Whole-grain intake may reduce the risk of ischemic heart disease death in postmenopausal women: the Iowa Women s Health Study 1 3

Transcription

1 Whole-grain intake may reduce the risk of ischemic heart disease death in postmenopausal women: the Iowa Women s Health Study 1 3 David R Jacobs Jr, Katie A Meyer, Lawrence H Kushi, and Aaron R Folsom ABSTRACT Background: A recent review of epidemiologic literature found consistently reduced cancer and heart disease rates in persons with high compared with low whole-grain intakes. Objective: We hypothesized that whole-grain intake was associated with a reduced risk of ischemic heart disease (IHD) death. Design: We studied postmenopausal women aged y and free of IHD at baseline in There were 438 IHD deaths between baseline and Usual dietary intake was determined with use of a 127-item food-frequency questionnaire. Results: Whole-grain intake in median servings/d was 0.2, 0.9, 1.2, 1.9, and 3.2 for quintiles of intake. The unadjusted rate of IHD death was 2.0/ person-years in quintile 1 and was 1.7, 1.2, 1.0, and 1.4 IHD deaths/ person-years in succeeding quintiles (P for trend < 0.001). Adjusted for demographic, physiologic, behavioral, and dietary variables, relative hazards were 1.0, 0.96, 0.71, 0.64, and 0.70 in ascending quintiles (P for trend = 0.02). The lower risk with higher whole-grain intake was not explained by intake of fiber or several other constituents of whole grains. Conclusion: A clear inverse association between whole-grain intake and risk of IHD death existed. A causal association is plausible because whole-grain foods contain many phytochemicals, including fiber and antioxidants, that may reduce chronic disease risk. Whole-grain intake should be studied further for its potential to prevent IHD and cancer. Am J Clin Nutr 1998;68: KEY WORDS Diet, epidemiology, prospective study, whole grains, heart disease, women, Iowa Women s Health Study INTRODUCTION A variety of studies have shown that food constituents such as dietary fiber and vitamin E have health benefits (1 6). All of these constituents are in abundance in whole-plant foods, including whole grains, but are deficient in refined grains. Whole grains are nutritionally superior to refined grains in that they contain higher amounts of fiber (7), minerals (8), vitamins (9), and phenols, phytic acid, and phytoestrogens such as lignins (10). Conversely, consumption of refined grains, which are relatively poor in these dietary factors, may not reduce chronic disease risk. On See corresponding editorial on page 218. the other hand, a higher grain intake of any type may protect against chronic diseases insofar as it substitutes for saturated fat and cholesterol in the diet. The popular and scientific belief that refinement of grains may result in the loss of constituents of nutritional importance is underscored by the fact that refinedflour products are enriched with thiamine, riboflavin, niacin, and iron, and that folic acid will soon be added to the mix. Surprisingly, relatively few studies have examined whether whole grains decrease chronic disease risk. A review of epidemiologic reports of high compared with low intake of wholegrain foods found a consistently reduced risk of several cancers, but only 15 studies were identified that examined this question and for any given cancer site there were no more than 7 such studies (11). To our knowledge, only 2 studies have examined the association of whole-grain intake and ischemic heart disease (IHD) risk directly (12, 13). Because ascertainment of whole- or refined-grain intake has not usually been the focus of most studies of diet and chronic disease, few data are available to examine this potentially important association. Among plant foods, grains are particularly important to public health: according to food disappearance data, 90 kg (199 lb) grains per capita were consumed and constituted 23% of energy intake in 1994 (14). A single plant food wheat represented 15% of energy intake in that year. Yet there have been major changes in the past century in the pattern of consumption of grains so that at present only 4.5 kg ( 10 lb) whole grains are eaten per year in the United States ( 5% of total grain intake) (14; S Gerrior, personal communication based on Census of Manufacturers, Grain Mill Products, Industry Series, MC D, US Department of Agriculture, 1997). An increased intake of whole grains would benefit public health by increasing intakes of fiber, vitamin E, and many other healthful constituents, likely 1 From the Division of Epidemiology, School of Public Health, University of Minnesota, Minneapolis. 2 Supported by the National Institutes of Health (research grant CA ). 3 Address reprint requests to DR Jacobs Jr, University of Minnesota, Division of Epidemiology, 1300 South 2nd Street, Suite 300, Minneapolis, MN Jacobs@epivax.epi.umn.edu. Received July 30, Accepted for publication March 2, Am J Clin Nutr 1998;68: Printed in USA American Society for Clinical Nutrition

2 WHOLE-GRAIN INTAKE AND ISCHEMIC HEART DISEASE 249 including some yet to be discovered. The US Department of Agriculture s Dietary Guidelines for Americans (15) states, Choose a diet with plenty of grain products, vegetables, and fruits and goes on to recommend 6 11 daily servings of grain products (breads, cereals, pasta, and rice), including whole-grain products and several servings of whole grain breads and cereals. This guideline recognizes the importance of whole grain. Yet at the same time, it supports the consumption of a substantial amount of refined grain. The scientific community often ignores the distinction between whole and refined grain. For example, in a randomized feeding study in which systolic blood pressure was reduced by 5.5 mm Hg, the highly publicized final report characterizes the blood pressure lowering diet as being high in fruit, vegetables, and low-fat dairy products (16). An earlier publication had characterized