The Effect of Vitamin Therapy on the Progression of Coronary Artery Atherosclerosis Varies by Haptoglobin Type in Postmenopausal Women

The design and methods of the WAVE trial have been previously described (9). Briefly, WAVE was a prospective, randomized, double-blind trial that enrolled 423 postmenopausal women with at least one 15–75% coronary stenosis at baseline coronary arteriography. Patients were randomly assigned in a 2 × 2 factorial design to receive either 0.625 mg/day of conjugated equine estrogen (plus 2.5 mg/day of medroxyprogesterone acetate for women who had not had a hysterectomy), matching placebo and 400 IU of vitamin E twice daily plus 500 mg of vitamin C twice daily, or placebo. The mean interval between angiograms was 2.8 ± 0.9 years. The main outcome measure of the trial was the annualized mean change in the minimum luminal diameter (MLD) from baseline to the concluding angiogram of all qualifying coronary lesions averaged for each patient. WAVE qualifying segments were defined as segments with 15–75% stenoses at baseline or new lesions at follow-up. The annualized mean change in the MLD from baseline to concluding angiogram (or to intercurrent angiogram before revascularization) was available for all WAVE qualifying lesions and averaged for each patient. Stored plasma from the baseline exam was available on 403 of 423 (95.2%) of the initial WAVE cohort.

Haptoglobin phenotyping was performed from 10 μl of plasma by polyacrylamide gel electrophoresis according to established methods (24). A signature banding pattern is obtained from individuals who are homozygous for the 1 allele (Hp 1-1), homozygous for the 2 allele (Hp 2-2), or heterozygous at the haptoglobin locus (Hp 2-1). We have established 100% concordance between the haptoglobin phenotype as determined from plasma and the haptoglobin genotype as determined from genomic DNA by the polymerase chain reaction (25). Haptoglobin phenotyping was performed with no knowledge of the patient’s treatment status.

The Hardy-Weinberg principle of population genetics is used to determine whether two alleles are in a balanced equilibrium, indicating a lack of survival benefit for one allele relative to the other. To determine whether a population is in Hardy-Weinberg equilibrium, it is necessary to use the Hardy-Weinberg formula to calculate the genotype frequencies in the population. In this equation (p² + 2pq + q² = 1), p is defined as the frequency of the 1 allele and q is defined as the frequency of the 2 allele for a trait or phenotype controlled by a pair of alleles (1 and 2). In this equation, p² is the frequency of homozygous (1-1) individuals in a population, 2pq is the frequency of heterozygous (2-1) individuals, and q² is the frequency of homozygous 2 (2-2) individuals. To actually determine whether our study population was in Hardy-Weinberg equilibrium, the actual distribution of the three haptoglobin types in this study was compared with the distribution of the three haptoglobin types expected if the population was in Hardy-Weinberg equilibrium according to χ² test (26).

All analyses were performed using SAS statistical software (version 8.0; SAS Institute, Cary, NC). ANCOVA or χ² tests (as appropriate) compared baseline characteristics of patients segregated according to haptoglobin phenotype. Student’s t test was used to compare change in HDL and LDL. The effect of haptoglobin type on change in MLD was analyzed by an ANCOVA model. The differential effect of antioxidant therapy on changes in the MLD in individuals with the different haptoglobin phenotypes was assessed by an ANCOVA model, and the mean changes and SDs in each group were calculated. The baseline covariates, representing known cardiovascular disease risk factors and glycemic control, of age, hypertension, cholesterol level, diabetes status, and HbA_1c were included in all models. Univariate regression with MLD as a dependent variable against HRT as well as against the hormone-vitamin interaction was not statistically significant, and, therefore, HRT was not included in the models presented. We cannot distinguish between WAVE participants with type 1 and 2 diabetes because this information was not obtained upon enrollment in WAVE and is not in the WAVE database.

There were originally 423 women enrolled in the WAVE trial, 154 of which had diabetes. The analyses presented here were performed only on those 299 participants (113 of whom had diabetes) for whom both repeat angiography and a haptoglobin type was obtained. The number of participants for whom repeat angiography and/or haptoglobin phenotype was obtained is described in Table 1. The distribution of the haptoglobin types of those individuals who enrolled in WAVE but did not undergo repeat angiography was not different from those who did undergo repeat angiography.

