products and although several TF assays are available, their sensitivity and specificity are variable and different methods correlate poorly. This can potentially cause significant differences between laboratories even if the TF concentration is measured. Furthermore, there is no international reference standard for TF and different products may behave differently.

A final potential source of variation is related to the calibrator used. The relationship between thrombin generation and fluorescent activity is non-linear due to substrate consumption during the reaction process and non-linearity of fluorescent activity with increasing concentration of fluorescent molecules (inner filter effect). Thus, every fluorescent level requires a different calibration factor to calculate thrombin concentration (Hemker et al, 2003). As the fluorescence level is also influenced by the colour of the plasma, different coloured plasmas with the same thrombin generating potential can potentially give different results if this is not calibrated for (Hemker et al, 2003). There are currently two calibrator systems in use. The CAT assay uses a calibrator with known thrombin-like activity comparing thrombin generation in test plasma continuously to α2-Macroglobulin-thrombin activity of the calibrator in the same plasma. This system corrects for the above variables automatically. The Technothrombin-TGA® assay compares initial thrombin generation in 2:5 diluted plasma to a fixed calibration factor determined in buffer. It does not correct for substrate consumption or the inner filter effect. A preliminary report directly comparing both methods on thrombin generation in the same plasma (de Smedt et al, 2007), showed an underestimation of plasma activities of approximately 18% and increased intra-individual variation of approximately 10% excess CV due to variation in the colour of the plasma in the Technothrombin-TGA® method. The latter method may also not recognize substrate depletion and the contribution of α2-Macroglobulin-thrombin activity is not taken into account. The stability of the calibrator is paramount for both comparative and longitudinal studies and we reported data suggestive of significant batch-to-batch variation in calibrator potency of the CAT assay (van Veen et al, 2006).

All the above suggests that it may be possible to perform reproducible measurements between different centres provided that standardized reagents and reference plasmas are used.

Several variations of the assay have been reported, including:

  • 1TF concentration used to trigger the assay. The positive feedback activation of thrombin by the intrinsic pathway can only be demonstrated at low TF concentrations. TF concentrations of <1 pmol/l in PPP are necessary to demonstrate the effect of FXI (Keularts et al, 2001) and although the assay is sensitive to FVIII and FIX at TF concentrations up to 5 pmol/l, its sensitivity is reduced at the higher concentration (van Veen et al, 2008a). Small differences in TF concentration can therefore cause significant differences in the thrombin generating potential of plasma. For example, using 1 pmol/l in patients with severe haemophilia will result in almost absent thrombin generation in some patients whereas in others thrombin generation may be present. If the same plasmas are investigated at 5 pmol/l, significant amounts of thrombin generation are present (Lewis et al, 2007) and it is possible that the differences seen at 1 pmol/l between individuals are missed due to the decreased sensitivity. Conversely, thrombin generation measurements at low TF concentrations in haemophilia are difficult because of the low signal obtained and it may not always be possible to obtain full curves (Lewis et al, 2007), limiting the information gained from the measurements. The response of the assay at 5 pmol/l (except FXI, measured at 1 pmol/l) for different coagulation factors is shown in Fig 2. This illustrates a non-linear relationship with most coagulation factors, except for prothrombin, and significant thrombin generation even at unmeasurable FVIII and FIX levels. Low TF concentrations are also needed to demonstrate the protein C independent effect of protein S on TFPI (Sere et al, 2004; Hackeng et al, 2006) and may be preferable in measurements in the presence of inhibitor bypass agents in haemophilia (van Veen et al, 2008b). The determinants of thrombin generation in a normal population at two TF concentrations without or without thrombomodulin or APC were studied by Dielis et al (2008), who showed that thrombin generation was determined by different coagulation factors or their inhibitors, depending on experimental conditions.
  • 2The use of CTI to inhibit contact pathway activation. Given the significant influence of contact factor activation suggested in various studies (Luddington & Baglin, 2004; Dargaud et al, 2006a; Lewis et al, 2007; Tappenden et al, 2007; van Veen et al, 2008a), it is our opinion that inhibition of contact factor activation may be useful in measurements of plasma with low thrombin generating potential, such as haemophilic plasma, at low TF concentrations. Contact factor activation in this population could give falsely high levels of thrombin generation, obscuring actual (low) thrombin generation levels.
  • 3Phospholipid concentration. The phospholipid concentration used in PPP measurements is dependent on a requirement to be rate limiting. The effect of the phospholipid concentration approaches a plateau at concentrations of 3–5 μmol/l (Chantarangkul et al, 2003; Hemker et al, 2003; Gerotziafas et al, 2005) and is not rate-limiting. Most studies have used concentrations in this range but lower concentrations or no phospholipid addition is used when the effect of microparticles is to be studied.
  • 4Activated FIX (FIXa) to trigger thrombin generation. FIXa is required to measure the effect of FVIII on thrombin generation. This can be achieved by triggering the assay with low concentrations of TF or alternatively, directly by using FIXa as trigger. The latter was found to be sensitive to differences in thrombin generation at very low FVIII concentrations (McIntosh et al, 2003).
  • 5Protein C pathway sensitizing agents. Unmodified thrombin generation assays are not sensitive to the actions of the protein C pathway. In vivo activation requires complex formation between thrombin and thrombomodulin, activating protein C. Thrombomodulin is not present in the standard assays but the sensitivity to the protein C pathway ex vivo can be increased by adding truncated human recombinant thrombomodulin (sTM) (van Hylckama Vlieg et al, 2007), APC (Regnault et al, 2003a) or the snake venom Protac® (Gatt et al, 2007a; Hezard et al, 2007). The use of these agents makes the assay more sensitive to deficiencies in protein C (not detected if APC is used), protein S, F5 R506Q (factor V Leiden) and conditions associated with acquired protein C resistance. The ETP is the most frequently reported parameter in this setting. Results can also be expressed as ratios of ETP in the presence of activators to ETP without activators (Gatt et al, 2007a; Hezard et al, 2007).
Figure 2.

