SEARCH

SEARCH BY CITATION

Keywords:

  • recurrence;
  • thrombin generation;
  • unprovoked;
  • venous thrombosis

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Disclosure of Conflict of Interests
  8. References

Summary. Objective: To determine the predictive value of measurement of parameters of thrombin generation for unprovoked recurrent venous thrombosis. Methods: Measurements were made of thrombin generation in a prospective cohort study of 188 patients with a first episode of venous thrombosis that was unprovoked, or provoked by a non-surgical trigger. Results: The endogenous thrombin potential (ETP) was the only parameter associated with unprovoked recurrent thrombosis in a multivariate model [hazard ratio (HR) 1.3 per 100 nmol L min−1 increase, 95% confidence interval (CI) 1.0–1.6]. Patients with a high ETP had a significantly higher rate of unprovoked recurrence than those with a low ETP (HR 2.9, 95% CI  1.3–6.6, cumulative recurrence at 4 years 27% vs. 11%). Patients with an unprovoked first event had a significantly higher rate of unprovoked recurrence than those with a provoking factor (HR 2.7, 95% CI  1.2–6.1), and in these patients there was a significantly higher rate of unprovoked recurrence in association with a high ETP (HR 4.0, 95% CI  1.3–11.8). After adjustment for D-dimer, thrombophilia, sex, and whether or not the first event was unprovoked, a high ETP remained a significant predictor of recurrence (HR 2.6, 95% CI  1.2–6.0). Conclusions: This study demonstrates a high rate of unprovoked recurrent venous thrombosis in patients presenting with a first episode of venous thrombosis and a high ETP.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Disclosure of Conflict of Interests
  8. References

A test that predicts a high rate of recurrent venous thrombosis would be clinically useful. The clinical utility of such a test would be the identification of patients in whom the benefit of continued long-term anticoagulant therapy outweighed the risk. In this study, we have investigated the positive predictive value of a thrombin generation assay in a cohort of patients with a high clinical risk of recurrence after a first episode of venous thrombosis.

The risk of recurrence following a first episode of venous thrombosis, with or without clinically evident pulmonary embolus, is predicted by clinical risk factors. Patients with thrombosis associated with surgery or pregnancy have a very low rate of recurrence when the risk period is over, as do patients with clot confined to the calf veins [1]. In these patients, long-term anticoagulant therapy is not indicated. In contrast, patients with a first episode of proximal vein thrombosis that is unprovoked or provoked by a non-surgical trigger have a higher risk of recurrence, and some patients might benefit from long-term anticoagulation. Patients with a first unprovoked venous thrombosis are at highest risk of recurrence [2–5], but it is arguable whether the level of risk estimated from clinical risk factors alone justifies long-term anticoagulation [6,7]. A negative (low) D-dimer measurement following completion of an initial period of 3–6 months of anticoagulation after unprovoked venous thrombosis predicts a low risk of recurrence [8–11]. Measurement of thrombin generation was associated with a high risk of recurrence in the AUREC study [12], and we now present the predictive value of measurement of thrombin generation in the Second Cambridge Prospective Cohort Study (Cambridge II).

Materials and methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Disclosure of Conflict of Interests
  8. References

Study design

The cohort was derived from a previously reported study population of consecutive patients presenting with a first episode of venous thrombosis who were considered to be at risk of recurrence by virtue of their clinical risk profile. Low-clinical-risk patients were excluded; these were patients with postoperative venous thrombosis or pregnancy-associated thrombosis and those with clot confined to the calf veins. Patients were to be followed up for a minimum of 2 years after stopping anticoagulant treatment. Inclusion required objectively confirmed proximal lower limb deep vein thrombosis (DVT) or pulmonary embolism (PE). DVT was diagnosed by compression ultrasound. PE was diagnosed by computed tomographic pulmonary angiography or ventilation-perfusion lung scanning. Patients with both symptomatic PE and DVT were categorized as having symptomatic PE. The study was approved by the local research ethics committee, and patients were included after giving written informed consent. Recruitment commenced immediately after completion of the first Cambridge prospective cohort study (Cambridge I) [2], and there was no overlap in recruitment between the two studies.

