Peak thrombin generation and subsequent venous thromboembolism: the Longitudinal Investigation of Thromboembolism Etiology (LITE) study

Authors


Pamela L. Lutsey, Division of Epidemiology and Community Health, University of Minnesota, 1300 South 2nd Street, Suite 300, Minneapolis, MN 55454, USA.
Tel.: +1 651 270 1514; fax: +1 612 624 0315.
E-mail: lutsey@umn.edu

Abstract

Summary. Background: Thrombin is an enzyme that is essential for the acceleration of the coagulation cascade and the conversion of fibrinogen to clottable fibrin. Objectives: We evaluated the relationship of basal peak thrombin generation with the risk of future venous thromboembolism (VTE), and determined whether associations were independent of other coagulation markers. Methods: The Longitudinal Investigation of Thromboembolism Etiology (LITE) study investigated VTE in two prospective population-based cohorts: the Atherosclerosis Risk in Communities (ARIC) study and the Cardiovascular Health Study (CHS). Peak thrombin generation was measured on stored plasma in a nested case–control sample (434 cases and 1004 controls). Logistic regression was used to estimate the relationship of peak thrombin generation with VTE, adjusted for age, sex, race, center, and body mass index. Mediation was evaluated by additionally adjusting for factor VIII and D-dimer. Results: Relative to the first quartile of peak thrombin generation, the odds ratio (OR) of VTE for those above the median was 1.74 [95% confidence interval (CI) 1.28–2.37]. The association was modestly attenuated by adjustment for FVIII and D-dimer (OR 1.47, 95% CI 1.05–2.05). Associations appeared to be stronger for idiopathic than for secondary VTE. Elevated peak thrombin generation more than added to the VTE risk associated with FV Leiden or low activated partial thromboplastin time. Conclusions: In this prospective study of two independent cohorts, elevated basal peak thrombin generation was associated with subsequent risk of VTE, independently of established VTE risk factors.

Background

Deep vein thrombosis (DVT) and pulmonary embolism (PE), collectively referred to as venous thromboembolism (VTE), are major sources of morbidity and mortality in the elderly [1]. Hypercoagulability is an established risk factor for VTE [2]. Thrombin is essential for the acceleration of the coagulation cascade, because it serves as an activator for platelets, factor V, and FVIII, and is a critical component of a positive-feedback loop that results in the generation of large amounts of additional thrombin, the conversion of fibrinogen to fibrin, and, ultimately, clot formation [3].

Although the measurement of thrombin generation has been possible since 1953 [4,5], only recently have assays been developed with which thrombin generation can be efficiently measured [6]. These functional assays activate the coagulation cascade using tissue factor and phospholipids, and then monitor the concentration of thrombin generated over time. Using information from the thrombogram curve (thrombin generation plotted against time) produced by these assays, thrombin generation can be expressed in multiple ways [7]. Most pertinent to this article is peak thrombin generation, which is the maximal concentration of thrombin formed at a given point in time. Endogenous thrombin potential (ETP), which is the area under the thrombin generation curve, is the other measure most commonly reported in the literature.

There is currently great interest in whether measurement of basal thrombin generation predicts VTE occurrence or recurrence. Of the five prospective studies (two using Austrian Study on Recurrent VTE data, with different thrombin generation assays) that have assessed the relationship of thrombin generation with recurrent VTE, four have found greater peak thrombin generation [8,9] and/or ETP [9–11] to be predictive of VTE recurrence. The Leiden Thrombophilia Study (LETS), however, observed no association between peak thrombin generation or ETP and risk of recurrent DVT [12].

Using its case–control data, the LETS also assessed whether measures of thrombin generation were predictive of first DVT [12]. Among cases, thrombin generation values were assessed at least 3 months after discontinuation of anticoagulant therapy for a first DVT. Individuals with an ETP above the 90th percentile measured in control subjects had an increased relative odds of DVT [odds ratio (OR) 1.5, 95% confidence interval (CI) 0.9–2.3)], which was stronger when the analyses were restricted to idiopathic DVT (OR 1.7, 95% CI 1.0–2.8). However, a dose–response relationship was not demonstrated when ETP was modeled as quartiles, and no associations were observed between peak thrombin generation and risk of DVT.

