Homocysteine Is Associated with Future Venous Thromboembolism in Two Prospective Cohorts of Women

Abstract

Objective

Case-control studies have identified plasma homocysteine as a risk marker for venous thromboembolism (VTE). Prospective data, particularly among women, are sparse. We examined whether plasma homocysteine associates with incident VTE in two large prospective cohorts of women.

Approach and Results

In the Women’s Health Study (WHS), a prospective cohort study of 27,555 women ≥45 years old and free of cardiovascular disease and VTE, we assessed baseline homocysteine concentration along with other thrombotic biomarkers for association with future VTE (n=743), pulmonary embolism (PE) (n=363), and deep vein thrombosis (DVT) (n=545). We used a second cohort of 2,672 women (n=102 VTE events) in the Women’s Antioxidant and Folic Acid Cardiovascular Study (WAFACS) to corroborate our findings.

In age-adjusted analyses, elevated homocysteine, hsCRP, fibrinogen, and sICAM-1 were associated with incident VTE (p for extreme quartile comparisons and p-trend <0.05). In multivariable models adjusting for body mass index (BMI) and other traditional VTE risk factors, only the association for homocysteine persisted (HR_Q4 1.31; 95% CI, 1.06 to 1.63). Elevated homocysteine levels were associated with unprovoked PE (HR_Q4 2.13; 95% CI, 1.30 to 3.51) and DVT (HR_Q4 1.59; 95% CI, 1.05 to 2.40) but not provoked events. In WAFACS, elevated homocysteine levels were also associated with VTE events (p-trend 0.023).

Conclusions

Higher plasma homocysteine levels associate with VTE events in two cohorts of middle-aged and older women. Among VTE subtypes, homocysteine was associated with unprovoked, but not provoked, events. These data suggest a plausible biological role for homocysteine in the development of VTE.

Graphic Abstract

Introduction

Homocysteine is an amino acid synthesized during methionine metabolism via both vitamin B₆- and vitamin B₁₂-dependent pathways.¹ In vitro and in vivo experimental data demonstrate that homocysteine contributes to oxidative stress, endothelial dysfunction, inflammation, and thrombosis.^2–7 Elevations in homocysteine arise through a number of mechanisms, including genetic determinants of methionine metabolism, deficiencies in vitamin cofactors, and environmental exposures such as transition to high animal protein diets in industrializing nations, growing use of dietary protein supplements, and peroxisome proliferator-activated receptor (PPAR)-α agonist therapy in lipid management.^8–10

Epidemiologic studies have shown associations between elevations in plasma homocysteine and arterial atherothrombosis,^11–16 but the epidemiologic link between homocysteine and venous thromboembolism (VTE) has been more limited.^17–23 Few studies have been prospective in nature, and prospective data have shown weaker associations than in retrospective analyses.²² Prior data suggest elevated homocysteine concentration is a risk marker for recurrent, rather than first, VTE,^{24, 25} and may have a stronger effect among women than among men.^{17, 20} Analyses from three randomized placebo-controlled clinical trials of homocysteine lowering supplements did not show a reduction in VTE.^26–28 However, low event rates limited the ability to detect modest treatment effects and to evaluate idiopathic (unprovoked) events and VTE subtypes. Furthermore, whether concurrent elevations in other thrombotic risk markers account for previously observed risk associations has not been explored.

In light of these knowledge gaps, we examined the association between plasma homocysteine and total VTE, provoked and unprovoked VTE events, and VTE subtypes in the Women’s Health Study (WHS), a large, prospective cohort of more than 27,000 healthy middle-aged women with measured baseline homocysteine levels and among whom a total of 743 VTE events have accrued. We evaluated additional biomarkers, including high-sensitivity C-reactive protein (hsCRP), fibrinogen, soluble intercellular adhesion molecule 1 (sICAM-1), and lipoprotein(a), based on prior data suggesting a link between each biomarker and VTE. All markers were measured at baseline in the cohort. We also sought to validate our findings in the Women’s Antioxidant and Folic Acid Cardiovascular Study (WAFACS), a second independent prospective cohort of 2,672 middle-aged women and among whom risk associations with baseline homocysteine have not been previously analyzed.

Materials and Methods

Data Availability

The data will not be made available to other researchers for purposes of reproducing the results. However, further details of the methods used in the analysis are available upon reasonable request.

Study Population

For the primary analysis, we used participants in WHS, a previously completed randomized, placebo-controlled trial of low-dose aspirin and vitamin E for the primary prevention of cardiovascular disease.²⁹ Between 1992–1995, the study enrolled 39,876 female healthcare professionals in the US ≥ 45 years of age without prior cancer, myocardial infarction (MI), stroke, coronary revascularization, or peripheral artery revascularization. At enrollment, women provided demographic, anthropometric, medical, and lifestyle information. After completion of the trial, individuals were invited to participate in the ongoing longitudinal observational component of the WHS, and additional health outcomes data were collected using annual questionnaires.

Before randomization, 28,345 of the participants provided non-fasting blood samples, which were stored in liquid nitrogen (−150°C to −180°C) until analysis. Of these samples, 27,939 were of sufficient quality to be used for subsequent analyses. Subjects with pre-randomization VTE, including both pulmonary embolism (PE) and deep vein thrombosis (DVT), were excluded from the analysis. The final study population (n = 27,555) was followed for a median of 20.5 years.

