A. van Hylckama Vlieg, Albinusdreef 2, 2333 ZA Leiden, the Netherlands. Tel.: +31 71 526 1562; fax: +31 71 526 6994. E-mail: email@example.com
Summary. Background: Oral contraceptive use increases the risk of venous thrombosis as well as sex hormone-binding globulin (SHBG) levels. Furthermore, increased SHBG levels are positively associated with activated protein C (APC) resistance and thrombotic risk in oral contraceptive users.
Objectives: To determine whether increased SHBG levels are causally related to venous thrombosis in women not using hormonal contraceptives.
Methods: Premenopausal women were selected from a case–control study on venous thrombosis, the Multiple Environmental and Genetic Assessment of risk factors for venous thrombosis (MEGA) study (23 patients; 258 controls). Women using hormonal contraceptives were excluded. First, the risk of venous thrombosis with SHBG levels above the normal reference range (70 nm) was determined. Second, because multiple regulatory factors affect SHBG levels and residual confounding may remain, we determined six single-nucleotide polymorphisms (SNPs) in the SHBG gene and assessed the risk of venous thrombosis in a different case–control study, the Leiden Thrombophilia Study (LETS) (20 patients; 74 controls), and in the MEGA study. Finally, the association between SHBG levels and the normalized activated partial thromboplastin time-based APC resistance (an intermediate endpoint for venous thrombosis) was determined.
Results: Elevated SHBG levels (> 70.0 nm) were associated with venous thrombosis (odds ratio 1.92; 95% confidence interval [CI] 0.74–5.00). However, this finding can be explained by residual confounding. Two SNPs in the SHBG gene affected SHBG levels, but not venous thrombosis risk. Furthermore, SHBG levels in controls were not associated with APC resistance (SHBG level, > 70.0 vs. ≤ 70.0 nm: mean difference in normalized APC sensitivity ratio, 0.03; 95% CI −0.05 to 0.10). Exclusion of women with FV Leiden did not materially change these results.
Conclusions: Increased SHBG levels are not causally related to the risk of venous thrombosis.
Venous thrombosis is the formation of a blood clot in the veins, predominantly in the legs. The overall age-dependent incidence is 1–3 per 1000 persons per year , whereas the incidence in women of reproductive age is estimated to be 5–10 per 10 000 women-years . Both genetic and acquired risk factors are known to influence the risk of venous thrombosis. An important acquired risk factor is the use of hormonal contraceptives in women . The use of combined oral contraceptives (COCs) is associated with a four-fold to six-fold increased risk of venous thrombosis [4–7]. The risk of venous thrombosis is higher in so-called third-generation progestagen (e.g. desogestrel and gestodene) COC users than in second-generation (e.g. levonorgestrel) COC users [8–11].
Results from several previous studies have suggested that the effect of a COC on sex hormone-binding globulin (SHBG) levels reflects the risk of venous thrombosis. SHBG is a plasma glycoprotein that binds the sex steroid hormones testosterone and 17β-estradiol. Plasma SHBG is primarily produced in hepatocytes. In contraceptive users, estrogens such as ethinylestradiol increase the production of SHBG [12,13], whereas progestagens induce a decrease in SHBG levels, depending on the type and dose used [14,15]. Therefore, the effect of COCs on SHBG levels can be seen as the sum of the stimulatory effect of ethinylestradiol and the inhibitory effect of the progestagen, resulting in the total estrogenicity of the pill [15,16]. This so-called total estrogenicity of a COC crudely correlates with the risk of venous thrombosis, in the sense that levenorgestrel-containing COCs have a lower associated risk and lower SHBG levels than third-generation pills containing desogestrel or gestodene [15,17–20]. Furthermore, SHBG levels in COC users were positively associated with thrombin generation-based activated protein C (APC) resistance [16,21]. APC resistance is the relative inability of protein C to cleave factor Va or FVIIIa, leading to a prothrombotic state. APC resistance has been shown to predict venous thrombosis risk in both men and women [22,23].
In addition to COCs, many environmental risk factors affect SHBG levels, such as age , obesity [25,26], diabetes , liver diseases [28,29], and hyperthyroidism [30,31]. Regarding genetic variation in the SHBG gene, the single-nucleotide polymorphisms (SNPs) rs13894 (C/T) and rs727428 (G/A) decrease SHBG levels with an increasing number of minor alleles [32–34]. Minor alleles of SNP rs6259 (Asp356Asn) were associated with increasing SHBG levels. Furthermore, the combination of the SNPs rs6259 (Asp356Asn), rs858521 (C/G) and rs727428 (G/A) accounted for 24% of the variation in SHBG levels in postmenopausal women .
