Homocysteine, MTHFR and risk of venous thrombosis: a meta-analysis of published epidemiological studies

Authors


M. den Heijer, Department of Endocrinology (531), Radboud University Nijmegen Medical Centre, PO Box 9101, 6500 HB Nijmegen, the Netherlands.
Tel.: +31 24 3614599; fax: +31 24 3618809; e-mail: m.denheijer@endo.umcn.nl

Abstract

Summary. Context: It has been suggested that elevated total plasma homocysteine levels are associated with the risk of venous thrombosis. Objective: To assess the relationship of homocysteine and the MTHFR 677TT genotype and the risk of venous thrombosis by conducting a meta-analysis of all relevant studies. Data sources and selection: Studies (case–control or nested case–control) were identified by searches of electronic literature for relevant reports published before July 2003 on homocysteine and the MTHFR 677TT genotype and venous thrombosis as an end-point, by hand-searching reference lists of original articles (including meta-analyses) on this topic and by contact with investigators in the field. Data extraction: A meta-analysis of 24 retrospective (n = 3289 cases) and three prospective studies (n = 476 cases) was carried out to examine the association of homocysteine with venous thrombosis. A meta-analysis of 53 studies (n = 8364 cases) of the MTHFR 677TT genotype (that increases homocysteine) was carried out to assess if this association is causal. Data synthesis: A 5 µmol L−1 higher measured homocysteine level was associated with a 27% (95% CI: 1–59) higher risk of venous thrombosis in prospective studies and a 60% (95% CI: 10–134) higher risk in retrospective studies. The 677TT genotype was associated with a 20% (95% CI: 8–32) higher risk of venous thrombosis compared with the 677CC genotype. In contrast with non-American studies, the 677TT genotype had no effect on venous thrombosis in North America, due probably to the higher intake of folate and riboflavin in North America. Conclusion: This meta-analysis of prospective and retrospective studies demonstrates a modest association of homocysteine with venous thrombosis. The elevated risk associated with the MTHFR 677TT genotype provides some support for causality.

Introduction

Venous thrombosis, including deep-vein thrombosis and pulmonary embolism, is an important cause of morbidity and mortality, particularly in older people [1]. Most cases of venous thrombosis arise due to prolonged immobilization, major surgery, trauma or cancer, but genetic or acquired hemostatic abnormalities, including elevated plasma homocysteine levels, have also been implicated [2]. The initial epidemiological evidence that examined the association between homocysteine and venous thrombosis was derived from retrospective case–control studies (in which blood for homocysteine measurements was collected after onset of thrombotic events in cases) [3–7], but it was not possible to ascertain whether the higher homocysteine levels caused the thrombotic event or was a consequence of it. Subsequently, prospective studies (in which blood for homocysteine measurements was collected before the onset of the thrombotic events) appeared to confirm these findings, but the weaker results raised questions about causality [8–10].

Elucidation of an association between a genetic variant associated with elevated homocysteine levels and venous thrombosis might be informative about the hypothesis that higher levels of homocysteine plays a causal role in the occurrence of venous thrombosis [11]. Studies of genetic variants that affect homocysteine levels would reflect long-term differences in homocysteine, and be independent of confounding and concerns about reverse causality [11]. A common polymorphism exists in the gene that encodes the catalytic domain of the MTHFR enzyme, in which a C > T substitution at position 677 (referred to as 677TT) results in a substitution of alanine to valine [12]. The single amino acid substitution results in impaired folate binding and reduced activity of the MTHFR enzyme [12]. About 10–12% of Caucasians of Northern European descent carry the 677TT genotype for MTHFR and have about 25% higher homocysteine levels than those with the 677CC genotype. The effect of the MTHFR 677TT genotype on homocysteine levels varies according to folate status, whereby the difference in homocysteine levels between those with TT vs. CC genotypes is greater among those with low folate or riboflavin status than in those with normal or high levels of these vitamins [13,14].

Individual studies and a previous meta-analysis included too few cases to produce reliable evidence for or against an association of this polymorphism with venous thrombosis [15,16].