 Influence of single coagulation factor concentrations of FII, FV, FVII, FVIII, FIX, FX and FXI on the thrombogram parameters. (A) ETP, (B) peak. Measured at 5 pmol/l TF except FXI which was measured at 1 pmol/l.

In summary, thrombin generation measurements can be performed in several ways that have different sensitivities for various haemostatic or thrombotic defects. As small variations in pre-analytical variables can cause significant changes in the sensitivity for these defects, results cannot be interpreted without detailed information on how the assay was performed. The assays are reproducible within a single laboratory when the same reagents and method is used but multicentre studies, at this stage, are difficult in the absence of international thrombin and tissue factor standards. Both issues are currently being addressed by the relevant International Society on Thrombosis and Haemostasis/Scientific and Standardization Committee working groups. Finally, even when a single well standardized test is available it is unlikely that it will be suitable in all situations and separate modified tests designed for specific clinical situations may need to be developed.

Thrombin generation in clinical practice

Inherited bleeding disorders

Haemophilia.  Most information on thrombin generation in inherited bleeding disorders is available for haemophilia. Several studies have shown a correlation between FVIII coagulant activity (FVIII:C)/FIX coagulant activity (FIX:C), measured at baseline, and the ETP/peak thrombin generation (Chantarangkul et al, 2003; Beltran-Miranda et al, 2005; Dargaud et al, 2005a). All showed a good discrimination for the ETP and peak thrombin generation between normal subjects and patients with different degrees of severity of haemophilia (Siegemund et al, 2003; Luddington & Baglin, 2004). They also showed a high degree of variability in thrombin generation between patients with similar FVIII:C/FIX:C levels. Some also reported on the effect of factor replacement therapy and in vitro spiking of plasma samples with FVIII on thrombin generation (Siegemund et al, 2003; Beltran-Miranda et al, 2005; Dargaud et al, 2005a). These showed a clear effect but with different responses in patients with similar FVIII:C levels. Some spiking studies suggest that the ETP and peak thrombin generation reach a maximum at FVIII:C levels of 0·50–1·0 IU/ml (McIntosh et al, 2003; Beltran-Miranda et al, 2005) and that FVIII:C is no longer rate limiting at these concentrations. Lewis et al (2007) reported the relationship between thrombin generation and FVIII pharmacokinetics over 72 h after the treatment of 12 patients. They used blood samples taken directly into CTI to demonstrate a linear relationship between log FVIII:C and the log ETP/peak thrombin generation/rate of thrombin generation without the occurrence of a plateau. Despite inter-individual variability, the response to treatment in terms of thrombin generation at any FVIII:C level could be predicted for individual patients (Lewis et al, 2007). The discrepancy between the study by Lewis et al (2007) and Beltran-Miranda et al (2005) in terms of the occurrence of a plateau at FVIII:C 0·5–1·0 IU/ml may be explained by contact factor interference (van Veen et al, 2008a).

