Increasing levels of free thyroxine as a risk factor for a first venous thrombosis: a case-control study

Venous thrombosis (VT) is an important cause of morbidity and mortality in developed countries. The estimated incidence rates vary between 1 and 2 per 1000 person-years.^1,2  In the past decades, several risk factors for VT, both genetic and acquired, have been established.³  Still, in 25% to 50% of first episodes of VT no apparent risk factor can be identified.⁴  Identification of additional risk factors associated with VT will improve the understanding and prevention of this disease.

Hyperthyroidism has been associated with a hypercoagulable state and is thus hypothesized to increase the risk of VT.⁵  Although there have been several reports on sinus, cerebral, or deep VT (DVT) after thyrotoxicosis, the relation between thyroid function and the risk of VT is not fully explored.^6-10  Although high concentrations of factor VIII and von Willebrand factor contribute to a hypercoagulable state in overt hyperthyroidism, lower von Willebrand factor concentrations found in overt hypothyroidism may, at least in part, protect against VT.^5,11  Regarding these alterations in coagulation factors, similar findings have been described for subclinical thyroid disease.⁵  Subclinical thyroid disease has also been linked with arterial vascular disease, and there are good indications that variations in thyroid hormone levels within the physiologic range can modify the function of several organs.^12-15  Therefore, we hypothesized that increasing levels of thyroid hormone may be a risk factor for VT.

In a case-control design, we aimed to clarify the associations between different plasma levels of free thyroxine (FT₄), thyrotropin (TSH), thyroid peroxidase antibodies (antiTPOs) and the presence of VT. Because acute illness such as VT may in itself affect thyroid hormone concentrations by altered protein binding or by inhibition of the conversion of T₄ to triiodothyronine (T₃), T₃ levels were subsequently analyzed to explore whether our findings were influenced by this so-called nonthyroidal illness syndrome (NTIS).

Patients with objectively confirmed DVT, calf vein thrombosis, or superficial thrombophlebitis of the lower extremities and control subjects in whom leg vein thrombosis was objectively ruled out were derived from a larger study designed to investigate new risk factors for VT. In this study, all consecutive outpatients suspected of DVT and referred to the Academic Medical Center, Amsterdam, The Netherlands, between September 1999 and August 2006 were recruited (n = 944). Inpatients (n = 58), patients younger than 18 years (n = 3), patients with a previous DVT (n = 119), or patients already receiving vitamin K antagonists or heparin for more than 24 hours (n = 3) were excluded. Among the eligible patients, 7 declined to participate. A total of 754 patients were eligible for the present analysis (Figure 1).

The study was approved by the Academic Medical Center Institutional Review Board, and all patients provided written informed consent in accordance with the Declaration of Helsinki.

At presentation, all patients were asked to complete a detailed questionnaire about family and medical history, medication use, and the presence of predisposing risk factors for VT. Subsequently, venous blood was obtained in a nonfasting state. Blood was collected in 0.109 mol/L trisodium-citrated tubes and immediately centrifuged, and the supernatant was recentrifuged, for 20 minutes at 1600g at 4°C to obtain platelet-poor plasma, which was stored at −80°C.

All patients underwent routine workup for diagnosis of DVT according to an algorithmic management strategy combining clinical pretest probability and D-dimer testing (Tinaquant; Roche Diagnostics), followed by compression ultrasound (CUS) scanning of the lower extremities if indicated.^16,17  Failure to fully collapse the lumen of the deep or superficial veins during compression testing was the main criterion for the presence of VT, including DVT, calf vein thrombosis, and superficial thrombophlebitis. DVT was defined as proximal thrombosis of the iliac or superficial femoral vein, or thrombosis of at least the upper third part of the deep calf veins. Thrombosis was considered provoked if at least one of the following criteria was present: use of estrogen- or progesterone-containing agents, malignancy, pregnancy or puerperium, paralysis of the symptomatic leg, recent trauma (within past 60 days), surgery within the past 4 weeks or bedridden for more than 3 days, hospitalization within the previous 6 months, long-distance travel (≥ 6 hours) within the previous 3 months. In the absence of these acquired risk factors, VT was considered unprovoked.

