Effects of grains on glucose and insulin responses

Ph.D. Kay M. Behall and Ph.D. Judith Hallfrisch

Diet and Human Performance Laboratory
Beltsville Human Nutrition Research Center
Agricultural Research Service, USDA, Beltsville, MD 20705


Источник: Behall K.M. and Hallfrisch J., Effects of Grains on Glucose and Insulin Responses. Pages 269-281 in: Whole-grain foods in health and disease: Marquart L., Slavin J.L. and Fulcher R.G., ads, Am. Assoc. of Cereal Chemists: st. Paul, MN, 2002.



Whole grains, if consumed, could provide a substantial contribution to the improvement of the diets of Americans, since many whole-grain foods and grain-fiber sources have been shown to be beneficial in reducing insulin resistance and improving glucose tolerance. Dietary guidelines by the U.S. Department of Agriculture (USDA, 1995, 1996) recommend the consumption of 6-11 servings per day from the grains group (such as bread, cereals, rice, and pasta), with three servings per day from whole grains. However, in the 1994-1996 survey data (Cleveland et al, 2000), U.S. adults averaged 6.7 servings of grain products per day in their diet. Only 8% of Americans consume at least three servings of whole grains per day, with the average consumption in the United States being less than one serving per day (Cleveland et al, 2000). It seems clear that a large part of the U.S. population could benefit from eating more grain products, especially whole grains.

Whole grains, grain fractions, and grain extracts have been reported to control or improve glucose tolerance and reduce insulin resistance. The inability of the body to maintain normal glucose levels with normal concentrations of insulin production or to require excessive levels of insulin (hyperinsulinemia) to do so has been called “glucose intolerance”, “impaired glucose tolerance”, “insulin resistance” and “Syndrome X”, a disorder in which insulin resistance, hyperinsulinemia, or both are present. Insulin resistance has been reported to be a major factor in the development of type 2 diabetes mellitus and, for many people, is the first observed abnormality in the progression of the disease (Daly et al, 1997). In addition, increased insulin concentrations generally indicate insulin resistance in nondiabetic individuals. Obesity has been reported to be the most common condition associated with insulin resistance (Daly et al, 1997). Increasing whole-grain intake in the population could result in improved glucose metabolism and could delay or reduce the risk of developing type 2 diabetes mellitus.

There are several mechanisms by which grains may improve glucose metabolism and delay or prevent the progression of impaired glucose tolerance to insulin resistance and diabetes. The form, amount, and method of cooking of these foods, as well as the age, sex, and health characteristics of the group of subjects studied, are all important factors in the effectiveness of the foods in altering these responses. These mechanisms are related to the physical properties and structure of grains. The composition of the grain (including particle size, amount [page 269] and type of fiber, viscosity, and amylose and amylopectin content) all affect the metabolism of carbohydrates from grains.


Particle Size

Most grains consumed in developed countries are milled and processed before being used to manufacture food products. Milling whole grains significantly disrupts the grain structure and changes the nutrient content; i.e., most of the bran and germ are removed and the starch content increases (Weaver, 2001). For example, whole-wheat grains contain approximately 14% bran, while refined white wheat flour contains less than 0.1% bran (Table 1). Whole rice and corn consumed as whole grains without being chewed resulted in lower postprandial glucose levels than the same food after it had been thoroughly chewed (Read et al, 1986). Lower glucose and insulin responses were reported by Collier and O'Dea (1982) after whole brown rice was consumed compared to responses after ground brown rice or glucose were consumed by both normal and diabetic subjects. The authors concluded that the difference in glucose response was due to the lack of processing of the whole grains before or during consumption.

Grains are not commonly consumed whole but after varying degrees of processing before consumption. When compared with white bread, whole wheat bread that contained boiled kernels elicited lower blood glucose and insulin responses (Braaten et al, 1991). A comparison of four types of wheat (whole-grain, cracked-grain, and coarse and fine whole-meal flour) in 10 healthy subjects resulted in glucose responses to whole-grain flour of approximately one-third the response to the fine flour (Holt and Brand-Miller, 1994). Insulin responses were similar. The highest responses occurred after the fine-ground flour, followed by the coarse flour and cracked grain, with the lowest response after the whole-grain product. We also have compared breads made with different particle sizes (Behall et al, 1999a). Consumption of breads made with white flour, standard whole-wheat flour, and ultra-fine-ground whole-wheat flour by middle-aged men and women resulted in lower glycemic responses compared with glucose, but the responses to the whole-wheat breads, although lower than those for white bread, were not different. The standard whole-wheat bread had particles similar in size to the flour used by Holt and Brand-Miller (1994).

