Long-Term Homocysteine-Lowering Treatment With Folic Acid Plus Pyridoxine Is Associated With Decreased Blood Pressure but Not With Improved Brachial Artery Endothelium-Dependent Vasodilation or Carotid Artery Stiffness: A 2-Year, Randomized, Placebo-Controlled Trial

In April 1993, we started a hyperhomocysteinemia screening program among consecutive patients with clinically manifest peripheral, cerebral, or coronary occlusive disease with onset before the age of 56 years or a history of obstetric complications, as described in detail elsewhere.^10,11 From April 1993 to January 1995, we screened 737 patients, 167 of whom had postmethionine hyperhomocysteinemia: 96, 34, and 35 with peripheral, cerebral, and coronary arterial occlusive disease, respectively, and 2 with a history of obstetric complications. The siblings (n=652; 1 to 13 siblings per index patient; 628 of whom were alive at the time of our study) of these 167 patients were then invited for further studies; 475 (72.9%) agreed to participate. We excluded siblings who had a history of venous thrombosis, myocardial infarction, stroke, or peripheral arterial occlusive disease; those younger than 18 years or older than 65 years; pregnant women or women who were planning pregnancy; and those who had impaired renal function (serum creatinine >150 mmol/L) or liver function (abnormal serum transaminase concentrations or presence of physical signs). Thus, a total of 25 siblings (5.3%) were excluded. All other siblings (n=450; 167 families) underwent a methionine loading test. None used vitamin supplements. We subsequently invited all subjects with postmethionine hyperhomocysteinemia (n=125; 27.8%) to participate in the trial; 89 (71.2%) agreed and were available. Of those with a normal postmethionine plasma homocysteine level (n=325; 72.2%), we randomly invited 73 subjects, 50 (68.5%) of whom agreed to participate. These 139 participants were stratified according to family of origin and the presence of postmethionine hyperhomocysteinemia. Within each stratum, subjects were randomized by random computerized digit sampling (Figure). All participants were then treated orally with folic acid (5 mg daily) plus pyridoxine (vitamin B₆; 250 mg daily) or placebo medication for 2 years in double-blind fashion. The hospital pharmacist prepared placebo and vitamin capsules in such a way as to give them the same appearance and taste and executed the treatment assignment. The randomization code was stored at the Department of Pharmacy, to which conductors of and participants in the study had no access. At baseline and after 1 and 2 years, we collected demographic and clinical data and measured blood pressure, brachial artery endothelium-dependent dilatation, and stiffness of the common carotid artery and conducted exercise electrocardiography tests. The methionine loading test was repeated after 20 months.

The assessment of general demographic, clinical (including the exercise electrocardiography test according to the modified Bruce protocol¹²), and laboratory data (including the methionine loading test) has been described previously.^10,11

Hemodynamic studies and blood sampling were performed on separate days. The systolic, mean, and diastolic blood pressures (in mm Hg) were recorded in the left upper arm at 10-minute intervals with an oscillometric blood pressure-measuring device (Colin Press-Mate BP-8800). Pulse pressure was calculated as systolic minus diastolic blood pressure (in mm Hg). The mean of 4 blood pressure measurements was used in statistical analyses.

Endothelium-dependent and -independent vasodilation of the brachial artery (flow-mediated vasodilation and nitroglycerin-induced vasodilation, respectively) were determined with ultrasonography in combination with an arterial wall-movement detection system that measures arterial diameter in time¹³; the exact protocol has been described elsewhere.^4,14 In short, we used 4 minutes of forearm circulatory occlusion followed by a diameter recording after 45 to 60 seconds for measurement of endothelium-dependent vasodilation; for endothelium-independent vasodilation, we sublingually administered 400 μg of nitroglycerin and measured brachial artery diameter 5 minutes later. In our hands, short-term (1 week) coefficients of variations were 4.6% for baseline diameter, 5.5% for flow-mediated vasodilation, and 7.7% for nitroglycerin-induced vasodilation.¹⁵

