Meta-analysis of Hyperhomocysteinemia as a Risk Factor for Venous Thromboembolic Disease

Background  Elevated plasma homocysteine (Hcy) levels are implicated in the development of atherosclerotic and venous thromboembolic disease. A meta-analysis of the risk for venous thromboembolism (VTE) in the presence of hyperhomocysteinemia (hyper-Hcy) was performed.

Methods  Studies were identified through MEDLINE (January 1980 to August 1997) using search terms related to both Hcy and VTE. The bibliographies of all review articles and letters were searched for additional relevant articles. English-language studies were selected if they included 10 or more human subjects; a measurement of the plasma, serum, or whole-blood Hcy level; the presence of VTE; and primary data that were not published elsewhere. Seventy-two articles were retrieved, of which 9 met all inclusion criteria. Data were extracted on the study type, subject demographics, methods for matching control subjects with case patients, whether an objective method was used to diagnose VTE, and whether other causes of thrombophilia and elevated Hcy levels were considered. The mean Hcy levels, both in the fasting state and following methionine loading, if done, were recorded, as were the number of patients and control subjects with Hcy levels greater than 2 SDs or greater than the 95th percentile above the mean value of the control group.

Results  Nine case-control studies measured fasting plasma Hcy levels, and 5 also measured Hcy values after methionine loading. All 9 studies showed a similar qualitative trend in fasting levels in the associated risk for hyper-Hcy and VTE. The corresponding pooled odds ratio was 2.95 (95% confidence interval, 2.08-4.17; 2-sided P <.001), with no evidence for heterogeneity across the studies (P=.50). Following methionine loading, the trend was also toward increasing the risk of VTE with hyper-Hcy (odds ratio, 2.15; 95% confidence interval, 1.20-3.85; 2-sided P=.01). Again, no evidence of heterogeneity was found (P=.65). The pooled odds ratio for VTE in the presence of hyper-Hcy rose to 4.37 (95% confidence interval, 1.94-9.84) when studies with patients older than 60 years were excluded. Limitations of the individual studies included a lack of proper matching of patients with control subjects, a limited description of subject recruitment, and a failure to test for other hypercoagulable mechanisms and other causes of elevated Hcy levels, such as renal insufficiency or folate deficiency.

Conclusions  A significant risk for VTE in the presence of hyper-Hcy apparently exists among a spectrum of patients with first or recurrent venous thromboembolic events. This risk appears to be most significant for patients with VTE disease before age 60 years. A well-designed prospective study is needed to confirm these findings.

A POSITIVE association exists between hyperhomocysteinemia (hyper-Hcy) and the development of premature atherosclerosis.¹ A meta-analysis¹ of arteriosclerosis and hyper-Hcy included 25 case-control and cross-sectional studies. The summary odds ratios (ORs) from this overview¹ were 1.7 (95% confidence interval [CI], 1.5-1.9) for coronary artery disease, 2.5 (95% CI, 2.0-3.0) for cerebrovascular disease, and 6.8 (95% CI, 2.9-15.8) for peripheral vascular disease. Limitations of these findings include substantial heterogeneity between studies, the omission of both published and unpublished data,² and the inclusion of patients with chronic renal insufficiency, for example, a state that may dramatically elevate homocysteine (Hcy) levels.^3,4

Several mechanisms may lead to hyper-Hcy. Known causes include deficiencies of folate, vitamin B₆, and vitamin B₁₂^5,6; a reduced activity of methylenetetrahydrofolate reductase⁷; or other enzyme defects within the Hcy metabolic pathway.⁸ Whether hyper-Hcy behaves as a biological marker of, or a direct etiologic agent in, the development of thrombosis also remains to be established.

Some etiologic risk factors for the development of arterial thrombosis, such as the lupus anticoagulant, are also shared with venous thromboembolism (VTE). Furthermore, patients with classic homocystinuria are at high risk for both arterial and venous events. Hence, the purpose of this systematic overview is to determine to what degree VTE disease is associated with hyper-Hcy. Such knowledge might further our understanding of which patients are at risk for recurrent VTE and aid in the development of other strategies for the secondary prevention of VTE beyond long-term anticoagulation therapy. A second general aim is to improve the design of future research in this field by analyzing these studies for sources of bias and heterogeneity.

A systematic literature search was performed using MEDLINE. Major medical subject headings and text words were searched between January 1980 and August 1997 using the terms "homocysteine," "homocystine," "homocyst(e)ine," "folate," "folic acid," and "vitamin B₁₂." These terms were individually cross-searched with the medical subject heading terms and text words "thrombophlebitis," "deep vein thrombosis," "pulmonary embolism," and "thrombosis."