the diet as high in fruits, vegetables, whole cereal products, low-fat dairy products, fish, chicken and lean meats (17). The public continues to consume whole grain in limited quantities and refined grain in large quantities. The food industry continues to offer primarily refinedgrain products. The Iowa Women s Health Study, a prospective cohort study of postmenopausal women, assessed food intake among study participants in a manner that allowed estimation of whole-grain and refined-grain intakes. This paper focuses on the association of whole- and refined-grain intakes with risk of IHD death. SUBJECTS AND METHODS Subjects The study was approved by the Committee on the Use of Human Subjects in Research, University of Minnesota. The methodology for the Iowa Women s Health Study has been described elsewhere (6). Briefly, a random sample (n = 99826) of women aged y with a valid Iowa driver s license were mailed a 16-page survey in January, The cohort under study comprises women who responded to this questionnaire. Women were excluded from the analysis if they were not postmenopausal (n = 569), left 30 items blank on the foodfrequency questionnaire (n = 2782), or reported implausibly low or high energy intakes (<2.5 or 21 MJ/d) (n = 538). Women were also excluded if they reported at baseline that a physician had told them that they had suffered a heart attack or angina or had heart disease (n = 4115). This left women for analysis (numbers excluded were not mutually exclusive). Missing covariate data reduced the sample size to in multivariate analysis. Data collection The baseline questionnaire included questions on known or suspected risk factors for IHD as well as a 127-item food-frequency questionnaire similar to that used in the 1984 survey of the Nurses Health Study (18, 19). For each food, a commonly used serving size was specified and participants were asked to report their frequency of consumption of various foods from among 9 categories, ranging from never or <1 serving/mo to 6 servings/d. Servings per week were calculated from these categories. Questions were included regarding other regularly consumed foods, current use and dosage of vitamin supplements, and the brand name of multivitamin preparations. Women could specify any number of servings per week for each food item queried, including cold breakfast cereal. Women were asked to name the single breakfast cereal they usually ate. The list of breakfast cereals mentioned was evaluated for wholegrain and bran content. Breakfast cereals were considered to be whole grain if the product contained 25% whole grain or bran by weight, as determined either from the package label or from records shared by General Mills, Inc (Minneapolis). Bran cereals were included in the whole-grain category because findings were similar for bran cereals and nonbran, whole-grain cereals (data not shown). Of the 152 cereals mentioned by these women, 91 contained 25% whole grain or bran by weight. Of women who reported eating breakfast cereal, reported typically consuming a product that was 25% whole grain or bran. Most reported a product that was 50% whole grain or bran; only 747 of these women reported a product with moderate whole-grain or bran content (25 49%). A total of 7429 women reported consuming a refined-grain cereal; this number includes 68 women who reported consuming a product that contains a small amount of whole grain (<25%). Because most of the women who ate breakfast cereal mentioned a whole-grain cereal, the remaining 1089 women who did not specify any one cereal as usual were assumed to eat a whole-grain product; sensitivity analysis varying this assumption did not strongly affect the findings. The reliability of the questionnaire in this cohort was evaluated in a randomly chosen subgroup of 44 women by comparing nutrient consumption determined by responses on the food-frequency questionnaire on 2 occasions. Comparative validity was determined by comparison with intake estimated from the average of five 24-h dietary recalls (20). Moderate correlations, adjusted for total energy intake between these 2 methods, were found for nutrients commonly found in whole grains. These included 0.43 for folate and 0.55 for vitamin E. Correlations for food items and food groups were not examined. However, in another study of nurses in which a similar questionnaire was used, correlations with 28 d of food records were higher for food items themselves: 0.75 for cold breakfast cereals, 0.61 for white bread, and 0.66 for dark bread (21). Case ascertainment The vital status of cohort members was determined through December 31, 1995, via an annual linkage with the State Health Registry of Iowa, the National Death Index, and follow-up questionnaires mailed in 1988, 1990, and Women were considered to have died from IHD if the death certificate was coded as International Classification of Diseases (9th revision) codes or (22). Although we did not validate cause-ofdeath coding, other studies have indicated that the validity of death certificates for IHD is relatively high (23, 24). Of the women included in these analyses, 3320 died, including 438 from IHD. Data analysis Length of follow-up was calculated for each individual as the number of days elapsed from completion of the baseline questionnaire until either the date of death or December 31, Analyses examined the associations of whole- and refined-grain intake with mortality from IHD, both as categories of total whole- and refined-grain intake and as mutually exclusive subgroups of these categories. The total whole- and refined-grain groups and subgroups, their respective serving sizes, and the verbatim food items in each subgroup are listed in Table 1. Crack-