Taken together, all these reports have shown a marked heterogeneity in thrombin generation at similar FVIII:C levels whereby some individuals (at the extremes) with severe haemophilia can have normal ETP values at baseline whilst others have subnormal values even when treated to normal FVIII:C levels. Each individual however appears to have a typical and predictive response.

Most of the studies have used PPP and not all thrombin binding proteins are taken into account. Siegemund et al (2003) demonstrated that the influence of platelets seemed to be different between haemophilia A and B and Keularts et al (2000) demonstrated the that thrombin generation in von Willebrand disease (VWD) can be normalized by scrambling of the platelet membrane whereas this does not influence thrombin generation in haemophilia A. Inter-individual differences in thrombin generation are therefore not only due to differences in coagulation factors and inhibitors but also to differences in components, such as glycoprotein receptors, on platelets and fibrin. Overall however, the studies in haemophilia have raised the possibility of tailoring treatment on global haemostatic response rather than normalization of factor levels, but its clinical utility will depend on a correlation between thrombin generation and bleeding phenotype. Two studies have demonstrated a relationship between the severity of the bleeding tendency and the ETP, independent of factor levels. Dargaud et al (2005a) showed that in haemophilia patients, a severe phenotype was associated with a reduction of the ETP to less than 50% of the mean ETP of a normal population whereas Al Dieri et al (2002) demonstrated this in patients with rare coagulation factor deficiencies with an ETP <20% of normal. In contrast, Beltran-Miranda et al (2005) reported a good correlation between the severity of haemophilia (when assessed as the degree of FVIII deficiency) and peak thrombin generation but not with the clinical phenotype and concluded that there was no advantage of measuring thrombin generation over FVIII:C measurements.

Haemophilia with anti-FVIII antibodies.  Of particular interest is the possibility of monitoring inhibitor bypass therapy in haemophilia as their clinical effectiveness cannot be assessed by traditional coagulation tests and monitoring may rationalize treatment schedules and lead to cost savings. Both spiking studies and pharmacokinetic studies with Factor Eight Inhibitor Bypassing Activity (FEIBA®) showed a near linear dose–response relationship with peak thrombin generation and normalization of thrombin generation after treatment, suggesting that the method may be used to monitor treatment (Turecek et al, 2003; Varadi et al, 2003; van Veen et al, 2008b). Monitoring was also suggested in a patient with inhibitors undergoing surgery (Dargaud et al, 2005b).

Several studies examined the effect of recombinant (r)FVIIa on thrombin generation in different settings (Monroe et al, 1997; Kjalke et al, 2001; Butenas et al, 2002; Gerotziafas et al, 2002, 2004a,b; Turecek et al, 2003; Wegert et al, 2005; Allen et al, 2006; Livnat et al, 2006; van Veen et al, 2008b). Different techniques and methodologies were used and only some could demonstrate an effect of rFVIIa on thrombin generation whereas all showed a shortening of the lag phase. Of the studies using a fluorogenic method (Gerotziafas et al, 2002, 2004b; Turecek et al, 2003; Wegert et al, 2005; Livnat et al, 2006; van Veen et al, 2008b), three investigated the effect of rFVIIa on thrombin generation in haemophilia (Turecek et al, 2003; Livnat et al, 2006; van Veen et al, 2008a). Turecek et al (2003) showed a dose–response relationship in spiking studies using inhibitor plasma but concentrations higher than those used in a therapeutic setting were needed to normalize peak thrombin. We showed a similar dose–response relationship in spiking experiments and demonstrated an effect on thrombin generation after four treatment episodes, dependent on TF concentration (van Veen et al, 2008b). Livnat et al (2006) spiked plasma of FVIII-, FIX- and FXI-deficient patients and demonstrated a response that was dependent on phospholipid and TF concentration used as well as the underlying defect. Two other studies using different techniques investigated the effect of rFVIIa in haemophilic blood (Butenas et al, 2002; Allen et al, 2006). Butenas et al (2002) did not find an effect of rFVIIa on thrombin generation whereas Allen et al (2006) did show a positive effect and dependence on underlying prothrombin concentration. The variability in results emphasizes the need for optimization of the assay and further investigations into the relationship with clinical outcome.