Although not entirely consistent, the evidence suggesting that measures of thrombin generation may be related to VTE is provocative. To date, however, the relationship of peak thrombin generation with incident VTE, using plasma collected prior to VTE, is unknown. Thus, using data from the Longitudinal Investigation of Thromboembolism Etiology (LITE) study, we explored the association of basal peak thrombin generation with risk of future VTE. To build on previous studies, we further assessed whether this relationship was independent of other hemostatic factor levels. We hypothesized that basal peak thrombin generation would be associated positively with VTE risk, but that this association would be largely explained by the levels of other hemostasis markers that contribute to higher thrombin generation.

Methods

LITE study design

The LITE study is a prospective study of VTE occurrence in two population-based cohorts: the Atherosclerosis Risk in Communities (ARIC) study [13] and the Cardiovascular Health Study (CHS) [14,15]. Extensive cardiovascular disease risk factor information was collected in both studies, generally in a similar fashion, using standard measures. Local institutional review boards approved the study protocols, and all participants gave informed consent.

The ARIC study is a multicenter population-based prospective cohort study designed to investigate the etiology and natural history of atherosclerosis in middle-aged adults [13]. Participants were recruited from four US communities: Forsyth County, NC; Jackson, MS; suburbs of Minneapolis, MN; and Washington County, MD. The study cohort included 15 972 white and black men and women aged 45–64 years at baseline, in 1987–1989.

The CHS is a population-based longitudinal study of coronary heart disease and stroke in adults aged 65 years and older [14,15]. A total of 5201 men and women sampled from Medicare eligibility lists were recruited in 1989–1990 from four US communities: Forysth County, NC; Sacramento County, CA; Washington County, MD; and Pittsburgh, PA. An additional 687 African-American participants were recruited from 1992 to 1993.

VTE identification and classification

ARIC participants were contacted annually by telephone or through clinic visits, and CHS participants were contacted twice a year, by alternating clinic visits and telephone calls. Hospitalizations were identified by participants or proxy reports in both the ARIC study and the CHS, by surveillance of local hospital discharge lists in the ARIC study, and by a search of Medicare records in the CHS. For each hospitalization identified, all ICD-9-CM discharge codes were recorded.

All records with ICD-9-CM codes indicating possible VTE (415.1x, 451, 451.1x, 451.2, 451.8x, 451.9, 453.0, 453.1, 453.2, 453.8, 453.9, 996.7x, 997.2, and 999.2, and procedure code 38.7) were identified, copied, and independently reviewed by two physicians (ARF and MC) [16]. DVT was defined on the basis of duplex ultrasound or venogram or, in rare cases, by impedance plethysmography, computed tomography, or autopsy. Definite PE required ventilation/perfusion scanning showing multiple segmental or subsegmental mismatched perfusion defects, or a positive pulmonary angiogram, computed tomography or autopsy finding [16]. VTE events were further classified as idiopathic or secondary (occurring within 90 days of major trauma, surgery, hospitalization, or marked immobility, or associated with active cancer or chemotherapy).

LITE nested case–control sample

A nested case–control design was employed to study associations between some blood parameters and VTE incidence. All VTE cases were included in the nested study. Controls were randomly selected at a ratio of two per case. They were frequency matched to the cases by age (5-year groupings), gender, race (African-American or Caucasian), follow-up time (case event date within 2 years of control assigned date), and study (ARIC or CHS).

Hemostatic factor assessment

The same phlebotomy and sample-processing methods were applied across all the field centers for the CHS and ARIC study, with centralized training of technicians and quality control monitoring [17–19]. Blood samples were drawn into Vacutainer tubes with minimal stasis using atraumatic venipuncure, a tourniquet time < 2 min, and a 21-gauge butterfly needle connected to a Vacutainer outlet via a Luer adaptor. In both cohorts, 32 g L−1 citrate samples were spun at 3000 × g for 10 min at 4 °C to remove platelets [17,18]. Therefore, samples are appropriate for coagulation factor testing. We have previously shown good results for other assays that are very sensitive to sample handling/preanalytical issues, such as prothrombin fragment 1.2, and especially fibrinopeptide A in the CHS [20]. Serum, citrate and EDTA plasma and DNA were collected and stored at − 70 °C.

Some hemostatic factors were assessed in each cohort at baseline soon after blood drawing, using published methods: fibrinogen, FVII, FVIII, von Willebrand factor (VWF; ARIC only), and activated partial thromboplastin times (APTTs; ARIC only) [17,18]. Other factors were measured previously in the LITE nested case–control sample on stored blood: FIX, FX, FXI, D-dimer, and FV Leiden [21], prothrombin 20210A [22], and ABO blood type by genotyping [23]. FIX, FX and FXI were measured using enzyme immunoassay kits from Enzyme Research Laboratories (South Bend, IN, USA). D-dimer was measured with the STAR automated coagulation analyzer (Diagnostica Stago, Parsippany, NJ, USA), using an immunoturbidometric assay (Liatest D-DI; Diagnostica Stago).