Although an increased SHBG level in oral contraceptive users is a marker for the risk of venous thrombosis, the question remains of whether increased SHBG levels are a risk factor for venous thrombosis in a causal sense. The aim of this study was three-fold. First, the risk of venous thrombosis associated with SHBG levels was evaluated in non-contraceptive users. Second, to eliminate the influence of residual confounding, the effect on thrombotic risk of genetic variants in the SHBG gene that affect SHBG levels was assessed. A similar approach was used to study genetic variation in the SHBG gene and SHBG levels in association with diabetes . Third, we investigated the association between SHBG levels and APC resistance, which is an established intermediate endpoint for venous thrombosis.
Materials and methods
Participants were selected from two large case–control studies, i.e. the Leiden Thrombophilia Study (LETS) and the Multiple Environmental and Genetic Assessment of risk factors for venous thrombosis (MEGA) study. In the LETS, participants with a first episode of venous thrombosis in the leg, younger than 70 years and without a known malignant disorder, were enrolled between 1 January 1988 and 31 December 1992. As controls, friends and partners of the patients were asked to participate. Details of the study have been described elsewhere . In the MEGA study, participants with a first venous thrombosis in the leg or arm or pulmonary embolism were recruited between 1 March 1999 and 31 August 2004. Controls were either the partners of the patients or recruited through random digit dialing. Details of the study have been described elsewhere . Both studies included objectively verified venous thrombotic events. In both the LETS and the MEGA study, participants were asked to fill in a questionnaire within a few weeks after the thrombotic event, and subsequently to provide a blood or buccal swab sample 3 months after discontinuation of anticoagulant therapy. The LETS and the MEGA study differed slightly in their inclusion and exclusion criteria. To make both studies comparable, patients with venous thrombosis in the arm or pulmonary embolism were excluded from the MEGA study.
The population of interest consisted of premenopausal women with or without a first venous thrombosis event (nLETS = 337; nMEGA = 2657). For the current study, only idiopathic venous events were selected, so we excluded women who had any type of cancer (nMEGA = 63), were hospitalized (nLETS = 22; nMEGA = 357), had undergone surgery (nLETS = 29; nMEGA = 277), or had had bone fractures (nLETS = 2; nMEGA = 81) or injuries (nMEGA = 529) in the 12 months before the event. Furthermore, women who were pregnant (nLETS = 10; nMEGA = 65) or postpartum (nLETS = 4; nMEGA = 17), had a miscarriage (nLETS = 1; nMEGA = 10), used hormone replacement therapy (nMEGA = 14) or used hormonal contraceptives (nLETS = 213; nMEGA = 1013) in the 12 months before the event were excluded. Totals of 94 and 385 women were included from the LETS and the MEGA study, respectively. For these women, DNA was available, through either a blood sample or a buccal swab sample. Plasma was required for SHBG measurement; therefore, women with a buccal swab sample were excluded (nMEGA = 104). The amount of plasma left in the LETS was insufficient for measurement of SHBG levels.
Data regarding age and body mass index (BMI) were retrieved from the questionnaire. The BMI (kg m−2) of these women was calculated from their reported weight and height. For the association with SHBG levels in controls, age and BMI were divided into three categories, i.e. for age into ≤ 30, 30–40 and 40–50 years, and for BMI into normal (≤ 25 kg m−2), overweight (25–30 kg m−2), and obese (>30 kg m−2).
DNA preparation and SNP typing
Blood samples were taken at least 3 months after discontinuation of anticoagulant therapy. Blood was drawn after an overnight abstinence from intake of food, caffeine, and alcohol, and collected into vacuum tubes containing 0.106 m trisodium citrate as anticoagulant. Blood was centrifuged to retrieve cell-free, citrated plasma. Processing of blood samples and subsequent DNA isolation have been described previously [35,36].
To determine the haplotypes in the SHBG gene, the Genome Variation Server (GVS)  was used. The GVS incorporates information from HapMap, and is sponsored by SeattleSNPs. The SHBG gene showed six haplotypes, hA to hF (the frequencies in a European population of northern and western ancestry [CEU] according to HapMap data were 11%, 14%, 22%, 14%, 29%, and 10%, respectively). Only SNPs with a minor allele frequency of ≥ 5% were considered. The following haplotype-tagging SNPs were selected: rs13894, rs6259, rs8066665, rs2955617, rs858521, and rs727428. The combination of these six SNPs led to six haplotypes in the SHBG gene (Table S1). For four of these, an effect of SHBG levels was reported previously (we found no reports for SNPs rs8066665 and rs2955617). As the total number of known SNPs in the SHBG gene is relatively low, the selected SNPs were the only ones available for discrimination between the different haplotypes.