Von Willebrand disease.  Only limited data are available on other inherited bleeding disorders. One report showed impaired thrombin generation in PRP but not in PPP in two patients with type 1 VWD and demonstrated a response to treatment with desmopressin (Keularts et al, 2000). A more recent, larger study evaluated thrombin generation in 53 patients with VWD, and showed decreased and delayed thrombin generation in both PPP and PRP (Rugeri et al, 2007). Thrombin generation was primarily dependent on FVIII:C levels and a correlation with bleeding (as assessed by bleeding score) was demonstrated with the peak thrombin generation.

Other bleeding disorders.  Thrombin generation in PRP was impaired in a patient with Bernard Soulier syndrome (Beguin et al, 2004) and seven patients with Glanzmann thrombasthenia (measured by subsampling technique) (Reverter et al, 1996; Beguin et al, 1999). The possibility of monitoring rFVIIa was suggested in a patient with Glanzmann thrombasthenia undergoing surgery (Dargaud et al, 2006b).

Overall, thrombin generation in inherited bleeding disorders shows marked differences between individual patients with similar defects. The differences persist after treatment and individuals may have a predictive response, opening the possibility of individualized treatment based on thrombin generating capacity. The data on the correlation with bleeding phenotype however are based on indirect parameters for bleeding, such as the Petterson score, self reported bleeding frequency and the need for prophylaxis for severe recurrent bleeding. Tools that rely on self reported bleeding frequency are dependent on an individual’s perception of bleeding and may vary between patients. Furthermore, all data are retrospective and there is a significant possibility of recollection bias. Therefore, although the assay has the potential to alter current treatment practice, large prospective multi-centre studies with clinical bleeding endpoints are required before adaptation of the technique in routine clinical practice. Such studies are difficult in a population of patients with inherited bleeding disorders, such as haemophilia, where set treatment protocols exist to normalize factor levels prior to invasive procedures or who are on prophylaxis and who do not get bleeds. In non-prophylaxis patients, the arthropathy influences the bleeding frequency to such a degree that it can be difficult to study the influence of thrombin generation. Studies into the relationship between thrombin generation parameters and bleeding phenotype may be feasible in procedures associated with relatively high bleeding rates in a normal population, such as tonsillectomy, or procedures in fully anticoagulated patients on warfarin, such as dental extractions.

Monitoring anticoagulation.  Heparin and warfarin are the most commonly used anticoagulants worldwide. These agents have a narrow therapeutic window and require regular monitoring with the APTT and the PT respectively. These tests are far from ideal since they are non-physiological and cannot be used to monitor most of the newer anticoagulant drugs. It would be desirable to have a test that could be sensitive to all anticoagulants and emulate what happens in vivo. Several groups studied thrombin generation assays in monitoring anticoagulant therapy. Most studies involve in vitro spiking experiments and a few ex vivo data from subjects taking different anticoagulants. Several modifications of the test have been used, making direct comparisons difficult.

Warfarin: The ETP has a modest hyperbolic relationship and negative correlation with the International Normalized Ratio (INR) in patients on warfarin therapy (Jackson et al, 2003; Altman et al, 2007). There is however significant variation in ETP results for patients with the same INR and some patients appear to be over and others under-anticoagulated if monitored by a thrombin generation assay (Gatt et al, 2007b). This suggests that the use of ‘physiological’ TF concentrations in thrombin generation assays may provide a more sensitive tool for warfarin monitoring which could potentially equate in a greater safety margin in these patients but large prospective studies are required to confirm this.