Peak thrombin generation, for both the CHS and the ARIC study, was measured in platelet-poor citrate plasma, usually from the baseline examination but always before VTE, using the Technothrombin TGA kit (Technoclone, Vienna, Austria) at the Laboratory for Clinical Biochemistry Research (University of Vermont, Burlington, VT, USA). The reaction was triggered with the TGA RC Low reagent, which contained a low concentration of phospholipid micelles containing 71.6 pm recombinant human tissue factor (rhTF) in Tris–Hepes–NaCl buffer. The median within-sample coefficient of variation in the ARIC study was 10.9%, and in the CHS it was 3.6%. Samples had been previously thawed and refrozen, but there was no difference in the number of freeze–thaw cycles between cases and controls. We evaluated freeze–thaw effects in nine analyzed samples from local volunteers. After a second cycle of freeze–thaw, the mean thrombin generated decreased from 397 to 345 nm, but it did not change further after more cycles (third and fourth cycles mean 369 and 334 nm, respectively). Other parameters of thrombin generation were not available.

Statistical analysis

Participants were excluded from analysis if they had no plasma available, had unusable peak thrombin generation data, or were using warfarin at baseline. A flow chart of study exclusions is presented in Fig. 1. In the ARIC study, cases had a greater proportion of missing or unusable thrombin generation data than controls. For both the ARIC study and the CHS, subject characteristics were compared in participants missing peak thrombin generation data or with undetectable peak thrombin generation levels, and those with valid peak thrombin generation data, stratified by case–control status. No consistent or remarkable differences existed between those included and excluded from this study.

Figure 1.

 Flow chart of Longitudinal Investigation of Thromboembolism Etiology (LITE) participant exclusions by cohort and case–control status. *Invalid due to plasma clots (n cases = 2; n controls = 9) or thrombin generation undetectable (≤ 0.5 nm) (n cases = 20; n controls = 25).

Owing to differences in the distribution of peak thrombin generation levels between the ARIC study and the CHS, study-specific quartiles and standard deviations (SDs) of basal peak thrombin generation were derived according to the peak thrombin generation distribution among controls.

Frequencies and means of variables of interest were compared between VTE cases and controls, using chi-squared tests or Student’s t-tests. In controls, general linear regression was used to assess the relationships of peak thrombin generation levels with demographics, physiologic characteristics, genetic risk factors, and hemostatic factor levels. Associations between hemostatic factor levels and peak thrombin generation were age-adjusted. Linear trends were assessed by numbering the peak thrombin generation quartiles (1–4) and entering this variable into the models as a continuous term.

For the primary analysis, ORs and 95% CIs for the relationship between peak thrombin generation and VTE were estimated using unconditional logistic regression. Model 1 was adjusted for age (continuous), sex, race, center, and body mass index (BMI) (continuous). ORs were also estimated for idiopathic and secondary VTE. Peak thrombin generation was represented both as quartiles and per SD increment. In further models, we evaluated whether associations between thrombin generation and VTE were independent of other hemostatic factors by adjusting for these factors and determining whether adjustment altered the β-coefficient for peak thrombin generation by 10% or more.

Interactions of peak thrombin generation with other VTE risk factors were also explored. Relative excess risks due to interactions (RERIs) were used to assess additive interactions [24,25], and cross-product terms were used to evaluate multiplicative interactions. All authors had access to the data; analyses were conducted by PLL.

Results

In this nested case–control sample, the mean baseline age of ARIC participants was 55 years, whereas in the CHS the mean baseline age was 73 years. In the ARIC study, 43% of participants were men, and 31% were black. These proportions in the CHS were 44% and 21%, respectively. Included in this analysis were 255 VTE cases and 624 controls in the ARIC study during 16 years of follow-up (median = 13.2 years), and 179 cases and 380 controls in the CHS during 14 years (median = 11.8 years). In both studies, 42% of cases were classified as idiopathic.

Unadjusted baseline characteristics of ARIC and CHS participants stratified by VTE case–control status are reported in Table 1. In the ARIC study, as compared with controls, those who developed VTE had a higher mean BMI, a greater prevalence of the FV Leiden mutation and non-O blood type, lower APTT, and higher levels of FVIII, FIX, FXI, D-dimer, VWF, and peak thrombin generation. In the CHS, cases also had a higher prevalence of the prothrombin mutation than controls, but did not have higher levels of FIX. APTT and VWF were not assessed in the CHS.