Six patients with a combination of high pretest probability and a positive D-dimer assay, but normal initial ultrasonographic findings, failed to return for repeated CUS scan after 1 week. In total, 211 patients were diagnosed with VT (cases), whereas in 537 patients diagnosis of VT was objectively ruled out with the use of the above-mentioned diagnostic management strategy (controls; Figure 1).

Citrated plasma was available for 190 cases and 486 controls. Because we anticipated a control-to-case ratio of 2:1 to suffice, we randomly selected 380 of the 486 controls according to the alphabetical order of their initials. Selection was performed for men and women separately to match for frequency (Figure 1).

Levels of FT₄, T₃, TSH, and antiTPO were assessed in citrated plasma with the use of commercially available assays (ADVIA Centaur immunoassay system; Siemens Healthcare Diagnostics). The intraassay and interassay coefficients of variations (CVs) were below or equal to 4.7% and 4.6% for FT₄, 3.2% and 1.3% for T₃, 9.0% and 4.4% for TSH, and 4.1% and 8.0% for antiTPO, respectively. Because these tests have not been validated by the manufacturer for use with citrated plasma, serum/citrated plasma studies were performed to characterize the correlation. Results in plasma were corrected for dilution with citrate. On the basis of the analysis of at least 21 samples, small systematic differences between serum and citrated plasma were observed, but linear regression analysis showed a significant association between serum and plasma levels of thyroid hormones (regression coefficient β ≥ 0.92; P < .01).

Overt and subclinical primary hyperthyroidism were biochemically defined as TSH below the lower limit of the local reference range (0.32-4.32 mU/L) combined with a FT₄ concentration above or within the local reference range (10-24 pmol/L), respectively.^18,19  Primary hypothyroidism, overt and subclinical, was biochemically defined as elevated TSH and a FT₄ concentration below or within the local reference range, respectively.^19,20  AntiTPOs above 80 U/mL were considered elevated.

In one plasma sample, thyroid hormones could not be assessed because of insufficient volume. After exclusion of this control subject, FT₄, TSH, and antiTPO levels of 190 cases and 379 controls were used for final analysis (Figure 1). Because levels of T₃ were analyzed only after the initial measurements of FT₄, TSH, and antiTPO had been performed, T₃ levels were available for 186 cases and 375 control subjects; in 8 plasma samples the volume was insufficient for this additional measurement. Subjects taking thyroid medication or with known thyroid dysfunction were not excluded from the analyses.

Categorical variables measured in this study were expressed as number and percentage. FT₄ and T₃ levels were normally distributed in both cases and controls and were presented as means (95% confidence interval [95% CI]). Distributions of TSH and antiTPO levels were skewed and therefore presented as medians (95% CIs). Between-group comparisons were performed with the use of t tests or nonparametric tests, depending on the distribution of the data. Odds ratios (ORs) and 95% CIs for the risk of VT at different levels of FT₄, TSH, and antiTPO were calculated with the use of binary logistic regression, taking different percentiles of the values observed in the control subjects as cutoff levels. For each cutoff level below the 50th percentile, we compared subjects below the cutoff to subjects above this level with the use of the latter as reference, and vice versa for cutoff levels above the 50th percentile. Subsequently, the same analyses were performed for levels of T₃, but mostly to support any findings on FT₄ levels. A multivariate model was used to adjust for sex to take the frequency matching on sex into account and for age as a possible confounding factor. In the analysis for TSH, we additionally adjusted for FT₄ to explore the causal relation. Separate analysis was performed for patients with DVT, excluding those with calf vein thrombosis or superficial thrombophlebitis. Chi-square tests, or Fisher exact test in case cells had a count less than 5, were used to compare the 2 study groups with respect to the presence of thyroid abnormalities in comparison with euthyroidism. Statistical analysis was performed with the use of SPSS 15.0 software package (SPSS Inc).