To validate our findings from the primary analysis, we used participants in WAFACS. This was a randomized, placebo-controlled trial of combination therapy with folic acid, vitamin B₆, and vitamin B₁₂ in the prevention of cardiovascular disease.³⁰ The trial population consisted of high risk female healthcare professionals in the US ≥ 42 years of age with a history of cardiovascular disease or at least 3 cardiovascular risk factors. Cardiovascular disease was defined as history of MI, stroke, coronary or peripheral revascularization, angina pectoris, or transient ischemic attack. Cardiovascular risk factors included hypertension, hypercholesterolemia, diabetes mellitus, parental history of MI before age 60, body mass index (BMI) ≥ 30 kg/m², and active cigarette use. Women were excluded if they had a history of cancer within the previous ten years or were currently receiving anticoagulation. A total of 5,442 individuals were randomized and followed for 7.3 years. In addition to providing baseline demographic, anthropometric, medical, and lifestyle data, participants completed annual health questionnaires. Prior to randomization, 2,672 women in WAFACS also provided blood samples in which homocysteine levels were measured. All participants provided written informed consent, and the institutional review board at Brigham and Women’s Hospital approved both studies.

Outcome Ascertainment

In WHS, health outcomes were initially ascertained by self-report using health questionnaires at randomization, 6 months, 12 months, and annually. The primary outcome in our study was incident VTE, including both PE and DVT. VTE events were adjudicated by physician review of medical records in a blinded manner. Documentation required to confirm a PE included an invasive pulmonary angiogram, computed tomographic angiogram, or ventilation-perfusion scan with ≥2 mismatched defects. PE resulting in death was confirmed using autopsy reports, medical history, preceding symptoms, and circumstances surrounding death. Confirmation of DVT required either a lower extremity venous duplex ultrasound or an invasive venogram. Unprovoked VTE was defined as an event occurring in the absence of malignancy (diagnosed prior or ≤ 3 months following the VTE event), trauma, hospitalization ≥ 3 days, or surgery within the preceding 3 months. Provoked VTE included those that occurred in patients with documented malignancy or in the setting of trauma or surgery. Only confirmed endpoints were used in this analysis.

In WAFACS, health outcomes were similarly ascertained by self-report using annual health questionnaires. The primary outcome of interest was again incident VTE, including both PE and DVT. Events were adjudicated by physician review of medical records in a blinded manner, and the documentation required for both PE and DVT was identical to that of WHS.

Laboratory Analysis

A core laboratory certified by the National Heart, Lung, and Blood Institute/Centers for Disease Control and Prevention Lipid Standardization Program performed all laboratory analyses. Low-density lipoprotein cholesterol (LDL-C) was measured using a homogeneous direct method with a Hitachi 917 analyzer using reagents from Roche Diagnostics (Indianapolis, Ind). High-density lipoprotein cholesterol (HDL-C) was measured using a direct enzymatic colorimetric assay, and triglycerides were measured enzymatically with correction for endogenous glycerol. Apolipoproteins B₁₀₀ and A-1 were measured using immunoturbidimetric assays (DiaSorin, Stillwater, Minn). hsCRP was measured by a high-sensitivity immunoturbidimetric assay (Denka Seiken, Niigata, Japan).

In both cohorts, homocysteine was measured with an enzymatic assay (Catch Inc., Seattle, Wash; Roche Diagnostics, Hitachi 917 analyzer). Fibrinogen was calculated with an immunoturbidimetric assay (Kamiya Biomedial Co., Seattle, Wash). Lipoprotein(a) was measured with a latex-enhanced turbidimetric assay (Denka Seiken, Niigata, Japan). sICAM-1 was measured using quantitative sandwich ELISA (R&D Systems, Minneapolis, Minn). Hemoglobin A1c was calculated with a turbidimetric assay (Roche Diagnostics, Indianapolis, Ind).

Statistical Analysis

Continuous data are summarized as either mean ± standard deviation or median with interquartile range depending on normality of the distributions. Categorical data are reported as percentages. Between group differences were assessed by a t-test or the Wilcoxon rank-sum test for continuous data and the χ² test for categorical data. Biomarkers were divided into quartiles based on the population distribution of each biomarker. Cox proportional-hazards models were used to estimate the hazard ratio (HR) and 95% confidence interval (CI) for each biomarker quartile, and results are presented as top quartile compared to bottom (reference) quartile unless otherwise noted. Tests of linear trend across quartiles were performed using the median value from each quartile. Unadjusted event rates were using for each quartile and were reported for the full duration of follow-up.

For the primary analysis within WHS, all Cox regression models were adjusted for age followed by the addition of the following covariates: active smoking, BMI, and postmenopausal hormone therapy (Model 1). Fully-adjusted models (Model 2) were adjusted for age (years), active smoking (baseline smoking: yes/no), BMI (kg/m²), postmenopausal hormone therapy (baseline use: yes/no), physical activity level (rarely/never, <1 time weekly, 2–3 times weekly, 4+ times weekly), metabolic syndrome (baseline: yes/no), post-menopausal status (baseline: yes/no/unknown), and glomerular filtration rate. Metabolic syndrome was defined using a modified Adult Treatment Panel III definition that has been previously validated in WHS.^{31, 32} All regression results in the text are presented for Model 2 unless otherwise noted. Additionally, all models were adjusted for randomized treatment assignment. Covariates were selected based on a priori knowledge linking them to incident VTE.