Heparins/Heparinoids: Thrombin generation is a sensitive way to measure the effects of unfractionated heparin (UFH) (Al Dieri et al, 2004), low molecular weight heparins (LMWHs) (Al Dieri et al, 2006; Gerotziafas et al, 2007a) and danaparoid (Stief, 2007). UFH, tinzaparin and danaparoid have the most profound effect on thrombin generation because of their highest anti-FIIa effect. There is significant variability in thrombin generation between patients with ‘similar’ APTT and anti-FXa measurements (Al Dieri et al, 2004).

Factor Xa inhibitors: Both indirect (e.g. fondaparinux) and direct (e.g. rivaroxaban) FXa inhibitors have been assessed using thrombin generation assays (Gerotziafas et al, 2004c, 2007b). A dose-dependent effect is seen on all thrombin generation parameters. With these new agents the concentration of anticoagulant required to decrease the LT and ttPeak to 50% (IC50) is lower than that required to produce the same effect on the peak and total thrombin potential. The direct FXa inhibitors like rivaroxaban, are small enough to block clot bound prothrombinase and this is mirrored in the thrombin generation assay, because direct FXa inhibitors are able to suppress thrombin generation completely at high concentrations whereas fondaparinux does not.

Direct thrombin inhibitors: Thrombin generation assays have been performed with the direct thrombin inhibitors lepirudin, bivalirudin, argatroban and dabigatran. Lepirudin and bivalirudin considerably prolong the LT with no significant effect on the other thrombin generation parameters (Tanaka et al, 2007a). Smaller molecules, such as dabigatran and argatroban, have a concentration-dependent effect on the LT, ttPeak, Peak and ETP (Tanaka et al, 2004; Wienen et al, 2007).

Anti-platelet agents: The adoption of a fluorogenic thrombin substrate has made it possible to measure thrombin generation in PRP. Thrombin generation in PRP has been employed in assessing the effect of aspirin, clopidogrel and the glycoprotein (GP)IIbIIIa antagonist abciximab (Altman et al, 2005, 2006). These studies suggest that in TF- or arachidonic acid-triggered thrombin generation assays, the LT and time to peak are prolonged in subjects on aspirin and clopidogrel but the ETP is not affected.

In summary, thrombin generation assays could offer some advantages over the traditional coagulation tests as they are sensitive to all anticoagulants and allow more precise titration of each anticoagulant to the individual patient, but large prospective studies correlating thrombin generation suppression and thrombotic risk are required.

Reversal of anticoagulation. Warfarin: The INR is currently the test of choice for monitoring warfarin reversal. However, this test is only sensitive to factors I, II, V, VII and X. Hence INR reversal does not necessarily equate to reversal of all vitamin K-dependent factors. We showed that despite normalization of the INR with FFP in most over-anticoagulated patients, FIX concentration is suboptimal when compared with the use of prothrombin complex concentrates (PCCs) (Makris et al, 1997). Evans et al (2001) confirmed that the normalization of the INR after PCC use was also associated with thrombin generation normalisation. Furthermore Tanaka et al reported that unlike the effect seen with the PCCs, rFVIIa led to complete INR correction but had minimal effects on thrombin generation parameters (Tanaka et al, 2007b).

Heparins/Heparinoids: With the more widespread use of LMWH in the treatment of venous and arterial thromboembolism, it is not uncommon to be presented with the issue of reversal of anticoagulation induced by these agents. We observed complete reversal of UFH and tinzaparin by protamine sulphate whereas enoxaparin was only partially neutralized (Gatt et al, 2006). This is most probably due to the different degree of sulphation of these drugs.

Both in vitro (Gerotziafas et al, 2004b; Gatt et al, 2006) and limited ex vivo and clinical experience (Bijsterveld et al, 2002; Huvers et al, 2005) point to the beneficial use of rFVIIa in reversing the anticoagulant effect of fondaparinux. Our study (Gatt et al, 2006) had shown that rFVIIa does not seem to neutralize fondaparinux but probably ‘bypasses’ its anticoagulant activity in the presence of TF.

Anti-platelet agents: Thrombin generation has been used to measure the reversal of antiplatelet agents. This is important due to the increasing number of high risk cardiac patients, especially those with coronary artery stents, undergoing surgery on both aspirin and clopidogrel. Altman et al (2006) showed that rFVIIa could restore baseline thrombin generation in PRP collected from subjects taking aspirin or aspirin plus clopidogrel. More studies are required in order to establish a role, if any, for thrombin generation assays in these specific clinical scenarios.