Table 1
Composition Differences Between Whole- and Refined-Wheat Component

 Whole WheatRefined Wheat
Bran (%)14<0.1
Germ (%)2.5<0.1
Fat (%)2.71.4
Protein (%)14.213.5
Carbohydrate (%, starch and sugars)67.281.2
Total dietary fiber (%)12.62.9
Insoluble dietary fiber (%)11.51.9
Soluble dietary fiber (%)1.11.0

a - Adapted from Weaver (2001).

[page 270] Holm and Bjorck (1992) also compared responses of healthy subjects to white wheat breads with or without intact kernels. Breads with intact kernels resulted in lower glucose responses and higher satiety scores. Significantly lower glucose and insulin responses were also reported after consumption of coarse bread products containing kernels from wheat, rye, or barley (but not after consumption of breads with oat kernels) compared with white bread (Holm and Bjorck, 1992). Heaton et al (1988) observed higher insulin, but not glucose, responses after meals containing fine-ground flour compared to responses after meals with coarse-ground flour, cracked grain, or whole grain. Similar to those observed with the wheat, insulin responses were greater with fine-ground corn-meal than with coarse or whole corn. However, no differences in insulin response were observed between fine oatmeal, rolled oats, and whole groats (Heaton et al, 1988). Obese subjects with ileostomies fed coarse or fine whole-meal flours in a test meal had higher glucose and insulin responses after the fine-ground flour (O'Donnell et al, 1989). Starch content in the effluent was significantly higher after the test meal containing the coarse-ground grain, indicating the effect of the grain's particle size on starch digestion. A high correlation was observed between the percentage of starch hydrolysis in vitro of diets fed to pigs and the mean insulin areas under the curve.

Boiled whole kernels and larger particle sizes are also associated with lower glucose and insulin responses for a variety of grain sources.

Starch Structure

Starch is composed of long, straight chains of glucose (amylose) and highly branched chains of glucose (amylopectin). These differences have been most frequently studied in corn products because the range of amylose varies from 30 to 70% of the starch, whereas most grains average 25-30% amylose. The difference in starch structure has been shown to have a profound effect on the glucose and insulin responses. A study in our laboratory (Behall et al, 1988), reported insulin responses of 12 women and 13 men to be significantly lower after consumption of corn crackers containing 70% amylose compared to crackers with 70% amylopectin. Peak glucose responses were also lower after high amylose. A controlled dietary study (Behall et al, 1989) found significantly lower glucose and insulin responses to corn crackers in 12 men after they consumed a variety of foods high in amylose or amylopectin for five weeks each. A long-term study (each starch consumed 14 weeks) found hyperinsulinemic men to be more responsive to the beneficial effects of the high-amylose diet than were controls (Behall and Howe, 1995). Although hyperinsulinemic individuals are not considered to be in a diseased state, abnormal insulin and glucose levels have been used as the primary indicators of carbohydrate sensitivity and type 2 diabetics (Beebe and Rubenstein, 1987; Kuczmarski et al, 1994). Noaks et al (1996) fed either low- or high-amylose starch for four weeks each to overweight hypertriglyceridemic men and women. Plasma glucose response was 7% lower at 45 min and insulin response 28% lower at 75 min after the high-amylose versus low-amylose muffin challenge.

Other researchers have reported beneficial effects of high-amylose corn products after acute tests. Postprandial glucose and insulin were significantly lower after meals containing retrograded high-amylose cornstarch than after meals with pregelatinized standard cornstarch (Achour et al, 1997). [page 271] Weststrate and van Amelsvoort (1993) found significantly reduced insulin responses after high-amylose breakfasts, and both glucose and insulin responses were significantly reduced after high-amylose lunches. Granfeldt et al (1995a) and Semprun-Fereira et al (1994) tested healthy subjects using arepas (flat cornbread made from precooked cornmeal) made with standard or high-amylose corn flour. Glucose and insulin responses after high-amylose arepas were significantly lower than after ordinary cornmeal arepas. Krezowski et al (1987) observed significant reduction in glucose in type 2 diabetic subjects after consumption of high-amylose muffins compared with standard (low-amylose) muffins, cornflakes, or glucose. Insulin response was higher after the standard muffins than after the high-amylose muffins. These results show that amylose content and the level of processing (which affects resistant starch development) have a great influence on the responses to corn-containing foods.