Previous studies have shown a positive correlation between BMI and plasma homocysteine concentration^{33, 34} as well as a strong association between BMI and risk of VTE.^35–37 To evaluate the joint effects of plasma homocysteine concentration as well as BMI, individuals were classified into 4 groups based on the values of each measure relative to the population median.

In our confirmatory study conducting in the WAFACS, Cox regression analyses were adjusted for age, active smoking, elevated BMI (>26.70 kg/m²), height, physical activity level, diabetes, post-menopausal status, and history of hypertension. Because more than half of WAFACS participants did not have baseline lipid measures, this regression analysis additionally adjusted for diabetes (as a surrogate for hyperglycemia) and history of hypertension to account for potential residual confounding from these components of metabolic syndrome. Models were additionally adjusted for baseline history of venous thromboembolism (91.1% of subjects reported no prior VTE).

All statistical analyses were performed using SAS statistical software version 9.4 (SAS Institute, Cary, NC). All 95% CIs are 2-tailed, and the p-value cutoff for all analyses was 0.05.

Results

As shown in Table 1, women in WHS with incident VTE were older with a larger mean BMI. They also had a greater prevalence of metabolic syndrome and hypertension and were more likely to be post-menopausal at baseline. Baseline diabetes was rare in the overall study population, and there were no significant differences in active smoking, self-reported physical activity, or hormonal therapy between the two groups. Median plasma measures of total cholesterol, LDL-C, triglycerides, apolipoprotein B₁₀₀, hsCRP, fibrinogen, sICAM-1, and homocysteine were all greater among women with incident VTE, although absolute differences were small.

Table 1.

Baseline Characteristics of the WHS Population

	Women Remaining Free of VTE (n = 26,812)*	Women Developing VTE (n = 743)†	p-value
Age, mean (SD), y	54.6 (7.1)	56.9 (7.5)	<0.0001
BMI, mean (SD), kg/m2	25.9 (4.9)	27.4 (5.5)	<0.0001
Height, mean (SD), in	64.6 (2.5)	65.0 (2.6)	<0.0001
Non-Hispanic white, %	25,298 (95.2)	713 (96.7)	0.045
Current smoking, %	3,103 (11.6)	80 (10.8)	0.52
Alcohol abstinence, %	11,829 (44.1)	332 (44.7)	0.76
Diabetes, %	644 (2.4)	18 (2.4)	0.90
Metabolic syndrome, %	6,423 (24.2)	232 (31.5)	<0.0001
Hypertension, %	6,668 (24.9)	214 (28.8)	0.016
Treatment for hypercholesterolemia, %	851 (3.2)	21 (2.8)	0.67
Exercise ≥1 time/wk, %	11,596 (43.3)	302 (40.7)	0.16
Current HT use, %	11,395 (42.6)	329 (44.3)	0.35
Post-menopausal, %	14,465 (54.1)	482 (65.1)	<0.0001
Hemoglobin A1c, %	5.0 (4.8–5.2)	5.0 (4.9–5.2)	0.07
GFR, mL/min/1.73 m2	91.8 (79.6–105.5)	89.5 (77.2–103.7)	0.005
Standard chemical lipids, mg/dL
Total cholesterol	208 (184–235)	214 (189–240)	0.002
HDL-cholesterol	52 (43–62)	51 (42–62)	0.08
LDL-cholesterol	121 (100–144)	125 (103–147)	0.006
Triglycerides	118 (83–174)	132 (92–187)	<0.0001
Apolipoproteins, mg/dL
Apolipoprotein A-1	149 (132–168)	148 (132–170)	0.94
Apolipoprotein B100	100 (84–121)	107 (87–125)	<0.0001
hsCRP, mg/L	2.0 (0.8–4.3)	2.6 (1.2–5.2)	<0.0001
Fibrinogen, mg/dL	350.1 (307.0–401.9)	362.1 (319.3–415.9)	<0.0001
sICAM-1, ng/mL	342.2 (300.6–394.0)	351.4 (311.0–404.3)	0.0003
Homocysteine, μmol/L	10.4 (8.7–12.9)	10.9 (9.1–13.7)	<0.0001
Lipoprotein(a), mg/dL	10.6 (4.4–32.8)	10.5 (4.7–35.4)	0.82

Values are median (25th-75th percentile) unless otherwise indicated. P values for continuous

variables were obtained from the t-test (age, BMI, height) or the Wilcoxon rank sum test. P values for categorical variables were obtained using Chi-square test.

VTE, venous thromboembolism; SD, standard deviation; BMI, body mass index; HT, hormonal therapy; GFR, glomerular filtration rate; HDL, high-density lipoprotein; LDL, low-density lipoprotein; hsCRP, high-sensitivity C-reactive protein; sICAM-1, soluble intercellular adhesion molecule 1.