In the population undergoing repeat angiography, the distribution of the three haptoglobin phenotypes in the entire cohort and in those individuals with diabetes was in Hardy-Weinberg equilibrium. There were no significant differences in the entire cohort or in the diabetic subgroup with respect to baseline demographic, angiographic, or clinical characteristics in participants with the three different haptoglobin phenotypes (Table 2).

In the entire study population, there was no difference in the change in MLD based on haptoglobin phenotype (P = 0.19). The average (±SD) change in MLD was −0.047 ± 0.199 for Hp 1-1, −0.026 ± 0.110 for Hp 2-1, and −0.048 ± 0.153 for Hp 2-2.

In the diabetic study population, there also was no difference in the change in the MLD based on haptoglobin phenotype (P = 0.12). The average (±SD) change in MLD was −0.071 ± 0.207 for Hp 1-1, −0.037 ± 0.100 for Hp 2-1, and −0.110 ± 0.188 for Hp 2-2.

The means and SDs for each vitamin/haptoglobin type, both before and after adjustment for baseline covariates, are shown in Table 3. Vitamin supplementation had a differential effect on MLD after adding a vitamin use/haptoglobin type interaction into the model (P = 0.035 for the interaction term between haptoglobin type and vitamin use on the change in the MLD). There was a significant benefit acquired on the change in the MLD with vitamin therapy as compared to placebo in individuals with Hp 1-1 (P = 0.049), whereas there was no significant benefit seen in Hp 2-1 and Hp 2-2 individuals. In Hp 2-2, there was a nonsignificant trend suggesting a negative effect of vitamin therapy compared with placebo. After adjusting for baseline covariates (age, hypertension, cholesterol, and diabetes status), the relative benefit or harm of vitamin therapy in Hp 1-1 or Hp 2-2 individuals, as compared with placebo, was unchanged. The means and SDs for each vitamin/haptoglobin type in the diabetic cohort, both before and after adjustment for baseline covariates, are shown in Table 3. This interaction of haptoglobin phenotype and vitamin therapy upon coronary artery stenosis progression was even more pronounced in the diabetic cohort (P = 0.003 for the interaction term between haptoglobin type and vitamin use on the change in the MLD). Not only was there a significant benefit demonstrated from vitamin therapy compared with placebo in Hp 1-1 individuals, but there was also a significant deleterious effect on MLD in those individuals with Hp 2-2 both before and after adjustment for baseline covariates (age, hypertension, cholesterol, and HbA_1c) compared with placebo. Moreover, we found a statistically significant trend effect between the number of haptoglobin 1 alleles and the change in MLD in diabetic subjects who were on vitamin therapy. In this group, the change in MLD had a significant linear association with the number of haptoglobin 1 alleles as described by linear regression analysis with adjustment for baseline covariates (P = 0.005). HRT had no effect on these outcome measurements, and univariate regression with MLD as a dependent variable against the HRT-vitamin interaction was not statistically significant.

Previous studies have suggested that antioxidant therapy may have a negative impact on the lipid profile and on the activity of lipid-lowering drugs (8,11). In the entire cohort, there was no significant difference in the change in LDL or HDL during the course of the study in those individuals receiving vitamin therapy or placebo. Moreover, there was no significant difference in the net change in LDL or HDL in any haptoglobin subgroup with or without diabetes (Table 4).

Although antioxidant therapy in animals has shown promise in slowing the progression of atherosclerotic lesions (4,5), prospective angiographic studies in humans have failed to corroborate these findings (7–9). In the original report of the overall results from the WAVE study, antioxidant vitamins were not shown to provide cardiovascular benefit; instead, a potential for harm was suggested (9). We have demonstrated one possible explanation for this paradox here, namely that the benefit of antioxidant therapy with vitamin C and E on progressive coronary artery stenosis may be restricted to women with the Hp 1-1 phenotype. These data are consistent with recent studies showing that antioxidant therapy only had a significant effect in lowering of LDL in Hp 1-1 individuals (23). It should be stressed that these findings were most prominent in diabetic subjects. Furthermore, only ∼17% of the original cohort (Hp 1-1 individuals) could have derived benefit from vitamin C and E therapy, whereas ∼37% of the original cohort (Hp 2-2 individuals) could have experienced harm from vitamin C and E therapy.