All patients were tested for antiphospholipid antibodies by anticardiolipin assay and lupus anticoagulant activity (Dilute Russel Viper Venom and Silica Clot Time). Patients with antiphospholipid syndrome or cancer were excluded, as were other patients with an indication for prolonged anticoagulation, for example atrial fibrillation. Patients with heritable thrombophilic defects, including antithrombin, protein C and protein S deficiency, were not excluded.

Patients were treated with once-daily subcutaneous low molecular weight heparin (enoxaparin 1.5 mg kg−1 day−1). Oral anticoagulant therapy was started at the same time, and continued for 6 months to a target International Normalized Ratio of 2.5 according to the guidelines of the British Society for Haematology [13]. Patients with DVT were encouraged to wear a European grade I/II compression stocking on the symptomatic leg for up to 2 years. Patients with PE without symptomatic DVT were not advised to wear stockings.

Anticoagulant therapy was continued or restarted in all patients with elevated anticardiolipin titers or a lupus anticoagulant, and these patients were excluded from the analysis. Anticoagulant treatment was not restarted in any patient on the basis of evidence of heritable thrombophilia. All patients were advised of how to minimize or avoid acquired risks for venous thrombosis and the need for short-term thromboprophylaxis at times of high risk, for example surgery. Women using combined oral contraceptive pills at the time of thrombosis were advised not to use estrogen-containing contraceptive pills after stopping treatment with warfarin.

At the time of entry into the study, patients were classified as having unprovoked (no identifiable clinical risk factor, often termed idiopathic thrombosis) or provoked venous thrombosis. Some of these patients had absolute risk factors such as a fracture, application of a plaster cast, and use of estrogen-containing oral contraceptives, whereas others had putative risk factors such as immobilization (3 days or more), a non-specific transient illness (with immobilization for 3 days or more) or a history of travel (> 6 h of continuous air flight or road travel within a week of onset of symptoms). Provoked thrombosis did not include thrombosis within 6 weeks of surgery, which was an exclusion criterion because the risk of recurrence is low in these patients (see above).

Patients were followed up yearly, and were contacted at closure of the study. The study was closed on 30 December 2006, and the data were analyzed after closure. All clinical events were confirmed by review of the clinical records and radiologic evidence by two members of staff before samples were analyzed for thrombin generation. Only cases with new or extended clot, resulting in reintroduction of anticoagulant therapy, were recorded as recurrent thrombosis. Recurrent thrombosis was classified as unprovoked or provoked. The primary outcome of the study was unprovoked recurrent thrombosis, which was defined in the study protocol a priori. Patients were classified as having unprovoked thrombosis when there was no identifiable provoking factor.

Sample collection

Blood samples were taken between 2 and 3 months after completion of anticoagulant therapy. Venous blood was collected from the antecubital vein into Sarstedt Monovette tubes, in a 0.1 volume of 0.106 mol L−1 trisodium citrate. Platelet-poor plasma was prepared by centrifugation for 10 min at 3000 × g at room temperature, and stored at − 70 °C until being analyzed. Analysis was performed after closure of the study.

Calibrated automated thrombography (CAT)

Thrombin generation was measured with the ThrombinoscopeTM Assay (Thrombinoscope BV, Maastricht, The Netherlands) [14].

