Suzanne C. Cannegieter, Department of Clinical Epidemiology, C9-P, Leiden University Medical Center (LUMC), PO Box 9600, 2300 RC Leiden, the Netherlands. Tel.: +31 71 5261508; fax: +31 71 5266994. E-mail: email@example.com
Summary. Background: High body mass index (BMI) is associated with an increased risk of venous thrombosis (VT). Clotting factor VIII levels are increased in obese subjects, possibly because of a chronic inflammatory state, which increases activated protein C (APC) resistance. The APC resistance in FV Leiden carriers could be aggravated and further worsened by high FVIII levels in blood group non-O carriers. We hypothesized that an association exists between BMI and APC resistance, and that this is amplified by the presence of FV Leiden and/or blood group non-O.
Methods: We used the Leiden Thrombophilia Study (LETS) to determine whether an association exists between BMI and APC resistance, and whether the combination of high BMI and APC resistance increases the risk of VT. In a pooled analysis of LETS and a Norwegian case-cohort study (TROL), we verified whether FV Leiden modified the risk of the occurrence of VT with increasing BMI, and whether this risk was further increased by blood group non-O.
Results: APC resistance increased linearly with increasing BMI, partly because of a concurrent rise in FVIII. A BMI in the median or upper tertile was associated with a 1.9-fold (95% confidence interval [CI] 1.0–2.5) and 2.2-fold (95% CI 1.4–3.4) increased risk as compared with the lowest tertile. Both relative risks decreased slightly after FVIII and APC resistance adjustments. The effect of BMI on VT risk was enhanced two-fold to 10-fold in FV Leiden or blood group non-O carriers.
Conclusions: The increased risk of VT in individuals with high BMI is partly mediated by FVIII-related APC resistance. This risk is more pronounced when other causes of increased APC resistance are also present.
Venous thrombosis (VT) occurs at a rate of 1–2 per 1000 individuals per year [1,2], and is more frequent in overweight individuals (body mass index [BMI] of 25–30 kg m−2) and obese individuals (BMI ≥ 30 kg m−2) than in lean individuals [3,4]. Overweight or obesity is associated with a 2-3-fold increased risk of first VT as compared with normal weight [4,5]. It is not completely understood how high BMI predisposes to VT. Overweight or obese people tend to be more immobile, which may lead to clot formation through stasis. It is also possible that these individuals acquire a prothrombotic state. In 1997, Tosetto et al. reported that activated protein C (APC) resistance was significantly influenced not only by the presence of factor V Leiden , but also by BMI and ABO blood group . Since then, other studies have found a stronger effect of high BMI on the risk of first VT in FV Leiden carriers [5,8]. Another study also reported an association between increasing BMI and increasing APC resistance . It was therefore speculated that a joint effect of FV Leiden and increased APC resistance resulting from high BMI might explain the enhanced risk of VT in heavy individuals with FV Leiden .
High levels of FVIII are also associated with an increased risk of VT . This risk is partially genetically determined, as individuals with blood group non-O have higher levels of FVIII than individuals with blood group O . A high BMI is also associated with an elevation in the level of FVIII [4,11–13], which may be a result of dysfunctional adipose tissue that influences hepatic metabolism and possibly the production of coagulation factors in the liver . As obesity is considered to be an inflammatory disease, the association between obesity and FVIII could also be a consequence of chronic inflammation, as FVIII is reported to be an acute-phase reactant . As FVIII increases APC resistance , the mechanism by which high BMI leads to VT could be FVIII-induced APC resistance. If this is so, a reinforcement of this risk in the presence of FV Leiden, blood group non-O or a combination of both genetic factors can be expected .
We used the data from two well-characterized population-based case-control studies, one from The Netherlands (Leiden Thrombophilia Study [LETS]) and one from Norway (TROL), to study the four hypotheses graphically illustrated in Fig. 1. We set out to determine, first, whether an association exists between BMI and APC resistance in LETS, and whether this association is mediated by FVIII levels (Fig. 1). Second, we studied whether the risk of VT associated with high BMI was influenced by APC resistance or FVIII levels in LETS (Fig. 1). Third, in a pooled analysis of LETS and TROL, we examined whether blood group non-O modified the effect of increasing BMI on the risk of VT (Fig. 1), and whether this risk was further increased by the presence of FV Leiden (Fig. 1).
