点击显示 收起
1 From the Division of Epidemiology, School of Public Health, University of Minnesota, Minneapolis, MN (D-HL, ARF, and DRJ); the Department of Preventive Medicine, School of Medicine, Kyungpook National University, Daegu, Korea (D-HL); and the Department of Nutrition, University of Oslo, Oslo, Norway (DRJ).
2 The Iowa Women's Health Study was funded by a grant (RO1 CA39742) from the National Cancer Institute. 3 Reprints not available. Address correspondence to DR Jacobs Jr, University of Minnesota, Division of Epidemiology, School of Public Health, 1300 South 2nd Street, Suite 300, Minneapolis, MN 55454. E-mail: jacobs{at}epi.umn.edu.
ABSTRACT
Background: The relation between iron status and atherosclerosis has long been a topic of debate.
Objective: We examined associations of cardiovascular disease (CVD) mortality with dietary intakes of iron (a possible prooxidant), zinc (a possible antioxidant), and alcohol (a disruptor of iron homeostasis).
Design: Postmenopausal women (n = 34 492) aged 55–69 y at baseline, who completed a food-frequency questionnaire, were followed for CVD mortality over 15 y.
Results: Among women who consumed 10 g alcohol/d, after adjustment for CVD risk factors in a model that contained dietary heme iron, nonheme iron, and zinc intakes, dietary heme iron showed a positive association, dietary nonheme iron showed a U-shaped association, and dietary zinc showed an inverse association with CVD mortality. For example, the relative risks (RRs) for categories of dietary heme iron were 1.0, 1.46, 1.52, 1.73, and 2.47 (P for trend = 0.04); corresponding RRs for dietary nonheme iron were 1.0, 0.93, 0.63, 0.83, and 1.20 (P for quadratic term = 0.02). The corresponding RRs for dietary zinc were 1.0, 0.61, 0.59, 0.57, and 0.37 (P for trend = 0.07). In an analysis restricted to those who consumed 30 g alcohol/d, the risk gradients strengthened.
Conclusions: Our results suggest that a higher intake of heme iron might be harmful, whereas a higher intake of zinc might be beneficial in relation to CVD mortality in the presence of a trigger that can disturb iron homeostasis, such as alcohol consumption.
Key Words: Iron • zinc • alcohol • cardiovascular diseases • oxidative stress
INTRODUCTION
The relation between iron status and atherosclerosis has long been a topic of debate (1, 2). Iron is an essential trace element that has been proposed to participate in free radical–producing reactions, in particular the hydroxyl radical–producing Fenton reaction (3, 4). Nevertheless, a meta-analysis (5) reported that published prospective studies do not support any strong epidemiologic association between iron status and ischemic heart disease (IHD).
We have hypothesized (6) that bound-form iron markers, common in epidemiologic studies, are suboptimal for investigating iron overload and health because the hydroxyl radical is produced only by the redox active free form of iron. Nearly all body iron is protected by sequestering catalytically free iron (7) in its location in the heme-containing proteins; in the iron-binding proteins, such as transferrin, lactoferrin, ferritin, and hemosiderin; or in enzymes. Protein-bound iron does not catalyze reactive oxygen species formation (8).
A general belief is that free iron rarely exists in the body, except in diseases with 100% transferrin saturation (9). However, nontransferrin bound iron or intracellular labile iron has been found when triggers disturb iron homeostasis (10, 11). Consequently, we hypothesized (6) that high amounts of stored body iron might be harmful in the presence of triggers disturbing iron homeostasis. Alcohol consumption might be a trigger because nontransferrin-bound iron, a redox active form of iron, was detected among alcohol abusers with normal transferrin saturation (11). In agreement with our hypothesis, we recently reported that dietary iron intake was positively associated with the risk of developing type 2 diabetes (12), colon cancer (13), and breast cancer (DH Lee, KE Anderson, LJ Harnack, and DR Jacobs, unpublished observations, 2004) in alcohol drinkers; the association was stronger as the amount of alcohol consumed increased. Moreover, zinc was inversely associated with the risk of colon and breast cancer, this apparently beneficial effect was confounded with heme iron because of similar food sources. There is limited evidence that zinc has antioxidant and antiatherosclerotic effects (14, 15). Therefore, we performed this study on the basis of the existing hypothesis that dietary iron intake is positively related to cardiovascular disease (CVD) mortality, whereas dietary zinc intake is inversely associated in alcohol drinkers.
SUBJECTS AND METHODS
Iowa Women's Health Study cohort
Methods for recruitment and data collection in the Iowa Women's Health Study were published previously (16). Briefly, this study was designed to examine associations between several host, dietary, and lifestyle factors and the incidence of cancer and mortality in 41 836 postmenopausal women aged 55–69 y, who completed a 16-page self-administered questionnaire at baseline in January 1986. Follow-up questionnaires were mailed in 1987, 1989, 1992, and 1997 to identify emigrants from Iowa and deaths. The study was conducted under the auspices of the University of Minnesota Institutional Review Board.