^*Number missing: 22 for BMI; 17 for Height; 226 for Race; 24 for Current smoking; 6 for Alcohol abstinence; 15 for Diabetes; 216 for Metabolic syndrome; 7 for Hypertension; 18 for Treatment for hypercholesterolemia; 9 for Exercise; 49 for Current HT use; 48 for Post-menopausal status; 432 for HbA_1c; 377 for GFR; 378 for Total cholesterol; 377 for HDL-cholesterol, LDL-cholesterol, and Triglycerides; 505 for Apolipoprotein A-1; 509 for Apolipoprotein B₁₀₀; 377 for hsCRP; 504 for Fibrinogen; 512 for sICAM-1; 509 for Homocysteine; 514 for Lp(a).

^†Number missing: 1 for BMI; 1 for Height; 6 for Race; 6 for Metabolic syndrome; 1 for Treatment for hypercholesterolemia; 1 for Current HT use; 2 for Post-menopausal status; 12 for HbA_1c; 12 for GFR; 12 for Total cholesterol, HDL-cholesterol, LDL-cholesterol, and Triglycerides; 14 for Apolipoproteins A-1 and B₁₀₀; 12 for hsCRP; 14 for Fibrinogen; 16 for sICAM-1; 15 for Homocysteine; 14 for Lp(a).

Table 2 shows the results of Cox regression analyses for each biomarker and incident VTE based on quartiles of the population distribution for each biomarker. In age-adjusted analyses, hsCRP had the strongest positive risk association (HR_Q4 1.84; 95% CI 1.47–2.30; p-trend <0.001) followed by fibrinogen, homocysteine, and sICAM-1 (all p-trend <0.05). In multivariable-adjusted models, the associations for hsCRP, fibrinogen, and sICAM-1 no longer reached statistical significance. However, there remained a strong association between plasma homocysteine concentration and incident VTE in multivariable-adjusted models after adjustment for BMI and other risk factors (HR_Q4 1.31; 95% CI 1.06–1.63; p-trend 0.006). Incidence rates for VTE increased across homocysteine quartiles: 1.18, 1.34, 1.53, and 1.80 per 1,000 person-years, respectively. Each 5 μmol/L increase in homocysteine concentration was also associated with an increased risk of VTE (HR 1.09; 95% CI 1.03–1.16; p=0.005 [data not shown]). There was no significant association between plasma concentrations of lipoprotein(a) and incident VTE in age-adjusted or multivariable-adjusted regression models.

Table 2.

Associations of Biomarkers with Incident VTE in the WHS

	Quartile 1	Quartile 2	Quartile 3	Quartile 4	p for Linear Trend
hsCRP
Range, mg/L	≤0.80	0.81–2.00	2.01–4.34	≥4.35
Cases, n	117	184	202	228
Age-adjusted HR (95% CI)	1.00	1.50 (1.19–1.89)	1.61 (1.28–2.02)	1.84 (1.47–2.30)	<0.001
Model 1 HR (95% CI)	1.00	1.32 (1.05–1.68)	1.28 (1.00–1.63)	1.27 (0.98–1.64)	0.45
Model 2 HR (95% CI)	1.00	1.34 (1.06–1.70)	1.29 (1.01–1.65)	1.30 (1.00–1.68)	0.37
Fibrinogen
Range, mg/dL	≤307.3	307.4–350.6	350.7–402.5	≥402.6
Cases, n	130	177	199	223
Age-adjusted HR (95% CI)	1.00	1.31 (1.05–1.64)	1.43 (1.14–1.78)	1.58 (1.27–1.96)	<0.0001
Model 1 HR (95% CI)	1.00	1.22 (0.97–1.53)	1.24 (0.99–1.56)	1.21 (0.96–1.53)	0.18
Model 2 HR (95% CI)	1.00	1.21 (0.96–1.52)	1.22 (0.97–1.53)	1.20 (0.95–1.51)	0.22
sICAM-1
Range, ng/dL	≤300.8	300.8–342.5	342.6–394.2	≥394.3
Cases, n	149	168	198	212
Age-adjusted HR (95% CI)	1.00	1.06 (0.85–1.33)	1.22 (0.99–1.52)	1.34 (1.09–1.66)	0.002
Model 1 HR (95% CI)	1.00	1.01 (0.81–1.25)	1.08 (0.87–1.34)	1.10 (0.88–1.37)	0.34
Model 2 HR (95% CI)	1.00	1.01 (0.81–1.26)	1.08 (0.87–1.35)	1.11 (0.88–1.39)	0.30
Homocysteine
Range, μmol/L	≤8.6	8.7–10.5	10.5–12.9	≥12.9
Cases, n	151	169	190	218
Age-adjusted HR (95% CI)	1.00	1.08 (0.87–1.35)	1.20 (0.97–1.49)	1.39 (1.13–1.71)	<0.001
Model 1 HR (95% CI)	1.00	1.05 (0.85–1.31)	1.14 (0.92–1.42)	1.33 (1.07–1.64)	0.004
Model 2 HR (95% CI)	1.00	1.04 (0.83–1.29)	1.14 (0.91–1.41)	1.31 (1.06–1.63)	0.006
Lipoprotein(a)
Range, mg/dL	≤4.4	4.5–10.6	10.7–32.9	≥33.0
Cases, n	174	196	165	194
Age-adjusted HR (95% CI)	1.00	1.13 (0.92–1.39)	0.93 (0.75–1.15)	1.11 (0.90–1.36)	0.47
Model 1 HR (95% CI)	1.00	1.15 (0.94–1.41)	0.93 (0.75–1.16)	1.12 (0.91–1.38)	0.43
Model 2 HR (95% CI)	1.00	1.15 (0.94–1.42)	0.94 (0.76–1.17)	1.13 (0.92–1.39)	0.40

VTE, venous thromboembolism; hsCRP, high-sensitivity C-reactive protein; sICAM-1, soluble intercellular adhesion molecule 1; HR, hazard ratio; CI, confidence interval.