The SNPs were determined with the MassARRAY platform (Sequenom, San Diego, CA, USA), according to the manufacturer’s protocols (Sequenom). Genotyping determination was performed blinded to the case–control status. Five per cent of the samples were repeated for allele-calling consistency; no discrepancies were found.
As the amount of plasma left from the LETS samples was not sufficient for an SHBG measurement, SHBG levels were measured only in samples from the MEGA study (n = 281). SHBG levels (nm) were measured with an immunometric assay (Immulite; Siemens Healthcare Diagnostics, Tarrytown, NY, USA). The sensitivity is 0.2 nm, and the assay has a log-term variation of 6% at levels of both 5 nm and 80 nm. The within-assay variation is 3–4%, and the between-assay variation is 3.5–6%. The samples were analyzed in a single series in random order. SHBG levels were measured without knowledge of any of the participant’s characteristics.
APC resistance was determined in samples from the MEGA study. APC resistance was measured with Cephotest (Nycomed Pharma, Oslo, Norway). The normalized APC sensitivity ratio (nAPCsr) was defined as the activated partial thromboplastin time (APTT) in the presence of APC divided by the APTT in the absence of APC in participants divided by the same ratio determined in normal pool samples, i.e. (APTT + APCparticipants/APTT − APCparticipants) / (APTT + APCnormalpool/APTT − APCnormal pool).
First, the association between SHBG levels and the risk of venous thrombosis was assessed in the MEGA study. SHBG levels were dichotomized with a cut-off value of 70 nm SHBG, which is above the normal reference range. With logistic regression analysis with robust standard errors (SEs), the risk of venous thrombosis associated with SHBG levels of > 70 nm as compared with ≤ 70 nm was assessed by calculating odds ratios (ORs) with 95% confidence intervals (CIs) with adjustment for age and BMI.
Second, the relationship between genetic variation in the SHBG gene and the risk of venous thrombosis was studied in the LETS and repeated in the MEGA study. For SNPs, Hardy–Weinberg equilibrium was assessed with a chi-squared test in controls. Regarding haplotypes, the posterior probabilities of the individual haplotype combinations as estimated by plink (version 1.07)  were used as weights in the statistical analyses. No underlying genetic model was assumed for the SNPs or haplotypes (i.e., SNPs and haplotypes were defined categorically in the regression model). Linear regression analysis with robust SEs was used to determine the effect of a SNP or haplotype in the SHBG gene on SHBG levels in controls from the MEGA study. To determine the presence of a relevant effect of the genetic variation in the SHBG gene on SHBG levels, we used a ≥ 20 nm difference between carriers of two copies of a minor allele or haplotype and non-carriers of the given allele or haplotype. This value was based on the minimal difference in SHBG levels between users of a second-generation oral contraceptive (who have the lowest levels of all pill-users) and non-users. Logistic regression with robust SEs was used to estimate the relative risk of venous thrombosis associated with different SNPs or haplotypes. The risks were determined in the LETS and replicated in the MEGA study.
Finally, the association between SHBG levels and nAPCsr was determined in the MEGA study. The association between nAPCsr and risk of venous thrombosis was assessed by calculating the mean difference in nAPCsr between SHBG levels of > 70 nm and ≤70 nm. The 95% CI was calculated with a robust SE. The analysis was repeated without women with FV Leiden, which is known to lead to APC resistance.
Statistical analyses were performed with stata, version 12.0 (StataCorp LP, College Station, TX, USA).
From the LETS, 20 patients and 74 controls were included. A total of 23 patients and 258 controls were included from the MEGA study. Baseline characteristics of the participants from the LETS and the MEGA study are shown in Table 1. In the MEGA study, the median SHBG level in the patients was 55.6 nm (interquartile range [IQR] 37.0, range 21.2–458.2 nm) and that in controls was 58.4 nm (IQR 37.2, range 16.1–524.3 nm). The influence of age and BMI on SHBG levels was assessed in the controls from the MEGA study. SHBG levels were higher in women aged 30–40 years than in women aged ≤ 30 years (mean difference, 20.6 nm; 95% CI −2.4 to 43.6). However, levels did not increase much with age after this: in women aged 40–50 years, SHBG levels were only 6.9 nm higher than in women aged ≤ 30 years (95% CI −15.3 to 29.2). SHBG levels were lower in obese women (BMI of > 30 kg m−2) than in women with normal weight (BMI of ≤ 25 kg m−2) (mean difference, 28.6 nm; 95% CI 9.1–48.1). There was no difference in SHBG level between women with a BMI of 25–30 kg m−2 and those with a BMI of ≤ 25 kg m−2 (mean difference, 2.5 nm; 95% CI −11.2 to 16.3).
