The contribution of obesity to the occurrence of cardiovascular events may not be wholly related to its influence on traditional risk factors. Coagulation and fibrinolysis may also influence cardiovascular risk, but the relationship of adiposity with these processes is unclear. The aim of the present study was to investigate the relationships of BMI (body mass index), waist circumference, hip circumference and WHR (waist-to-hip ratio) with VIIc (factor VII activity), plasma markers of thrombin generation [F1+2 (prothrombin fragment 1+2)], fibrin formation [SF (soluble fibrin)] and fibrin turnover (D-dimer), and PAI-1 (plasminogen activator inhibitor-1; a marker of fibrinolytic inhibitory capacity). The study cohort was 80 healthy postmenopausal women who were not diabetic, current smokers or taking hormone therapy and who had a fasting sample of blood collected. VIIc, F1+2, SF and PAI-1 were all positively correlated with BMI, waist circumference and WHR, whereas D-dimer was positively correlated with waist circumference and WHR, but not BMI. WHR was the strongest correlate of all the markers except for PAI-1, which was most closely related to BMI. Hip circumference became a negative correlate of F1+2 and D-dimer after adjusting for waist circumference. The relationships of WHR with F1+2 and SF, but not with VIIc and D-dimer, were independent of traditional risk factors. The positive association between waist circumference and markers of thrombin generation, fibrin production and fibrin turnover suggests that abdominal adiposity may contribute to atherothrombosis by activating intravascular coagulation. In contrast, a larger hip circumference appears to have a protective affect against coagulation activation.
- body mass index (BMI)
- hip circumference
- waist-to-hip ratio (WHR)
Excess body adiposity, assessed most commonly by BMI (body mass index), has been found in epidemiological studies to be associated with an increased risk of coronary heart disease, ischaemic stroke and mortality . The degree of abdominal adiposity, assessed by waist circumference or WHR (waist-to-hip ratio), also predicts cardiovascular risk independently of BMI . Although waist circumference is a better marker of abdominal obesity than WHR , there is accumulating evidence that WHR may be a better marker of cardiovascular risk [2–4]. This advantage of WHR occurs because hip circumference becomes an independent inverse predictor of cardiovascular risk after adjustment for waist circumference [2–4].
The mechanisms underlying the association of adiposity with coronary heart disease and mortality are incompletely understood. Although obesity is associated with the coronary risk factors of hypertension, hyperlipidaemia and Type 2 diabetes, a portion of the increased risk accompanying obesity appears to be independent of traditional coronary risk factors [3,5–7]. A possible independent mechanism by which obesity may contribute to cardiovascular risk is by potentiation of thrombosis in the arterial circulation. Indeed, an association between obesity and arterial thrombosis is suggested by evidence that BMI is a predictor of acute coronary events in patients with coronary artery disease independently of the number of coronary lesions . Furthermore, an elevated BMI is a recognized risk factor for venous thrombo-embolism [8,9], suggesting a prothrombotic effect of obesity in the venous circulation. A mechanism underlying these observations in the arterial and venous circulations could be the abnormalities of coagulation and fibrinolysis which also occur in obesity.
Most consistently associated with obesity and a potential contributor to thrombosis is an increase in PAI-1 (plasminogen activator inhibitor-1), the main inhibitor of the fibrinolytic system . Activity of the vitamin K-dependent clotting factor VII (VIIc), a component of the extrinsic coagulation pathway, is also increased in obesity [11–13]. F1+2 (prothrombin fragment 1+2), a marker of thrombin generation, was associated with BMI in one study , but other studies have not confirmed a relationship between BMI and F1+2 [12–15]. However, interpretation of these negative studies is potentially confounded by methodological issues, such as smoking [11,12], oral contraceptive pill use , combined analysis of premenopausal and postmenopausal women  and the collection of blood samples at different times of the day . Furthermore, previous studies have only measured BMI, but, in view of the closer relationship between abdominal adiposity than BMI with cardiovascular events  and the apparent protective effect of lower body adiposity against cardiovascular events [2–4], the relationship of waist and hip circumference with coagulation and fibrinolysis may be of greater relevance.