The aim of this study was to examine the association between homocysteine and venous thrombosis by conducting a meta-analysis of epidemiological studies on this topic and to assess if such associations are causal by conducting a meta-analysis of studies of the MTHFR 677TT genotype and venous thrombosis.

Methods

Data sources and study selection

Eligible studies were identified by searching the electronic literature (MEDLINE) for relevant reports published between 1990 and July 2003 (using the terms homocysteine, hyperhomocysteinemia, MTHFR, methylenetetrahydrofolate reductase and venous thrombosis, pulmonary embolism, deep vein thrombosis or venous thromboembolism), by hand-searching reference lists of original articles (including meta-analyses) on this topic and by personal contact with relevant investigators in the field. The MEDLINE search yielded 443 studies and 70 retrospective or prospective studies were eligible for inclusion [4,6,8–10,17–81]. Studies that primarily involved cerebral vein thrombosis or Budd–Chiari syndrome were excluded, as were studies carried out in diseased populations (e.g. inflammatory bowel syndrome, systemic lupus erythematosis or Behçet's syndrome). Studies with incomplete data or zero observations in cases or controls were excluded [3,5,26,82–88].

Among the 70 individual studies that were eligible, 27 provided data on homocysteine and venous thrombosis, 53 additional studies provided data on the MTHFR C677T genotype and venous thrombosis and 10 studies that had data on homocysteine also provided data on MTHFR. It was not possible to distinguish studies that included first-time or recurrent venous thrombosis from the published reports.

Data extraction and data synthesis

For the meta-analysis of homocysteine and venous thrombosis, information was extracted from each report on: the odds ratio (or risk ratio) and its confidence interval; the study design — prospective (or nested case–control) or retrospective (case–control) — and the number of cases and controls. If the odds ratios (and 95% CI) were not provided, they were calculated from the number of cases and controls above and below a particular cut-off point using Woolf's method [89].

Different studies reported odds ratios on the basis of different cut-off levels, including comparisons of top and bottom thirds, quarters, fifths and top 95th vs. the rest. To estimate the log odds ratio for 5 µmol L−1 increase in homocysteine, information about the normal distribution was used to calculate the expected mean level of homocysteine in the groups being compared. For example, the mean value in the top fifth of a normal distribution is 1.4 standard deviations (SD) above the mean, and the mean in the bottom fifth is 1.4 SD below the mean. Hence, the odds ratio comparing the top vs. the bottom fifth is also the odds ratio for a 2.8 SD difference in homocysteine. Most of the studies reported odds ratios for the top 95th percentile vs. the rest, giving an odds ratio for a 2.17 SD difference in homocysteine levels [90,91]. Wherever possible, the SD in controls was used to estimate the difference in µmol L−1 between the groups being compared in a study, but if this was not reported then a weighted average of the SDs from the studies that did report it was used. Assuming the association was log-linear, it was then possible to calculate the odds ratio for a unit change in homocysteine and hence for a 5 µmol L−1 increase in homocysteine.

None of the analyses were corrected for the effects of regression dilution; that is, they relate risk to the measured values of homocysteine, rather than to the long-term average (i.e. ‘usual’) homocysteine level [92,93], because of the limited data reported on the time between the thrombotic event and blood collection. Thus, all analyses estimate the odds ratio for a 5 µmol L−1 increase in measured homocysteine, which is approximately equivalent to a 3.5 µmol L−1 increase in usual homocysteine (i.e. after correction for regression dilution) for prospective studies (assuming a mean time to event of 5 years) and about 4.5 µmol L−1 for retrospective studies (assuming an interval between the event and blood collection of less than 1 year) [93].

For analyses of MTHFR C677T and venous thrombosis, data were collected on the frequency of CC, CT and TT genotypes in cases and controls (or calculated from the allele frequency), and the odds ratios (and 95% CI) for TT compared with CC genotypes for MTHFR were estimated using Woolf's method [89]. The studies were classified by continent as a surrogate for folate status (because the folate status of the North American population is more favorable compared with that in Europe or elsewhere) [94,95]. For all analyses, summary estimates were obtained by taking an inverse variance weighted average of the log odds ratios from individual studies. Heterogeneity was assessed using standard χ2 tests.