Rice can vary substantially in amylose content, although it does not have as wide a range in content as corn. Comparison of responses to three rice varieties with varying amylose content (served as brown rice, puffed rice cakes, rice pasta, and rice bran) found only high-amylose varieties to lower glucose and insulin responses (Brand-Miller et al, 1992). Holt and Brand-Miller (1995) compared standard and quick-cooking rice and high- and low-amylose rice puffs and found responses to quick-cooking rice to be 60% higher than to standard rice. Low-amylose rice resulted in a 50% higher glucose response than high-amylose rice. However, Panlasigui et al (1991) concluded after testing three high-amylose varieties of rice that amylose content alone cannot predict the glycemic response and that gelatinization is also a factor.


Viscosity and Solubility

Fiber of oats, barley, and rye is, on average, about one-third soluble fiber capable of forming a gel and two-thirds insoluble (non-gel-forming) fiber. Wheat contains less soluble fiber and rice virtually none. Refined grains are typically low in total dietary fiber; refining primarily decreases insoluble fiber, which increases the percentage of soluble fiber remaining. The gel-forming property of soluble fiber sources has been proposed as the mechanism by which oats and barley reduce both glucose and insulin responses and cholesterol. Jenkins and coworkers were among the first to investigate the beneficial effects of soluble fibers, starting with isolated fiber sources. Jenkins et al (1978) reported that oat gum reduced glucose and insulin responses of healthy adults when added to a glucose solution. The high viscosity of the solution containing oat gum was concluded to be the property that delays gastric emptying and/or intestinal absorption, resulting in these lower responses (Wood et al, 1989). Braaten et al (1991) also tested responses to glucose and glucose with oat gum and found reductions in glucose and insulin when the nine healthy subjects consumed solutions to which oat gum had been added. In diabetics and controls, Braaten et al (1994) found both oat gum and oat bran added to farina to reduce glucose and insulin responses below those of farina alone. An oat-based soup was used as a means of weight reduction in 31 subjects (Rytter et al, 1996). After 23 weeks, subjects lost an average of 6 kg, and both glucose and insulin responses declined; however, it is difficult to separate the effects of the soup from the weight loss caused by reduced energy intake. [page 272] In our own research, we have found two levels of highly viscous oat extracts to lower glucose and insulin responses of middle-aged men and women (Hallfrisch et al, 1995); both levels were lower than that used by Wood et al (1994). Lower glucose and insulin responses were observed whether the oat extract consumed was uncooked, boiled, or baked (van der Sluijs et al, 1999).

Granfeldt et al (1995b) tested responses of nine older men to raw rolled oats, boiled rolled oats, boiled intact oat kernels, and white bread. Boiled intact oat kernels resulted only in glucose and insulin response reductions below the response to white bread. Preliminary data using rolled oats or oat flour indicate reductions in glucose response of 15-30% in moderately overweight women (Behall et al, 1999b). Comparisons of glucose and insulin responses of 24 hypercholesterolemic men to wheat, rice, and oats found no significant differences (Kestin et al, 1990). A smaller study of six men found no differences among glucose and insulin responses to 10 g of fiber from oat bran, wheat bran, wheat fiber, or wheat germ (Cara et al, 1992). These sources of fiber, however, are predominantly insoluble and would not result in highly viscous intestinal contents. The test subjects in these studies were also relatively young, lean subjects.