Why might vitamin E and vitamin C therapy be harmful in some haptoglobin types? Our results do not support a mechanism based on vitamin-induced differences in the lipid profile (8,11). One potential explanation is related to the pro-oxidant capacity of vitamin C in the presence of iron (12,14,16,17). Vitamin C promotes the reduction of Fe³⁺ to Fe²⁺ and consequently promotes the formation of hydroxyl radicals by the Fenton reaction (12,14). In the process, vitamin C is oxidized to dehydroascorbate and becomes a pro-oxidant. Serum levels of iron are highest in Hp 2-2 individuals (27). Serum levels of vitamin C are also dependent on haptoglobin type; the lowest levels occur in Hp 2-2 individuals (28). It has been proposed that the lower levels of vitamin C in Hp 2-2 individuals are due to an increased oxidation of vitamin C (28). Supplemental vitamin C may not be beneficial to individuals with Hp 2-2 because it may result in increased production of oxidized vitamin C. The fundamental reason why vitamin C is more rapidly consumed and converted into a pro-oxidant in Hp 2-2 individuals is not altered by vitamin therapy.

Why should the haptoglobin-dependent effect of vitamin therapy on angiographic stenosis be more pronounced in the WAVE diabetic subgroup? Hyperglycemia results in a marked increase in oxidative stress and directly enhances LDL oxidation (29). Mowri et al. (30) have shown that this glucose-enhanced oxidation of LDL is strictly metal-ion dependent. Similar to vitamin C, glucose auto-oxidation results in the reduction of Fe³⁺ to Fe²⁺ (30), thereby enhancing the ability of Fe to serve as a Fenton reagent and produce hydroxyl radicals. As haptoglobin type–dependent differences in oxidative stress and redox active iron seem to be most prominent in the diabetic state (20) and vitamin C therapy may be synergistic with hyperglycemia in promoting LDL oxidation via the Fenton reaction as discussed above, vitamin C supplementation in individuals with hyperglycemia and Hp 2-2 may be particularly deleterious.

Limitations in the WAVE study suggest caution in extrapolating these results immediately to the clinical arena at the present time. First, ∼20% of the original cohort recruited did not undergo follow-up angiography and only 84% of the patients assigned to the vitamin group actually took the drug, which may have produced a selection bias (9). Second, the primary end point of the WAVE study was coronary artery dimensions and not coronary events. Although coronary progression has been shown in previous angiographic trials to be a strong, independent predictor of future coronary events (31), the culprit coronary artery lesion in coronary events does not correlate well with the lesion that was most stenosed on an antecedent angiogram. Third, systemic markers of oxidative stress were not measured in WAVE; therefore, the efficacy of the antioxidant vitamins in reducing oxidation cannot be assessed (1). Fourth, the formulation of antioxidant vitamins used in WAVE was quite different from that used in other antioxidant trials (32,33). Whereas WAVE participants received 400 IU of vitamin E and 500 mg of vitamin C twice daily, in the Heart Outcomes Prevention Evaluation (HOPE) study (32), participants received only 400 mg of vitamin E once daily, and in the Heart Protection Study (HPS) (33), participants received 600 mg of vitamin E, 250 mg of vitamin C, and 20 mg of β-carotene once daily. A final limitation is that this is a subgroup analysis, resulting in significantly smaller groups for comparison, especially in the diabetes subgroup, which has a small sample size for Hp 1-1. Despite these limitations, the current study suggests that it would be prudent to reexamine several of the recent large antioxidant cardiovascular disease prevention studies for possible benefit of antioxidant therapy to certain patient subgroups based on the haptoglobin type (32,33). If the findings presented here are validated in such studies, haptoglobin typing may become a useful tool to identify individuals who will benefit from antioxidant therapy.

This study was supported by grants from the Kennedy Leigh Charitable Trust (to A.P.L.), NHLBI Contracts NO1-HV-68165 to N01-HV-68170, and General Clinical Research Center Grant MO1-RR02715 for the WAVE study.