Model 1 adjusted for age, smoking, BMI, hormonal therapy, and randomization status.

Model 2 adjusted for age, smoking, BMI, hormonal therapy, randomization status, activity level, metabolic syndrome, post-menopausal status, and GFR.

As homocysteine was the only biomarker positively associated with incident VTE in both age-adjusted and multivariable-adjusted regression models, we then assessed the association between homocysteine and different sub-categories of VTE (Table 3). Homocysteine concentration was associated with total PE (HR_Q4 1.44; 95% CI 1.07–1.94; p-trend 0.005) and unprovoked PE (HR_Q4 2.28; 95% CI 1.41–3.69; p-trend <0.001) cases in age-adjusted models. In multivariable-adjusted models, however, homocysteine was only associated with incident unprovoked PE (HR_Q4 2.13; 95% CI 1.30–3.51; p-trend 0.002). There was no significant association between plasma homocysteine concentration and provoked PE in age-adjusted or multivariable-adjusted models.

Table 3.

Risk Associations Between Homocysteine and Incident Provoked and Unprovoked Pulmonary Embolism and Deep Vein Thrombosis in the WHS

	Quartile 1	Quartile 2	Quartile 3	Quartile 4	p for Linear Trend
Total PE
Homocysteine
Range, μmol/L	≤8.6	8.7–10.5	10.5–12.9	≥12.9
Cases, n	74	78	95	110
Event rate (per 1,000 person-years)	0.58	0.62	0.76	0.90
Age-adjusted HR (95% CI)	1.00	1.02 (0.74–1.40)	1.23 (0.91–1.67)	1.44 (1.07–1.94)	0.005
Model 1 HR (95% CI)	1.00	0.97 (0.71–1.34)	1.14 (0.84–1.55)	1.32 (0.98–1.78)	0.03
Model 2 HR (95% CI)	1.00	0.96 (0.70–1.33)	1.17 (0.86–1.59)	1.36 (1.00–1.85)	0.02
Unprovoked PE
Homocysteine
Range, μmol/L	≤8.6	8.7–10.4	10.5–12.8	≥12.9
Cases, n	24	35	42	55
Event rate (per 1,000 person-years)	0.19	0.28	0.34	0.45
Age-adjusted HR (95% CI)	1.00	1.43 (0.85–2.41)	1.71 (1.03–2.83)	2.28 (1.41–3.69)	<0.001
Model 1 HR (95% CI)	1.00	1.38 (0.82–2.32)	1.57 (0.95–2.61)	2.08 (1.28–3.38)	0.002
Model 2 HR (95% CI)	1.00	1.37 (0.81–2.31)	1.60 (0.96–2.66)	2.13 (1.30–3.51)	0.002
Provoked PE
Homocysteine
Range, μmol/L	≤8.6	8.7–10.4	10.5–12.8	≥12.9
Cases, n	48	43	52	54
Event rate (per 1,000 person-years)	0.38	0.34	0.42	0.44
Age-adjusted HR (95% CI)	1.00	0.85 (0.57–1.29)	1.02 (0.69–1.51)	1.07 (0.72–1.58)	0.52
Model 1 HR (95% CI)	1.00	0.81 (0.54–1.23)	0.96 (0.65–1.43)	0.98 (0.66–1.46)	0.79
Model 2 HR (95% CI)	1.00	0.80 (0.53–1.22)	0.98 (0.66–1.46)	1.02 (0.68–1.53)	0.62
Total DVT
Homocysteine
Range, μmol/L	≤8.6	8.7–10.4	10.5–12.8	≥12.9
Cases, n	117	116	139	159
Event rate (per 1,000 person-years)	0.92	0.92	1.12	1.31
Age-adjusted HR (95% CI)	1.00	0.95 (0.74–1.23)	1.13 (0.88–1.44)	1.30 (1.02–1.65)	0.009
Model 1 HR (95% CI)	1.00	0.93 (0.72–1.21)	1.09 (0.85–1.39)	1.27 (1.00–1.62)	0.02
Model 2 HR (95% CI)	1.00	0.92 (0.71–1.19)	1.07 (0.83–1.37)	1.24 (0.96–1.59)	0.03
Unprovoked DVT
Homocysteine
Range, μmol/L	≤8.6	8.7–10.4	10.5–12.8	≥12.9
Cases, n	40	45	58	64
Event rate (per 1,000 person-years)	0.31	0.36	0.47	0.53
Age-adjusted HR (95% CI)	1.00	1.10 (0.72–1.68)	1.41 (0.94–2.11)	1.56 (1.05–2.32)	0.01
Model 1 HR (95% CI)	1.00	1.10 (0.72–1.68)	1.39 (0.93–2.09)	1.60 (1.07–2.40)	0.01
Model 2 HR (95% CI)	1.00	1.10 (0.71–1.68)	1.39 (0.92–2.09)	1.59 (1.05–2.40)	0.01
Provoked DVT
Homocysteine
Range, μmol/L	≤8.6	8.7–10.4	10.5–12.8	≥12.9
Cases, n	74	68	79	90
Event rate (per 1,000 person-years)	0.58	0.54	0.64	0.74
Age-adjusted HR (95% CI)	1.00	0.88 (0.63–1.22)	1.00 (0.73–1.37)	1.14 (0.84–1.55)	0.22
Model 1 HR (95% CI)	1.00	0.85 (0.61–1.18)	0.95 (0.69–1.31)	1.09 (0.80–1.49)	0.34
Model 2 HR (95% CI)	1.00	0.82 (0.59–1.14)	0.93 (0.67–1.28)	1.05 (0.76–1.45)	0.46