Among the 190 cases and 379 controls there were 80 (42%) men in the cases and 160 (42%) in the control subjects. Mean age in the cases (57 years; range, 19-90 years) was similar to that in the control subjects (56 years; range, 18-93 years). Of all patients with VT, 155 (82%) had a DVT, 12 (6%) had an isolated calf vein thrombosis, and 23 (12%) had a superficial thrombophlebitis of the lower extremities. In 69 (36%) VT cases, and in 51 (33%) of the patients with DVT, VT was considered unprovoked.

At inclusion, 5 patients with provoked DVT, 1 with thrombophlebitis, and 7 control subjects were on thyroxine substitution therapy for previously diagnosed hypothyroidism. Only 1 patient, diagnosed with thrombophlebitis, was on thiamazole treatment for known hyperthyroidism. All patients with known hypothyroidism or hyperthyroidism had plasma FT₄ levels within the reference range combined with either TSH levels within reference range or marginally decreased or marginally elevated TSH levels at the time of inclusion.

Baseline characteristics for cases and controls are summarized in Table 1.

Mean FT₄ in all patients with VT was 16.0 pmol/L (95% CI, 15.6-16.4 pmol/L), in patients with DVT it was 16.2 pmol/L (95% CI, 15.8-16.7 pmol/L), and in control subjects it was 15.4 pmol/L (95% CI, 15.1-15.6 pmol/L; Table 2). We found ORs for each cutoff level that clearly increased with the percentiles used (adjusted for the matching factor sex). These ORs were below unity for cutoff levels below the 50th percentile (ie, indicating a protective effect of lower FT₄ levels) but above unity for higher cutoff levels (ie, indicating an increased risk of higher FT₄ levels). Further adjustment for age slightly reduced the crude ORs. Similar results were found when we analyzed only patients with DVT of the leg. In the absence of traditional acquired risk factors for VT, FT₄ above reference range (> 24 pmol/L) yielded an OR of 9.6 (95% CI, 0.8-109.1) for VT and 13.0 (95% CI, 1.1-154.1) for DVT, adjusted for sex and age (Figure 2; Table 3).

Mean T₃ levels were higher in patients with VT (1.94 nmol/L; 95% CI, 1.87-2.00 nmol/L) than in control subjects (1.79 nmol/L; 95% CI, 1.75-1.82 nmol/L), indicating that the association between higher FT₄ and VT as found in the present study is very unlikely to reflect the NTIS because low serum T₃ is the hallmark of NTIS (Table 2). Similar to FT₄, we found the risk of VT to gradually rise with increasing levels of T₃, although the relation was slightly less linear (Figure 2).

Median TSH in all patients with VT was 1.41 mU/L (95% CI, 1.22-1.51 mU/L), in those with DVT it was 1.37 mU/L (95% CI, 1.17-1.51 mU/L), and in control subjects it was 1.21 mU/L (95% CI, 1.11-1.30 mU/L; Table 2). TSH levels were not associated with the overall risk of VT. The risk of unprovoked DVT was almost 3-fold increased for TSH below 0.02 mU/L (OR, 2.9; 95% CI, 0.7-12.0) but returned to unity after adjustment for FT₄ (OR, 0.9; 95% CI, 0.2-5.2; data not shown).

Median antiTPO in all patients with VT was 29.0 U/mL (95% CI, 27.3-30.8 U/mL), in those with DVT it was 28.9 U/mL (95% CI, 27.2-30.7 U/mL), and in control subjects it was 29.8 U/mL (95% CI, 29.0-30.7 U/mL; Table 2). No association between different levels of antiTPO and VT was observed (data not shown).

Three patients with DVT had thyroid hormone concentrations consistent with primary hyperthyroidism (1.6% of all patients with VT; 1.9% of patients with DVT only), whereas this was not observed in any of the control subjects. Statistical analysis confirmed that hyperthyroidism and (deep) VT co-occurred more frequently than expected by chance (P = .04, all patients with VT; P = .02, patients with DVT only, 2-sided Fisher exact test). No such association was found for subclinical hyperthyroidism or for overt or subclinical primary hypothyroidism (Table 2).

