Marked elevation of thrombin generation in patients with elevated FVIII:C and venous thromboembolism


Dr James O'Donnell, Department of Haematology, Hammersmith Hospital/ICSM, DuCane Road, East Acton, London W12 ONN, UK. E-mail: james.o'


Elevated plasma factor VIII coagulant activity (FVIII:C, > 150 IU/dl) is a risk factor for venous thromboembolism (VTE). We hypothesized that increased FVIII:C may exert a prothrombotic effect by increasing basal thrombin generation. To test this hypothesis we have measured prothrombin fragment 1 + 2 (F1 + 2) and thrombin–antithrombin complex (TAT) in three groups: (i) patients with objectively confirmed VTE and elevated FVIII:C; (ii) patients with VTE and no detectable thrombophilia; and (iii) healthy age- and sex-matched control subjets. In the group of patients with elevated FVIII:C, TAT and F1 + 2 levels were increased in 85% and 78% of individuals respectively. This frequency of coagulation activation is dramatically higher than that reported for other recognized constitutional thrombophilias. In the group of patients with VTE but no proven thrombophilia, increased thrombin generation was present in 30% of individuals. Basal thrombin generation was significantly higher in patients with elevated FVIII:C compared with individuals with VTE but no documented thrombophilia (median TAT = 8·65 µg/l versus 2·95 µg/l, median F1 + 2 = 1·5 nmol/l versus 0·87 nmol/l; P < 0·0001, P < 0·001). Overall FVIII:C levels were strongly correlated with levels of thrombin generation (r= 0·5, P < 0001). The clinical significance of such markedly increased F1 + 2 and TAT levels in patients with high FVIII:C levels remains unclear.

An association between high levels of plasma factor VIII coagulant activity (FVIII:C) and arterial disease was suggested many years ago and its role in coronary thrombosis was demonstrated in the Northwick Park Heart Study (Meade et al, 1994). More recently, it has emerged as a powerful and highly prevalent risk factor for venous thrombosis (Koster et al, 1995; Kraaijenhagen et al, 2000). Three independent groups have each reported that elevated plasma FVIII is found in ≈ 25% of patients with venous thrombosis (Koster et al, 1995; O'Donnell et al, 1997; Kraaijenhagen et al, 2000). Using multivariate analysis of the Leiden Thrombophilia study, an adjusted odds ratio of 4·8 was determined for first deep vein thrombosis (DVT) in those individuals with FVIII:C levels > 150 IU/dl compared with those with FVIII:C < 100 IU/dl (Koster et al, 1995). In this study and in the prospective study of post-operative DVT (Lowe et al, 1999a) the risk was found to be associated with FVIII:C but not von Willebrand factor. The increased FVIII:C levels observed in these patients is persistent and independent of any post-thrombotic acute phase reaction (O'Donnell et al, 2000). Early data suggest that unlike many other risk factors for first DVT, high levels of FVIII (> 234 IU/dl) also identify patients at high risk of recurrence (Kyrle et al, 2000). Interestingly, although the Leiden study reported a dose–response relationship between FVIII and risk of thrombosis, Kyrle et al (2000) observed a threshold FVIII:C level below which risk of recurrence was not increased. Neither study was able to determine whether the risk of thrombosis continues to rise with ever higher levels of FVIII, or whether the effect reaches a plateau.

The cause of high FVIII:C levels in these patients remains unexplained. In particular, it is unclear whether elevated FVIII:C is a constitutional risk factor, an acquired risk factor or merely a reactive consequence of venous thrombosis. Early family studies reported familial clustering of FVIII:C plasma concentrations and suggested a contribution from an X-linked locus (Kamphuisen et al, 1998). However, a search for regulatory polymorphisms in the FVIII gene promoter was unsuccessful (Mansvelt et al, 1998).

The mechanism by which elevated FVIII might increase the risk of thrombosis also remains unclear. Although it may seem obvious that elevation of a coagulation factor should increase the likelihood of thrombosis, this relationship appears to be true for only some (e.g. FVIII, FXI and prothrombin; Koster et al, 1995; Poort et al, 1996; Meijers et al, 2000) and not all factors (FV; Kamphuisen et al, 2000). However, mathematical modelling of the coagulation system does suggest that plasma FVIII levels can significantly influence the rate of thrombin generation (Jones & Mann 1994). The effect of varying FVIII:C levels on thrombin generation has also been investigated using an in-vitro coagulation system. Butenas et al (1999) demonstrated that the duration of the initiation phase of coagulation was reduced, and maximum thrombin generation increased, when FVIII:C levels were at the upper limit of the normal range. However, the effects of FVIII:C levels > 150% on thrombin generation were not reported.