The baseline questionnaire included self-reported questions concerning several risk factors pertinent to cardiovascular disease, including history of diabetes or high blood pressure, smoking, physical activity, alcohol consumption, and hormone replacement therapy. Blood glucose and lipids were not measured. A 127-item food-frequency questionnaire (FFQ) similar to that used in the 1984 survey of the Nurses' Health Study was used to assess typical food intake over the previous year (17). Nutrient intake was computed by multiplying the frequency response by the nutrient content of the specified portion sizes. The content of heme iron was calculated by applying a factor of 0.4 to the total iron content of all meat items; the remainder of iron in meat was allocated to nonheme iron.
Deaths in Iowa were identified annually through the State Health Registry of Iowa. Deaths for subjects who did not respond to follow-up questionnaires or had emigrated from Iowa were found through the National Death Index. Underlying cause of death was assigned by state vital registries via the International Classification of Disease (ICD). We defined 1) IHD by codes 410 through 414 or 429.2 in the 9th ICD revision or codes I20 through I25 or I51.6 in the 10th revision, 2) stroke as 430 through 438 in the 9th revision or I60 through I69 in the 10th revision, and 3) all CVD as 390 through 459 in the 9th revision or I00 through I99 in the 10th revision.
Data analysis
We excluded from analysis those subjects who reported implausibly high (>5000 kcal) or low (<600 kcal) energy intakes (n = 536), left 30 items blank on the FFQ (n = 2782), were pre- or perimenopausal (n = 569), or reported angina, heart disease, or heart attack at baseline (n = 4115). After exclusions, 34 492 women remained eligible for the study.
Participants were categorized according to quartiles of dietary heme iron, nonheme iron, or zinc intake (supplements not included). The last quartile for each of the 3 nutrients was additionally divided into 2 groups at its median because its range was very broad. Associations were explored among all subjects and were also stratified by alcohol consumption (0, 1-9, and 10 g/d). Because the associations of nondrinkers and drinkers who consumed 1-9 g/d were similar, we combined these 2 strata. In the Iowa Women's Health Study, information on alcohol intake was collected 2 times, in 1986 and in 1992. In the present analyses, for women who had an event or were last followed up before 1992, we used alcohol consumption based on the 1986 questionnaire and for women who had an event or were last followed up after 1992, we averaged alcohol intake for 1986 and 1992. In a sensitivity analysis, the women were further restricted to those who reported an intake of 30 g alcohol/d.
We used proportional hazards regression to control for confounding by other risk factors for CVD mortality. Multivariate models adjusted simultaneously for age (continuous), total energy intake (continuous), history of hypertension, BMI (continuous), waist-hip ratio (WHR) (continuous), physical activity score (low, medium, or high), cigarette smoking (none, 1–19 pack-years, 20–39 pack-years, or 40 pack-years), smoking status (current smoker, exsmoker, or nonsmoker), alcohol consumption (g/d), and hormone replacement therapy (current, former, or never). An additional model adjusted for intake of saturated fat, trans fat, polyunsaturated fat, folate, ß-carotene, vitamin C, and vitamin E. In tests for trend, median values of the 4 quartiles and subdivided highest quartile categories were used. For all relative risks, we calculated 95% CIs using SAS version 8.2 (SAS Institute, Cary, NC). All P values were two-tailed.
RESULTS
During 15 y of follow-up, 1767 deaths attributed to CVD (1020 deaths to IHD, 348 deaths to stroke, and 399 deaths to other categories of CVD) were identified. Women with higher dietary heme iron intake tended to be younger, be more obese, consume less alcohol, be hypertensive, use postmenopausal hormones less frequently, and be less physically active (Table 1). These women also had higher energy intakes and, even adjustment for energy intake, consumed more saturated fat and trans fat, but less polyunsaturated fat, folate, ß-carotene, vitamin E, and vitamin C. On the contrary, women who consumed more dietary nonheme iron reported consuming less saturated fat, trans fat, and polyunsaturated fat but more folate, ß-carotene, vitamin E, and vitamin C. In addition, they were older, had a lower mean BMI and WHR, consumed less alcohol, were physically active, and were less likely to be hypertensive or smokers. Baseline characteristics according to dietary zinc intake differed only slightly from those for heme iron. Women with higher dietary zinc intakes were less likely to smoke, engaged in more physical activity, consumed less trans fat, and consumed more folate and ß-carotene than did women with lower dietary zinc intakes (data not shown). After adjustment for total energy intake, correlation coefficients were –0.02 between heme iron and nonheme iron (P < 0.01 given large sample size), 0.68 between heme iron and zinc (P < 0.01), and 0.38 between nonheme iron and zinc (P < 0.01). Among the FFQ items included in this study, "beef, pork, or lamb as a main dish" was most strongly associated with heme iron and zinc intake (energy-adjusted correlation coefficients of 0.65 and 0.52, respectively), whereas cold breakfast cereal was most strongly associated with nonheme iron intake (energy-adjusted correlation coefficient of 0.44).