Tissue factor (TF) was mixed with phospholipids (PLs) and added to polypropylene 96-well microtiter plates (Greiner Bio-one Ltd, Stonehouse, UK). Plasma was then added, and thrombin generation was triggered and monitored by addition of premixed fluorophore and calcium solution. For the calibrator wells, the TF–PL mix was replaced with a thrombin solution (660 nmol L−1). Innovin TF was at a final concentration of 5 pmol L−1 (Dade Behring, Milton Keynes, UK). TF concentrations were measured with an Xa-based chromogenic assay (Actichrome TF; American Diagnostica, Stamford, CT, USA). PLs were at a final concentration of 4 μmol L−1 (Avanti Polar Lipids, Alabaster, AL, USA) and consisted of 20 mol% phosphatidylserine, 20 mol% phosphatidylethanolamine and 60 mol% phosphatidylcholine prepared by repeated extrusion (Avestin extruder, Mannheim, Germany). Eighty microliters of platelet-poor plasma was tested in triplicate in parallel with a calibrator. Twenty microliters of Z-Gly-Gly-Arg-AMC reagent at a final concentration of 0.417 mmol L−1 per well (Bachem AG, Bubendorf, Switzerland) suspended in buffer with 0.1 mol L−1 CaCl2, 20 mmol L−1 HEPES, 60 mg mL−1 and bovine serum albumin (pH 7.35) (all Sigma Inc., St.Louis, MO, USA) was added by the automatic dispenser. For the internal calibrator, PLs were mixed with 20 μL of thrombin calibrator solution [α2-macroglobulin bound to thrombin (660 nmol L−1); Thrombinoscope BV]. To increase sensitivity to the protein C pathway, CAT was also performed after addition of rabbit lung thrombomodulin (American Diagnostica Inc., Stamford, CT, USA) to the starting TF–PL reagent at a final concentration of 8 nmol L−1.

Fluorescence was measured with a Fluoroscan reader (Thermo Labsystems, Helsinki, Finland). The fluorescence filters used were an excitation filter at 390 nmol L−1 and an emission filter at 460 nmol L−1, as supplied by Thrombinoscope BV. The Thrombinoscope™ software (Thrombinoscope BV, Maastricht, the Netherlands) was used to convert the fluorescence signal into nmol L−1 thrombin activity, with correction for the inner filter effect and thrombin bound to α2-macroglobulin activity, as described by Hemker et al. [14].

Four parameters were derived from the thrombograms: lag-time, time to peak, peak thrombin activity, and area under the curve, referred to as the endogenous thrombin potential (ETP) [15].

This was not a case–control study, and so a normal range or upper limit of normal was not determined for each parameter of thrombin generation. Test results were interpreted as high or low on the basis of the 50th percentile (P50) for the patient cohort, a cut-off that was prespecified. For ETP and peak thrombin, a value greater than the P50 was considered to indicate relative hypercoagulability, whereas for lag-time and time to peak, a value less than the P50, that is, a short time, was considered to indicate relative hypercoagulability.

Thrombophilia testing

Determination of levels of antithrombin, protein C and protein S, genotyping for factor V Leiden and the F2G20210A mutation and detection of antiphospolipid activity as measured by anticardiolipin antibody and lupus anticoagulant were performed as previously described [2].

Statistics

spss software was used to compute descriptive statistics and for comparative analysis (Statistical Package for the Social Sciences; SPSS Ltd, Woking, UK). The study was powered on an assumption of the predictive value of a measurement of hypercoagulability as previously reported [16].

The test performance characteristics were examined by analysis of receiver operator curves (ROCs). Cox proportional hazards modeling was used to examine the association between thrombin generation and time to recurrence, with Kaplan–Meier estimates of cumulative recurrence rates reported at 5 years. Averaged recurrence rates were calculated as events per 100 patient-years calculated from the total follow-up of patients with calculation of 95% confidence intervals (CIs).

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Disclosure of Conflict of Interests
  8. References

Patient cohort

Recruitment commenced in January 2001 and was completed in December 2003. After exclusion of patients who died during initial anticoagulant therapy (eight), had a subsequent diagnosis of cancer (27) or antiphospholipid syndrome (19), or continued anticoagulant therapy (11), 272 patients fulfilled clinical entry criteria. The predictive value of D-dimer measurement in this cohort has been reported [16]. In 83 of these patients, there was insufficient plasma for measurement of thrombin generation. In one patient, a valid ETP result could not be obtained.