The aim of the present study was to determine the relationship between markers of the coagulation and fibrinolytic pathways with BMI, waist circumference, hip circumference and WHR in a selected group of healthy non-smoking postmenopausal women without diabetes and not on hormone therapy. As well as the measurement of PAI-1, VIIc and F1+2, we also assessed downstream activation of the coagulation pathway by measuring SF (soluble fibrin). In addition, we measured D-dimer, a marker of fibrin turnover/breakdown, which is a predictor of cardiovascular events independent of traditional cardiovascular risk factors [16–22].
MATERIAL AND METHODS
The study group comprised 80 postmenopausal women who were enrolled in two randomized controlled trials [23,24]. Women aged 50–75 years were enrolled from the community after responding to an advertisement in community newspapers. The women had not consumed antibiotics, soya products or supplements (for 3 months), nor had they taken oestrogen therapy (for 12 months) prior to entry. Postmenopausal status was defined as 12 months of amenorrhoea and FSH (follicle-stimulating hormone) >20 international units/l. Exclusion criteria included smoking within the past 10 years, diabetes, body weight >100 kg, alcohol or elicit drug abuse, uncontrolled hypertension [BP (blood pressure) >180/100 mmHg], history of venous thrombosis, breast or endometrial cancer, abnormal uterine bleeding, abnormal cervical smear or mammogram results, presence of a major illness or alcohol consumption >30 g/day. The Monash Medical Centre Human Research and Ethics Committee approved the studies, and all participants gave written informed consent.
Weight and height were measured, and BMI was calculated by dividing weight (in kg) by height squared (in m2). Waist circumference was measured over the unclothed abdomen at the midpoint of the lower thoracic cage and iliac crest in the mid-axillary line using the method recommended by the World Health Organization . Hip circumference was measured at the level of the widest diameter around the buttocks. BP was measured with subjects in a recumbent position after 10 min of rest using a Dinamap (CRITIKON 1846 SX). A total of six BP recordings were performed over 15 min, and the last five readings were averaged.
Blood sampling and assays
Fasting morning blood samples were collected, by a single technician trained in non-traumatic phlebotomy, prior to measurement of lipids and haemostatic variables (VIIc, SF, F1+2, D-dimer and PAI-1). Venepuncture was performed with a 19-gauge needle directly into plain tubes (Vacutainer; Becton Dickson) for lipid and hormone assays, and then into two 3.8% citrate tubes (9:1 ratio) for coagulation and fibrinolysis studies. The samples were immediately centrifuged at 2500 g for 12 min, plasma was separated into 200 μl aliquots, stored at −80°C and only thawed immediately prior to analysis.
Hormone and lipoprotein assays
Total cholesterol and triacylglycerols (triglycerides) were measured using enzymatic reagents (Dade Diagnostics), and HDL-C (high-density lipoprotein cholesterol) was measured by homogeneous HDL-C assay techniques (HDLC-Plus; Dade Diagnostics) adapted to a Dade Dimension RXL chemistry analyser (Dade Diagnostics), whereas LDL-C (low-density lipoprotein cholesterol) was calculated using the Friedewald equation adapted to S.I. units.
Assays of haemostatic markers
All variables were assayed by a single trained medical scientist, with all assays performed in duplicate and a mean of two values reported. VIIc was measured in a one-stage clotting assay with factor VII-deficient plasma (Helena Laboratories) and Innovin thromboplastin (Dade Behring). F1+2 was measured by enzyme immunoassay (Enzygnost F1+2; Behring). SF was measured using Enzymum FM (Boehringer Mannheim). The PAI-1 activity assay (Berichrom®PAI) was automated on ACL Futura Instrumentation Laboratory equipment. D-dimer was measured with the Dimertest Gold EIA kit (Biomedical Ltd).