3 250 JACOBS ET AL TABLE 1 Descriptions and mean intake of grain food groups for postmenopausal Iowan women free of ischemic heart disease, 1986 Grain item No. of servings/wk 1 Verbatim wording of food items included Total whole grains ± Dark bread 6.43 ± 6.76 Dark bread Whole-grain breakfast cereal 1.86 ± 2.40 Cold breakfast cereal 3 Other whole grains 2.47 ± 3.53 Brown rice, popcorn, wheat germ, bran, cooked oatmeal, other grains (eg, bulgar, kasha, couscous 4 ) Total refined grains ± White bread 4.46 ± 6.61 White bread including pita bread Refined-grain breakfast cereal 0.59 ± 1.55 Cold breakfast cereal 3 Other refined grains 2.74 ± 2.31 English muffins, bagels, or rolls; pancakes or waffles; white rice; pasta (eg, spaghetti, noodles, etc); pizza Sweets or desserts 7.36 ± 7.74 Cookies, home baked; cookies, ready made; doughnuts; brownies; sweet roll, coffee cake, other pastry, home baked; sweet roll, coffee cake, other pastry, ready made; cake, home baked; cake, ready made; pie, home baked; pie, ready made; muffins or biscuits Unspecified grain type 3.42 ± 6.38 Crackers, Triscuits, 5 Wheat thins 5 1 Serving sizes were as follows: white or dark bread, 1 slice; pizza, 2 slices; cold breakfast cereal, white or brown rice, pasta, other grain, popcorn, cooked oatmeal, 1 cup; bran, wheat germ, 1 tablespoon; English muffins, pancakes, muffins or biscuits, cookies, doughnuts, brownies, pastries, pies, cakes, crackers, 1 item or 1 serving. 2 x ± SD. 3 The one brand name or generic cereal specified was coded for whole-grain content; breakfast cereals with 25% whole-grain or bran content by weight were classified as whole grain. 4 The item Other grains was categorized as whole, although couscous is a refined grain. 5 Nabisco, East Hanover, NJ. ers, which can be made from either whole or refined grain, were analyzed separately. The association of grains with IHD death was examined primarily by proportional hazards regression analysis. Individual and grouped dietary variables were categorized by quintiles (or a smaller number of categories when the range of intake was too small to justify use of quintiles) and the IHD mortality rate in each category was compared with that in the lowest intake category. Initial analyses examined associations adjusted for age and total energy intake. Analyses were also adjusted for other risk factors that were significant predictors of IHD death, including demographic factors (marital status and education) and physiologic or pathologic factors [self-reported history of hypertension, self-reported history of diabetes mellitus, self-measured body mass index, and waist-tohip ratio measured by a friend or spouse (25)]. Additional covariates represented lifestyle behaviors (cigarette smoking, physical activity, estrogen replacement therapy, alcohol intake, and use of dietary supplements) and aspects of diet other than grains [nutrients (sucrose), food groups (red meat, fish and seafood, fruit and vegetables), and Keys score, a summary measure incorporating saturated fat, polyunsaturated fat, and cholesterol (26)]. Finally, we examined the possibility that dietary factors found in whole grains may explain any association observed between whole grains and IHD death. Intake from supplements was omitted. For these analyses, the association of whole grains with IHD death was adjusted for several constituents of whole grains shown or hypothesized to be associated with IHD. These constituents included dietary fiber, vitamin E, folate, phytic acid, magnesium, manganese, and iron. The relative risk for a given category of intake was estimated by exponentiation of the proportional hazards regression coefficient for that level of intake. A test for trend was determined across a vector of indicator variables for the grain of interest, with each level of exposure weighted by its median value. The SAS statistical analysis package was used (27). RESULTS The mean weekly intake of whole-grain foods reported was 10.8 servings/wk (Table 1), consisting of 6.4 slices of dark bread, 1.9 cups of whole-grain breakfast cereal, and 2.5 servings of other whole-grain foods. The women reported higher intakes, 15.2 servings/wk, of refined-grain foods, consisting of 4.5 slices of white bread, 0.6 cups of refined breakfast cereal, 2.7 servings of other refined grains, and 7.4 servings of sweets and desserts. In addition, the women reported eating an average of 3.4 crackers/wk, which were unspecified as to refined- or whole-grain content. The median intake of whole-grain foods ranged from 1.5 to 22.5 servings/wk from the lowest to highest categories of whole-grain intake, and was 4.0 and 30.0 servings/wk in comparable refined-grain intake categories (Tables 2 and 3). The distribution of covariates according to these categories of whole- and refined-grain intake is also given in Tables 2 and 3. Higher whole-grain intake was associated with having more education, a lower body mass index and waist-to-hip ratio (even after adjustment for body mass index), being a nonsmoker, doing more regular physical activity, and using vitamin supplements and hormone replacement therapy. Diabetes was reported somewhat more frequently and hypertension less frequently in those who ate whole grain more often. Higher whole-grain intake was also associated with greater energy intake; consumption of less refined grains, sucrose, and red meat; a lower Keys score; and consumption of more fruit and vegetables. Findings were generally reversed in those with higher intakes of refined grains, except for the sharp increase in energy intake and the lack of gradient in smoking status or Keys score across the refined-grain categories. In addition, both diabetes and hypertension were