We hypothesized that FVIII elevation may exert a prothrombotic effect by increasing the basal rate of thrombin generation as has been observed in a proportion of patients with other prothrombotic traits. We have therefore measured two markers of thrombin formation [plasma thrombin–antithrombin (TAT) complexes and prothrombin fragment 1 + 2 (F1 + 2) levels respectively] in patients with high FVIII:C levels who have had an objectively confirmed VTE. We expect that if the relationship between FVIII:C and thrombin generation is causal, the rate of thrombin generation should correlate with the FVIII:C level. Weak associations have previously been found in normal populations (Lowe et al, 1997). To control for the possibility that elevated thrombin generation is a feature of thrombotic patients in general, we measured the same parameters in a group of patients with venous thrombosis but who had no detectable thrombophilia trait and had normal FVIII:C. A group of age- and sex-matched normal control subjects were recruited to establish normal ranges.

Patients and methods

Patient population

Patients were identified from 249 individuals referred to the Hammersmith Hospital for investigation of objectively confirmed venous thrombosis. Seventy-two patients had suffered recurrent episodes of deep vein thrombosis (DVT) and/or pulmonary embolism (PE). For each patient, plasma samples were obtained a minimum of 3 months after their acute thrombosis and at least 2 weeks after discontinuation of oral anticoagulant therapy. Investigations performed comprised full blood count, biochemistry screen, C-reactive protein, ABO blood group, coagulation screening tests (prothrombin time, activated partial thromboplastin time, thrombin time, fibrinogen), antithrombin, proteins C and S, activated protein C resistance, factor V Leiden genotype, FVIII:C and lupus anticoagulant by dilute Russell's Viper venom time). Individuals with any thrombophilic trait other than elevation of FVIII:C were excluded as these might also increase thrombin generation.

Using the above investigations, we identified a group of 54 patients (Group A) in whom the only abnormal finding was an elevated FVIII:C level (geqslant R: gt-or-equal, slanted 150 IU/dl). All these patients were < 70 years of age with no evidence of underlying malignancy, and none of the female patients were pregnant or using oral contraceptives at the time of study. A second group (Group B) consisting of 37 patients was identified in whom we could not determine any inherited or acquired thrombophilic trait. A third group (Group C) of 37 healthy volunteers was used as the normal control group. These individuals were recruited from routine blood donors at the Wessex Regional Transfusion Centre, Southampton (UK). The donors selected were age- and sex-matched for those individuals comprising group B. Each donor provided written informed consent.

Laboratory analysis

Blood collection Blood was collected from the antecubital vein into Becton-Dickinson (Oxford, UK) Vacutainer® tubes containing 0·105 ML trisodium citrate. Citrated plasma was obtained by centrifugation at 2000 g for 20 min and then aliquotted and stored at −70°C before testing. Samples from patients with thrombosis were obtained at a routine clinic visit, and from controls at a normal donor session.

Screening for deficiency of antithrombin, protein C (functional) and protein S (free and total), activated protein C resistance and factor V Leiden were carried out using standard methods as previously described (Mansvelt et al, 1998).

FVIII:C and factor VIII antigen (FVIII:Ag) assays FVIII:C was measured by the one-stage clotting method, using FVIII-deficient substrate (Immuno, Vienna, Austria): Kaolin/platelet substitute mixture was from Diagnostic Reagents Ltd., Thames, England. Appropriate dilutions were performed to ensure the test clotting times fell within those of the reference curve. The inter assay coefficient of variation (CV) was 6% for pooled normal plasma. In group A patients (FVIII:C > 150 IU/dl), plasma FVIII:Ag levels were also assayed using a commercial enzyme-linked immunosorbent assay (ELISA) kit (Immunozym, Technoclone, Dorking, UK) according to the manufacturer's recommendations. All samples were measured against a reference plasma calibrated in IU/ml.

F1 + 2 and TAT assays Plasma F1 + 2 and TAT were also measured using commercially available ELISA kits (Enzygnost F1 + 2 and Enzygnost TAT, Behring, Marburg, Germany). F1 + 2 and TAT values were normally distributed within the normal control population. The 95th percentile of the control group was taken as the upper limit of the normal range. (F1 + 2 < 1·07 nmol/l; TAT < 3·8 µg/ml).

Statistical analysis

Because the TAT and F1 + 2 levels were not normally distributed in groups A and B, and failed to normalize by transformation, all results have been expressed as median and range unless stated otherwise, and non-parametric tests for significance performed. Statistical significance was assigned at P < 0·05.