LETS patients and controls
The inclusion of patients and controls in LETS has been extensively described in the past [6,15,16]. In short, cases were consecutive patients treated for deep vein thrombosis (DVT) at the Anticoagulation Clinics in Leiden, Amsterdam and Rotterdam during the period 1 January 1988 to 30 December 1992. Patients had to be without a diagnosis of active cancer, be < 70 years of age at entry, and have a diagnosis of DVT established with objective diagnostic methods. With these criteria, 90% participated, resulting in a case group with 474 consecutive DVT patients. A control group was assembled, in which patients were to find their own healthy control subjects of the same sex, of the same age (± 5 years), with no biological relationship, and without a diagnosis of active cancer, a history of VT, or a history of having used coumarin therapy in the last 3 months before inclusion. In this way, 474 controls were found [6,15,16].
TROL patients and controls
Between August 1995 and June 1997, all inhabitants > 20 years of age (n = 94 194) in Nord-Trøndelag County (in Middle Norway) were asked to be part of a population-based health survey called ‘Helseundersøkelsen i Nord-Trøndelag’ (HUNT2) . All participants (n = 66 140; participation rate, 71%) underwent a physical examination, donated 7.5 mL of blood, and filled in a questionnaire. Participants had a median age of 46 years (range, 19–103 years) . The follow-up of HUNT2 has been described previously . In brief, cases with DVT and/or pulmonary embolism (PE) were identified by screening the computerized diagnosis registries of all departments in the only two local hospitals in the region (Levanger and Namsos hospitals) until 1 January 2002. The search was supplemented by crosslinking with all positive diagnostic procedures for venography, duplex ultrasound and Doppler ultrasound within the registries of the radiology departments at the two hospitals. Finally, patients from Nord-Trøndelag County discharged from St Olav University Hospital with diagnostic codes of VT were included. This search identified 2136 cases with a diagnostic code of VT, and their hospital records were reviewed to obtain information for validation of the diagnosis.
We used the following criteria as a confirmation of the diagnosis of DVT: an intraluminal filling defect or no venous filling on ascending contrast venography; or no compressible venous segment or no venous flow in the popliteal, femoral or axillary veins on duplex ultrasound; or a positive computed tomography scan or a positive autopsy. We used the following criteria as a confirmation of the diagnosis of PE : a ventilation/perfusion scan with one or more segmental or subsegmental perfusion defects with normal ventilation; or a contrast defect on pulmonary computed tomography scan; or a positive autopsy.
We found 1271 cases with a validated diagnosis, of whom 798 (63%) were within the HUNT2 cohort. We excluded 283 patients with eye vein thrombosis or previous VT, i.e. enrollment in the HUNT2 cohort after the event, leaving 515 cases with a definite diagnosis of VT. Subsequently, we sampled 1505 controls from the baseline of the HUNT2 cohort, frequency-matched to the cases by sex and 5-year age strata.
LETS Blood drawn from the participants was placed in 0.106 m trisodium citrate, and centrifuged at 2000 × g for 10 min before storage at − 70 °C in a 1.5-mL container [6,13,15]. A standard salting-out method was used to extract DNA .
APC resistance and the FV 1691 (G→A) variant (FV Leiden) were measured according to methods described previously in LETS [6,16].
Plasma activated partial thromboplastin time sensitivity to APC was measured according to the technique used by Dahlbäck et al. , and is expressed as normalized APC ratio. APC resistance measurements were missing or unsuccessful in 51 cases and five controls. DNA samples were missing or the measurement failed in three cases but in none of the controls. FVIII activity was measured with a one-stage coagulation assay, and blood group with PCRs [10,20].