4 WHOLE-GRAIN INTAKE AND ISCHEMIC HEART DISEASE 251 TABLE 2 Description of various risk factors for ischemic heart disease, according to categories of whole-grain intake for postmenopausal Iowan women, Whole-grain intake category Risk factor SEM 2 P 3 Grain intake Whole-grain intake (servings/wk) Median Range Mean refined-grain intake (servings/wk) < Demographics Mean age (y) < Education (% >high school) < Anthropometry Mean body mass index (kg/m 2 ) < Mean waist-to-hip ratio < Illness Self-reported diabetes (%) < Self-reported hypertension (%) < Lifestyle behaviors Current smoker (%) < Vitamin supplement use (%) < Alcohol (% never drink) < Physical activity (% engaging in regular activity) < Hormone replacement therapy (% ever) < Diet Mean energy intake (MJ/d) < Mean fruit and vegetable intake, no juice (servings/wk) < Mean red meat intake (servings/wk) < Mean fish and seafood intake (servings/wk) < Mean Keys score (mmol/l) < Mean sucrose intake (g/d) < Mean dietary fiber intake (g/d) < Mean vitamin E intake (supplements excluded) (IU/d) < Mean folate intake (supplements excluded) (µg/d) < All tabulated means and percentages were adjusted for age and energy intake, except age, which was adjusted only for energy intake; energy intake, which was adjusted only for age; and waist-to-hip ratio, which was adjusted for age, energy intake, and body mass index. 2 SEM within a whole-grain intake category, computed as the root mean square error, derived from linear regression by holding adjustment factors constant, and divided by 83.06, the square root of the average number per category. 3 Tests any difference among whole-grain intake categories. reported somewhat less frequently in high than in low refined-grain eaters. Red meat was eaten slightly less frequently in the highest refined-grain intake quintile. Other characteristics that were unrelated to either whole- or refined-grain intake included alcohol intake, marital status, and prior use of oral contraceptives (data not shown). Relations of grain constituents such as dietary fiber, vitamin E, and folate were steep and direct with whole-grain intake and inverse and shallow with refined-grain intake (Tables 2 and 3). Various other dietary factors were increased in whole-grain eaters and decreased in refinedgrain eaters, including calcium, iron, magnesium, manganese, copper, zinc, and phytic acid (data not shown). Grain constituents were highly correlated in these data; for example, correlations with dietary fiber were 0.53 for vitamin E and 0.83 for manganese. There was a striking inverse association of whole-grain intake with risk of death from IHD (Table 4). For total whole-grain intake, ageand energy-adjusted relative risks from lowest to highest category of intake by quintiles were 1.0, 0.84, 0.58, 0.45, and 0.60 (P for trend = ). After adjustment for other potentially confounding variables, this inverse association was attenuated but still remained (relative risks: 1.0, 0.96, 0.71, 0.64, and 0.70; P for trend = 0.018). Dietary fat and cholesterol intakes, in the form of the Keys score, were also included in these models; thus, the inverse association of whole-grain intake with IHD death could not be attributed to inverse associations of Keys score with whole-grain intake. Significant inverse age- and energy-adjusted associations with IHD death were also seen for dark bread and whole-grain breakfast cereals. After adjustment for other confounding variables, these associations were somewhat attenuated, with relative risks comparing high and low categories of intake of 0.7 for these food items. Intake of other whole grains was not associated with risk of IHD death. In contrast, there was little evidence of an association between total refined-grain intake and risk of IHD death (Table 5). The multivariable-adjusted relative risks from lowest to highest quintile of intake were 1.0, 0.99, 1.14, 1.04, and 1.12 (P for

5 252 JACOBS ET AL TABLE 3 Description of various risk factors for ischemic heart disease, according to categories of refined-grain intake for postmenopausal Iowan women, Refined-grain intake category Risk factor SEM 2 P 3 Grain intake Refined-grain intake (servings/wk) Median Range Mean whole-grain intake < (servings/wk) Demographics Mean age (y) Education (% >high school) < Anthropometry Mean body mass index (kg/m 2 ) Mean waist-to-hip ratio < Illness Self-reported diabetes (%) < Self-reported hypertension (%) Lifestyle behaviors Current smoker (%) < Vitamin supplement use (%) < Alcohol (% never drink) < Physical activity (% engaging in regular activity) < Hormone replacement therapy < (% ever) Diet Mean energy intake (MJ/d) < Mean fruit and vegetable intake, no juice (servings/wk) < Mean red meat intake (servings/wk) < Mean fish and seafood intake (servings/wk) < Mean Keys score (mmol/l) < Mean sucrose intake (g/d) < Mean dietary fiber intake (g/d) < Mean vitamin E intake (supplements excluded) (IU/d) < Mean folate intake (supplements excluded) (µg/d) < All tabulated means and percentages were adjusted for age and energy intake, except age, which was adjusted only for energy intake; energy intake, which was adjusted only for age; and waist-to-hip ratio, which was adjusted for age, energy intake, and body mass index. 2 SEM within a refined-grain intake category, computed as the root mean square error, derived from linear regression by holding adjustment factors constant, and divided by 83.06, the square root of the average number per category. 3 Tests any difference among refined-grain intake categories. trend = 0.57). Subgroups of refined-grain intake showed associations with risk of IHD death that were different from those of the whole-grain subgroups. After adjustment for age and energy intake, white bread was positively associated, and intake of sweets and desserts and other refined grains was inversely associated, with risk of IHD death. After adjustment for potential confounders, these associations were attenuated and not significant. Refined-grain breakfast cereals were not associated with IHD death. An inverse association of IHD with intake of crackers, which can be made from either whole or refined grains, was also weakened and no longer significant after adjustment for confounding variables (Table 6). The association of whole-grain intake with IHD death was independent of intake of refined grain. Risk for increasing total whole-grain intake was not