Table 1.   Baseline characteristics of Longitudinal Investigation of Thromboembolism Etiology (LITE) participants stratified by cohort [Atherosclerosis Risk in Communities (ARIC) study or Cardiovascular Health Study (CHS)] and venous thromboembolism (VTE) case–control status
 ARICCHS
ControlsCasesControlsCases
N624255380179
  1. APTT, activated partial thromboplastin time; BMI, body mass index; CHD, coronary heart disease; SD, standard deviation; VWF, von Willebrand factor. *Geometric mean.

Demographics (matching factors)
 Age [years (SD)]55.9 (5.6)55.9 (5.5)73.1 (5.6)73.0 (5.5)
 African-American (%)31.331.819.523.5
 Male gender (%)42.842.843.445.3
Physiologic characteristics
 BMI (SD)28.0 (5.2)29.7 (5.8)26.5 (4.5)28.3 (5.3)
 Prevalence diabetes (%)10.214.913.418.4
 Prevalence of CHD (%)5.27.213.710.6
VTE genetic risk factors
 FV Leiden (%)2.910.93.47.9
 Prothombin 20210A (%)2.02.11.76.1
 Non-O blood type (%)53.867.052.364.6
Hemostatic factors
 FVII, [% (SD)]118 (27)122 (30)123 (29)124 (30)
 FVIII [% (SD)]131 (39)149 (48)122 (37)130 (43)
 FIX [% (SD)]140 (30)146 (31)150 (31)154 (30)
 FX [% (SD)]139 (29)141 (35)138 (30)139 (33)
 FXI [% (SD)]135 (34)141 (37)126 (32)133 (33)
 Fibrinogen [mg dL−1 (SD)]305 (61)308 (66)325 (66)313 (65)
 D-dimer* (μg mL−1)0.350.420.520.66
 VWF [% (SD)]118 (47)140 (57)
 APTT [s (SD)]29.1 (2.8)28.0 (3.3)
Peak thrombin generation
 Mean [nm (SD)]107 (91)123 (97)245 (121)272 (149)
 Median (nm)8098232253

Table 2 presents characteristics of ARIC and CHS controls according to study-specific peak thrombin generation quartiles. Overall, there was no relationship of peak thrombin generation levels with demographics, physiologic characteristics, or VTE genetic risk factors. Following age adjustment, in the ARIC study there were associations between peak thrombin generation and FVIII, VWF, and APTT. The Spearman’s age-adjusted partial correlations between peak thrombin generation and these hemostatic factors were r = 0.17 (P < 0.0001), r = 0.13 (P = 0.0002), and r = − 0.22 (P < 0.0001), respectively. In the CHS, after age adjustment, peak thrombin generation was positively related to FVII, FVIII, FIX, and FX. The age-adjusted Spearman’s correlations between peak thrombin generation and these factors were as follows: FVII, r = 0.14 (P = 0.002); FVIII, r = 0.18 (P = 0.0001); FIX, r = 0.15 (P = 0.0009); and FX, r = 0.12 (P = 0.01).

Table 2.   Characteristics (mean or %) of Longitudinal Investigation of Thromboembolism Etiology (LITE) control participants according to quartiles* of peak thrombin generation and cohort
 ARICCHS
  1. APTT, activated partial thromboplastin time; ARIC, Atherosclerosis Risk in Communities; BMI, body mass index; CHD, coronary heart disease; CHS, Cardiovascular Health Study; VTE, venous thromboembolism; VWF, von Willebrand factor. *Study-specific, based on the distribution among controls. P-value for chi-squared test, other P-values by anova. Adjusted for age. §Geometric mean.

Peak thrombin generation
 Quartiles1234Ptrend1234Ptrend
 N15615615615695959595
 Median (nm)2659107205110200267378
 Range (nm)(1–45)(46–80)(81–143)(144–665) (4–155)(156–232)(233–316)(317–693)
Demographics
 Age (years)55.655.556.056.40.1873.473.672.673.10.47
 African-American (%)27.634.024.439.10.0216.719.022.120.00.83
 Male gender (%)45.545.545.534.60.1354.743.239.036.80.06
Physiologic characteristics
 BMI (kg m−2)27.628.127.828.30.3326.825.926.526.90.68
 Prevalence of diabetes (%)12.411.28.98.20.1717.94.213.717.90.54
 Prevalence of CHD (%)3.34.16.56.90.1014.614.58.716.90.95
VTE genetic risk factors
 FV Leiden (%)2.03.43.42.80.884.44.62.32.20.72
 Prothrombin 20210A (%)4.051.41.41.40.260.01.23.52.20.31
 Non-O blood type (%)51.454.255.554.10.9150.650.651.852.30.99
Hemostatic factors
 FVII (%)1191161171210.491181201201320.003
 FVIII (%)1251271341390.00041141211211320.004
 FIX (%)1381431371430.371501421511580.02
 FX (%)1381381341450.081351361371460.01
 Factor XI (%)1341331311410.091281211281300.38
 Fibrinogen (mg dL−1)3132993053040.363293163143390.33
 D-dimer§ (μg mL−1)0.520.460.470.480.490.640.710.670.680.80
 VWF (%)1151141191260.02
 APTT (s)29.629.528.828.5< 0.0001