Thrombin generation in thrombotic disorders

Venous thrombosis

Venous thrombosis is associated with significant morbidity and mortality. After a first idiopathic event, treatment with anticoagulation is often discontinued after 6 months even though up to 19% of patients are reported to have a recurrence in the following 2 years (Baglin et al, 2003). Although various genetic and environmental risk factors for venous thrombosis are known, these are only weakly associated with recurrence (Christiansen et al, 2005). A predictive parameter for recurrence would be desirable for deciding on continuing anticoagulation in the group of patients who are likely to develop a second thrombosis. Thrombin generation, measuring the cumulative effect of pro-thrombotic tendencies, may be suitable for this purpose. In fluorogenic thrombin generation measurement, the presence of FV R506Q, the F2 G20210A (prothrombin G20210A) mutation and hereditary deficiencies of protein C, protein S and antithrombin have all been associated with increased thrombin generation in a variety of studies (Andresen et al, 2004; Luddington & Baglin, 2004; Regnault et al, 2004; Dargaud et al, 2006c; Hezard et al, 2006, 2007; Hron et al, 2006; Castoldi et al, 2007; Gatt et al, 2007a; van Hylckama Vlieg et al, 2007). However, similar to the studies in haemophilia patients, there are large differences in thrombin generating potential between patients with the same defect. Increased thrombin generation has been shown in patients with high levels of:

These studies have used a variety of methods to demonstrate the effect of different prothrombotic states, including the use of PRP, frozen-thawed PRP (ft-PRP) and PPP, different TF and phospholipid concentrations and agents to improve the sensitivity of the assay for defects in the protein C pathway. The latter appear to be of particular importance in demonstrating a prothrombotic phenotype. Several studies have shown that patients with F5 R506Q, protein C or protein S deficiency, using the COCP/HRT or with high levels of FVIII have an increased thrombin generation/protein C resistance compared to normals when an activator of the protein C pathway is used in the assay (Regnault et al, 2003b, 2004; Dargaud et al, 2006c; Eilertsen et al, 2007; Lecompte et al, 2007). Failure to use an activator is a possible reason why some studies have not been able to demonstrate a clear relationship between FVIII, F5 R506Q, protein C or protein S and increased thrombin generation (Andresen et al, 2004; Hron et al, 2006). One study, which did not use a protein C pathway activator, compared thrombin generation in patients with venous thromboembolism (VTE) and healthy controls in PPP, PRP and WB (Tappenden et al, 2007). This study showed increased thrombin generation in patients with VTE that was clearer in the WB assay than using PRP or PPP. Demonstration of increased thrombin generation related to the F2 G20210A mutation and antithrombin deficiency does not require a protein C pathway activator. Thrombin generation/protein C resistance is increased in women using the COCP. CAT measurements with APC can differentiate between the effects of second and third generation COCP as compared to normals (ETP+APC third generation >ETP+APC second generation >ETP+APC no COCP) suggesting that the effects on the protein C system are likely to determine the differences in thrombotic risk between the different preparations (Tchaikovski et al, 2007).

Lupus anticoagulant

Activated protein C resistance is thought to be one of the ways through which LA is associated with a prothrombotic phenotype in the antiphospholipid syndrome (APS) (Esmon et al, 1999; Male et al, 2001). Fluorogenic thrombin generation measurements have shown that the presence of LA is associated with an increased initiation time of thrombin generation but also a decreased APC-induced inhibition of the ETP and peak thrombin (Regnault et al, 2003a,b, 2004; Lecompte et al, 2007). Decreased APC-induced inhibition of thrombin generation in these studies was clearer when measurements were performed in the presence of the patients own platelets (PRP or ft-PRP) than with exogenous added phospholipids (PPP). One study, using PPP, demonstrated inhibition of APC-induced thrombin generation independent of warfarin usage (Liestol et al, 2007).