attenuated by adding total refinedgrain intake to the models, whereas risk for increasing total refined-grain intake was further attenuated by adding total whole-grain intake to the models (data not shown). To examine whether the association of whole-grain intake with IHD death could be attributed to various constituent nutrients that have been hypothesized to be associated with risk of IHD, we conducted further analyses in which we adjusted for dietary fiber and other factors. When adjusted for behaviors and other dietary factors (not including whole grains), the relative risks (95% CI) for the highest compared with the lowest quintile of intake were as follows: fiber, 0.48 (0.26, 0.88); phytic acid, 0.63 (0.40, 0.99); vitamin E (supplements excluded), 0.85 (0.53, 1.38); folate, 0.83 (0.53, 1.31); iron, 0.88 (0.57, 1.37); manganese, 0.67 (0.42, 1.06); and magnesium, 0.76 (0.46, 1.28). All of these relative risks lost significance (or remained statistically insignificant) after further adjustment for whole-grain intake. Of

6 WHOLE-GRAIN INTAKE AND ISCHEMIC HEART DISEASE 253 TABLE 4 Relative risk of ischemic heart disease death by categories of whole-grain consumption in postmenopausal Iowan women, Whole-grain intake category Grain item P for trend Total whole grain Servings/wk Cases Person-years RR 3 (95% CI) (0.65, 1.09) 0.58 (0.43, 0.78) 0.45 (0.33, 0.62) 0.60 (0.45, 0.81) Adjusted RR 4 (95% CI) (0.71, 1.28) 0.71 (0.51, 0.98) 0.64 (0.45, 0.90) 0.70 (0.50, 0.98) Dark bread Servings/wk Cases Person-years RR (95% CI) (0.53, 0.88) 0.52 (0.40, 0.68) 0.63 (0.47, 0.83) Adjusted RR (95% CI) (0.62, 1.06) 0.62 (0.46, 0.82) 0.67 (0.49, 0.91) Whole-grain breakfast cereal 5 Servings/wk Cases Person-years RR (95% CI) (0.58, 0.97) 0.62 (0.47, 0.82) 0.57 (0.43, 0.76) Adjusted RR (95% CI) (0.62, 1.08) 0.78 (0.58, 1.06) 0.77 (0.56, 1.04) 0.23 Other whole grains Servings/wk Cases Person-years RR (95% CI) (0.89, 1.66) 0.91 (0.66, 1.25) 0.84 (0.57, 1.26) 0.14 Adjusted RR (95% CI) (1.01, 2.02) 1.17 (0.81, 1.68) 1.26 (0.81, 1.95) Number of ischemic heart disease deaths. 2 Sum of the number of years of follow-up for all women at risk. 3 Relative risk. Proportional hazards regression model (n = 34492) adjusted for age (y) and total energy intake. 4 Proportional hazards regression model (n = ) adjusted for age, total energy intake, educational attainment (did not graduate from high school, high school graduate, some college or vocational school, or college graduate), marital status (currently married, never married, separated or divorced, or widowed), high blood pressure (yes or no), diabetes (yes or no), body mass index (quintiles), waist-to-hip ratio (quintiles), physical activity (low, moderate, or high), pack-years (number of packs of cigarettes smoked per year times the number of years smoked; none, 1 19, 20 39, or 40), alcohol intake (none, <4 g/d, 4 to <10 g/d, or 10 g/d), use of vitamin supplements (yes or no), use of oral contraceptive pills (never or ever), use of estrogen replacement therapy (current, former, or never), Keys score (quartiles), intake of fruit and vegetables except juice (quartiles), intake of red meat (quartiles), intake of fish and seafood (0 0.5, , or >1.5 servings/wk), and sucrose intake (quintiles). 5 The 1089 women who did not specify a usual breakfast cereal were included as whole-grain cereal eaters; this assumption did not strongly affect the findings. these, the lowest relative risk, comparing quintile 5 with quintile 1 after whole-grain adjustment, was for dietary fiber [0.62 (0.32, 1.20)]; the corresponding relative risk for quintile 4 compared with quintile 1 of dietary fiber intake was 0.55 (0.32, 0.95). On the other hand, after adjustment for dietary fiber intake, the association of whole grains with risk of IHD death was attenuated slightly, with relative risks (and 95% confidence limits) from lowest to highest category of intake of 1.0, 0.99 (0.74, 1.33), 0.76 (0.54, 1.06), 0.71 (0.49, 1.01), and 0.77 (0.54, 1.10) (P for trend = 0.10). There are close plant physiologic relations of fiber intake with content of vitamin E, folic acid, phytic acid, iron, magnesium, and manganese. Adjustment for these other constituents of whole grain yielded similar relative risks. After simultaneous adjustment for all of these constituents, relative risk of IHD death from lowest to highest category of whole-grain intake was 1.0, 0.97 (0.72, 1.31), 0.72 (0.51, 1.02), 0.66 (0.45, 0.98), and 0.74 (0.50, 1.10) (P for trend = 0.09). DISCUSSION We found an association between higher intakes of wholegrain products and lower risk of IHD death in postmenopausal women who initially reported being free of IHD. After adjustment for potentially confounding variables, the risk of IHD death over the 9 y of follow-up was reduced by about one-third in those eating 1 serving of a whole-grain product each day, compared with those who reported rarely eating any whole-grain products. These findings held equally for dark bread and whole-grain breakfast cereals, whereas the combination of rarely consumed items, such as brown rice, popcorn, wheat germ, cooked oatmeal, and other grains (eg, bulgar, kasha, and couscous) was unrelated to IHD death. In contrast, the associations of refined-grain intake and its subgroups with risk of IHD death were inconsistent and were attenuated and lost statistical significance after adjustment for other risk factors. Risk associated with whole-grain intake was not affected by adjustment for refined-grain intake, but risk associated with refined-grain intake was further attenuated by adjustment for whole-grain intake. Relatively little has been published that specifically addresses whether whole-grain intake is associated with IHD. One study of Seventh-day Adventists reported a 50% reduced risk of nonfatal IHD and a 25% reduced risk of fatal IHD for those who ate whole-wheat bread compared with those who ate only white bread (12). However, no association was found by Gramenzi et