In the ARIC study, the CHS and the LITE study as a whole, peak thrombin generation levels were positively associated with VTE risk (Table 3). In the LITE study, after adjustment for demographics and BMI, there was a positive linear trend (P = 0.0003). Those with thrombin generation values above the median were at at 74% greater risk of VTE than those in the lowest quartile. The odds of VTE increased by 20% (95% CI 7–34%) per one SD increment of peak thrombin generation level. FVIII and D-dimer were identified statistically as potential ‘mediators’ of the LITE thrombin generation–VTE association; after adjustment for these factors, the association of peak thrombin generation with VTE was modestly attenuated [OR per one SD: 1.16 (95% CI 1.03–1.32)].

Table 3.   Odds ratios (ORs) with 95% confidence intervals (CIs) of venous thromboembolism (VTE) according to peak thrombin generation quartile and per one standard deviation (SD) increment in peak thrombin generation*: the Longitudinal Investigation of Thromboembolism Etiology (LITE) study
 Peak thrombin quartilesPer one SD
Q1Q2Q3Q4Ptrend
  1. *Adjusted for age, sex, race, center, and body mass index. Quartile and SD based on the study-specific control distribution.

LITE
 Total VTE
  Cases (n)7995126134  
  Controls (n)251251251251  
  OR (95% CI)1.001.24 (0.87–1.76)1.70 (1.21–2.40)1.78 (1.26–2.51)0.00031.20 (1.07–1.34)
 Idiopathic VTE
  Cases (n)29385858  
  OR (95% CI)1.001.33 (0.79–2.24)2.10 (1.28–3.42)2.05 (1.25–3.37)0.0011.24 (1.07–1.44)
 Secondary VTE
  Cases (n)50576876  
  OR (95% CI)1.001.19 (0.78–1.82)1.47 (0.97–2.23)1.64 (1.08–2.49)0.011.17 (1.02–1.33)
ARIC
 Total VTE
  Cases (n)44498280  
  Controls (n)156156156156 SD = 91
  OR (95% CI)1.001.09 (0.68–1.75)1.82 (1.17–2.81)1.77 (1.14–2.75)0.0021.17 (1.01–1.35)
 Idiopathic VTE
  Cases (n)17163342  
  OR (95% CI)1.000.92 (0.45–1.90)1.88 (1.00–3.53)2.33 (1.26–4.31)0.00071.28 (1.07–1.53)
 Secondary VTE
  Cases (n)27334938  
  OR (95% CI)1.001.18 (0.67–2.08)1.79 (1.05–3.03)1.40 (0.80–2.44)0.111.06 (0.88–1.28)
CHS
 Total VTE
  Cases (n)35464454  
  Controls (n)95959595 SD = 121
  OR (95% CI)1.001.48 (0.86–2.55)1.52 (0.87–2.65)1.81 (1.04–3.15)0.051.26 (1.05–1.50)
 Idiopathic VTE
  Cases (n)12222516  
  OR (95% CI)1.001.99 (0.91–4.34)2.42 (1.11–5.28)1.52 (0.65–3.54)0.321.18 (0.91–1.52)
 Secondary VTE
  Cases (n)23241938  
  OR (95% CI)1.001.23 (0.63–2.40)1.00 (0.49–2.03)2.06 (1.08–3.93)0.041.33 (1.08–1.63)

In the ARIC study, after demographic and BMI adjustments, participants with peak thrombin generation values above the median were at nearly 80% greater risk of VTE than those in the lowest quartile. Per one SD increment of peak thrombin generation, the odds of VTE increased by 17% (95% CI 1–35%). FVIII and APTT were statistical ‘mediators’ of the ARIC thrombin generation–VTE association; after adjustment for these factors, the association of peak thrombin generation with VTE was modestly attenuated [OR per one SD: 1.12 (95% CI 0.96–1.31)].