Recurrent thrombosis

Although thrombin generation is increased in thrombophilic states the test will only be clinically important if it can be used to predict recurrent thrombosis. This has been investigated by two studies. Hron et al (2006) used the Technothrombin® TGA assay and found a relationship between deep vein thrombosis recurrence and peak thrombin generation in 914 patients with a median follow-up of 47 months. The risk of recurrence at 4 years was 6·5% when peak thrombin was <400 nmol/l and 20% if peak thrombin was >400 nmol/l. The authors felt that thrombin generation could be used to identify patients at low risk for recurrence. However, these findings were not supported by van Hylckama Vlieg et al (2007), who used one in four diluted PPP, a relatively high concentration TF (15 pmol/l) and thrombomodulin, in 360 patients and 404 controls of the Leiden Thrombophilia Study. Although an increased ETP was predictive for a first VTE, it did not predict for recurrence at 7·6 years follow-up. Table II summarizes the data of studies into the relationship between thrombin generation and hypercoagulable states, limited to studies including patients with VTE and using a fluorogenic thrombin generation method.

Table II.   Studies into the relation between thrombin generation and hypercoagulable states, limited to those including patients with VTE and using a fluorogenic method.
(Tissue factor/ phospholipid)Protein C pathway sensitizing agentPlasma
  1. VTE, venous thromboembolism; AT, anti-thrombin; LA, lupus anticoagulant; FVIII:C, FVIII coagulant activity; COCP, combined oral contraceptive pill; DVT, deep vein thrombosis; APS, anti-phopsholipid syndrome; PPP, platelet poor plasma; (ft-) PRP, (frozen-thawed-) platelet rich plasma; WB, whole blood; CTI, corn trypsin inhibitor; ETP, endogenous thrombin potential; OR, odds ratio; APC, activated protein C.

van Hylckama Vlieg et al, 2007360 patients after a first VTE (187 idiopathic and 173 provoked) of Leiden Thrombophilia Study and 404 controls15 pmol/l/4 μmol/lThrombomodulinPPPOR 1·5 for 1st VTE overall, OR 1·7 in idiopathic VTE if ETP >90th percentile of control group. Not predictive for recurrence
Hron et al, 2006914 patients with 1st VTE of which 100 had a recurrence. Protein C/S and AT deficiency, LA and FVIII:C > 230 IU/dl excludedApproximately 6 pmol/l/3·2 μmol/lPPPRecurrence risk 6·5% if peak <400 nmol/l and 20% if peak >400 nmol/l
Dargaud et al, 2006c71 normal subjects, 11 women on the COCP, 89 patients after first DVT (50 with heritable thrombophilia or high FVIII)5 pmol/l/4 μmol/lThrombomodulinPPPETP >mean + 2SD of normal was associated with OR 40·5 in patients with congenital thrombophilia and OR 5·85 without thrombophilic makers for 1st DVT
Lecompte et al, 200740 patients with antiphospholipid antibodies (24 with thrombosis) and 19 controls0·5 pmol/lActivated protein CPRPSignificant activated protein C resistance with OR for thrombosis up to 7
Liestol et al, 200781 patients with LA (52 APS, 34 on warfarin) and 91 controls5 pmol/l/4 μmol/lActivated protein CPPPThrombin generation in LA plasma is increased independent of warfarin use and the degree of APC appears to be associated with thrombotic events
Hezard et al, 2006169 patients with personal (71%) or family history of VTE0·5 pmol/lActivated protein Cft-PRPSignificant APC resistance in patients with protein S deficiency and F5 R506Q compared to patients without heritable thrombophilia. Can be used to screen for these defects
Hezard et al, 2007102 patients with personal or family history of VTE0·5 pmol/lProtacft-PRPSignificant APC resistance in patients with protein C/S deficiency and F5 R506Q compared to patients without heritable thrombophilia. The test can be used to screen for these defects
Castoldi et al, 2007302 individuals from unrelated families. 167 heterozygous (39 with previous thrombosis) and 3 homozygous for the F2 G20210A mutation vs. 132 non carriers (14 with previous thrombosis)6·8 pmol/l/30 μmol/lActivated protein CPPPSignificantly increased thrombin generation in patients with increased FII concentration due to the F2 G20210A mutation compared to non carriers and increased FII not associated with this mutation
Gatt et al, 2007a56 patients with inherited thrombophilia, 34 negative family members and 32 controls5 pmol/l/4 μmol/lProtacPPPSignificant protein C resistance with high interindividual variation in the presence of thrombophilia
Regnault et al, 2003b8 patients with LA (3 with previous thrombosis)0·5 pmol/l/4 μmol/l (PPP only)Activated protein C or thrombomodulinPPP/(ft-)PRPPatients with LA show protein C resistance, more marked when own platelets are present
Tappenden et al, 200750 patients with VTE (56% idiopathic) and 31 controlsPPP: 5 pmol/l/4 μmol/l PRP/WB: 1 pmol/lPPP/PRP/WB ± CTIThrombin generation was significantly increased in VTE patients in the WB assay only
Andresen et al, 200424 patients with heritable thrombophilia (7 ≥ 1 previous VTE) and 24 controls5 pmol/l/4 μmol/lPPPMean ETP in patients > controls (P = 0·083). Patients with AT deficiency (n = 6) >upper limit normal
Luddington & Baglin, 200420 patients with idiopathic VTE and 10 controls2 pmol/l/4 μmol/lPPP ± CTISignificantly increased thrombin generation in VTE