Barley is also high in soluble fiber and has the potential to improve insulin sensitivity and glucose metabolism; however, little barley is consumed by Americans. Yokoyama et al (1997) compared responses of five subjects to pastas containing wheat or wheat and 12 g of β-glucans from barley. Consumption of barley-containing pasta resulted in lower glycemic and insulin indices. Bourdon et al (1999) compared glucose and insulin responses of 11 healthy men (28-42 years). Subjects consumed traditional wheat pasta or a pasta in which 40% of the wheat flour had been replaced with either a highly viscous barley cultivar (Prowashonupana), which naturally has about 15% soluble β-glucans, or Waxbar barley enriched in soluble fiber by repeated milling. Insulin responses were lower after the barley pastas than after the traditional wheat pasta during the first hour. Although the glucose areas under the curves did not differ, the decline in glucose after barley pastas was more gradual than after the wheat pasta. Substantial reductions in insulin and glucose responses were found in seven young subjects after consumption of varying amounts of boiled barley compared with white bread (Wolever and Bolognesi, 1996a). Breads made with 10% whole barley or 15% pearled barley lowered glucose responses of 15 diabetics compared with responses to white bread (Urooj et al, 1998). Another study of diabetics compared long-term (24 weeks) effects of barley bread to white bread in 11 men with type 2 diabetes (Pick et al, 1998). The glycemic index was reduced, but insulin responses were increased when the men consumed the barley bread. However, the insulin regimen was changed for some subjects during the study. Comparison of barley and oat foods with white bread in nine healthy subjects (Liljeberg et al, 1996) found no effect of porridges, but consumption of high-fiber barley breads resulted in glycemic indices of 57-72% of the white-bread index and insulin indices of 42-72% of the white bread index. Whole kernels of wheat, rye, and barley added to breads resulted in glycemic indices lower than those of white bread, but adding whole oat kernels to the bread had no effect on the responses of 10 healthy subjects (Liljeberg et al, 1992). However; insulin indices for all breads to which boiled kernels had been added were lower than those for white bread. Preliminary data from our laboratory indicate that both flakes and flour of the high-soluble-fiber barley (Prowashonupana) lower glucose and insulin in moderately overweight middle-aged women (Behall et al, 1999b). [page 273]

Rye also is relatively high in soluble fiber, but few human studies report responses to rye consumption. Normal-glycemic men and women had significantly lower insulin responses after eating whole-kernel rye bread compared to wheat bread, with no difference in glucose response (Leinonen et al, 1999). Consumption of whole-meal rye bread resulted in significantly higher glucose response and lower insulin response compared to responses from whole-meal rye crisp-bread (Leinonen et al, 1999). Whole-kernel rye has been tested in a few groups of subjects with diabetes (Jenkins et al, 1986; Brand et al, 1990; Liljeberg et al, 1992). Compared to white bread, the whole-kernel rye bread produced significantly lower glucose response in diabetics, both insulin and noninsulin-depend-ent diabetics, with a glycemic index of 42-56 compared to the bread. However, rye crispbreads had glucose responses very similar to those of white bread. No differences in the glucose responses of 14 young diabetics were reported after they consumed dark rye bread and white bread (Birnbacher et al, 1995).

Wood et al (1994) concluded that reductions in plasma glucose and insulin (79-96%) after consumption of oat gum were primarily the result of viscosity changes. Granfeldt et al (1995b), however, found neither degree of gelatinization nor viscosity to affect glucose levels. Grains that contain highly viscous fibers, and foods made with these grains, can be effective in lowering blood glucose and insulin responses. These effects are most likely to be found in subjects for which lowering glucose and insulin is an improvement, that is, older, overweight subjects and those with type 2 diabetes. The effects on glucose and insulin responses are less significant if subjects are young, fit, and have normal glucose and insulin responses. However, to be effective, whole kernels or the soluble-fiber components of the grains must be consumed, and currently the typical American appears to prefer highly refined flours and cereals or their brans.


Glycemic Index

The glycemic index is a calculation based on the postprandial blood glucose response (area under the curve) of a specific weight of carbohydrate-containing food compared to the response after the same weight of a reference food consumed by the same subject (Reaven and Miller, 1968; Wolever, 1991, Brand-Miller and Foster-Powell, 2000). The resulting glycemic index can be used to rank foods (Foster-Powell and Brand-Miller, 1995; Brand-Miller and Foster-Powell, 2000) (Table 2). Glucose and white wheat bread have both been used as the reference food. It is important to remember that white wheat bread is low in fiber content, uses finely ground flour, and (when compared to glucose) has a glycemic index of 70. This is substantially lower than many other foods, including cornflakes, some rice foods, potatoes, cereals, crackers, and sports drinks. Key factors that influence the observed glycemic index include cooking methods, physical form of the food, type of starch, fiber, and sugar. An insulin index has also been calculated utilizing the areas under the curve as described above for the glycemic index. A much more extensive database exists for the glycemic index of foods than for the insulin index (Foster-Powell and Brand-Miller, 1995; Brand-Miller and Foster-Powell, 2000). [page 274]