Likewise, in the CHS, the OR for extreme quartiles of peak thrombin generation, adjusted for demographics and BMI, was 1.81 (95% CI 1.04–3.15), whereas the odds of VTE increased by 26% (95% CI 5–50%) per one SD increment in peak thrombin generation. FX and D-dimer were statistical ‘mediators’ of the association; adjustment for these factors attenuated the peak thrombin generation–VTE relationship in the CHS to a larger extent than in the ARIC study [OR per one SD: 1.14 (95% CI 0.93–1.39)].

The median peak thrombin generation level of the CHS Pittsburgh center (181 nm) was lower than that of the other three CHS centers (258 nm). In sensitivity analyses in which we excluded Pittsburgh data, the results were somewhat stronger. ORs of VTE per one SD increment of peak thrombin generation were 1.32 (95% CI 1.09–1.61) following adjustment for demographics and BMI, and 1.21 (95% CI 0.96–1.54) after additional adjustment for FX and D-dimer. Center differences were also present in the ARIC study, but were less pronounced. Stratifying ARIC analyses by centers with high and low means yielded similar results (data not shown).

In the LITE study as a whole, and in both individual cohorts, associations of peak thrombin generation with risk of idiopathic VTE and secondary VTE were similar, although, at times, these subgroup associations did not achieve statistical significance. Associations tended to be larger for idiopathic than for secondary VTE in the ARIC study, but the opposite was observed in the CHS (albeit with smaller numbers).

Table 4 presents ORs for combinations of peak thrombin generation greater than the median and other VTE risk factors, adjusted for demographics and BMI. For FV Leiden, as compared with those without both risk factors, those with both factors had an OR for VTE of 6.93 (95% CI 3.38–14.22) (P = 0.04). The RERI suggested that 61% of the excess VTE risk for FV Leiden and high peak thrombin generation was due to the combination of these two factors. The APTT also interacted synergistically with peak thrombin generation, with a RERI% of 67% (P = 0.03). As compared with those with low APTT and low peak thrombin generation, those with both factors had an OR for VTE of 3.17 (95% CI 2.06–4.89). No other additive or multiplicative interactions between peak thrombin generation and VTE genetic risk factors or hemostatic factors were observed (data not shown).

Table 4.   Odds ratios (ORs) and relative excess risk due to interactions (RERI) for venous thromboembolism (VTE) in relation jointly to peak thrombin generation and other factors*: the Longitudinal Investigation of Thromboembolism Etiology (LITE) study
VariableVariable levelPeak thrombin generation (median cut-point) OR (95% CI)RERI RERI%P-value§
  1. APTT, activated partial thromboplastin time. *Adjusted for age, sex, race, center, and body mass index. Median based on the study-specific control distribution. RERI% estimates the proportion of VTE caused by two factors that can be attributed to their additive interaction. §P-values for additive interaction.

VTE genetic risk factors
 FV Leiden
  YesHigh6.93 (3.38–14.22)3.64610.04
  NoHigh1.56 (1.20–2.02)
  YesLow2.73 (1.25–5.94)
  NoLow1.00 (Reference)
 Prothrombin mutation
  YesHigh2.93 (1.13–7.63)0.41210.81
  NoHigh1.60 (1.24–2.06)
  YesLow1.93 (0.66–5.60)
  NoLow1.00 (Reference)
 ABO blood type
  YesHigh2.66 (1.85–3.81)0.52310.20
  NoHigh1.51 (1.01–2.24)
  YesLow1.63 (1.11–2.38)
  NoLow1.00 (Reference)
Biomarkers
 FVIII (top 25%)
  HighHigh2.38 (1.65–3.42)0.76550.14
  LowHigh1.39 (1.04–1.85)
  HighLow1.23 (0.77–1.95)
  LowLow1.00 (Reference)
 D-dimer (> median)
  HighHigh2.53 (1.78–3.61)0.0960.83
  LowHigh1.71 (1.22–2.39)
  HighLow1.73 (1.21–2.48)
  LowLow1.00 (Reference)
 APTT (bottom 25%)
  LowHigh3.17 (2.06–4.89)1.45670.03
  HighHigh1.36 (0.94–1.97)
  LowLow1.36 (0.75–2.44)
  HighLow1.00 (Reference)

Sensitivity analyses were conducted in which we excluded individuals with a self-reported history of VTE at baseline. In the ARIC study, 23 cases and 17 controls reported a history of VTE, whereas in the CHS, 22 cases and 12 controls reported prior VTE. Results of these analyses were similar to those obtained when the entire LITE case–control sample was included (data not shown). Additional sensitivity analyses were conducted in which: (i) we included individuals with levels below the manufacturer’s 0.5 nm lower limit of detection (n = 45); and (ii) we excluded individuals with extremely low (but detectable) peak thrombin generation values of ≤ 4 nm (n = 19). In both instances, the results were virtually identical to those obtained with the entire cohort.