WHS, Women’s Health Study; PE, pulmonary embolism; DVT, deep vein thrombosis; HR, hazard ratio; CI, confidence interval.

Event rates were calculated using the Kaplan-Meier estimator and are reported for the full duration of follow-up.

Model 1 adjusted for age, smoking, BMI, hormonal therapy, and randomization status.

Model 2 adjusted for age, smoking, BMI, hormonal therapy, randomization status, activity level, metabolic syndrome, post-menopausal status, and GFR.

We found similar associations for DVT subtypes (Table 3). Homocysteine was positively associated with total DVT (HR_Q4 1.30; 95% CI 1.02–1.65; p-trend 0.009) and unprovoked DVT (HR_Q4 1.56; 95% CI 1.05–2.32; p-trend 0.01) in age-adjusted models. These associations were slightly attenuated for total DVT (HR_Q4 1.24; 95% CI 0.96–1.59; p-trend 0.03) but persisted for unprovoked DVT (HR_Q4 1.59; 95% CI 1.05–2.40; p-trend 0.01) after adjusting for additional VTE risk factors. As with provoked PE, there was no significant association between homocysteine and provoked DVT.

Women were classified as having high/low BMI and high/low homocysteine concentration based on the population median of each measure to evaluate the joint role of these risk factors on incident VTE, PE, and DVT (Figure 1; Supplemental Table I). Even among women with a low BMI, an elevated plasma homocysteine concentration identified women at heightened risk for VTE, PE, and DVT, although the association was strongest for PE (HR 1.61; 95% CI 1.10–2.34). Women with elevated BMI were at heightened risk for all VTE events regardless of plasma homocysteine concentration, although elevations in homocysteine further identified women in the highest risk group (HR 2.25; 95% CI 1.78–2.84; HR 2.99; 95% CI 2.10–4.27; and HR 2.20; 95% CI 1.68–2.89 for incident VTE, PE, and DVT respectively). Across all categories, both BMI and homocysteine concentration were associated with a greater risk of PE than VTE or DVT.

Figure 1. Risk of Incident VTE, PE, and DVT by Baseline BMI and Homocysteine Concentration.

Participants are divided into low/high categories based on the study population median for body mass index (24.9 kg/m²) and plasma homocysteine (10.5 μmol/L). VTE indicates venous thromboembolism; PE, pulmonary embolism; DVT, deep vein thrombosis; BMI, body mass index; Hcys, homocysteine; CI, confidence interval. Adjusted for age, smoking, hormonal therapy, activity level, metabolic syndrome, post-menopausal status, glomerular filtration rate, and randomized treatment.

Finally, we examined the risk association between plasma homocysteine concentration and incident VTE in WAFACS (Figure 2). In WAFACS, women in the highest quartile of homocysteine had an increased risk of VTE (HR_Q4 2.15; 95% CI 1.17–3.96; p-trend 0.02). There was no statistically significant increased risk for women in the second or third quartiles.

Figure 2. Association Between Homocysteine and Incident VTE in WHS and WAFACS.

Hazard ratios and 95% confidence intervals for homocysteine in both WHS and WAFACS. WHS indicates Women’s Health Study; WAFACS, Women’s Antioxidant and Folic Acid Cardiovascular Study. Models for WHS adjusted for age, smoking, BMI, hormonal therapy, activity level, metabolic syndrome, post-menopausal status, glomerular filtration rate, and randomized treatment. Models for WAFACS adjusted for age, smoking, BMI, height, activity level, diabetes, post-menopausal status, history of hypertension, randomized treatment, and baseline venous thromboembolism.

Discussion

In two prospective cohorts of women, we found that plasma homocysteine concentration was independently associated with a heightened risk for incident VTE. In the primary prevention WHS, these findings for homocysteine differed in comparison to other biomarkers of atherothrombosis, including hsCRP, fibrinogen, sICAM-1, and lipoprotein(a), which were not associated with incident VTE in multivariable-adjusted models. Homocysteine concentration was associated with unprovoked, but not provoked, cases of incident PE and DVT. Our data also support a joint association of BMI with homocysteine, as women with elevations in both BMI and plasma homocysteine concentration were at highest risk.