Glycemic Response

Most high-carbohydrate foods are based on grains (especially wheat, corn, and oats) or sugars. After processing to increase commercial palatability, most cereals have high glycemic indexes. Differences in responses observed after consumption of wheat and wheat products appear to be the result of processing and the particle size differences discussed previously. Wheat pastas are processed with minimal mechanical disruption of the starch granule during manufacture. Compared to bread, pasta meals have been reported to significantly reduce plasma glucose and insulin (35 and 39%, respectively) (Jarvi et al, 1995). The wide variety of pasta types exhibits a broad range in glycemic index, from 32 to 55 compared to glucose. Protein appears to interact with the starch to reduce gelatinization and slow stomach emptying. Wheat contains low to average amounts of amylose and relatively low levels of soluble fiber compared to oats.

Corn and corn products have a wide range in glucose and insulin responses depending on cultivar, form, processing, and levels of amylose and amylopectin. Cornflakes, by prediction and actual measurement, had higher glycemic and insulin indexes (139 and 149, respectively) than white bread when determined in eight nondiabetic university students (Wolever and Bolognesi, 1996b) and 14 diabetic children (Birnbacher et al, 1995). Wheeler et al (1996) compared responses of 16 young diabetics (14-25 years old) to cornflakes, sweetened cornflakes, glucose, and sucrose. Finding no differences between sweetened and unsweetened cornflakes and either cornflakes and glucose or cornflakes and sucrose, the authors concluded that cornflakes, whether sweetened or unsweetened, are not detrimental to these patients. While an occasional meal of cornflakes may be no more detrimental than glucose when consumed by a diabetic, cornflakes are not beneficial in reducing plasma glucose and insulin requirement, as is observed after consumption of cereals with lower indices. Golay et al (1992) fed 14 overweight type-2 diabetic patients cornflakes or muesli for breakfast for two weeks, each with all other dietary components the same. After a glucose tolerance test at the end of each period, plasma glucose responses were not different. However, insulin was significantly lower at fasting (17%) and 2 hr (21%) after the muesli compared to that after cornflakes.

Table 2
Glycemic Index (GI) of Selected Grains and Grain Products
Compared to the Response after a Glucose Tolerance Test *

Whole GrainGIProcessed GrainGI
 Barley, including pearled25 Rolled barley66
 Corn, sweet kernel55 Corn flakes84
 Oatmeal, old-fashioned49 Oatmeal, quick66
 Rice, brown55 Krispy rice88
 Rice, sweet, low-amylose88 Rice, high-amylose59
 Rye, whole-kernel34 Rye flour bread65
 Wheat, whole-kernel41 Puffed wheat74
 Bulgur wheat48 Wheat flakes75

* Adapted from Foster-Powell and Brand-Miller (1995).

These authors concluded that the change of cereal type, from rapid- to slow-release starch, improved carbohydrate metabolism and probably insulin sensitivity in the diabetic patient. [page 275]

Responses of healthy and diabetic South Africans to refined maize, rice, and bread found the glucose response to maize to be higher than to bread and similar to glucose, but the insulin response to maize was about two-thirds the response to bread (Segal et al, 1991; Wolever et al, 1994). Some Native American corn products have much lower glycemic indices – e.g., tortillas (~38) and hominy (~40)– than white bread (~70 when glucose = 100) (Foster-Powell and Brand-Miller, 1995). Even boiled sweet corn and popcorn have glycemic indices (both ~55) lower than white bread (~70) compared to glucose (100). However, maize porridge, whether refined or unrefined and taco shells, have glycemic indices (~71, ~74, and ~68, respectively) comparable to that of white bread (Segal et al, 1991; Foster-Powell and Brand-Miller, 1995).