7 254 JACOBS ET AL TABLE 5 Relative risk of ischemic heart disease death by categories of refined-grain consumption for postmenopausal Iowan women, Refined-grain intake category Grain item P for trend Total refined grain Servings/wk Cases Person-years RR 3 (95% CI) (0.58, 1.10) 0.90 (0.57, 1.01) 0.86 (0.56, 1.09) 1.06 (0.70, 1.36) 0.60 Adjusted RR 4 (95% CI) (0.71, 1.38) 1.14 (0.82, 1.57) 1.04 (0.77, 1.47) 1.12 (0.77, 1.62) 0.57 White bread Servings/wk Cases Person-years RR (95% CI) (0.67, 1.28) 1.48 (1.12, 1.91) 1.43 (1.07, 1.81) Adjusted RR (95% CI) (0.64, 1.27) 1.43 (1.08, 1.89) 1.24 (0.94, 1.64) 0.13 Refined-grain breakfast cereal 5 Servings/wk Cases Person-years RR (95% CI) (1.04, 1.84) 1.05 (0.71, 1.54) 1.54 (1.08, 2.19) 0.14 Adjusted RR (95% CI) (0.81, 1.53) 0.82 (0.52, 1.28) 1.45 (0.99, 2.13) 0.14 Sweets or desserts Servings/wk Cases Person-years RR (95% CI) (0.48, 0.81) 0.72 (0.53, 0.98) 0.69 (0.51, 0.91) 0.51 (0.37, 0.71) Adjusted RR (95% CI) (0.70, 1.26) 1.13 (0.78, 1.64) 1.17 (0.81, 1.67) 0.86 (0.56, 1.31) 0.56 Other refined grains Servings/wk Cases Person-years RR (95% CI) (0.70, 1.09) 0.87 (0.65, 1.16) 0.67 (0.45, 0.99) Adjusted RR (95% CI) (0.74, 1.20) 0.90 (0.65, 1.25) 0.79 (0.52, 1.21) Number of ischemic heart disease deaths. 2 Sum of the number of years of follow-up for all women at risk. 3 Relative risk. Proportional hazards regression model (n = 34492) adjusted for age (y) and total energy intake. 4 Proportional hazards regression model (n = 31284) adjusted for age, total energy intake, educational attainment (did not graduate from high school, high school graduate, some college or vocational school, or college graduate), marital status (currently married, never married, separated or divorced, or widowed), high blood pressure (yes or no), diabetes (yes or no), body mass index (quintiles), waist-to-hip ratio (quintiles), physical activity (low, moderate, or high), pack-years (number of packs of cigarettes smoked per year times the number of years smoked; none, 1 19, 20 39, or 40), alcohol intake (none, <4 g/d, 4 to <10 g/d, or 10 g/d), use of vitamin supplements (yes or no), use of oral contraceptive pills (never or ever), use of estrogen replacement therapy (current, former, or never), Keys score (quartiles), intake of fruit and vegetables except juice (quartiles), intake of red meat (quartiles), intake of fish and seafood (0 0.5, , or >1.5 servings/wk), and sucrose intake (quintiles). 5 The 1089 women who did not specify a usual breakfast cereal were included as whole-grain cereal eaters; this assumption did not strongly affect the findings. al (13) in a study of myocardial infarction in 287 case and 649 control women aged y. Compared with those who reported never consuming whole-meal bread or pasta, the odds ratio was 0.9 for those who ate 1 3 servings/wk and 1.1 for those who consumed 4 servings/wk. Whole grains are not frequently eaten in northern Italy: 71% of the control women never ate whole-meal bread or pasta. Although few studies have addressed whole-grain intake per se, several other studies support our findings. In a Finnish study, Pietinen et al (28) reported reduced risk of IHD death, but not of all coronary events, associated with increased intake of total rye products; rye is usually consumed in whole form in Finland. Consumption of other cereal products, unspecified as to degree of refinement, was not associated with reduced risk of IHD in the Finnish study. Rimm et al (29) reported a relative risk of 0.83 in men who consumed cold breakfast cereal 2 4 times/wk compared with those who ate no cold breakfast cereal, but did not separate grains by degree of processing. A study of patrons of health food shops, subscribers to health food magazines, and members of vegetarian and health food societies found that those who ate whole meal daily and those who did not did not differ from each other in their IHD death rates, but that daily consumers did have a reduced death rate from cerebrovascular disease (30, 31). However, the total IHD death rate in this study was only about half of that in the general population, and in this sense the study (30, 31) may indicate that whole-grain intake is part of a generally healthful diet. Knekt et al (32) reported an inverse association of IHD with total grain intake in women, but not in men. Rimm et al (29) also reported reduced risk of IHD with increasing cereal fiber intake, a finding echoed by Salmerón et al (33, 34) for a related endpoint, diabetes. Several other studies have found reduced IHD rates in those who consume high

8 WHOLE-GRAIN INTAKE AND ISCHEMIC HEART DISEASE 255 TABLE 6 Relative risk of ischemic heart disease by categories of cracker consumption in postmenopausal Iowan women, Cracker intake category Grain item P for trend Crackers Servings/wk Cases Person-years RR 3 (95% CI) (0.66, 1.18) 0.80 (0.58, 1.09) 0.83 (0.61, 1.11) 0.68 (0.49, 0.93) Adjusted RR 4 (95% CI) (0.66, 1.25) 0.99 (0.71, 1.40) 0.87 (0.63, 1.22) 0.81 (0.57, 1.14) Number of ischemic heart disease deaths. 2 Sum of the number of years of follow-up for all women at risk. 3 Relative risk. Proportional hazards regression model (n = 34492) adjusted for age (y) and total energy intake. 