TROL DNA from peripheral blood leukocytes (whole blood or blood clots) was extracted with the Puregene kit (Gentra Systems, Inc., Minneapolis, MN, USA), manually or with an Autopure LS (Gentra Systems). FV Leiden and blood group were measured with PCR according to the Taqman method . In nine cases (1.7%) and 50 controls (3.4%), DNA samples were missing or the measurements failed. Technicians were not aware of the status of cases or controls, as all samples were stored anonymously before the selection of a control group from the baseline cohort. In the TROL study, citrated plasma was not obtained, and therefore FVIII levels and APC resistance could not be determined in this study setting.
Linear regression was used to determine the relationships between increasing APC resistance and BMI, increasing FVIII levels and BMI, increasing APC resistance and FVIII levels, and increasing APC resistance and BMI adjusted for FVIII levels.
The cut-off points needed to create tertile categories of APC resistance were derived from the control group of the LETS population after the exclusion of FV Leiden carriers and individuals who used oral contraceptives at time of blood draw. Cut-off points to create tertile categories for BMI were derived from the control groups of the LETS and TROL populations separately, irrespective of FV Leiden or oral contraceptive use. Logistic regression was used to calculate odds ratios (ORs) and their 95% confidence intervals (CIs), adjusted for age and sex.
The effect of the combination of a high BMI with FV Leiden or blood group was calculated in the LETS and the TROL populations, both separately and combined. To create similar groups for the combination of both TROL and LETS populations, we restricted the analysis in the TROL subjects to those who were < 70 years of age and to those who had a DVT only in the current study, according to the LETS inclusion criteria (n = 183 cases and n = 696 controls). Including only patients with DVT in this study has the additional benefit that findings cannot be disturbed by the FV Leiden paradox, i.e. the fact that FV Leiden is associated with a higher risk of DVT than PE .
The statistical software used was spss version 16.0 (SPSS, Chicago, IL, USA).
Approval for this study was obtained from the Medical Ethics Committee of the Leiden University Medical Center (for LETS) and from the National Data Inspectorate and the Regional Committee for Medical Research Ethics of Central Norway (for TROL). All participants provided written informed consent according to the Declaration of Helsinki.
LETS and TROL study populations
The descriptive characteristics of the patients and control individuals of both study populations are listed in Table 1. The TROL population is an older population, as the upper limit of inclusion in LETS was 70 years, whereas there was no age limit in the TROL study. Hence, LETS cases and controls had a median age of 45 years, whereas this was 54 years for the TROL cases and 56 years for the controls. There were slightly more women than men in both studies.
Table 1. Baseline characteristics of subjects in the Leiden Thrombophilia Study (LETS) and the TROL study
Cases n (%)
Controls n (%)
Cases n (%)
Controls n (%)
Age groups (years)
Correlation of BMI, APC resistance, and FVIII
Figure 2A shows that, in the LETS control individuals, the normalized APC ratio decreased linearly with increasing BMI. This association was found in both men and women, but appeared to be slightly stronger in men (for men, β = − 0.0088, 95% CI − 0.0133 to − 0.0042; for women, β = − 0.0036, 95% CI −0.0068 to − 0.0004). Figure 2B shows that increasing BMI was associated with increased FVIII levels in both men and women, the relationship again being somewhat stronger in men than in women. The APC ratio decreased with increasing FVIII levels (Fig. 2C). FVIII levels explained part of the relationship between BMI and decreased APC ratio, as the slope of the regression line of APC ratio and BMI levels decreased after adjustment for FVIII in both men and in women (for men, adjusted β = − 0.0042, 95% CI − 0.0080 to − 0.0003; for women, adjusted β = − 0.0013, 95% CI − 0.0039 to 0.0013).
Effect of increasing BMI and decreased APC ratio on the risk of VT
To examine the effect of increasing BMI on the risk of VT, separate from APC resistance caused by other factors, we restricted this analysis to individuals from LETS without FV Leiden or oral contraceptive use. In these individuals, the risk of VT increased 1.9-fold for those with a BMI in the median tertile (OR 1.9, 95% CI 1.0–2.5) and 2.2-fold for those with a BMI in the upper tertile (OR 2.2, 95% CI 1.4–3.4), as compared with individuals in the lowest tertile (Table 2). Adjustment for decreased APC ratio or FVIII levels in a logistic regression model led to a slight decrease in these relative risk estimates (Table 2).