View this table:
TABLE 1. Age- and energy intake–adjusted baseline means or percentages by median dietary heme and nonheme iron intakes in the Iowa Women's Health Study 1
In the total sample, dietary heme iron, dietary nonheme iron, and dietary zinc intake were not associated with the risk of CVD mortality, even after adjustment for each other (data not shown). However, among alcohol drinkers who consumed 10 g alcohol/d, dietary heme iron showed a positive association, dietary nonheme iron showed a U-shaped association, and dietary zinc showed an inverse association with CVD mortality (Table 2) The RRs for categories of dietary heme iron were 1.0 in quartile 1, 1.46 in quartile 2, 1.52 in quartile 3, 1.73 in the lower half of quartile 4, and 2.47 in the upper half of quartile 4 (P for trend = 0.04); corresponding values for dietary nonheme iron were 1.0, 0.93, 0.63, 0.83, and 1.20 (P for quadratic term = 0.02), and corresponding values for dietary zinc were 1.0, 0.61, 0.59, 0.57, and 0.37 (P for trend = 0.07). Findings were similar in minimally adjusted models, which included only heme iron, nonheme iron, zinc, age, and energy intake (data not shown). Because we restricted the analysis to those who consumed higher amounts of alcohol, the risk gradient strengthened. For example, among drinkers who consumed 30 g alcohol/d (a subset of the 10 g/d group), the adjusted RRs for categories of dietary heme iron were 1.0, 1.29, 2.62, 3.11, and 4.11 (P for trend = 0.02); corresponding values for dietary nonheme iron were 1.0, 0.76, 0.35, 0.78, and 1.67 (P for quadratic term < 0.01), and corresponding values for dietary zinc were 1.0, 0.45, 0.35, 0.42, and 0.21 (P for trend = 0.10; P < 0.01 from the chi-square statistic when a model with all variables was compared with a model without the heme iron, nonheme iron, zinc, and quadratic terms of nonheme iron, df = 4). Directionally similar trends were observed when the outcomes of IHD or stroke were separately analyzed, although, consistent with smaller sample sizes, the outcome-by-iron-by-alcohol interaction was not statistically significant (data not shown).
View this table:
TABLE 2. Relative risks (and 95% CIs) of cardiovascular disease mortality according to categories of dietary heme iron, nonheme iron, and zinc intakes stratified by alcohol consumption in postmenopausal women in the Iowa Women's Health Study (1986–2000) 1
On the other hand, supplemental iron showed a positive trend, whereas supplemental zinc showed an inverse trend among alcohol drinkers who consumed 10 g alcohol/d (NS). Compared with nonusers, the adjusted RR for users of iron supplements was 1.37 (95% CI: 0.82, 2.31), whereas the corresponding RR for users of zinc supplements was 0.58 (95% CI: 0.33, 1.03). There was no dose-response relation when users were additionally categorized by their usage dose.
Alcohol consumption itself (mean consumption: 2.3 g/d) showed a U-shaped association with CVD mortality (P < 0.01 for quadratic term). Compared with nondrinkers (n = 1013), adjusted relative risks for CVD mortality were 0.71 (95% CI: 0.64, 0.79), 0.78 (95% CI: 0.64, 0.96), and 0.83 (95% CI: 0.66, 1.05) for drinkers who consumed 1-9 g/d (n = 552), 10-29 g/d (n = 111), or 30 g/d (n = 91).
DISCUSSION
In this large prospective cohort study, dietary heme iron intake was positively and dietary zinc intake was inversely associated with the risk of CVD mortality among women who consumed 10 g alcohol/d. Because we restricted the analysis to those who consumed higher amounts of alcohol, the risk gradients strengthened. These patterns of association were similar to those in diabetes (12), colon cancer (13) and breast cancer (DH Lee et al, unpublished observations, 2004) in this cohort. Because oxidative stress is implicated as a common mechanism in the pathogenesis of various diseases, including cancer, diabetes, and CVD (18, 19), these similar patterns with different outcomes are not surprising but rather support our hypothesis.