4 Proportional hazards regression model (n = 31284) adjusted for age, total energy intake, educational attainment (did not graduate from high school, high school graduate, some college or vocational school, or college graduate), marital status (currently married, never married, separated or divorced, or widowed), high blood pressure (yes or no), diabetes (yes or no), body mass index (quintiles), waist-to-hip ratio (quintiles), physical activity (low, moderate, or high), pack-years (number of packs of cigarettes smoked per year times the number of years smoked; none, 1 19, 20 39, or 40), alcohol intake (none, <4 g/d, 4 to <10 g/d, or 10 g/d), use of vitamin supplements (yes or no), use of oral contraceptive pills (never or ever), use of estrogen replacement therapy (current, former, or never), Keys score (quartiles), intake of fruit and vegetables except juice (quartiles), intake of red meat (quartiles), intake of fish and seafood (0 0.5, , or >1.5 servings/wk), and sucrose intake (quintiles). amounts of dietary fiber (35 40) or starch (41), both of which tend to increase with increased whole grain in the diet. A clinical trial of increased grain fiber intake, achieved through recommendations to increase consumption of wholemeal bread, high-fiber breakfast cereals, and wheat bran (42), was undertaken in 2033 survivors of myocardial infarction. In the intervention group, grain fiber intake increased from 9 to 17 g/d during the 2 y of follow-up, with little change in the control group. Contrary to expectation, there was a higher rate of total deaths in the intervention group (14%) than in the control group (12%) (43). The implications of these findings to those that we report here are unclear, in part because this was a secondary prevention trial among subjects who had a previous myocardial infarction. The length of the intervention was also relatively short, in contrast with the long period of atherogenesis. In contrast with the negative overall findings, simultaneous changes in multiple dietary areas were observed to be more efficacious than changes in whole grains alone: among the 8 subgroups given advice to modify fat, fish, and grain fiber intake in this trial, the 251 participants who received advice to modify all 3 dietary components had the second lowest all-cause mortality rate (8.0%). The 257 participants receiving advice only to increase their fish intake had the lowest rate (7.8%), whereas the group of 257 given advice on both fish and fat intake, but not grain intake, had a total death rate of 8.9%. In sum, the findings from this clinical trial (43) suggest caution in interpreting the results of our study. The hypothesis that whole-grain intake may reduce the risk of IHD death arose from consistent findings of an inverse association of dietary fiber intake with IHD and from a general review of the literature, which found that whole-grain intake was associated with reduced risk of several cancers (11). Reviews by Slavin et al (10) and by Thompson (44) suggested that many constituents of whole grains may reduce risk of chronic diseases such as IHD. Dietary fiber and resistant starch slow the absorption of nutrients, possibly modulating glucose and insulin responses, and enhance production of short-chain fatty acids. Antioxidants such as vitamin E and phenolic compounds can slow the rate of oxidation. Lignins are phytoestrogens that have hormonal effects in humans; for example, they may increase menstrual cycle length and relieve menopausal symptoms. Antinutrients such as phytic acids, lectins, phenolics, amylase inhibitors, and saponins have been found to lower plasma glucose, insulin, or plasma cholesterol and triacylglycerol. In addition, many constituents of whole grains and their potential health effects have not yet been identified (10). The inverse association of whole-grain intake with risk of IHD death was somewhat weakened after adjustment for various constituent dietary factors, such as dietary fiber, phytic acid, and vitamin E. The persistence of most of the whole grain and IHD association suggests that other unidentified dietary factors may also play a role in the etiology of IHD. Although it is important to identify such factors, doing so is only indirectly relevant for making public health recommendations concerning foods and dietary patterns for disease prevention. Because the physiologic effects of various dietary constituents are likely interactive, and many such compounds are either unidentified or poorly quantified in specific foods and the food supply, observations related to intake of foods or food groups may be of more relevance to the formulation of public health recommendations than studies of specific nutrients or dietary compounds per se. As a corollary to the decrease in whole-grain intake in recent decades in the United States, we speculate that a return to eating greater quantities of whole grain may be beneficial in combating chronic diseases such as IHD and cancer. In these Iowan women, whole grains were eaten largely as breads or cereal (on average >70% of servings). Refined grains were eaten in diverse forms as breads, cereal, pasta, rice, pizza, and high-sugar products such as cakes. Many of these refined-grain products could be eaten with whole grain substituted. If further research finds that whole grains do protect against important chronic diseases, modification of the form in which grains are eaten (whole compared with refined) would be a powerful public health tool. A strength of this study is that it corrects deficiencies in previous studies identified by Jacobs et al (11). The Iowa Women s Health Study analyses specifically defined level of intake so that high intake could be examined for women eating 1 serving daily for all foods studied. The food-frequency questionnaire allowed the study of internal consistency and relative frequency