All statistical analyses were performed using SPSS for Windows software (version 12.0). Triacyglycerols, Lp(a) [lipoprotein (a)] and all of the haemostatic variables were log-transformed to normalize their distributions. Pearson correlation coefficients (r) were used to assess the associations between the different anthropometric measures and the haemostatic variables VIIc, F1+2, SF, D-dimer and PAI-1 with BMI, waist circumference, WHR, age, BP and lipids. Multiple linear regression models were used to test two hypotheses: (i) that body adiposity is associated with coagulation activation independent of traditional risk factors, and (ii) that hip circumference has an independent negative association with haemostatic markers after adjusting for waist circumference. The partial correlation coefficient (β) value is provided in those multivariate analyses where the aim was to determine the degree of contribution of various independent variables to the dependent variable. The partial regression coefficient (B) value is presented in those analyses of haemostatic markers where the aim was to determine the changes in contribution of waist and hip when they were both included in the model. The coefficient of determination (r2) was adjusted for increases in the numbers of independent variables in a statistical model (adjr2). Statistical significance was defined as P<0.05.
The subject characteristics at baseline are shown in Table 1. Of the cohort, 27 (34%) were overweight (BMI, 25.0–29.9 kg/m2) and 19 (24%) were obese (BMI>29.9 kg/m2). The relationship between the various anthropometric measures of adiposity were determined, and waist circumference was moderately correlated with hip circumference (r=0.83, P=0.0001) and strongly correlated with BMI (r=0.93, P=0.0001). Hip circumference was moderately correlated with BMI (r=0.86, P=0.0001), and WHR was moderately correlated with BMI (r=0.67, P=0.0001). Age was not related to BMI, waist or hip circumference by univariate analysis, but was positively correlated with WHR (r=0.24, P=0.03) and negatively correlated with hip circumference (P=0.001) after adjusting for waist circumference.
Anthropometric measures and lipids
The relationships of age, BMI, waist circumference, hip circumference and WHR with the lipid variables were analysed. Age was not associated with any of the lipids. Waist circumference, hip circumference, BMI and WHR were all positively correlated with triacylglycerols (r=0.40 to 0.47, P<0.001) and negatively correlated with HDL-C (r=−0.39 to −0.46, P<0.001). However, hip circumference was not independently related to any of the lipids after adjusting for waist circumference.
Correlations of haemostatic markers with age, anthropometric measures and lipids
The relationships of the log-transformed variables VIIc, F1+2, SF, D-dimer and PAI-1 with age, BMI, waist circumference, WHR and lipids are shown in Table 2. VIIc was positively associated with waist circumference, but more strongly correlated with WHR. VIIc was also positively associated with total cholesterol, LDL-C and triacylglycerols, and the correlation of VIIc with WHR was no longer significant after adjusting for LDL-C and triacylglycerols. The best multivariate model of VIIc using anthropometric measures and lipids included WHR (β=0.294, P=0.007) and total cholesterol (β=0.263, P=0.014), and this model explained 17% of the variability of VIIc.
F1+2 was positively associated with BMI and waist circumference, but most strongly related to WHR (Figure 1A). WHR remained significantly associated with F1+2 (P=0.04) after adjusting for age, cholesterol and systolic BP. The best multivariate model of F1+2 (adjr2=0.29) included age (β=0.413, P<0.0001) and WHR (β=0.260, P=0.011).
SF was positively correlated with BMI and waist circumference, but marginally better correlated with WHR (Figure 1B). The relationship of SF with WHR remained significant (P=0.003) after adjusting for age and systolic BP. The combination of age (β=0.310, P=0.003) and waist circumference (β=0.356, P=0.001) was the best model of SF (adjr2=0.22).
D-dimer was positively correlated with waist circumference and more strongly correlated with WHR (Figure 1C), but was not correlated with BMI. The relationship of WHR with D-dimer was no longer significant after adjusting for age, systolic BP, LDL-C and HDL-C. The best multivariate model of D-dimer (adjr2=0.14) included age (β=0.325, P=0.003) and HDL-C (β=−0.241, P=0.025).