All abstracts were reviewed in an unblinded manner. Review articles and letters to editors were read, and their bibliographies were searched for additional articles. Primary studies were considered if they included all of the following: data collected on 10 or more human subjects; a measurement of the plasma, serum, or whole-blood Hcy level; the presence of VTE; and data that had not been published elsewhere. Only articles published in the English language were included.

No formal quality rating was used in evaluating studies meeting the inclusion criteria. Both Shapiro⁹ and Petitti¹⁰ have argued against scoring nonexperimental studies because this technique has not been formally validated. Instead, the important features from each study were extracted and their methodological limitations identified.^9,10

A separate meta-analysis was performed for fasting Hcy levels and for Hcy levels after methionine loading. Hyperhomocysteinemia, in both the fasting and postmethionine-loading states, was defined as the level greater than either 2 SDs above the mean value in the control group or the 95th percentile. Defining a single upper percentile figure as the cutoff aids in reducing the variation between studies that might occur due to differences in their Hcy assays or in their laboratories' interpretation of an abnormal Hcy level. Therefore, whenever necessary, the original data were reanalyzed using the 95th percentile cutoff.

Unadjusted ORs were calculated for each study and then pooled using a random-effects model based on the DerSimonian and Laird technique.¹¹ A 95% CI was calculated for the summary OR using the same method. Heterogeneity across the studies was assessed by visual inspection, and a formal statistical test was based on the method described by Breslow and Day.¹² The hypothesis that the studies were not heterogeneous was rejected at P<.10.

Because most subjects in the previous meta-analysis of hyper-Hcy and atherosclerosis were younger than 60 years,¹ a similar "subgroup" analysis was performed herein.

Seventy-two studies were retrieved, of which 10^13-22 met all of the inclusion criteria. In 1 study,²² however, it was not possible to distinguish the results from patients with isolated venous thrombosis and those with mixed arterial and venous disease. In addition, the fasting Hcy results were not described for 49 of 107 patients with VTE. Therefore, this study was excluded from further analysis.

A description of the 9 case-control studies and their subjects is presented in Table 1. Case patients were selected on the basis of a history of a first VTE episode^17-21 or possibly multiple VTE episodes.^13-16 The methods used for subject recruitment and the number who refused to participate or were ineligible were rarely disclosed.

A total of 856 patients were included in this analysis, with a male-to-female ratio of approximately 1:1.4 and a mean age of 56.7 years (range, 17-91 years). The control subjects numbered 1553, with a higher male-to-female distribution of 2.3:1 and a similar mean age (57.4 years; range, 16-87 years). Proper matching of patients with controls, at least for age and sex, was performed in 3 studies.^18,19,21 Seven studies^13,15,17-21 used objective methods to diagnose VTE, but the other 2^14,16 did not describe any formal technique. Most patients had deep vein thrombosis of the leg or pulmonary embolism, and a smaller proportion had thrombi of the retinal, cerebral, mesenteric, or hepatic veins. Five studies^{14,15,17,18,20} included an additional workup for other congenital thrombophilic disorders, including protein C, protein S, and antithrombin III deficiency. Testing was done in 4 studies^15,18,20,21 for factor V Leiden or activated-protein C resistance. The presence of a concomitant thrombophilic disorder was adjusted for in the analysis of only 1 study.²⁰

Most studies measured Hcy levels, using high-performance liquid chromatography, at least several months after the last VTE episode. One study¹³ adequately assessed red blood cell folate levels, whereas serum folate and vitamin B₁₂ levels were determined in 6 studies.^{13,15-17,19,20} A deficiency of vitamin B₁₂ or folate did not usually lead to any adjustments or subject exclusion, although these levels were rarely abnormal. Most investigators had ensured that renal function was normal during subject recruitment.

The pooled individual ORs from the 9 studies in which the Hcy level was measured in the fasting state are presented in Table 2 and Figure 1. The OR of each study was greater than 1.0, and statistical significance for a positive association was seen in 5 of the 9 studies. The pooled OR was 2.95 (95% CI, 2.08-4.17; 2-sided P<.001) for VTE in the presence of fasting hyper-Hcy. The test for heterogeneity supported the qualitative trend (P=.50). For the aforementioned study²² excluded from this meta-analysis, the approximate OR was 1.80 (95% CI, 0.41-7.84), which would not have significantly altered the summary OR if it had been included.