Results

Homocysteine and venous thrombosis

Data were obtained from 24 retrospective studies (n = 3289 cases) [4,6,15,21–41] and three prospective or ‘nested’ case–control studies (n = 476 cases) [8–10]. Figure 1 shows the odds ratios of venous thrombosis associated with a 5 µmol L−1 increase in measured plasma total homocysteine, separately for individual studies and subtotals for prospective and retrospective studies when taken together. There was significant heterogeneity between the results of retrospective studies (inline image = 45.2; P = 0.004), but there was no heterogeneity between the results of prospective studies (inline image = 1.87; P =0.39), although with only three studies there was limited power with which to detect this. A 5 µmol L−1 increase in measured homocysteine was associated with a 27% (OR 1.27; 95% CI: 1.01–1.59) increased risk of venous thrombosis in prospective studies and a 60% (OR 1.60; 95% CI: 1.10–1.34) increased risk in retrospective studies.

Figure 1.

The odds ratio of venous thrombosis for a 5 µmol L−1 increase in the measured concentration of plasma total homocysteine in individual studies. The combined odds ratio for the subtotals for each study design and their 95% CI are indicated by the diamonds. While data for three studies has been excluded in the graph of individual studies [31,35,37], the data from these studies are included in the subtotals. The size of the diamond is inversely proportional to the variance of the log odds ratio and the width of the diamond represents the 95% CI.

MTHFR and venous thrombosis

Data on the C677T polymorphism were obtained from 53 studies involving 8364 cases with venous thrombosis and 12 468 controls [10,27,28,30,35–37,40–81]. Figure 2 shows the results of individual studies and when taken together in groups of studies classified by continent of origin as Europe, North America and elsewhere. The individual studies within each subgroup are ordered by the number of cases to evaluate publication bias. The absence of asymmetry in the distribution of the odds ratios in each continent (whereby smaller studies might yield more extreme results) excludes evidence of publication bias.

Figure 2.

The odds ratio of venous thrombosis for TT vs. CC genotype for MTHFR. The symbols and convention are the same as for Fig. 1.

Among the 53 studies, 30 were carried out in Europe, 11 in North America and 12 were carried out elsewhere. Overall, the 677TT genotype was associated with a 20% (OR 1.2; 95% CI 1.08–1.32) higher risk of venous thrombosis compared with the 677CC genotype, with no significant differences between the 53 studies (inline image = 55.2; P = 0.36). However, although not statistically significant there was an apparent difference between the studies carried out in Europe, North America or elsewhere.

The 677TT genotype was associated with a 15% (OR 1.15; 95% CI: 1.02–1.30) increased risk of venous thrombosis in Europe and a 60% (OR 1.60; 95% CI: 1.27–1.02) increased risk in studies carried out elsewhere, but had no effect on venous thrombosis in North America (OR 1.03; 95% CI: 0.82–1.29).

Discussion

This meta-analysis of prospective and retrospective epidemiological studies demonstrates an association between an elevated plasma total homocysteine level and venous thrombosis. The odds ratio of 27% higher risk of venous thrombosis for a 5 µmol L−1 increase in homocysteine observed in prospective studies was only half as extreme as that in retrospective studies. As is the case for coronary heart disease and stroke [92], the results of prospective (or nested case–control) studies are likely to provide more reliable estimates of the risks of venous thrombosis associated with differences in homocysteine levels than the retrospective studies. However, it is likely that the 27% higher risk of venous thrombosis for a 5 µmol L−1 increase in homocysteine observed in prospective studies may be underestimated due to the effects of regression dilution (where the measured homocysteine levels may not reflect the usual homocysteine level during an observation period of about 10 years) to a greater extent compared with retrospective studies [92–94].