PAI-1 was positively correlated with BMI, waist circumference and WHR, but the strongest relationship was with BMI. PAI-1 was also positively associated with LDL-C and triacylglycerols, and negatively associated with HDL-C. BMI remained a significant correlate of PAI-1 (P=0.007) after including LDL-C, HDL-C and triacylglycerols in the model. The combination of BMI (β=0.359, P=0.0001) and HDL (β=−0.411, P=0.0001) was the best multivariate model of PAI-1, and explained 41% of the variability.
Linear regression models for haemostatic markers
Table 3 demonstrates the regression coefficients and adjr2 of the haemostatic markers with hip circumference (Model 1), waist circumference (Model 2) and multivariate models, including waist and hip circumference (Model 3) and waist circumference, hip circumference and age (Model 4). Hip circumference was positively correlated with SF and PAI-1, but not related to VIIc, F1+2 or D-dimer (Model 1). Waist circumference was positively correlated with VIIc, F1+2, D-dimer and PAI-1 and positively correlated with borderline significance with D-dimer (Model 2). For F1+2, SF and D-dimer, the inclusion of waist circumference and hip circumference in the model (Model 3) resulted in an increase in the regression coefficient for waist circumference, an increase in the adjr2, and the regression coefficient for hip circumference becoming negative. The contributions of hip circumference in the models were significant for F1+2 and D-dimer, but not for SF. After inclusion of hip and waist circumference in the model of PAI-1 (Model 3), there was no increase in adjr2 and the regression coefficient of hip circumference remained positive. Hip circumference was still inversely related to F1+2 and D-dimer after including age in the model (Model 4), but was no longer significant.
Correlations between haemostatic markers
The relationships between the various haemostatic variables were examined. F1+2 was positively associated with SF (r=0.55, P<0.001) and D-dimer (r=0.30, P=0.007), and SF was positively associated with D-dimer (r=0.62, P<0.001). After including F1+2 in models of SF, the contribution of waist circumference (P=0.011) and WHR (P=0.045) remained significant. After including SF in models of D-dimer, the contributions of WHR and HDL-C were no longer significant.
In this cross-sectional study, performed in a highly selected group of healthy non-smoking postmenopausal women without diabetes and not on hormone therapy, we found positive correlations between haemostatic markers and standard anthropometric measures of adiposity. There were differences in the degree of these relationships, with VIIc, F1+2, SF and D-dimer more closely correlated with WHR than BMI, whereas PAI-1 was most closely related to BMI. These findings not only suggest that adiposity contributes to activation of intravascular coagulation and fibrinolysis, but also provide evidence that there is heterogeneity in the effects of regional adipose stores on specific components of the coagulation and fibrinolytic pathways.
There has been limited previous investigation of the relationship between adiposity and intravascular thrombin generation as reflected by F1+2 concentrations and the results have not been consistent. An increase in F1+2 with higher BMI was reported in one study , but four other studies found no significant relationship between F1+2 and BMI [12–15]. However, all previous studies included subjects with characteristics which could have confounded the interpretation of the results. Several of the studies performed a combined analysis of men and women [11,12], which may not be appropriate, as there is evidence that women have higher F1+2 concentrations than men [11,15,26,27]. Most of the studies included smokers [11–15], despite consistent findings that smokers have higher F1+2 concentrations than non-smokers [11,15,28]. In one study, the cohort included premenopausal and postmenopausal women together, and some of the women were using the oral contraceptive pill, which is recognized to increase F1+2 . In two of the studies, the blood sample was collected at variable times during the day [12,14], yet there is recognized to be a diurnal rhythm of F1+2 [26,30], with F1+2 concentrations lower after 15.00 hours. Although it may be possible to statistically correct for the above factors, such correction was not performed in all studies. Furthermore, such correction could reduce the power of a study to detect a relationship between BMI and F1+2, particularly if the numbers were small to begin with or if there were interactions between the variables (e.g. smoking and BMI ). In the present study, we avoided these confounding factors by including only healthy women who were postmenopausal non-smokers not diabetic and not on hormone-replacement therapy, and collecting all blood samples in the morning in fasted subjects. Consistent with our finding of a positive correlation between adiposity and F1+2, in a study of non-diabetic non-smoking women with morning sample collection F1+2 was higher in obese subjects .