Table 2. Risk of venous thrombosis according to body mass index (BMI) tertiles, adjusted for activated protein C (APC) resistance and factor VIII*
BMI (kg m −2)
Case n (%)
Control n (%)
OR (95% CI)
OR (95% CI) adjusted for APC
OR (95% CI) adjusted for FVIII
CI, confidence interval; OR, odds ratio. *Data from LETS, excluding individuals with FV Leiden and women who used oral contraceptives.
Lower tertile (< 24.0)
Median tertile (24.0–27.3)
Upper tertile (≥ 27.3)
Effect of the combination of increasing BMI, FV Leiden and blood group non-O on the risk of VT
Table 3 shows the combined effects of FV Leiden and blood group non-O within increasing BMI categories on the risk of VT in the LETS and TROL populations. When BMI increased, APC ratio decreased and FVIII levels increased in a dose-responsive manner in non-FV Leiden carriers with blood group O. The risk of VT was higher when they had a BMI in the upper tertile rather than in the lowest tertile (adjusted OR 1.9, 95% CI 1.2–3.1). A similar dose–response relationship of decrease in APC ratio and increase in FVIII levels with higher BMI was observed in non-FV Leiden carriers with blood group non-O. VT was modestly increased, in a dose-responsive manner, when these individuals were compared with the reference group. FV Leiden carriers with blood group O also showed a dose-responsive decrease in APC ratio and an increase in FVIII level when BMI increased. The risk of VT was strongly increased, again in a dose-responsive manner, within the BMI tertiles as compared with the reference group. Carriers of FV Leiden with blood group non-O had the highest FVIII levels and the highest risk of VT as compared with the reference group. In this subset, where all ORs were high and varied between 23.3 and 40.6, a dose–response relationship of BMI and VT risk within the BMI tertiles could no longer be observed.
Table 3. Leiden Thrombophilia Study (LETS) and TROL combined: risk of venous thrombosis according to the combinations of factor V Leiden body mass index (BMI) tertiles and blood group*
BMI (kg m−2)
Mean normalized APC ratio†
Mean FVIII (IU dL−1)†
Cases n (%)
Controls n (%)
OR (95% CI)
OR‡ (95% CI)
APC, activated protein C; CI, confidence interval; OR, odds ratio. LETS: upper tertile, BMI ≥ 26.9; median tertile, BMI ≥ 23.7 < 26.9; lowest tertile, BMI < 23.7. TROL: upper tertile, BMI ≥ 27.9; median tertile, BMI ≥ 24.9 < 27.9; lowest tertile, BMI < 24.9. *Tertiles derived from the LETS and TROL control populations, respectively. Cases and controls < 70 years of age and deep vein thrombosis cases only (pulmonary embolism excluded). †Data from LETS. ‡OR adjusted for age and sex.
In this study, we have shown that increasing BMI is related to increasing APC resistance. Part of this relationship could be explained by increased FVIII levels, as the slope of the regression line of APC ratio and BMI levels decreased after adjustment for FVIII. There was a dose–response relationship between increasing BMI and increasing VT risk. This may be partly explained by a decreased APC ratio, as ORs were reduced by adjustment for APC ratio or for FVIII (Table 2), and we also observed an interaction between APC ratio-related genes and high BMI and risk of VT (Table 3). These latter findings support evidence from other (but not all ) studies, which also showed that increasing BMI is associated with APC resistance  and elevation of FVIII levels [4,11,12,24,25].
When individuals had other causes of APC resistance, such as blood group non-O or FV Leiden, we saw a gradual decrease in APC ratio and FVIII levels with increasing BMI, and a joint effect on the VT with increasing BMI. This supports our hypothesis that FV Leiden and/or blood group non-O modify the risk of the occurrence of venous thrombosis with increasing BMI through additionally increased APC resistance. However, this finding did not apply to individuals who had both blood group non-O and FV Leiden, as these individuals had the highest risk of VT (ORs > 20) irrespective of their BMI tertile. Apparently, the combination of APC resistance resulting from FV Leiden and high FVIII levels was sufficient to generate high thrombosis risks, and BMI itself did not add further to the risk in these individuals. That FV Leiden carriers with blood group non-O are at high risk of VT has been reported previously [20,26,27], and may deserve attention in future studies of the management of primary or secondary prevention of VT in such individuals. It should be noted that the numbers supporting this reasoning, particularly in the control group, were small.