Arterial thrombosis

In arterial thrombosis the ETP, measured by subsampling technique in PRP, was associated with the risk of stroke (Faber et al, 2003). Whether thrombin generation measurements in arterial disease are useful in predicting outcome, as suggested for some in vivo markers of thrombin generation (Ardissino et al, 2003; Barber et al, 2004; Furie et al, 2004; Paramo et al, 2004; Elmas et al, 2006; Tanne et al, 2006; Anzej et al, 2007), requires further investigation.

Overall, thrombin generation is increased in patients with thrombophilic defects and appears to be correlated with the risk of a first thrombosis. The data on the risk of recurrence however is conflicting and further large prospective studies are required. Studies exploring the relationship between thrombin generation and recurrence could also incorporate other variables that have been associated with the risk of recurrence such as D-dimers, FVIII:C, residual thrombosis and right ventricular enlargement after pulmonary embolism. Such studies could lead to the establishment risk stratification strategies for recurrence.

Thrombin generation in other coagulopathies

The value of thrombin generation has been examined in a number of other settings including liver disease, sepsis and chemotherapy-induced thrombocytopenia. Tripodi et al (2006) showed that in patients with cirrhosis the most important factor influencing thrombin generation was the platelet count rather than the concentration of the coagulation factors. In patients with sepsis-associated coagulopathy, although activation of coagulation is delayed, thrombin generation is normal or enhanced when initiated (Collins et al, 2006). Cauwenberghs et al (2007) demonstrated that thrombin generation improved following platelet transfusion in patients with chemotherapy-induced thrombocytopenia. Although thrombin generation estimation can be used in many different settings, it is at this stage difficult to say that the information it provides is superior to that of the current widely available tests. Only large prospective studies with good clinical endpoints, such as bleeding, will be able to document the place of the new assays.


Fluorescent thrombin generation measurements are automated, easy to perform and have a good intra-laboratory variation. There are however many variations of the assay depending on the coagulation defect to be investigated and there is considerable inter-laboratory variation. These make the interpretation and comparison between studies difficult. Furthermore, it seems likely that separate tests designed for specific clinical situations may need to be developed. Thrombin generation testing in both haemorrhagic and thrombotic disorders all show a significant variation in thrombin generating potential between individuals with similar defects. In patients with bleeding tendencies or who are on anticoagulant treatment, it is hoped that these variations may be translated into individualized treatment options. A high thrombin generating potential in patients with thrombotic disorders may be related to risk of recurrence but the data to date are inconclusive. Even though the results of the studies into the use of thrombin generation suggest possible applications in routine clinical practice, its precise role will only be defined through large prospective multicentre studies with clinically relevant endpoints of bleeding or thrombosis. This however needs further standardization of the test conditions. Until this is done thrombin generation testing should remain a research tool.


This work was supported by grants from the British Heart Foundation (FS/05/001), the Sheffield Hospitals Charitable Trust (Grant no. 7838) and the British Society for Haemostasis and Thrombosis Novo Nordisk UK Research Fellowship. [Correction added after online publication 6 August 2008: The preceding Acknowledgements section was added to the article.]