Although we found F1+2 to be positively correlated with BMI, a marker of general adiposity, F1+2 was more strongly associated with WHR. The relationship with WHR remained after adjustment for age, cholesterol and systolic BP, and 29% of the variability of F1+2 was explained by the combination of age and WHR. We also investigated the relationship between hip circumference and F1+2 by constructing linear regression models including hip circumference, waist circumference and age. Although there was no significant relationship between hip circumference and F1+2 on univariate analysis, the combination of waist and hip not only increased the contribution of waist circumference to the model, but resulted in hip circumference becoming a significant negative correlate of F1+2. The improved contribution of waist circumference to the model of F1+2 after including both waist and hip circumference indicates a benefit of adjusting waist for hip circumference, the development of a negative correlation of hip with F1+2 in the combined model with waist suggests an ‘anticoagulant’ effect of hip adiposity, whereas the lack of a relationship between hip and F1+2 on univariate analysis can be explained by colinearity between waist and hip circumference.
An increase in F1+2 with age was observed in the present study and has been consistently demonstrated in previous studies [15,26–28,33–35]. Therefore we examined the relationship of F1+2 with the various anthropometric measures after adjustment for age. Although WHR and waist circumference remained significant predictors of F1+2 (and SF) after adjusting for age, hip circumference was no longer a significant contributor in the model after including waist and age. Nevertheless, the interpretation of this finding is not clear for several reasons. First, age was negatively correlated with hip circumference after adjusting for waist circumference. Therefore in our cohort the lack of significance of hip circumference in a model of F1+2 including age and waist circumference may be due to colinearity between age and hip circumference. Secondly, although there is evidence that aging does not result in diminished clearance of F1+2 , the mechanism by which aging results in increases in F1+2 is not understood. Given that percentage body adiposity increases with age even in the absence of an increase in BMI , at least part of the correlation observed previously between age and F1+2 could be due to the relationship between older age and increased adiposity. The relationship between age and F1+2 also probably accounted for the positive association of F1+2 with systolic BP, the latter was also associated with age in our present study. The positive association of WHR with F1+2 observed in our non-smoking cohort remained after adjustment for age, cholesterol and systolic BP, indicating that it was independent of traditional cardiovascular risk factors.
In the present study, there was not only a positive correlation between F1+2 and SF, but SF was also positively correlated with anthropometric markers of adiposity, indicating that increases in fat mass are accompanied by elevations in intravascular fibrin production as well as thrombin generation. Although an increase in both thrombin and fibrin production is not unexpected given that thrombin formation is a precursor of SF generation, increased SF is not an automatic consequence of increased thrombin due to the existence of natural circulating inhibitors of thrombin . Indeed, a contrast between thrombin and fibrin generation was observed in the univariate analysis, as SF was positively correlated with hip circumference and waist circumference, whereas F1+2 was only correlated with waist circumference. Moreover, waist circumference remained a predictor of SF even after adjusting for F1+2, raising the possibility that adiposity predisposes to fibrin production independently of its effects on thrombin generation. After adjustment for waist circumference, we observed a similar inverse correlation of hip circumference with SF to that seen with F1+2, although this relationship was not statistically significant.
D-dimer is a marker of fibrin production and breakdown and, as expected, we found a positive correlation between D-dimer and SF, a marker of fibrin production, as well as between D-dimer and F1+2, a marker of thrombin generation. D-dimer was positively associated with waist circumference and negatively associated with hip circumference, also consistent with the findings for F1+2 and SF. The positive association of D-dimer with waist circumference is consistent with a previous report in children which found that D-dimer was associated with various measures of body fat , and supports the hypothesis that activation of coagulation is related to abdominal adipose tissue. The relationship between adiposity and D-dimer is of clinical interest, as D-dimer is a predictor of venous thrombo-embolism , prevalent atherosclerotic disease independent of traditional risk factors  and atherothrombotic events in subjects with [18,19] and without [20,21] known vascular disease. Furthermore, in healthy elderly subjects over 65 years of age, D-dimer was a predictor of myocardial infarction and coronary death, but not angina, suggesting a relationship between atherothrombosis and D-dimer independent of the relationship between D-dimer and atherosclerotic disease .