Our data must be placed in the context of previous, at times conflicting, analyses of homocysteine and VTE. The first published case-control analysis found no association between homocysteine and VTE.³⁸ However, subsequent retrospective analyses have shown more consistent associations. A meta-analysis of case-control studies found that every 5 μmol/L increase in plasma homocysteine associates with a 60% increased risk of VTE (odds ratio [OR] 1.60; 95% CI 1.10–2.34).²² Furthermore, in a case-control analysis of 269 patients with first DVT in the Leiden Thrombophilia Study, a homocysteine concentration above the 95^th percentile (>18.5 μmol/L) was associated with a 2.5-fold increased odds of VTE (95% CI 1.2–5.2).¹⁷ This risk association persisted even with a lower cutoff of the 90^th percentile value (>16.6 μmol/L). Notably, in subgroup analyses, the risk association did not materially increase until reaching a threshold >18 μmol/L, with a dramatically increased risk among those with a concentration >22 μmol/L. This is in contrast to our present analysis, which showed not only a dose-response relationship between homocysteine and incident VTE but also a heightened risk association at lower concentrations (lower limit for top quartile ≥12.9 μmol/L).

Retrospective case-control analyses do not account for the potential impact of acute VTE events as well as associated medical interventions on homocysteine levels. In a prospective nested case-control analysis of initially healthy men in the Physicians’ Health Study, investigators measured plasma homocysteine concentration in 145 individuals who subsequently developed incident VTE as well as 646 controls.²¹ Over 10 years of follow-up, men with homocysteine levels >95^th percentile (17.25 μmol/L) were at increased risk of idiopathic (unprovoked), but not total, VTE (relative risk [RR] 3.4; 95% CI 1.6–7.3; p=0.002 vs. RR 1.6; 95% CI 0.8–3.3; p=0.2, respectively). In the present analysis from WHS, we similarly found the strongest risk associations for unprovoked DVT and PE compared to either total or provoked events. Notably, our analysis showed a risk association at even more modest homocysteine levels (top quartile ≥12.9 μmol/L; HR 1.31; 95% CI 1.06–1.63), potentially identifying important sex differences for homocysteine risk.

Until the present analysis, the largest prospective study examining homocysteine and VTE was a nested case-control analysis of 303 VTE cases and 635 matched controls in the Longitudinal Investigation of Thromboembolism Etiology (LITE) study.²⁰ Homocysteine levels in the top quintile (>15.9 μmol/L) were not associated with a statistically significant increased risk of a composite of incident and recurrent VTE compared to those in the lowest quintile (OR 1.55; 95% CI 0.93–2.58). However, when restricted to individuals age 45–64, a range closer to that of WHS, the risk association reached statistical significance (OR 2.05; 95% CI 1.10–3.83). This may have been due to a stronger link between homocysteine and competing arterial thromboembolic events among older participants.²⁰

Despite these epidemiologic findings, the null results from randomized clinical trials of homocysteine lowering therapy have posed the greatest challenge to the “homocysteine hypothesis” in VTE (Table 4). The Vitamins and Thrombosis (VITRO) study enrolled 701 individuals with a history of VTE.²⁶ 341 participants had homocysteine levels ≥75^th percentile (≥8.5–10.6 μmol/L depending on sex and the local population homocysteine distribution), and 360 had homocysteine levels <75^th percentile. Participants were randomized to either a combination of folic acid, vitamin B₆, and vitamin B₁₂ or placebo. After a planned follow-up of 2.5 years, vitamin supplementation led to 46% and 33% reductions in homocysteine in the hyperhomocysteinemic and normohomocysteinemic groups, respectively. However, there was no reduction in recurrent VTE events, although with only 43 and 50 events in the intervention and placebo arms, respectively, this study may have been underpowered. VITRO was also limited to individuals with a prior history of VTE, and it is possible that medical interventions following the initial events could have both altered baseline homocysteine levels as well as individual response to vitamin therapy. It is also possible that the intervention occurred too late in the disease process to yield any benefit.

Table 4.

Summary of Prior Placebo-Controlled Clinical Trials Evaluating Folic Acid and B Vitamins for VTE Prevention

	VITRO Hyperhomocysteinemic26	VITRO Normohomocysteinemic26	HOPE-228	WAFACS39
Study Population	Secondary VTE Prevention	Secondary VTE Prevention	High CV Risk	High CV Risk
No. of Subjects	360	341	5,522	2,672
No. of Women	152	187	1,559	2,672
Geometric Mean Homocysteine (μmol/L)	15.9*	9.0	11.5	12.2
Folic Acid Dose	5 mg/d	5 mg/d	2.5 mg/d	2.5 mg/d
Vitamin B6 (Pyridoxine) Dose	50 mg/d	50 mg/d	50 mg/d	50 mg/d
Vitamin B12 (Cyanocobalamin) Dose	0.4 mg/d	0.4 mg/d	1.0 mg/d	1.0 mg/d
Duration of Follow-Up	2.5 years	2.5 years	5 years	7.3 years
No. of Total VTE Events	50	43	88	132
No. of Unprovoked VTE Events	Not Classified	Not Classified	42	Not Classified
Treatment Effect (μmol/L)†	−8.5	−2.5	−2.2	−2.4
Treatment Effect (Hazard Ratio [95% Confidence Interval])	1.14 (0.65–1.98)	0.58 (0.32–1.08)	1.01 (0.66–1.53)	0.82 (0.59–1.16)
Key Secondary Analyses	Per 5 μmol/L higher baseline homocysteine: 1.13 (1.05–1.20)		Unprovoked VTE: 1.21 (0.66–2.23)	Unprovoked VTE: 0.68 (0.44–1.04) BMI ≥ 30 kg/m2: 0.57 (0.36–0.91)