The same ultrasound technique was used for the determination of common carotid artery diastolic diameter and the difference between systolic and diastolic diameter (ie, distension), which were then combined with brachial artery pulse pressure measurements to calculate compliance and distensibility coefficients. Compliance coefficient is defined as (π× diameter)× [distension/(266.6×pulse pressure)], in mm²/kPa, and distensibility coefficient as (2×distension)/(diameter×pulse pressure×0.01333), in 1/Pa.⁴ In our hands, intrameasurement intraobserver variability (coefficients of variation) in the 3 recordings of 1 ultrasound measurement was 1.0% for diameter and 3.8% for distension. Intraobserver intersession (50 days; n=9) variability was 2.3% for diameter, 9.7% for distension, 9.7% for distensibility coefficient, and 8.3% for compliance coefficient.

Assessment of all outcome variables was done without knowledge of treatment allocation or of any clinical data. The trial was approved by the ethics committee of the University Hospital Vrije Universiteit, and written informed consent was obtained from all participants.

Linear generalized estimating equations^16,17 were used to analyze the influence of therapy with folic acid plus pyridoxine on the longitudinal development of the outcome variables. This technique is suitable for the analysis of longitudinal data because it takes into account that repeated observations within 1 subject are correlated. Generalized estimating equations are “pooled” analyses of within-patient and between-patient relationships, both of which represent important aspects of any treatment effect. The magnitude of the regression coefficient thus obtained is usually somewhat higher than the average difference in the changes in the outcome measure between the treatment and placebo group, because the latter only reflects the between-patient relationship.

Analyses were performed on an intention-to-treat basis with absolute values of continuous variables. Primary analyses used treatment (vitamin/placebo) as the key independent variable (model 1). Adjustments were made for the baseline value of the outcome variable only (model 2); as in model 2 with age and sex added, and mean arterial pressure in the case of common carotid artery properties and pulse pressure (model 3); and as in model 3 with baseline pyridoxine, folate, and homocysteine (fasting or after methionine) concentrations added (model 4). Model 5 was based on model 3 with further adjustment for LDL and HDL cholesterol and triglyceride concentrations at baseline and first and second follow-up, as well as baseline tobacco smoking habit (yes/no) and body mass index. Mean arterial pressure was entered into models with common carotid artery properties as an outcome because of the direct physical effects of blood pressure on arterial properties.¹⁸ Additional analyses were performed to investigate whether the use of antihypertensive medication or the baseline presence of hypertension was associated with any of the outcomes, first by adjustment for the presence of hypertension or the use of antihypertensive medication, and second by exclusion of all subjects with hypertension. For these analyses, hypertension was defined according to the World Health Organization/International Society of Hypertension 1989 definition, which was in use when this trial was designed, ie, with 160 and 95 mm Hg as cutoffs for systolic and diastolic blood pressure, respectively, or by the use of antihypertensive medication. We also analyzed whether adjustment for baseline serum lipoprotein(a) affected the results. Finally, analyses were done on an on-treatment basis. With respect to these analyses, compliance with vitamin treatment was defined as an increase at the first and second follow-ups of plasma folate and pyridoxine by >500% and >300%, respectively. Compliance with placebo treatment was defined as a maximal increase of folate and pyridoxine of <2 SDs above the baseline mean. To investigate the need for separate analyses of subjects with and without baseline postmethionine hyperhomocysteinemia, we also added the interaction term “treatment times presence of baseline postmethionine hyperhomocysteinemia” to the main analyses.

The study had 80% power at α=0.05 to detect, between the placebo- and vitamin-treated groups, a 2.2-mm Hg difference in systolic blood pressure, a 1.6-mm Hg difference in diastolic blood pressure, a 2.3-mm Hg difference in pulse pressure, a 3.1% point difference in brachial artery flow-mediated vasodilation, a 0.09-mm²/kPa difference in the compliance coefficient, and a 2.3 1/Pa difference in the distensibility coefficient of the common carotid artery.

