Professional Documents
Culture Documents
Gaesser 2007
Gaesser 2007
Review
A
n estimated 66% of adult Americans are considered
ABSTRACT overweight, with a body mass index (BMI; calcu-
The increased prevalence of overweight and obesity in the lated as kg/m2) of 25 to 29.9, or obese, with a BMI
United States since approximately 1980 is temporally ⱖ30 (1). Average weights of American workers aged ⬎18
associated with an increase in carbohydrate intake, with years increased at a rate of 0.61% per year from 1986 to
no appreciable change in absolute intake of fat. Despite 1995; from 1997 to 2002, the rate increased 0.95% per
speculation that both carbohydrate quantity and quality year (2). Contributing causes of this weight gain are
have contributed significantly to excess weight gain, the largely environmental (3). Per capita food energy in the
relationship between carbohydrate intake and body mass US food supply has increased steadily over the past 40
index (BMI) is controversial. A review of relevant litera- years, from about 3,100 kcal/day in 1965 to an all-time
ture indicates that most epidemiologic studies show an high of 3,900 kcal/day in 2000 (Figure 1) (4). This coin-
inverse relationship between carbohydrate intake and cides with National Health and Nutrition Examination
BMI, even when controlling for potential confounders. Surveys data that indicated an increase in reported daily
These observational studies are supported by results energy intake between 1971 and 2000 of 168 kcal in men
from a number of dietary intervention studies wherein and 335 kcal in women (Figure 2) (5). Because levels of
modest reductions in body weight were observed with an leisure-time physical activity among US adults remained
stable or increased slightly between 1990 and 2004 (6,7),
ad libitum, low-fat, high-carbohydrate diet without em-
increased energy intake likely explains much of the
phasis on energy restriction or weight loss. With few
weight gain experienced by Americans during this time.
exceptions, high glycemic load is associated with lower The increase in reported energy intake since 1970 oc-
BMI, even when adjusted for total energy intake. Data on curred in conjunction with an increase in the consump-
the association between glycemic index and BMI are not tion of carbohydrates, which rose by 60 to 70 g/day (5). By
as consistent, with more studies showing either no asso- contrast, reported intake of fat remained relatively stable
ciation or an inverse relationship, rather than a positive between 1971 and 2000 (Figure 3) (5). Consequently, fat
relationship. Whole-grain intake is generally inversely intake, as a percentage of total energy, actually decreased
associated with BMI; refined grain intake is not. Because slightly during this period. Because the increase in obe-
overall dietary quality tends to be higher for high-carbo- sity prevalence occurred during a period in which public
hydrate diets, a low-fat dietary strategy with emphasis health messages encouraged low-fat eating, proponents of
on fiber-rich carbohydrates, particularly cereal fiber, may low-carbohydrate diets attributed Americans’ weight
be beneficial for health and weight control. gain to carbohydrates. As a result, a variety of low-car-
J Am Diet Assoc. 2007;107:1768-1780. bohydrate diets became popular in the past 10 years
(8-11). However, popularity of low-carbohydrate diets has
waned considerably since their peak in 2004, in much the
same manner that interest in these diets all but disap-
peared in the mid-1970s after nearly a decade of popu-
G. A. Gaesser is a professor of exercise physiology, De- larity (12-14). Loss of enthusiasm for carbohydrate-re-
partment of Human Services, University of Virginia, strictive eating plans most likely is attributable to their
Charlottesville. lack of sustainability (15,16).
Address correspondence to: Glenn A. Gaesser, PhD, The role of diet composition in weight control and obe-
210 Emmet St S, PO Box 400407, Charlottesville, VA sity remains controversial (17,18). Carbohydrate quality
22904-4407. E-mail: gag2q@virginia.edu and quantity has received considerable attention (19-23).