9 256 JACOBS ET AL of intake by considering a variety of products separately. There was a wide range of self-reported grain intake in these postmenopausal women. The data show that eating of whole grains is associated with an otherwise healthier diet and also with healthier behaviors. Unlike most of the studies reviewed by Jacobs et al (11), adjustments could be made in the Iowa Women s Health Study for both other risk behaviors and aspects of diet unrelated to constituents of whole grain, although we bear in mind that there may be residual confounding because other risk factors were self-reported. There are limitations to this study. Because data were collected through mailed, self-administered questionnaires, information on blood lipids, an important risk factor for IHD, was not available. The dietary measure may not be robust, given that it is a single assessment of late-life diet only. A perplexing finding is the upturn in observed risk in the highest quintile of whole-grain intake compared with the fourth quintile; alternatively, one might interpret that the risk of IHD reaches a threshold with an intake of whole grain >1 serving/d. One might think that the food-frequency questionnaire was less reliable in those who endorse the most items, but exclusion of participants in the top 5% of energy intake (>12 MJ/d) did not alter the findings (data not shown). The accuracy of food-frequency questionnaires may not be sufficient for judging optimal whole-grain intake. Women who had a disease at baseline might have changed their diets after a diagnosis, creating a bias; however, exclusion of those who reported having cancer or diabetes at baseline did not materially alter the findings (data not shown). Also, the Iowa Women s Health Study does not have validated incident IHD events; it may be that associations with IHD incidence differ from those with IHD death. Although the health effects of grains may differ from each other, it is not possible to study specific grains with these data because the data do not allow a specific and reliable judgment to be made about the type of grain used in the breads, cereals, and other products typically eaten. Asking about a single, usual cold breakfast cereal limits the flexibility of reporting when people eat multiple cereals or are otherwise inconsistent in their consumption pattern. The rubrics dark bread and other whole grains, for example, may include refined-grain products such as pumpernickel bread and couscous. The dominant grain in the US diet is wheat, and these findings may therefore pertain largely to whole wheat. However, oats (45) and combined whole grains (46), which have been shown to have a small hypocholesterolemic effect, are also widely eaten. In summary, whole-grain intake was inversely associated with risk of death from IHD. The inverse association was not attributable to other self-reported risk factors for IHD, including saturated fat or cholesterol intake as summarized by the Keys score. In these data, this finding was only partially explained by constituents of whole grains such as dietary fiber, phytic acid, or vitamin E. In contrast with whole-grain intake, refined-grain intake was inconsistently associated with risk of death from IHD. The findings reported here are generally corroborated by other studies indicating inverse associations of dietary fiber (7), vitamin E (6, 9), folic acid (47 49), or other constituents of whole grains with risk of IHD. They are also consistent with public health dietary recommendations to make bread and cereal a foundation of one s diet and to emphasize whole grains in this context (15, 50). We acknowledge Leonard Marquart and Kathryn Wiemer, both of the Nutrition Division, General Mills, Inc, for contributions to the conceptualization of this investigation and for assistance with coding the whole- and refined-grain content of brand name breakfast cereals. REFERENCES 1. Potter JD. Food and phytochemicals, magic bullets and measurement error: a commentary. Am J Epidemiol 1996;144: Kritchevsky D. Diet in heart disease and cancer. Adv Exp Med Biol 1995;369: Steinmetz KA, Potter JD. Vegetables, fruit and cancer. I. Epidemiology. Cancer Causes Control 1991;2: Steinmetz KA, Potter JD. Vegetables, fruit and cancer. II. Mechanisms. Cancer Causes Control 1991;2: Prineas RJ, Kushi LH, Folsom AR, Bostick RM, Wu Y. Walnuts and serum lipids. N Engl J Med 1993;329:359 (letter). 6. Kushi LH, Folsom AR, Prineas RJ, Mink PJ, Wu Y, Bostick RM. Dietary antioxidant vitamins and death from coronary heart disease in postmenopausal women. N Engl J Med 1996;334: Meyskens FL Jr. Strategies for prevention of cancer in humans. Oncology 1992;6: Strain JJ. Putative role of dietary trace elements in coronary heart disease and cancer. Br J Biomed Sci 1994;51: Schalch W, Weber P. Vitamins and carotenoids a promising approach to reducing the risk of coronary heart disease, cancer and eye diseases. In: Armstrong D, ed. Free radicals in diagnostic medicine. New York: Plenum Press, 1994: Slavin J, Jacobs D, Marquart L. Whole-grain consumption and chronic disease: protective mechanisms. Nutr Cancer 1997;27: Jacobs DR, Slavin J, Marquart M. Whole grain intake and cancer: a review of the literature. Nutr Cancer 1995;24: Fraser GE, Sabaté J, Beeson WL, Strahan TM. A possible protective effect of nut consumption on risk of coronary heart disease: the Adventist Health Study. Arch Intern Med 1992;152: Gramenzi A, Gentile A, Fasoli M, Negri E, Parazzini F, LaVecchia C. Association between certain foods and risk of acute myocardial infarction in women. Br Med J 1990;300: Putnam JJ, Allshouse JE. Food consumption, prices, and expenditures, Annual data, Beltsville, MD: US Department of Agriculture, (Food and Economics Division, Economic Research Service, US Department of Agriculture, statistical bulletin #928.) 15. US Department of Agriculture. Nutrition and your health: dietary guidelines for Americans. 4th ed. Hyattsville, MD: US Department of Agriculture Human Nutrition Information Service, (Home and Garden Bulletin 232.) 16. Appel LJ, Moore TJ, Obarzanek E, et al. A clinical trial of the effects of dietary patterns on blood pressure. N Engl J Med 1997;336: Sacks FM, Obarzanek E, Windhauser MM, et al. Rationale and design of the Dietary Approaches to Stop Hypertension Trial (DASH). A multicenter controlled-feeding study of dietary patterns to lower blood pressure. Ann Epidemiol 1995;5: Willett WC, Sampson L, Browne ML, et al. The use of self-administered questionnaire to assess diet four years in the past. Am J Epidemiol 1988;127: Feskanich D, Rimm EB, Giovannucci EL, et al. Reproducibility and validity of food intake measurements from a semiquantitative food frequency questionnaire. J Am Diet Assoc 1993;93:790 6.