All 3 patients with biochemical hyperthyroidism appeared to have clinically manifest hyperthyroidism during follow-up that was diagnosed 2, 54, and 65 months after presentation for DVT, respectively. For a description of these patients, see supplemental Document 1 (available on the Blood Web site; see the Supplemental Materials link at the top of the online article).

Using laboratory assessment of thyroid function in 190 patients with confirmed VT of the lower extremities and 379 control subjects, we were able to show a clear gradual relation between plasma FT₄ levels and the risk of VT. Notably, the thrombotic risk was substantially increased for FT₄ levels well within the physiologic range. Moreover, FT₄ levels were particularly associated with the risk of unprovoked DVT, indicating FT₄ as a potential novel risk factor. Similar to FT₄, the risk of VT also increased with higher levels of T₃, but the relation was slightly less linear. No clear association was found for TSH or antiTPO.

The present analysis is to our knowledge the first to study the effect of increasing levels of thyroid hormones on the risk of VT. Over the past decades we have witnessed a multiplicity of genetic or acquired risk factors for VT to be identified. The estimated magnitude of each varies widely. To illustrate, thrombosis risk after surgery is 6- to 15-fold increased; hospitalization is associated with an 8- to 11-fold increased risk, whereas in pregnancy the VT risk is 1- to 5-fold increased, in puerperium 14- to 60-fold, in users of oral contraceptives 1.4- to 5-fold, and in heterozygotes of factor V Leiden 3- to 8-fold increased.^21-23  Generally, newly identified genetic or acquired risk factors tend to confer less thrombotic risk than the well-established ones. Although confirmation is required, the here reported risk of increasing FT₄ levels, varying from 1.7- to 13-fold, appears to be of a similar magnitude.

It is probable that the risk of VT associated with FT₄ levels reflects thyroid hormone-induced alterations in factor VIII synthesis and secretion, high concentrations of which are an independent risk factor for VT, and possibly other procoagulant yet unknown changes.²⁴  Indeed, our findings corroborate clinical data suggesting both high concentrations of factor VIII in thyroid hormone excess and low factor VIII concentrations in overt hypothyroidism.^5,11  However, the reduced risk of VT for lower FT₄ levels, yet still within the physiologic range, is in contrast with some studies, including one of our own, that suggest the hypercoagulable state to extend into subclinical hypothyroidism.^25-27  In a previous pilot study, we found an increased frequency of subclinical hypothyroidism in patients with unprovoked DVT (OR, 6.8; 95% CI, 0.7-64.5). The different outcome is best explained by both the limited sample size and the fact that, in this pilot study, thyroid function was assessed only several months after the thrombotic event.²⁷  Although in the present study we did not observe any association between biochemical diagnosis of subclinical thyroid disease and thrombotic risk, it is now clearly shown that FT₄ levels and risk of VT constitute a continuum.

Putative mechanisms by which thyroid hormone may influence coagulation proteins are ill defined, yet it is most likely that it does so by thyroid hormone receptor–mediated regulation of gene transcription at the hepatic or endothelial cell level, or both.^28,29  Other hypotheses include indirect effects mediated through β-adrenergic receptor function.^30,31 

VT is a multicausal disease in which a combination of more than one genetic or acquired risk factor is needed to pass the thrombosis threshold.³  Thus, the clinical utility of our findings may be of particular relevance for subjects with additional risk factors for VT, eg, women on oral contraceptive agents or patients undergoing surgery. If confirmed, screening for thyroid function in high-risk patients may improve our ability to predict and prevent this disease. For patients with VT, knowledge of thyroid function could be of importance in decisions about the duration of anticoagulant treatment in those 25% to 50% in whom no additional risk factor is identified. In addition, screening for thyroid function in patients presenting with a new VT could help us to early detect thyroid disease, especially because clinical symptoms and signs of (subclinical) hyperthyroidism might easily go unnoticed during the early years of the disease. For patients with overt hyperthyroidism, physicians should be aware that these patients might be at increased risk of VT.