Group A: VTE and elevated FVIII:C > 150 IU/dl

We investigated thrombin generation in 54 patients (36 women and 18 men) with objectively confirmed VTE and FVIII:C levels > 150 IU/dl. The median age of these individuals was 45 years (range 24–69 years). Elevated FVIII:Ag levels were present in 23/36 patients (64%).

TAT and F1 + 2 levels were increased relative to the normal range in 46/54 (85%) and 42/54 (75%) of individuals respectively. The median TAT and F1 + 2 levels were 8·65 µg/l and 1·5 nmol/l respectively.

Group B: VTE and no proven thrombophilia (FVIII:C < 150 IU/dl)

Thirty-seven patients (23 women and 14 men) were identified with confirmed VTE but no indication of thrombophilia. The median age of these patients was 38 years (range 17–57 years). TAT and F1 + 2 levels were increased in 10/37 (27%) and 11/37 (30%) of individuals respectively (median TAT and F1 + 2 levels were 2·95 µg/l and 0·87 nmol/l respectively).

FVIII:C and thrombin generation

As expected, there was a highly significant correlation between TAT and F1 + 2 (P < 0·0001; r = 0·76; Spearman's rank correlation).

TAT and F1 + 2 levels were both significantly higher in patients with VTE and elevated FVIII:C compared with individuals with VTE but no documented thrombophilia (P < 0·0001 and P < 0·001 respectively; Mann–Whitney) (Fig 1A and B). Also, TAT and F1 + 2 levels were significantly higher in patients with VTE but no identified thrombophilia compared with healthy age- and sex-matched controls (P = 0·002 and P = 0·015 respectively; Mann–Whitney) (Fig 1A and B).

Figure 1.

(A) Plasma prothrombin fragment 1 + 2 and (B) thrombin–antithrombin complex levels. Plasma F1 + 2 and TAT levels in patients with VTE and high FVIII:C levels (●); patients with VTE and no identified thrombophilia (▴); and age-and sex matched controls (▵). The dashed lines represent the upper limit of the normal ranges at 1·1 nmol/l (F1 + 2) and 3·8 µg/ml (TAT). The differences are significant as shown.

When data from groups A and B were pooled, there was a significant correlation between FVIII:C and thrombin generation (P < 0·001; r= 0·52; and P < 0·001; r= 0·51 for TAT and F1 + 2, respectively, Spearman's rank correlation) (Fig 2). However, when group A (high FVIII) was analysed alone, this relationship was no longer apparent (P = 0·4; r= 0·12; and P = 0·13; r= 0·21 for TAT and F1 + 2, respectively, Spearman's rank correlation).

Figure 2.

F1 + 2 against FVIII:C levels in patients with venous thrombosis. Group A: patients with VTE and high FVIII:C levels (●); Group B: patients with VTED and no identified thrombophilia (▴). Taken together there was a significant relationship between F1 + 2 and FVIII (P < 0·0001; r= 0·51). However, this correlation was lost when the two groups were analysed independently. (P = 0·13; r= 0·21 and P = 0·37; r= 0·15: Spearman's rank correlation). The dashed lines represent the upper limits of the normal range (FVIII:C < 1·5IU/ml and F1 + 2 < 1·1 nmol/l). Essentially the same results were obtained with TAT (data not shown) (P = 0·4; r= 0·12 and P = 0·7; r= 0·05: Spearman's rank correlation).


Elevated levels of F1 + 2 and TAT have been sought in many prothrombotic states to detect evidence of increased thrombin generation. These studies have had limited success. Approximately one third of individuals with heterozygous protein C or protein S deficiency have F1 + 2 levels greater than the upper limit of the normal range (Bauer et al, 1988; Mannucci et al, 1992). Similarly, in patients with the factor V Leiden mutation Martinelli et al (1996) demonstrated high levels of F1 + 2 and TAT complexes in 32% and 23% respectively. On the other hand, several studies have shown only a small increase in F1 + 2 levels in antithrombin-deficient patients (Bauer et al, 1991; Simioni et al, 1996) and a population-based study failed to find an association between F1 + 2 or TAT and factor V Leiden (Lowe et al, 1999b). Moreover, Kyrle et al (1997a) have shown clearly that elevated levels or F1 + 2 and TAT are not predictive of future thrombosis. Thus, the relationship between increased thrombin generation and the prothrombotic state is unclear.

