We found that long-term homocysteine-lowering with folic acid, vitamin B6 and vitamin B12 did not reduce CIMT or increase FMD in individuals with a history of stroke. When we included our results with data from similar randomised trials in a meta-analysis the estimated pooled effect was a modest improvement in both CIMT and FMD.
The 95% confidence intervals for the effect of B-vitamin supplementation on CIMT (-0.04 mm, 0.06 mm) and FMD (-0.73%, 1.73%) in our study overlap with confidence intervals from the meta-analyses of -0.25 mm to -0.01 mm and 0.73% to 2.23% (current data excluded), indicating that our results are consistent with pooled data from similar randomised studies. In a recent systematic review and meta-analysis, Lorenz et al. report that a difference of 0.1 mm in CIMT predicts a 10% to 15% reduction in the risk of future MI and a 13% to 18% reduction in the risk of future stroke [16
]. To the best of our knowledge, a similar calculation for the predicative value of an absolute difference in FMD has not been reported. A recent study by Yeboah et al. reports a 9% reduction (95% CI 1%, 17%) in the risk of a future cardiovascular event (cardiovascular disease death, myocardial infarction, stroke, congestive heart failure, claudication, angioplasty or cardiac bypass graft surgery) per unit SD in FMD, an absolute difference of approximately 1.2% [17
]. Our meta-analyses estimates for the effect of B-vitamin supplementation on CIMT (-0.1 mm) and FMD (+1.4%) thus appear to be consistent with a significant clinical benefit, a result that conflicts with randomised trial evidence that homocysteine-lowering treatment does not significantly reduce mortality or morbidity [12
]. There are several possible explanations for this discrepancy.
One explanation is that the results of the CIMT meta-analysis are unreliable. Only a small number of studies have measured the effect of B-vitamin treatment on CIMT (n = 768 subjects, including the current study), the results show considerable heterogeneity (I2
= 85%), are strongly influenced by a single study and are likely to have been influenced by publication biases. A large study (n = 819) that has not yet published detailed data has recently reported finding no difference between placebo and treatment groups in mean CIMT after three years of treatment with folic acid [23
]. Adding this study to the CIMT meta-analysis would double the subject numbers and significantly reduce the estimated effect of homocysteine-lowering treatment.
By contrast with the CIMT data, the FMD meta-analysis results appear to be relatively robust with no clear evidence of a publication bias. A similar meta-analysis has also recently reported a positive effect of homocysteine-lowering treatment on FMD (1.08%, 95%CI 0.57, 1.59%, n = 732) [24
]. A possible explanation for the discrepancy between the apparent improvement in FMD in the meta-analyses and the lack of improvement in clinical trial event rates is that the laboratory studies and the intervention trials have been conducted in different subject groups. Most clinical trials have recruited subjects with existing vascular disease, either coronary artery [19
] or cerebrovascular disease [18
], or with strong vascular risk factors such as diabetes [19
] or renal impairment [22
]. While the effect of homocysteine-lowering on FMD has also been assessed most frequently in subjects with existing disease or vascular risk factors, many studies have been conducted in healthy individuals.
It is possible that homocysteine-lowering treatment may prove more effective in individuals without established vascular disease, as a recent meta-analysis of clinical trial data suggests that the incidence of new ischemic stroke is significantly reduced by B-vitamin supplementation whilst recurrent stroke is not [11
]. However, our sub-group analysis indicates that B-vitamin treatment improves FMD in people with diabetes and coronary disease but not in healthy subjects. The discrepancy between the FMD data and the clinical trial results is thus not explained by the laboratory studies having been undertaken in healthy individuals and the clinical trials in subjects with established disease.
Our sub-group analysis also indicates that the significant improvements in FMD associated with short-term folic acid treatment are not sustained in longer-term studies (Figure ). Whether this observation reflects a true biological effect, is due to publication biases, or to systematic differences between short and long term trials with respect to the characteristics of participants or dose of folic acid dose is not clear. However, it raises the interesting question of whether long-term exposure to B-vitamins might have adverse effects on the vasculature that counteract any short-term improvement in endothelial function. The Heart Outcomes Prevention Evaluation (HOPE) 2 study found that vitamin-treated subjects were more likely to be admitted to hospital with unstable angina than those on placebo (RR 1.24, 95% CI 1.04, 1.49) [19
]. This finding might be attributed to chance were it not for evidence that B-vitamin therapy increases the risk of in-stent stenosis and the need for revascularisation following angioplasty [20
]. The Norwegian Vitamin Trial (NORVIT) in 3749 men and women with a recent myocardial infarction also found a small increase in the risk of the composite endpoint (recurrent myocardial infarction, stroke or sudden death attributed to coronary artery disease) in the vitamin group (RR 1.22, 95% CI 1.00, 1.50, p = 0.05) [27
]. Excess folic acid, through its growth-promoting effects, could conceivably accelerate intimal hyperplasia and smooth muscle cell proliferation in existing vascular lesions.
A more controversial explanation for the discrepancy between the improvements in FMD and CIMT in our meta-analysis and the negative results from the clinical trials is that FMD and CIMT are not particularly reliable or sensitive indicators of cardiovascular risk. Although FMD and CIMT are widely and enthusiastically accepted as surrogates for vascular events, the evidence linking these measurements with hard clinical endpoints is relatively weak and the small changes in FMD and CIMT associated with B-vitamin treatment in our meta-analyses may actually have limited independent prognostic significance [30