Median age at presentation of the 188 remaining patients was 66 years (range 20–100) and there were 91 males (48%) and 97 females (52%). The distribution of the initial thrombotic events was symptomatic PE in 81 (43%) and proximal leg DVT in 107 (57%). There were 99 (53%) unprovoked events and 89 (47%) events associated with a provoking factor.

The median duration of anticoagulant therapy was 6.1 months (range 3.0–8.5) and was the same in patients with unprovoked and provoked thrombosis (mean difference − 0.1 months, 95% CI  −0.4 to 0.3). Median follow-up after stopping anticoagulant therapy was 42 months (range 3–71 months). Twenty-one of the 188 patients died during follow-up. Three had a post-mortem, and PE was not considered to be the cause of death of any patient. In the remainder, PE was not quoted as a factor on the death certificate.

Unprovoked recurrent venous thrombosis in relation to thrombin generation

There were 29 episodes of unprovoked recurrent thrombosis during follow-up of the 188 patients with a cumulative rate of unprovoked recurrent thrombosis at 4 years of 19% (4.9/100 patient-years, 95% CI  3.3–6.9).

For the standard assay (without the addition of thrombomodulin), the ROCs for unprovoked recurrent thrombosis in relation to each parameter of thrombin generation for the complete patient cohort (n = 188) were plotted. The area under the curve was significantly different to 0.5 (null hypothesis) for ETP (area = 0.66) and peak thrombin (area = 0.64). In a Cox proportional hazards multivariate model, the ETP was the only parameter of thrombin generation associated with unprovoked recurrent thrombosis [hazard ratio (HR) 1.25 per 100 nmol L−1 min increase, 95% CI  1.01–1.55]. The interassay and intra-assay coefficients of variation for this parameter were 9.7% and 4.7%.

When the thrombin-generating capacity was determined in the presence of added thrombomodulin (modified assay), the area under the ROCs was not significantly different to 0.5 (null hypothesis) for any parameter of thrombin generation. In a Cox proportional hazards multivariate model, no parameter was associated with unprovoked recurrent thrombosis.

Recurrence rates and number of events occurring in different categories of patients and in relation to the prespecified ETP P50 threshold are shown in Table 1. Kaplan–Meier analysis was performed with reference to the prespecified P50 threshold for ETP. Patients with a high ETP had a significantly higher rate of unprovoked recurrence than those with a low ETP (HR 2.9, 95% CI  1.3–6.6, cumulative recurrence at 4 years 27% vs. 11%, Fig. 1). Patients with an unprovoked first event had a significantly higher rate of unprovoked recurrence than those with a provoking factor (HR 2.7, 95% CI  1.2–6.1). In patients with a first unprovoked event, there was a significantly higher rate of unprovoked recurrence in patients with a high ETP (HR 4.0, 95% CI  1.3–11.8).

Table 1.   Rates of recurrence (per 100 patient-years) in relation to endogenous thrombin potential (ETP) (high > P50, low ≤ P50) and whether the first episode of venous thrombosis was unprovoked or not
VariableRecurrent VTE
First episode of venous thrombosisETPRate (95% CI)Events/patientsPatient % of cohort% of recurrent events
  1. VTE, venous thromboembolism.

  2. The rate of recurrence in the total cohort was 5.2/100 patient-years [95% confidence interval (CI) 3.5–7.4]. The numbers of events in each group are indicated, followed by each patient group as a percentage of the total cohort. In addition, unprovoked recurrences as a percentage of the total recurrent events are presented; the total number of patients was 188, and the number of events was 29. For example, 54 patients had an unprovoked first event and a high ETP, and 17 had a recurrence. These 54 patients constituted 29% of the total cohort, and 59% of all the recurrences occurred in this group.

5.2 (3.5–7.4)29/188100100
Unprovoked7.6 (4.8–11.4)21/995372
Provoked2.8 (1.2–5.5)8/894728
High8.1 (5.1–12.0)21/945072
Low2.7 (1.2–5.2)8/945028
UnprovokedHigh12.1 (7.3–18.6)17/542959
UnprovokedLow2.9 (0.8–7.4)4/452414
ProvokedHigh3.3 (0.9–8.2)4/402114
ProvokedLow2.5 (0.7–6.2)4/492614
image

Figure 1.  Cumulative proportions of recurrent thrombosis after completion of anticoagulant therapy after a first thrombosis for patients with a high endogenous thrombin potential (ETP) [bold line (ETP > P50 (1489 nmol L−1 min)] and for patients with a low ETP [thin line (ETP ≤ P50)].