Table 1. Baseline characteristics from the LETS and MEGA study
LETS (n = 94)
MEGA study (n = 385)
Cases (n = 20)
Controls (n = 74)
Cases (n = 29)
Controls (n = 356)
BMI, body mass index; IQR, interquartile range; LETS, Leiden Thrombophilia Study; MEGA, Multiple Environmental and Genetic Assessment of risk factors for venous thrombosis; SD, standard deviation; SHBG, sex hormone-binding globulin. *No data on ancestry were available for nine women. †No data on weight or height were available for 18 women. ‡Measured in 281 women (23 cases; 258 controls).
Age (years), mean (SD)
Caucasian*, no. (%)
BMI (kg m−2)†, mean (SD)
SHBG levels (nm)‡, median (IQR)
Nine cases (39%) and 90 controls (35%) from the MEGA study had SHBG levels above the normal reference range (> 70 nm) (Table 2). After adjustment for age and BMI, SHBG levels > 70 nm were associated with a 1.9-fold increased risk of venous thrombosis (OR 1.92; 95% CI 0.74–5.00).
Table 2. Sex hormone-binding globulin (SHBG) levels and the risk of venous thrombosis in the Multiple Environmental and Genetic Assessment of risk factors for venous thrombosis (MEGA) study
SHBG level (nm)
Cases, n (%)
Controls, n (%)
OR (95% CI)
Adjusted* 95% CI
CI, confidence interval; OR, odds ratio. *Adjusted for age and body mass index.
To assess the association between SHBG levels and venous thrombosis, a total of six SNPs were selected in the SHBG gene, tagging six haplotypes (Table S1). In both the LETS and the MEGA study, all SNPs were found to be in Hardy–Weinberg equilibrium (P > 0.05) as measured in the controls. The effect of genetic variation in the SHBG gene on SHBG levels was substantial, making these SNPs informative for a ‘Mendelian randomization’ analysis. The predefined increase or decrease of 20 nm is substantial relative to the SHBG levels of the majority of the women in this study. In our population, 95% of women had SHBG levels of ≤ 128.5 nm.
Homozygosity for the minor allele (genotype TT) of SNP rs13894 was associated with a decrease in SHBG levels of 50 nm as compared with homozygosity for the major allele (genotype CC). However, this was based on only one control with genotype TT (Fig. 1; Table 3). The G allele of SNP rs2955617 and the A allele of SNP rs727428 affected SHBG levels (Fig. 1; Table 3). Genotype GG of SNP rs2955617 decreased SHBG levels by 20.7 nm (95% CI −8.2 to 33.1) as compared with genotype TT. The same decrease was observed for genotype AA of SNP rs727428 (20.9 nm; 95% CI 4.9–36.8). For SNP rs727428, a linear association was observed; with each increase in the number of minor alleles, the SHBG level increased by 9.9 nm (95% CI 1.0–18.8).
Table 3. Effect of single-nucleotide polymorphisms (SNPs) in the sex hormone-binding globulin (SHBG) gene on SHBG levels (nm)
Mean SHBG levels (95% CI)
Mean difference (95% CI)
CI, confidence interval.
15.3 (−15.0 to 45.6)
−49.8 (−55.9 to – 43.7)
−11.1 (−20.8 to – 1.3)
−15.5 (−32.7 to 1.6)
6.6 (−7.3 to 20.4)
1.2 (−15.2 to 17.5)
1.1 (−12.9 to 15.2)
−20.7 (−33.1 to – 8.2)
11.9 (−1.4 to 25.1)
−2.7 (−14.0 to 8.6)
−8.4 (−23.8 to 7.0)
−20.9 (−36.8 to – 4.9)
None of the cases in the LETS was homozygous for the minor allele of SNP rs2955617 (genotype GG); therefore, no risk could be calculated. In the MEGA study, genotype GG of SNP rs2955617 was not associated with a risk of venous thrombosis (OR 1.32; 95% CI 0.34–5.12) (Table 4). In the LETS, a decrease in the risk of venous thrombosis was observed in women with genotype AA of SNP rs727428 (OR 0.24; 95% CI 0.03–2.19); however, this risk was not confirmed in the MEGA study (OR 1.05; 95% CI 0.33–3.33) (Table 4). Although we observed a linear trend in SHBG levels with an increasing number of minor alleles of this SNP, no association was observed with the risk of venous thrombosis in either the LETS or the MEGA study. Haplotype analysis gave the same results (data not shown).