Discussion

In this prospective study which combined data from two independent cohorts, elevated basal peak thrombin generation was associated with an increased risk of subsequent VTE. In the combined LITE data, participants with peak thrombin generation values above the median were at 74% greater risk of VTE than those in the lowest quartile. In the ARIC study, the associations were more apparent for idiopathic VTE than for secondary VTE, suggesting a possible genetic basis for the association. Peak thrombin generation was positively associated with some other hemostatic factors related to VTE risk, and the association of peak thrombin generation with VTE was partly attenuated after accounting for levels of some hemostatic factors.

The relationship of thrombin generation with first DVT has been explored in the LETS case–control study [12]. The LETS investigators reported that participants with high levels of ETP (90th percentile measured in controls) were at increased risk of incident DVT, with associations being stronger for idiopathic DVT. However, a dose–response relationship was not demonstrated when ETP was modeled as quartiles, and no associations were observed between peak thrombin generation and incident DVT. It is unclear why our results were more robust than those of the LETS, although methodological differences may provide some explanation. Although similar, the assays used to measure peak thrombin generation in these studies were not identical; the LETS used the Thrombinoscope assay [7] (Synapse, Maastricht, The Netherlands), whereas the LITE study used the Technothrombin TGA assay (Technoclone). Also, the LITE study enrolled participants and collected plasma samples before the onset of VTE, whereas in the LETS, case participants were sampled after their first VTE and at least 3 months after the discontinuation of anticoagulant therapy [12]. Blood storage duration also varied between studies. Most LITE findings for other hemostatic factors have been similar to LETS findings, but these design differences might still be a partial explanation for the different findings.

Elevated peak thrombin generation interacted synergistically with both FV Leiden and low APTT. Participants with both elevated peak thrombin generation and FV Leiden were at greatly increased risk of VTE (OR: 6.93). Those with low APTT and elevated peak thrombin generation had a two-fold greater VTE risk. Screening for these risk factor combinations may be clinically useful in identifying individuals at greatest risk of VTE.

In the ARIC study and the CHS, demographic variables, BMI and genetic VTE risk factors were unrelated to peak thrombin generation levels. Published data evaluating the relationship of thrombin generation with these variables have been inconclusive. Most [8,26,27], but not all [9,10], prior studies have reported a positive association with age. Sex has generally been unrelated [8,10,12], although one study found higher peak thrombin generation in females [26]. The FV Leiden mutation had no relationship with thrombin generation in studies using assays similar to the one used here [8,10,12], but a positive relationship in another that utilized an assay measuring the effect of thrombomodulin on thrombin generation [28]. Likewise, relationships of thrombin generation with the prothrombin mutation have been mixed, with reports showing both a positive relationship [10,12] and no relationship [8,28]. The absence of an association of peak thrombin generation with BMI, race or ABO blood group has not previously been reported. Notably, it is difficult to compare studies, because assay methodology differences, such as the presence or absence of thrombomodulin or activated protein C, and phospholipid or tissue factor concentration, probably explain the differences in the findings. Any clinical application of peak thrombin generation testing requires assay standardization across methods and further clinical research data comparing different assays for clinical utility.

Thrombin generation plays a central role in hemostasis and thrombosis, as the entire coagulation system is engaged in its generation and subsequent inactivation [29]. Consistent with published findings [26,30], in the ARIC study and the CHS, peak thrombin generation levels were positively correlated with levels of other hemostatic factors. Specifically, in the ARIC study, there was a positive association between peak thrombin generation levels and FVIII and VWF, and an inverse association with APTT, whereas in the CHS, peak thrombin generation was positively related to concentrations of FVII, FVIII, FIX, and FX.