In contrast, dietary nonheme iron showed various associations with different outcomes in this cohort: null with diabetes (12) and colon cancer (13), positive with breast cancer (DH Lee et al, unpublished observations, 2004), and U-shaped with CVD mortality. The effect of nonheme iron may depend on other beneficial or detrimental constituents also in the main food sources of nonheme iron: vegetables, grains, or iron-fortified commercial foods such as cold breakfast cereal. In this cohort, cold breakfast cereal was most strongly associated with nonheme iron intake (energy-adjusted correlation coefficient of 0.44, correlations with various vegetables or grains of 0.1). Thus, the inverse association of moderate intakes of nonheme iron with CVD mortality might reflect vegetable and grain intakes, whereas its positive association with CVD mortality might reflect iron-fortified commercial foods.
Sullivan (20) proposed that iron depletion attributable to menstruation could explain the low risk of CVD in women. Experimental and animal studies also support a potentially adverse effect of iron overload on atherosclerosis (21–23). Epidemiologic studies have, nevertheless, reported conflicting results concerning iron and IHD (1, 2, 5). Although free iron is a strong oxidant and catalyzes LDL oxidation in vitro, in vivo iron is bound to proteins, such as ferritin or transferrin, which effectively prevents tissue damage from iron-free radicals (7, 8). However, a trigger such as alcohol consumption can affect human iron homeostasis, which leads to nontransferrin-bound iron (10, 11). Any effect of iron intake on CVD mortality would be expected among drinkers, as we observed. At the same time, our observed U-shaped association of alcohol with CVD mortality is consistent with the literature (24).
Zinc has antioxidant properties that could stabilize macromolecules against radical-induced oxidation, inhibit metal-catalyzed radical formation, participate in the induction of synthesis of certain antioxidant proteins, and limit excess radical production by biological systems (14, 15). In one experimental study (25), zinc antagonized atherosclerotic lesions when free radicals were produced by iron, which suggests opposite roles for iron and zinc.
In the current study, supplemental iron was not significantly associated with the increased risk of CVD mortality. As a potential explanation, women may take supplemental iron because of low iron stores resulting from iron deficiency anemia, minor pathologic blood loss (eg, from hemorrhoids), or vegetarianism. This supplemental iron may result in adequate iron concentrations. In addition, supplemental iron tends to be consumed through multivitamins, such that various other antioxidants could counteract any adverse effect of iron. Supplemental zinc was also not significantly associated with a decreased risk of CVD mortality. Supplemental micronutrients might have effects different from those of the same micronutrients in food, ie, an antioxidant paradox: persons with diets rich in fruit and vegetables have a decreased risk of getting chronic diseases, but the same vitamin supplements do not have a clear beneficial effect, rather the opposite among some groups (26).
In this study, we hypothesized that high amounts of stored body iron might be harmful combined with alcohol consumption—a possible trigger that disturbs iron homeostasis. Theoretically, other triggers, such as inflammation, might disturb iron homeostasis. This concept remains to be shown in vivo; in vitro experimental studies have reported that free iron can be released from ferritin by superoxide or nitric oxide, which are produced by stimulated polymorphonuclear leukocytes and macrophages during inflammation (27, 28).
The limitations of our study warrant consideration. First, although the reproducibility and validity of the FFQ used in the Iowa Women's Health Study was evaluated in 44 study participants (29), the accuracies of heme iron, nonheme iron, or zinc were not separately examined. Although problems related to measurement error undoubtedly resulted in some women being misclassified, generally, nondifferential misclassification of exposure variables leads to a null rather than to a spurious association. Second, although the study of possible triggers of free iron formation was performed on the basis of an a priori biological mechanism, we cannot totally rule out the possibility that some of the findings in small subgroups occurred by chance. However, this pattern has been repeatable for heme iron and zinc with diabetes, colon cancer, breast cancer, and now CVD mortality in the Iowa Women's Health Study, consistent with oxidative stress caused by iron as a unifying hypothesis. Third, we did not have CVD incidence data. However, associations of risk factors with CVD mortality endpoints generally parallel those for incidence (30, 31). Fourth, the CVD risk factors adjusted for in this study, especially high blood pressure, were self-reported. Fifth, our study may be generalizable only to postmenopausal women. Finally, it is possible that the finding for heme iron reflects nonspecific adverse properties of meat, whereas that for zinc reflects nonspecific positive properties of plant foods.
In summary, the results suggest that a high dietary heme iron intake is associated with an increased risk of CVD mortality, whereas a high dietary zinc intake is associated with a decreased risk of CVD mortality among alcohol drinkers of 10 g alcohol/d, within the context of a U-shaped association of alcohol with CVD mortality. In this study, relatively few of the women in the Women's Health Study were affected by these relations; the importance of these findings is, therefore, primarily in the potential etiologic insight they provide.
ACKNOWLEDGMENTS
D-HL was responsible for the data analysis, conception, and the first draft of the manuscript. ARF provided intellectual input and edited the manuscript. DRJ carried out the data analysis, provided intellectual input, and edited the manuscript. None of the authors had a financial or personal interest involved in the sponsorship of this research study.
REFERENCES