Table 4. Effect of single-nucleotide polymorphisms (SNPs) in the sex hormone-binding globulin (SHBG) gene on venous thrombosis risk
Cases, n (%) (N = 20)
Controls, n (%) (N = 74)
LETS, OR (95% CI)
Cases, n (%) (N = 29)
Controls, n (%) (N = 356)
MEGA, OR (95% CI)
CI, confidence interval; LETS, Leiden Thrombophilia Study; MEGA, Multiple Environmental and Genetic Assessment of risk factors for venous thrombosis; OR, odds ratio.
Finally, the association between SHBG levels and APC resistance was evaluated in the MEGA study. SHBG levels of > 70 nm were associated with a 0.03 (95% CI −0.05 to 0.102) increase in nAPCsr as compared with levels of ≤ 70 nm. Adjustment for age and BMI did not alter the result (mean difference between SHBG levels of > 70 vs. ≤70 nm in nAPCsr, 0.02; 95% CI −0.06 to 0.10). Exclusion of women with FV Leiden did not materially change the results (mean difference of 0.03, 95% CI −0.04 to 0.10; and adjusted for age and BMI, mean difference of 0.03, 95% CI −0.05 to 0.11).
We set out to study whether a high SHBG level is causally associated with the risk of venous thrombosis. First, we showed that there was a mild increase in risk associated with SHBG levels above normal (i.e. 70.0 nm) after adjustment for age and BMI in non-users of hormonal contraceptives (OR 1.92; 95% CI 0.74–5.00). However, as SHBG levels are affected by many regulatory factors, residual confounding may remain. Therefore, we performed a Mendelian randomization analysis: here, genetic variants are used that are associated with levels that, by definition, cannot have been affected by potential confounding factors. We showed that several SNPs were associated with SHBG levels, but not with thrombotic risk. Finally, no association could be found in non-users of hormonal contraceptives between SHBG levels and APC resistance, an established intermediate endpoint for venous thrombosis.
Our results in non-users are in contrast with the results observed in oral contraceptive users, where an increase in SHBG levels is associated with an increase in nAPCsr (endogenous thrombin potential-based) . Apparently, both SHBG levels and nAPCsr are affected by oral contraceptive use (Fig. 2A). In non-users, there is no common factor that influences both SHBG levels and APC resistance (Fig. 2B).
Four of the SNPs described in our study had previously been associated with SHBG levels, i.e. SNPs rs13894, rs6259, rs858521, and rs727428. In contrast to the current results, SNP rs6259 was previously associated with an increase in SHBG levels [32,33]. This difference may be explained by a difference in study population. We included premenopausal women, whereas previous studies included only women with hirsutism or postmenopausal women. As SNP rs13894 was present in only one woman, no effect on SHBG levels could be demonstrated. SNPs rs858521 and rs727428 were associated with a decrease in SHBG levels in our study, which was also reported in two other studies [33,34].
A limitation of our study is the small number of patients included. An explanation is that we restricted our study population to women who did not use hormonal contraceptives at the time of thrombosis. The strength of our study is that we used a three-pronged approach to evaluate a possible association between SHBG levels and venous thrombosis. All analyses consistently showed no association between SHBG levels and venous thrombosis, strengthening the conclusion that SHBG levels are not associated with the risk of venous thrombosis.
Although an association with venous thrombosis for the highest levels of SHBG cannot be excluded, SHBG levels within the range observed in this study are not causally related to an increased risk of venous thrombosis. This does not imply that SHBG level may not be a marker for venous thrombosis in oral contraceptive users, which in all likelihood it is, but that the level is only a marker, and not a cause. The situation is different for APC resistance, which is an intermediate, i.e. both a marker and a cause. The explanation is that APC resistance is a global read-out of the coagulation system, whereas SHBG level is not.
This study was funded by a grant from the Netherlands Organization for Scientific Research (NWO-TOP Grant no. 40-00812-98-07-045).
Disclosure of Conflict of Interests
The authors state that they have no conflict of interest.