Both prospective and retrospective studies are influenced by confounding due to age, sex, smoking and body mass index. In this meta-analysis, the unadjusted odds ratios were abstracted from each report because the adjusted odds ratios were not routinely provided. Hence, it was not possible to control for confounding factors (as data were not available for individuals), but most of the studies had selected controls that were approximately age- and sex-matched. Analysis of individual studies where this information was provided suggests that the extent of confounding due to body mass index or smoking was not substantial. For the same reason it was not possible to distinguish between type of thrombosis (deep-vein thrombosis or pulmonary embolism) or even first-time and recurrent thrombosis. While most studies focused on first-time thrombosis, some included recurrent venous thrombosis or both recurrent and first-time thrombosis.

Genetic variants affecting homocysteine levels precede the onset of disease and reflect long-term differences in homocysteine levels that are not influenced by reverse causality. Moreover, apart from effect modification by dietary intake of folate and riboflavin, it is unlikely that the effects of these genotypes on venous thrombosis will be influenced by lifestyle or other confounders [11,95]. This meta-analysis demonstrated that the 677TT genotype was associated with a 20% increased risk of venous thrombosis. If the studies just from outside North America are combined (to minimize the effect of folate and riboflavin supplementation), the overall result is little altered [OR (95% CI) 1.23 (1.11–1.37)].

Most studies included in this meta-analysis involved less than a few hundred cases, and so the confidence intervals around the odds ratios were wide. The present meta-analysis includes more than twice the number of cases with venous thrombosis compared with those which were included in previous meta-analyses on this topic [15,16] and provides more reliable evidence about the importance of the 677TT genotype for venous thrombosis.

The odds ratios of venous thrombosis for the approximately 2.5 µmol L−1 (i.e. 25%) difference in measured levels of homocysteine associated with the TT vs. CC genotypes are approximately equivalent to the estimates derived from prospective or nested case–control studies for a 3.5 µmol L−1 difference in usual homocysteine levels. The concordance of the risk estimates from prospective (nested case–control) studies and genetic studies of venous thrombosis support the hypothesis that the association of homocysteine with venous thrombosis is causal.

The risk estimates for 677TT genotype and venous thrombosis obtained from studies carried out in North America differed from those carried out in Europe and elsewhere, which may be explained by the higher dietary intake of folate and riboflavin in North America compared with Europe [20,95,96]. Hence, in addition to providing support for a causal association between homocysteine and venous thrombosis, the genetic studies demonstrating effect modification in North America suggest that higher vitamin intake may have a protective effect, but these inferences require confirmation in large randomized trials. A meta-analysis of the short-term trials of folic acid-based vitamin supplements showed that a daily dose of folic acid of between 0.5 and 5 mg was associated with a reduction in homocysteine levels of 25% and that vitamin B12 was associated with a further reduction of about 7%[97]. This difference in homocysteine of 25% that is typically achieved with folic acid supplements is comparable to the difference in homocysteine between 677TT and 677CC genotypes. If it could be shown that folic acid supplementation reduced the risk of venous thrombosis by 20%, this could be of substantial relevance for clinical practice and for public health. In clinical practice, measurement of homocysteine levels (and subsequent determination of folate and vitamin B12 levels if homocysteine levels are elevated) may be indicated for individuals with unexplained idiopathic venous thrombosis or recurrent venous thrombosis, or venous thrombosis occurring at early age or at an unusual site [98]. Vitamin supplements may be indicated for those individuals who are identified with evidence of folate or vitamin B12 deficiency [98], but whether such supplements may reduce the risk of recurrent venous thrombosis is not known. Large trials of folic acid-based vitamin supplements are currently testing whether lowering homocysteine levels may reduce the risk of vascular disease and further evidence is required to assess the relevance of folic acid supplementation for the prevention of venous thrombosis [99,100].

Acknowledgements

We thank Paul Sherliker for producing the figures. Robert Clarke and Sarah Lewington are supported by grants from the British Heart Foundation and Medical Research Council, European Union BIOMED (BMH4-98–3549). Martin den Heijer is recipient of a VENI-grant from the Netherlands Organization for Scientific Research.

Ancillary