Probability values <0.05 were considered statistically significant. The Statistical Package for Interactive Data Analysis (SPIDA, version 6.05, North Ryde) was used for generalized estimating equations analyses and SPSS (SPSS 9.0 for Windows, SPSS Inc) for other analyses.

Table 2 shows that vitamin treatment was associated with an increase in plasma folate and pyridoxine (both 7-fold versus placebo at the second follow-up). Fasting and postmethionine homocysteine concentrations decreased by 40.1% (95% CI 27.3% to 53.0%) and 29.7% (95% CI 19.2% to 40.1%) versus placebo (all P<0.001). Compliance with vitamin treatment, as defined in Methods, was 100%. Compliance with placebo medication was 94% at the first follow-up and 92.5% at the second follow-up, which suggests additional vitamin use in the placebo-treated group.

Baseline blood pressure was slightly higher in the placebo-treated group than in the vitamin-treated group (Table 3). Model 1 (Table 4) thus may overestimate the effects of vitamin treatment. Models 2 to 5, therefore, were adjusted for baseline blood pressure. Vitamin treatment, compared with placebo, was associated with decreased systolic (−3.7 mm Hg; 95% CI −6.8 to −0.6 mm Hg; P=0.02) and diastolic (−1.9 mm Hg; 95% CI −3.7 to −0.02 mm Hg; P=0.04) blood pressure (Table 3; Table 4, model 2). Additional adjustment for age, sex, and baseline pyridoxine, folate, or homocysteine concentrations gave similar results (Table 4, models 3 and 4). The point estimates of these treatment effects increased to −4.6 mm Hg for systolic and −2.4 mm Hg for diastolic blood pressure after further adjustment for LDL cholesterol, HDL cholesterol, and triglyceride concentrations; tobacco smoking; and body mass index (Table 4, model 5). Vitamin treatment was also associated with decreased pulse pressure (−2.3 mm Hg; 95% CI −4.2 to −0.4 mm Hg; P=0.02 versus placebo; Table 4, model 2), but this was not independent of mean arterial blood pressure (P=0.16 after adjustment).

Vitamin treatment, compared with placebo, was not associated with a significant change in brachial artery endothelium-dependent or -independent vasodilation (Tables 3 and 4). Further adjustments gave similar results. The increase in peak systolic flow velocity during reactive hyperemia at the first and second follow-ups did not differ between the placebo and vitamin-treatment groups (94.0% versus 94.3% [P=0.96] and 95.1% versus 99.6% [P=0.53], respectively).

Compared with placebo, vitamin treatment was not associated with a significant change in common carotid artery compliance coefficient or distensibility coefficient or in common carotid artery diameter or distension (Tables 3 and 4). Analyses gave similar results when adjusted for lipoprotein(a), the use of antihypertensive medication, or the presence of hypertension at baseline. In addition, after exclusion of individuals with hypertension, the point estimates of the effect of vitamin-treatment compared with placebo on blood pressure in models 2 through 5 or on any of the other outcomes were not materially different. Results of the on-treatment analysis were similar to those of the intention-to-treat analysis for all outcome variables (data not shown). Analyses with the interaction term “treatment times presence of baseline postmethionine hyperhomocysteinemia” were not materially different from those without the interaction term. The interaction term was not statistically significant in any of the analyses (data not shown).

We performed an additional analysis with the occurrence of abnormal exercise electrocardiography tests as the dependent variable and systolic or diastolic blood pressure as time-dependent covariates to test whether the vitamin-treatment-associated reduction of the occurrence of abnormal exercise electrocardiography tests¹⁰ was influenced by changes in blood pressure (Table 5). Adjustment for the change in systolic blood pressure changed the estimate of the effect of vitamin treatment (odds ratio) from 0.41 to 0.53. A decrease in systolic blood pressure was itself associated with a decreased rate of abnormal exercise electrocardiography tests (odds ratio, 0.98 per 1-mm Hg systolic blood pressure decrease). Adjustment for the change in diastolic blood pressure did not affect the estimate of the effect of vitamin treatment.