In this study, we have shown that increased thrombin generation is present in ≈ 80% of patients with confirmed venous thrombosis and increased plasma FVIII:C levels. This is a much higher prevalence than has been found in the other prothrombotic states. Furthermore, the increase in thrombin generation is also of much greater degree than previously reported. The clinical significance of this observation remains unclear, but it may be useful in understanding the nature of the prothrombotic state and the aetiology of elevated FVIII. The mean age of the high FVIII group (A) was higher than the other two groups, and FVIII increases slowly with age but the effect is small (0·05iu/ml per decade) compared with the elevation seen in these patients. (Kamphuisen et al, 1998).

If elevated FVIII:C is responsible for the increased thrombin generation observed in these individuals, we would expect to find a correlation between these two variables. Indeed, a highly significant correlation was apparent when the full range of FVIII levels was analysed, supporting this hypothesis. This relationship between FVIII and thrombin generation was lost when groups A and B were analysed separately. The lack of correlation may be attributed to smaller numbers and restricted range of FVIII with consequent loss of power in the subgroup analysis but other explanations are possible. One possible explanation is that at high levels of FVIII, beyond a certain threshold the plasma concentration of FVIII is no longer a limiting factor in thrombin generation. As other factors then determine the rate of thrombin generation, the correlation with FVIII levels is lost.

An alternative explanation is that increased thrombin generation does not result directly from the increased levels of FVIII:C. Instead, FVIII and thrombin generation may be independent results of some other unidentified abnormality that affects them differentially. A possible link between FVIII:C, thrombin generation and thrombosis would be the acute phase reaction. Although we have reported previously that elevation of FVIII:C is not associated with elevated C reactive protein (CRP) (O'Donnell et al, 1997), we also reviewed CRP measurements in these patient groups. There was no relationship between CRP and TAT or F1 + 2 in either of the thrombotic groups A or B (data not shown).

It might be argued that because the one-stage assay for FVIII:C is susceptible to preactivation, then the elevated FVIII:C and TAT/F1 + 2 may be measurements of the same phenomenon. We therefore also measured FVIII:Ag in 36 individuals with raised FVIII:C levels. Elevation of FVIII:Ag levels was present in 23out of 36 patients. Repeating the analysis using FVIII:Ag did not significantly alter the results.

Nonetheless it is unlikely that the elevation of FVIII:C is simply an epiphenomenon. The dosage effect seen in the Leiden thrombophilia study, and the effect of ABO blood group suggest that FVIII:C has a direct causal effect on the risk of thrombosis (Koster et al, 1995). This need not necessarily be via increased thrombin generation. Although antithrombin deficiency confers a much higher relative risk for venous thrombosis than elevated FVIII:C (> 150 IU/dl), it is associated with significantly less basal thrombin generation (Simioni et al, 1996). Furthermore heterozygous factor V Leiden, which confers a relative risk for venous thrombosis similar to that of elevated FVIII:C, is associated with intermediate levels of thrombin generation (Martinelli et al, 1996; Zoller et al, 1996). These results suggest that the different thrombophilias may be exerting their thrombotic effects via different mechanisms. The models provided by Butenas et al (1999) suggest that effects on the initiation and propagation phases of coagulation may also be important mechanisms, on which FVIII is known to exert an effect.

In patients with previous venous thrombosis in whom no thrombophilic trait was identified, thrombin generation was also increased in approximately 30%. Although the frequency and intensity of thrombin generation were significantly less than that observed among patients with elevated FVIII:C levels (P < 0·001), they were significantly higher than age- and sex-matched controls (P < 0·01). These results are similar to those reported by Kyrle et al (1997a) who found that 20% of patients without a defined clotting defect had at least one F1 + 2 level above the upper limit of the normal range during a 12-month period of follow-up. The mechanism underlying coagulation activation in these patients remains unclear. Clearly a proportion of these individuals may have some undetected underlying thrombophilic tendency. Hyperhomocysteinaemia, which was not included in our routine thrombophilia screen, has been shown recently to increase F1 + 2 levels (Kyrle et al, 1997b).

The clinical significance of such markedly increased F1 + 2 and TAT levels in patients with high FVIII:C levels is not established. Although, Kyrle et al (1997a) have shown that F1 + 2 levels cannot be used to identify patients at higher risk of recurrent venous thrombosis, elevated FVIII:C is clearly associated with a very marked disturbance of coagulation homeostasis.


The authors would like to thank Dr Frank Boulton and the staff of the Wessex Regional Transfusion Centre, Southampton for their help. James O'Donnell was sponsored by a Medical Research Council Training Fellowship award.