Rice products have a wide range of glycemic indices, and rice is considered to be either a high- or low-glycemic-index food based on cooking method, cultivar, form of food, and subject group studied (Brand-Miller et al, 1992). Baking rice after boiling, compared to boiling only, resulted in a significantly lower glucose area under the curve 60 min after the meal as well as differences in viscosity and in vitro hydrolysis rates (Gatti et al, 1987). Larsen et al (2000) fed diabetic subjects white bread or polished rice of the same variety that was either not parboiled, mildly parboiled, or severely pressure-parboiled. All three rice meals resulted in significantly lower postprandial glucose and insulin responses compared to those of the bread meal. The glycemic indices were 55, 48, and 39, respectively, for the rice not parboiled, mildly parboiled, or severely pressure-parboiled. However, the treatment of rice to produce “instant” rice results in a product with a glycemic index higher than that of the bread (~87 vs. 70) (Foster-Powell and Brand-Miller, 1995). The amylose content (0-40% depending on cultivar) and final gelatinization temperature have been reported to be negatively correlated with the rate of glucose found in the plasma after rice consumption and to decrease with processing, while resistant starch generally increases with processing such as parboiling and noodle extrusion (Juliano, 1992).

The results for rice are inconsistent, with some rice-based foods having higher glycemic indices than white bread while brown rice may have a lower glycemic index than white rice (Foster-Powell and Brand-Miller, 1995). Rice bran has been reported to have a glycemic index of 27 compared to that of bread (Foster-Powell and Brand-Miller, 1995). However, addition of 10 g of rice bran to a liquid meal did not lower glucose or insulin responses of six healthy males (Cara et al, 1992). Comparison of brown rice and barley responses in 10 healthy subjects resulted in 30% lower glucose responses to the barley meal than to the rice (Thorburn et al, 1993). The higher level of fermentation of the barley, as measured by hydrogen expiration, was thought to be the mechanism by which barley improved glucose response. The South African study found higher glucose and insulin responses to refined white rice than to white bread (Segal et al, 1991). Rasmussen et al (1992a) compared responses of seven diabetics to 25 and 50 g of carbohydrate from white rice and white bread and found significantly lower glucose and insulin responses after the 50 g of rice compared to the 50 g of white bread but no differences in the responses to 25 g of either carbohydrate, indicating that the amount of food consumed affects glucose and insulin responses. Similar reductions were found in responses of men and women to 100 g office compared to white bread (Rasmussen et al, 1992b). [page 276]

These results show that the assumption that all carbohydrates are equal (exemplified by the current concept of carbohydrate exchange) is no longer accurate. Many factors, including obesity and diabetes, affect the glucose and insulin responses resulting from consumption of grain-based carbohydrate-containing foods. Consumption of diets high in foods with high glycemic indices promote insulin resistance, obesity, and (in susceptible segments of the population) noninsulin-dependent diabetes mellitus (Brand-Miller, 1994). Obesity is associated with decreased ability of the body to control blood glucose with normal levels of insulin (Expert Committee, 1998). This may also be an early step in the development of noninsulin-dependent diabetes mellitus (Expert Committee, 1998). Consumption of grain sources that reduce the level of insulin required to maintain normal blood glucose is a means of improving insulin resistance or increasing insulin sensitivity (Gannon and Nuttall, 1987; Kiens and Richter, 1996). Other tests that measure changes in insulin resistance are more time-consuming or invasive; examples are glycemic clamps (Elahi, 1996) and Bergman's minimal model (Bergman et al, 1985).


Summary

Various grains and grain products can be beneficial in lowering glucose and insulin responses. Although this review is not by any means complete, several generalities can be expressed. 1) The greater the particle size, the lower the glucose and insulin response. Or, inversely, the greater the level of processing and refining, the higher the response. 2) Higher amylose content results in lower glucose and insulin responses. Corn and rice can have either high or low glycemic indices because their amylose and amylopectin contents vary. 3) Grains with high levels of soluble β-glucans such as oats, rye, and barley are generally more effective in improving insulin sensitivity than wheat, which contains predominantly insoluble dietary fiber. The high viscosity of these soluble fibers is partially responsible for these beneficial effects. 4) The type and characteristics of the subjects tested are important in determining the level of reduction that can be achieved. Therefore, older, less slim, more glucose-intolerant subjects have the capacity for greater improvement in glucose and insulin responses than do young, fit, slim subjects. The present American diet has great room for improvement if the amount of daily whole-grain servings would be increased from less than one serving to the recommended three serving per day. Replacing foods from low-fiber grains such as cornflakes or white bread with whole-grain products having higher fiber or higher amylose content will reduce the risk of developing insulin resistance and obesity and improve the health of the American population. [page 277]