VTE, venous thromboembolism; VITRO, VItamins and ThROmbosis; HOPE, Heart Outcomes Prevention Evaluation; WAFACS, Women’s

Antioxidant and Folic Acid Cardiovascular Study; CV, cardiovascular

^*Geometric mean in placebo group

^†Not placebo adjusted as these were not available in all trials

The Heart Outcomes Prevention Evaluation 2 (HOPE-2) trial enrolled 5,522 men and women age 55 or older with either a history of atherosclerotic cardiovascular disease or diabetes mellitus plus an additional risk factor for cardiovascular disease.²⁷ Participants were randomized to a combination of folic acid, vitamin B₆, and vitamin B₁₂ or placebo for a mean duration of 5 years. Although vitamin supplementation reduced plasma homocysteine levels compared to placebo, it did not reduce the rates of DVT, PE, or unprovoked VTE.²⁷ In subgroup analyses, there were also no differences based on sex, age, baseline oral anticoagulant use, or baseline homocysteine level. In the WAFACS, the combination of folic acid, vitamin B₆, and vitamin B₁₂ did not lower VTE compared to placebo in the total trial population, though once again the event rate was low.³⁹ There was, however, a significant interaction by obesity such that among 2,690 obese women (BMI ≥ 30 kg/m²), the treatment-associated risk reduction was 43% (95% CI 9–67%; p=0.019). Our data from WHS complement this finding by demonstrating the highest incidence rates in overweight women with high levels of homocysteine.

Despite the large sample size and prospective nature of our study, there are several potential limitations that warrant discussion. Our analysis was restricted to women, and WHS participants were largely Caucasian and healthy at baseline. Therefore, our results may not be generalizable to a more diverse population. Participants in WAFACS, the secondary prevention cohort used to validate our findings, had a higher prevalence of cardiovascular disease and cardiovascular risk factors at baseline, thus mitigating some of this limitation. Variability in preanalysis sample processing conditions, such as time between blood draws and freezing, number of freeze-thaw cycles, and the presence or absence of hemolysis, may limit our ability to compare results across different studies, although prior data suggest homocysteine concentration is minimally impacted by multiple freeze-thaw cycles.⁴⁰ We also did not utilize fasting homocysteine concentration in our analysis. However, data suggest the impact of fasting is likely minimal, and fasting homocysteine levels may not accurately reflect the typical physiologic state of individuals.⁴¹ Measures of renal function were unavailable in WAFACS, and renal function is known to strongly correlate with homocysteine concentration.⁴² However, adjusting for renal function in WHS did not significantly alter any of our findings. Finally, during the course of WHS and WAFACS, the Food and Drug Administration implemented mandatory folic acid supplementation, which began in 1996 and was largely completed by mid-1997.⁴³ Although this may have influenced homocysteine concentration subsequent to participants’ blood draws, this would have biased our results toward the null. The results of VITRO and HOPE-2 further suggested that folic acid supplementation had no influence on VTE risk.

In summary, in what is to our knowledge the largest prospective analysis of homocysteine and VTE, higher plasma homocysteine concentrations were associated with a heightened future risk of VTE. Specifically, homocysteine was associated with unprovoked, rather than provoked, PE and DVT. The risk association between homocysteine and VTE was also seen in a second validation cohort. Although prior clinical trials of homocysteine reduction have not been effective in reducing VTE risk, our data suggest homocysteine remains an important risk marker. Rather than focusing on homocysteine reduction, future studies may use circulating biomarkers like homocysteine and other markers of procoagulability,⁴⁴ as well as additional markers like polygenic risk scores, to help clinicians implement primary and secondary prevention strategies to reduce the risk of VTE.

Highlights.

• We found that baseline plasma concentration of homocysteine is associated with future risk of venous thromboembolism in two independent cohorts of women.

• In the Women’s Health Study, this risk was further increased among women with elevated body mass index.

• Elevated homocysteine was associated with unprovoked, but not provoked, deep vein thrombosis and pulmonary embolism

Nonstandard Abbreviations and Acronyms

BMI

body mass index

DVT

deep vein thrombosis

HDL-C

high-density lipoprotein cholesterol

HOPE-2

Heart Outcomes Prevention Evaluation 2

hsCRP

high-sensitivity C-reactive protein

LDL-C

low-density lipoprotein cholesterol

MI

myocardial infarction

PE

pulmonary embolism

sICAM-1

soluble intercellular adhesion molecule 1

VITRO

Vitamins and Thrombosis

VTE

venous thromboembolism

WAFACS

Women’s Antioxidant and Folic Acid Cardiovascular Study

WHS

Women’s Health Study