Download figure to PowerPoint

Relationship between thrombin generation and D-dimer levels

When adjusted for D-dimer, sex and whether or not the first event was unprovoked, a high ETP remained a significant predictor of recurrence (HR 2.6, 95% CI  1.2–6.0). In this multivariate model, D-dimer was not significantly associated with risk of recurrence (HR 1.3, 95% CI  0.6–3.0) when adjusted for ETP, sex and type of first event.

When ETP and D-dimer were compared as logistic variables (high ETP > 1489 nmol L−1 min and positive D-dimer > 500 ng mL−1), there was no correlation (χ2 = 0.6, P = 0.4). Thirty percent of patients with a high ETP had a negative D-dimer and 65% of patients with a low ETP had a positive D-dimer.

Relationship between thrombin generation and thrombophilia

Heritable thrombophilic defects were present in 35/188 (19%): 28 FV Leiden, seven F2G20210A mutation, one protein C deficiency and two protein S deficiency. The rate of recurrence was the same in patients with and without thrombophilia (log rank = 1.1, P = 0.3). The ETPs in patients with and without thrombophilia were not statistically different (mean difference without thrombomodulin 122 nmol L−1 min−1, 95% CI  −23 to 268, P = 0.1, and mean difference with thrombomodulin 148, 95% CI  −27 to 323, P = 0.1). When thrombophilia was added to the multivariate model, the predictive value of ETP was unchanged (HR 2.6, 95% CI  1.1–6.0) and the adjusted HR for thrombophilia was 1.1 (95% CI  0.5–2.5).

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Disclosure of Conflict of Interests
  8. References

There is a high rate of unprovoked recurrence after a first episode of venous thrombosis in patients with an unprovoked first event and those with minor provoking factors. All patients who suffer an episode of venous thrombosis should receive thromboprophylaxis at times of high risk. Therefore, the purpose of continuing long-term anticoagulation is to prevent unpredictable recurrence, which, by definition, is usually unprovoked. The clinical utility of a test with positive predictive value would be in relation to its ability to predict unprovoked recurrence. With this in mind, the primary outcome of this study was unprovoked recurrence, rather than all recurrences, as the clinical relevance of the study was in relation to identifying patients who might be considered for long-term anticoagulation.

The main finding was a higher rate of unprovoked recurrence in patients with an ETP above the 50th percentile (P50) of the patient group. The rate was highest in patients presenting with a first episode of unprovoked venous thrombosis and a high ETP. In these patients, the recurrence rate was in excess of 10 per 100 patient-years, with a cumulative recurrence rate at 4 years of almost 30%. On the assumption that 95% of events can be prevented by continued anticoagulant therapy, approximately 70% of unprovoked recurrent events might have been prevented in this high-risk group of patients by continued anticoagulant therapy.

The methodology of a thrombin generation assay will determine its clinical utility [17–19]. For this study, we chose a TF preparation with relatively high sensitivity for detection of hypercoagulability combined with insensitivity to contact factor activation [18]. When a low concentration of TF is used as the activator, the measured thrombin generation is subject to contact factor activation. This can be abolished by taking blood into tubes containing corn trypsin inhibitor (CTI) [20]. The samples available from this study were citrated without CTI, and so we chose to use Innovin as the source of TF, as we have previously found that ETPs measured with Innovin-derived TF at 5 pmol are relatively unaffected by contact factor activation. The THE-VTE study (Thrombophilia-Hypercoagulability-Environmental Risk in Venous-Thrombo-Embolism) has recruited more than 600 consecutive patients with samples taken with and without CTI, and will report on the need and value of taking blood samples into CTI in 2–3 years when there are sufficient outcome data.