Given the shared pathway of hemostatic factors and thrombin generation, we explored whether peak thrombin generation was predictive of VTE risk, independently of hemostatic factors related to thrombosis. In the LITE study, FVIII and D-dimer were identified statistically as potential ‘mediators’ of the association between peak thrombin generation and VTE. FVIII is activated by thrombin [3], whereas D-dimer is a marker of fibrin formation; both are VTE risk factors [31–38]. Levels of these ‘mediators’ only partly explained associations of peak thrombin generation with VTE, thereby suggesting that peak thrombin generation measurement may add independent information to current thrombotic risk assessment. Findings from two recent VTE recurrence cohorts also showed that thrombin generation and D-dimer were independently associated with risk of recurrence [10,11]. Furthermore, the odds of VTE were qualitatively elevated among LITE participants with high peak thrombin generation in conjunction with high FVIII or D-dimer levels, as compared with those with a single elevated hemostatic factor. This, too, indicates that thrombin generation assessment may provide complementary clinical information.

The peak thrombin generation distributions in the ARIC study and the CHS were different enough that we were reluctant to combine the cohorts using absolute thrombin generation values. The amount of missing data also differed between cohorts. Nevertheless, as hypothesized, peak thrombin generation was positively associated with VTE incidence in both cohorts, and in the data combined using relative values. Thus, whereas the absolute levels of peak thrombin generation in the LITE study may be questioned, peak thrombin generation was moderately effective in rank-ordering participants according to future VTE risk.

Although the study methods in the ARIC study and the CHS were very similar, minor differences in blood collection, processing or storage may have influenced thrombin generation levels. Peak thrombin generation levels may also have been influenced by prior thawing and refreezing of the samples, or the long duration of storage at − 70 °C (up to 20 years). Any impact of preanalytical or analytical factors for this assay on our results, however, would presumably be distributed comparably in cases and controls, and would most likely weaken associations between peak thrombin generation and VTE, thereby biasing our results towards the null hypothesis. This suggests that the risk estimates that we observed are underestimates of the true risk of VTE associated with peak thrombin generation levels.

Given the novelty of this assay, little has been published regarding how factors such as storage and processing may affect values. Our difference in mean values between cohorts is not without precedent. A review of published clinical studies that measured thrombin generation indicated large variability in results, probably owing to assay differences, making comparisons among studies difficult [39]. It has been suggested that between-study variability may be improved through normalization via use of reference plasma, by standardizing protocols, and by using standard reagents [39,40]. If clinical applications are considered, these issues require resolution.

There are limitations of this study, aside from those already discussed pertaining to assay factors. Inclusion as a case in this study required a clinical diagnosis of VTE. Cases of VTE that were asymptomatic or not clinically recognized would not have been detected. Likewise, some cases of fatal PE may have been missed. Symptomatic DVT may also have been missed if it was treated in an outpatient setting, although this would have been rare during the time period of this study [41]. Strengths of this study include the prospective investigation of VTE with assessment of peak thrombin generation before VTE onset, systematic case-validation, relatively large sample size and number of hemostatic factors assessed, and the population-based sample from a wide geographic distribution in the USA. Furthermore, this study is unique in its exploration of whether the VTE–peak thrombin generation association was independent of other hemostatic factors, and its assessment of additive interactions.

To conclude, in these prospective LITE data, elevated basal peak thrombin generation was associated with a 74% increased risk of VTE. Although additional research is needed, the observation that the association of peak thrombin generation with VTE was only partly explained by other measured hemostatic factors is intriguing, and suggests that thrombin generation assessment may have clinical utility in stratifying individuals according to their risk of incident VTE. FV Leiden and low APTT were found to interact synergistically with elevated peak thrombin generation. Patients with relevant combinations of these adverse risk factors may be worthwhile targets for intervention. Overall, our findings on risk of first VTE are consistent with previous studies of recurrent VTE [8,9], which reported that thrombin generation potential was predictive of future thrombosis.

Acknowledgements

The LITE study was funded by National Heart, Lung, and Blood Institute (NHLBI) grant R01 HL59367. The ARIC study is carried out as a collaborative study supported by NHLBI contracts N01-HC-55015, N01-HC-55016, N01-HC-55018, N01-HC-55019, N01-HC-55020, N01-HC-55021, and N01-HC-55022. The CHS was funded by NHLBI contracts N01-HC-85079 to N01-HC-85086, N01-HC-35129, N01 HC-15103, N01 HC-55222, N01-HC-75150, and N01-HC-45133, and NHLBI grant U01 HL080295, with additional contributions from the National Institute of Neurological Disorders and Stroke. P. L. Lutsey was supported as a predoctoral fellow on NHLBI training grant T32 HL07779. The authors thank the staff and participants of the ARIC study and the CHS for their important contributions. We have no conflicts of interest to declare.

Disclosure of Conflict of Interests

The authors state that they have no conflict of interest.

Ancillary