VIIc was positively correlated with BMI in a number of studies [11–13], but limited information has been available about the relationship of VIIc with regional adiposity. In our present study, WHR was the strongest correlate of VIIc, although hip circumference was not a significant negative correlate of VIIc after including waist circumference in the model. However, the significance of an increase in VIIc to thrombotic risk remains unclear. Thus, although factor VII is an important component of the extrinsic coagulation pathway , we found no relationship between VIIc and the markers of coagulation activation (F1+2 and SF) measured in the present study.
Similar to previous reports, we found a positive correlation between PAI-1 activity and anthropometric measures of adiposity . In contrast with our findings with the coagulation markers, there was also a positive association of PAI-1 with hip circumference, which accounted for a similar percentage of the variability of PAI-1 as waist circumference, and no sign of any protective effect of hip circumference after adjusting for waist circumference. The relationship between adiposity and PAI-1 is likely to have a direct component as adipocytes produce PAI-1 [39,40], although the degree of contribution of adipose production of PAI-1 to circulating PAI-1 concentrations remains unclear. Most previous studies have emphasized the relationship between PAI-1 and abdominal obesity , and there is experimental evidence suggesting that visceral fat may play a more important role than subcutaneous fat to increased PAI-1 concentrations . However, this is not supported by our findings in postmenopausal women, as BMI was the best predictor of PAI-1 and the correlation of PAI-1 with waist and hip circumference was of a similar degree. The significance of increased PAI-1 remains unclear as some observational studies have found high concentrations of PAI-1 to be a risk factor for cardiovascular disease, but the significance of PAI-1 as a predictor often disappears after adjusting for lipids and BMI .
Our finding that waist circumference and hip circumference have independent and opposite associations with markers of coagulation activity adds to accumulating evidence of similar inverse relationships between waist and hip circumference with coronary risk factors [42–46] and cardiovascular events. Of particular interest is a 24-year follow-up of Swedish women, where hip circumference was a significant inverse predictor of myocardial infarction after adjusting for BMI , a 6.8-year follow-up of Danish men and women, in which waist and hip circumferences had independent and opposite associations with all-cause mortality , and the INTERHEART study, which found hip circumference to be an independent inverse predictor of myocardial infarction after adjusting for waist circumference . The mechanism by which hip circumference is protective remains theoretical , as does the mechanism underlying the activation of coagulation associated with increased abdominal adiposity .
A limitation of the present study is its relatively small numbers. Furthermore, our study only included postmenopausal women and, therefore, the findings may not apply to other populations. Nevertheless, the strict selection criteria for our cohort is also a strength of our study as it avoided a number of the potential confounding factors present in previous studies investigating the relationship between haemostatic factors and adiposity.
In conclusion, we have found a positive association between waist circumference and markers of thrombin generation, fibrin production and fibrin turnover, suggesting that abdominal adiposity may contribute to thrombosis by activating intravascular coagulation. In contrast, the inverse association between hip circumference and coagulation activity after adjusting for waist circumference raises the possibility of a protective or anticoagulant effect of hip adiposity. Further investigation is required to better understand the mechanisms underlying these relationships and to determine whether there are similar relationships between waist and hip circumference with coagulation activity in men.
We thank the Department of Clinical Biochemistry, Southern Health, Clayton, Victoria, Australia for performance of the lipid assays.
Abbreviations: adjr2, adjusted r2; BMI, body mass index; BP, blood pressure; F1+2, prothrombin fragment 1+2; HDL-C, high-density lipoprotein cholesterol; LDL-C, low-density lipoprotein cholesterol; Lp(a), lipoprotein (a); PAI-1, plasminogen activator inhibitor-1; SF, soluble fibrin; VIIc, factor VII activity; WHR, waist-to-hip ratio
- © The Authors Journal compilation © 2007 Biochemical Society