For the modified assay, thrombomodulin was added to increase sensitivity to the protein C pathway [21,22]. We previously reported that measurement of thrombin generation in the patient cohort of the Leiden Thrombophilia Study (LETS) did not predict rate of recurrence [23]. However, the difference between the Leiden and Cambridge studies is readily explained. CAT was used in both studies to measure thrombin generation, but the assays were radically different. The sample volumes available from the Leiden patients were insufficient to measure thrombin generation both with and without added thrombomodulin, and so we elected to measure thrombin generation in the presence of thrombomodulin. This decision was based on the assumption that added thrombomodulin would increase the sensitivity of the assay for defects of the protein C pathway, including the FV Leiden mutation. In the Cambridge study, we were able to measure thrombin generation both with and without added thrombomodulin. In the Cambridge study, the only parameter that identified a group of patients at high risk of recurrence was the ETP, and then only in the absence of thrombomodulin. The ETP in the presence of thrombomodulin in the Cambridge cohort had no predictive value, just as it did not in the Leiden study. This may be because defects such as the FV Leiden mutation are not associated with a high risk of recurrence, and so the addition of thrombomodulin reduces the positive predictive value of the assay for recurrent venous thrombosis due to the dominant effect of the FV Leiden mutation on the assay result. There were additional methodological differences between the Leiden and Cambridge studies. The Leiden samples were of insufficient volume to perform CAT on neat plasma, and so samples were diluted one in four with buffer before analysis; this is likely to alter the dynamics of thrombin generation. The source and concentration of TF used in the two studies were different, as it is now apparent that these variables make a material contribution to the resulting thrombin-time integral and the influence of preanalytical contact factor activation [18].

Finally, for logistic reasons, a high ETP was defined in the Leiden study as a level above the 90th percentile of the control group and in the Cambridge study as a level above the 50th percentile of the patient cohort. Standardization of CAT is problematic [18]. The use of the P50 derived from the patient group has the advantage of internal calibration and avoids the problem of a defined ‘normal’ threshold. It will be informative for other centers to determine the performance characteristics of CAT and other thrombin generation assays using a P50 threshold derived from their own patient populations as well as other predefined thresholds, including an upper limit of normal.

An association between thrombin generation and recurrent venous thrombosis was previously reported in the Austrian Study of Recurrent Venous Thromboembolism (AUREC) [12]. In the Austrian study, thrombin generation was also measured without the addition of thrombomodulin. Using a different thrombin generation assay (Technothrombin TGA, Technoclone), there was a higher rate of recurrence in patients with a peak thrombin level above 400 nmol L−1 than in those with a lower level (probability of recurrence at 4 years 20% vs. 6.5%). The threshold for a high level (> 400 nmol L−1) was not prespecified, but when peak thrombin was analyzed as a continuous variable, there was a significant increase in the relative risk of recurrence for every 10 nmol L−1 increase in peak. The ETPs were not reported.

Clinicians will want to know the predictive value of a high ETP in patients with unprovoked venous thrombosis and how an ETP result relates to a D-dimer result. This study indicates that a high ETP, measured in the absence of thrombomodulin, identifies those patients presenting with an unprovoked thrombosis who have an annual risk of recurrence in excess of 10 per 100 patient-years. ETP and D-dimer results do not correlate, suggesting that the two assays are measuring different aspects of coagulability. This study was not large enough to determine the combined predictive value of ETP and D-dimer measurement or the predictive value in subgroups of patients such as males. However, the study indicates that measurement of ETP has a positive predictive value independent of D-dimer, and that it could be used in conjunction with an assessment of clinical risk factors and measurement of D-dimer to identify patients who would benefit from prolonged anticoagulation after a first episode of venous thrombosis.

Disclosure of Conflict of Interests

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Disclosure of Conflict of Interests
  8. References

M. Besser's research was supported by the Baxter Royal College of Pathologists Training Fellowship. The other authors state that they have no conflict of interest.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Disclosure of Conflict of Interests
  8. References