Several potential limitations should be addressed. First, the present analysis was limited to a population suspected of DVT. As mentioned, acute or chronic disease is known to affect thyroid function, and multiple alterations in thyroid hormone levels have been observed in patients with this NTIS.³²  The most common and earliest change is inhibition of T₄-to-T₃ conversion, with a resulting decrease in the circulating T₃ level. In general, levels of FT₄ are usually less affected by NTIS than T₃, but values may be higher than normal in mild or moderate forms of the syndrome.^32,33  T₄ is present in the circulation either free or bound to thyroxine-binding globulin, thyroxine-binding prealbumin, or albumin. As such, changes in protein binding probably decrease the total amount of T₄ (free and bound), whereas levels of the unbound hormone may increase. If, in the present study, decreased protein binding was more present in the patients with VT than in control subjects, this would have resulted in higher FT₄ levels in cases compared with controls and, thus, to an overestimation of our effect measures. However, to discriminate between NTIS (low T₃) or a direct association between high FT₄ (and T₃) and the risk of VT, we subsequently performed additional assessment of T₃ levels. Although a certain degree of NTIS was present in some patients with VT, or in some control subjects (eg, those with erysipelas), the higher T₃ levels in cases compared with controls, as well as the clear association between T₃ and VT, make it highly unlikely that our findings are solely a reflection of NTIS. Nonetheless, prospective epidemiologic studies are needed to further confirm our findings.

Second, the design of our study that included blood sampling in the acute phase of VT did not allow us to investigate the relationship between thyroid hormone and factor VIII, because factor VIII levels measured during the acute phase are unlikely to be representative for levels before the event: increased consumption of factor VIII has been reported during coagulation activation as well as up-regulation of factor VIII synthesis during the acute phase response.³⁴  Therefore, in the present study we cannot confirm thyroid hormone–induced alterations in factor VIII concentrations.

Third, not all patients suspected of DVT of the lower extremities underwent CUS scanning to exclude VT. Nevertheless, because the diagnostic management strategy combining clinical pretest probability and D-dimer testing has repeatedly proven to be equally safe in refuting the diagnosis of VT, mismatching seems unlikely.^17,35,36 

In conclusion, our data suggest that increasing levels of FT₄ are a risk factor for VT, whereas lower FT₄ levels are protective of VT. Future studies are needed to further broaden our knowledge on this issue and to evaluate whether implementation of this novel risk factor in daily practice can improve our ability to prevent and manage VT in terms of reducing its risks. In particular, epidemiologic studies in which thyroid hormone levels are measured before or after the thrombotic event may provide definitive answers on the influence of NTIS. Furthermore, it would be interesting to investigate whether patients with VT with higher FT₄ levels also have an increased risk of recurrent thrombosis, or, vice versa, whether subjects who have been treated for hyperthyroidism have a lower risk. In addition, further insights in the clinical relevance of our findings could be obtained by assessing the relation between thyroid hormone levels and VT in high-risk patients, such as those undergoing surgery.

We thank the “vasculisten,” a group of medical students at Academic Medical Center, University of Amsterdam, who had an essential role in the execution of this study, especially the recruitment and data collection. We thank Huib Bout and Jan van Veen, Department of Clinical Biochemistry, Slotervaart Hospital (Amsterdam, The Netherlands) for their laboratory efforts.

Contribution: B.v.Z. had full access to all of the data in the study and takes responsibility for the integrity of the data, the accuracy of the data analysis, and the decision to submit for publication. B.v.Z. helped to obtain data, perform statistical analysis, and write the report; A.S. helped to design the study and participated in writing the report; R.H. helped to design and coordinate the study, recruit participants, obtain data, and write the report; A.P.v.Z. performed laboratory analysis of thyroid function and participated in writing the report; E.F. participated in writing the report; S.C.C. helped to perform statistical analysis and write the report; H.R.B. helped to design and coordinate the study and write the report; V.E.A.G. helped to design the study, perform statistical analysis, and write the report; and D.P.M.B. participated in designing the study and writing the report.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: B. van Zaane, Department of Internal Medicine, Slotervaart Hospital, Louwesweg 6, 1066 EC, Amsterdam, The Netherlands; e-mail: b.vanzaane@amc.uva.nl.