The Hcy levels were examined in 5 studies^13,15-17,19 following methionine loading. All results trended in the same direction as those for the fasting state (Figure 2). The pooled OR for these studies was 2.15 (95% CI, 1.20-3.85; 2-sided P=.01) (Table 3), with both qualitative and statistical evidence against heterogeneity (P=.65).

The inclusion of only studies with patients with VTE before age 60 years^13-15,17 generated an even higher pooled OR of 4.13 (95% CI, 1.25-13.72) for fasting hyper-Hcy, without evidence for heterogeneity (P=.25).

This meta-analysis was designed to explore the relationship between hyper-Hcy and the risk for VTE. A significant positive association was found across all 9 case-control studies for both the fasting and postmethionine-loading states. The magnitude of association was high, with an OR of 2.95, which is statistically and perhaps clinically significant. This finding was even more apparent for patients younger than 60 years.

The application of a meta-analysis to nonexperimental studies is most useful when several small studies share consistent findings and is a way of exploring for heterogeneity between studies.²³ The present meta-analysis demonstrated the same trend for all 9 studies. A random-effects model for meta-analysis considers the variance both within and between studies. The latter is often conservative in its estimation of the pooled effect size, especially in the presence of significant heterogeneity between studies,²³ which was not observed in the studies analyzed here. Publication bias, therefore, is less likely to play a major role in the apparent association between Hcy levels and VTE.⁹

Several possible explanations for nonsignificant heterogeneity were identified in this meta-analysis, which was a second goal of this overview. One likely source was the poor matching of patients with controls, even for age and sex. Several studies, but not all, excluded patients with other thrombophilic causes, such as protein C, protein S, or antithrombin III deficiency, as possible confounders. This is probably most important for activated-protein C resistance (factor V Leiden) because 2 studies^18,21 observed a compounded risk for VTE in the presence of hyper-Hcy and activated-protein C resistance.

Although the summary risk estimate (OR, 2.95) is probably valid, correcting for unaccounted-for variables and sources of bias could potentially nullify this finding. Larger historical and nested-cohort studies using strict criteria for diagnosing VTE and measuring Hcy levels could establish whether this finding is valid.

The data on Hcy and the risk of atherosclerosis were collected primarily from studies of persons younger than 60 years.¹ In the present meta-analysis, the mean age of the patients was about 57 years. By including only studies of patients with VTE before age 60 years, the pooled OR for fasting hyper-Hcy rose to 4.13 (95% CI, 1.25-13.72). This suggests that hyper-Hcy may be particularly important in the development of VTE in younger adults, although such a hypothesis also requires further exploration.

In most studies, the Hcy level was analyzed using high-performance liquid chromatography, a technique that has become the most widely accepted and reliable method for Hcy measurement.^24,25 That the pooled ORs were similar for Hcy levels measured in either the fasting or the postmethionine-loading state is of interest. It has been argued that fasting Hcy measurement is a satisfactory method for testing for common Hcy defects, particularly because of its ease of performance and little loss of accuracy compared with measuring Hcy levels following methionine challenge.²⁶

Until better data are available, the measurement of the plasma Hcy level as part of the thrombophilic workup should depend on the local availability and reliability of the test. The laboratory responsible for Hcy analysis should have reference values established from its own control population (those without arterial or venous disease) and should be consistent in its use of the high-performance liquid chromatography technique.

Although the level of evidence implicating hyper-Hcy as a possible risk factor for VTE seems similar to that for the heterozygous states of antithrombin III²⁷ and protein C deficiency,²⁸ weighing their relative contributions to the risk of VTE is difficult. Clinicians need to decide whether identifying hyper-Hcy in young patients with VTE will alter their clinical management, both by determining the risk for VTE recurrence and in the use of anticoagulants for long-term prophylaxis or during high-risk states (eg, perioperatively or during pregnancy). If the answer is yes to these questions, then Hcy measurement is probably worthwhile.

Future observational studies should also match patients with control subjects by age, sex, and the setting in which the VTE arose (idiopathic vs situational [ie, postoperative or following prolonged immobilization]). Researchers should consider whether they wish to exclude patients with other known thrombophilic causes or to examine their interaction with hyper-Hcy. Secondary causes of hyper-Hcy, such as vitamin B₁₂ or folate deficiency and renal insufficiency, need also to be considered. A research focus on treatment is also needed to establish whether supplementation with folic acid and other B vitamins can substantially reduce Hcy levels and the risk of VTE recurrence.^29,30

Accepted for publication March 5, 1998.

William Geerts, MD, provided helpful comments during the preparation of the manuscript.

Reprints not available from the author.