The prevalence of overweight and obesity in the western world is 50–65% [5,28]. It has been calculated that almost one-third of all VT events could be prevented by weight loss, on the assumption that weight loss reduces VT risk [5,29]. For this reason, encouragement of overweight or obese individuals to lose weight seems reasonable.
Some limitations of our data merit consideration. First, height and weight were self-reported in both LETS and in TROL. In general, thin individuals tend to overreport their body weight, whereas obese individuals tend to underreport their body weight . The actual risks would be somewhat higher if this phenomenon had occurred. Second, control individuals in LETS were introduced by the cases (friends and acquaintances). It is therefore possible that, in LETS, cases and controls had similar lifestyles, which may have resulted in similar BMIs. If this were the case, it also would have led to lower risk estimates in our study. Finally, residual confounding is always a concern in observational studies such as ours. However, the main purpose of our analyses was not to show a relationship between BMI and VT, but to consider the mechanism. In Table 2, we show that the relationship decreases towards unity after adjustment for either APC resistance or FVIII. When a factor is in the causal pathway between an exposure and the outcome, adjustment for this factor leads to attenuation of the effect, so this can be performed to test whether a factor is in the causal pathway. The results in Table 2 show that the ORs did indeed decrease after this adjustment, so it can be concluded that these factors are at least partly a result of high BMI, leading in turn to VT. Furthermore, we observed that various combinations of three variables (BMI, FV Leiden, and blood group) led to increasing APC resistance and an accompanying dose-responsive increase in risk of VT. A similar strategy has been used previously to confirm that the risk of VT in oral contraceptive users is explained by increased APC resistance . It is unlikely that a confounding factor could explain the attenuation by APC resistance and FVIII, or the very strong dose-responsive increase in risk over the different risk factor combinations that we found.
We conclude that the increased risk of VT in individuals with high BMI may be partly mediated by FVIII-induced APC resistance. The effect of BMI on VT risk was enhanced two-fold to 10-fold in FV Leiden or blood group non-O carriers as compared with non-carriers. Future studies are needed to show whether these risks can be reduced by weight loss.
S. C. Christiansen, W. M. Lijfering, A. van Hylckama Vlieg, and S. C. Cannegieter: study concept and design; S. C. Christiansen, A. van Hylckama Vlieg, I. A. Naess, and S. C. Cannegieter: acquisition of data; S. C. Christiansen, W. M. Lijfering, and S. C. Cannegieter: analysis and interpretation of data; S. C. Christiansen, W. M. Lijfering, I. A. Naess, J. Hammerstrom, A. van Hylckama Vlieg, F. R. Rosendaal, and S. C. Cannegieter: drafting of the manuscript; J. Hammerstrom, F. R. Rosendaal, and S. C. Cannegieter: critical revision of the manuscript for important intellectual content; S. C. Christiansen, W. M. Lijfering, and S. C. Cannegieter: statistical analysis; J. Hammerstrom and F. R. Rosendaal: administrative, technical or material support; J. Hammerstrom, F. R. Rosendaal, and S. C. Cannegieter: study supervision.
The LETS study was funded by grant 89.063 from the Netherlands Heart foundation. The TROL study was supported by grants from the Research Council of Norway (grant no. 148037) and the Netherlands Heart Foundation (grant no. 2000B185). W. M. Lijfering is a Postdoc of the Netherlands Heart Foundation (2011T012). The funding organizations are public institutions, and had no role in the design and conduct of the study, the collection, management, analysis and interpretation of the data, or the preparation, review or approval of the manuscript.
Disclosure of Conflict of Interests
The authors state that they have no conflict of interest.