Copyright © 2007 by the American Dietetic Several reports suggest that diets with a high glycemic
Association. index or glycemic load, or that are high in refined carbo-
0002-8223/07/10710-0013$32.00/0 hydrates, increase risk of obesity and associated health
doi: 10.1016/j.jada.2007.07.011 problems (19-22), although the hypothesized link be-
1768 Journal of the AMERICAN DIETETIC ASSOCIATION © 2007 by the American Dietetic Association
Figure 1. Increase in daily food energy in the US food supply, per
capita per day: 1909 to 2000. (Reprinted from reference 4.)
Women
NHANESb (29) 26.3⫾0.2c 26.4⫾0.3c 26.1⫾0.4 25.9⫾0.2c 25.5⫾0.2c
n⫽6,125 35.6d 44.6d 50.4d 56.4d 66.8d
CSFIIe (30) 26.7⫾0.5c 26.0⫾0.2c 25.7⫾0.1c 25.2⫾0.2c
n⫽4,711 0 to 30d 30 to 45d 45 to 55d ⬎55d
NHSf I (31) 25.2⫾4.6g 25.2⫾4.6g 25.1⫾4.6g 24.9⫾4.6g 24.7⫾4.5g
n⫽71,919 34.8d 39.5d 42.2d 46.5d 55.1d
NHS II (34) 26 25 23
n⫽90,655 41.2d 50.1d 59.4d
WHSh (35) 26.7⫾5.4g 26.3⫾5.0g 26.1⫾5.0g 25.7⫾4.8g 25.2⫾4.7g
n⫽38,446 41.7d 46.8d 50.5d 54.6d 63.6d
PLCOi (39) 27.4⫾5.6g 27.2⫾5.6g 26.5⫾5.2g
n⫽18,341 44.1d 51.9d 59.0d
Canadian NBSSj (36) 25.2⫾10.2g 25.1⫾10.2g 24.7⫾10.2g 24.6⫾10.2g 24.3⫾10.2g
n⫽49,111 28.5d 35.3d 38.8d 42.1d 47.9d
Men
NHANES (29) 26.8⫾0.3c 26.5⫾0.2c 26.4⫾0.2c 26.3⫾0.2c 26.2⫾0.3c
n⫽5,730 33.2d 42.7d 48.5d 54.3d 64.3d
CSFII (30) 26.8⫾0.4c 26.8⫾0.2c 26.3⫾0.1c 26.0⫾0.2c
n⫽5,075 0 to 30d 30 to 45d 45 to 55d ⬎55d
HPFSk (31) 26.1⫾3.4g 25.9⫾3.3g 25.6⫾3.2g 25.3⫾3.2g 24.8⫾3.2g
n⫽39,926 37.5d 43.6d 47.7d 52.5d 61.5d
PLCO (39) 28.1⫾4.4g 27.6⫾4.1g 26.8⫾3.9g
n⫽20,172 39.5d 47.6d 55.9d
a
Not all studies reported standard error of the mean or standard deviation.
b
NHANES⫽National Health and Nutrition Examination Survey.
c
Body mass index⫾standard error of the mean.
d
Mean or range of carbohydrate intake as a percentage of total energy within each quintile/quartile.
e
CSFII⫽Continuing Survey of Food Intakes by Individuals.
f
NHS⫽Nurses’ Health Study.
g
Body mass index⫾standard deviation.
h
WHS⫽Women’s Health Study.
i
PLCO⫽Prostate, Lung, Colorectal, and Ovarian Screening Study.
j
NBSS⫽National Breast Screening Study.
k
HPFS⫽Health Professionals Follow-up Study.
questionnaire), it is unlikely that underreporting would ysis of combined NHS I and II cohorts was not attribut-
entirely undermine this relationship. able to either lower reported energy intake or higher
Although some studies report that high carbohydrate con- reported physical activity (32).
sumption is associated with lower total energy intake (29- High consumption of carbohydrates tends to be associ-
31), most studies suggest that lower BMI among high-car- ated with higher intake of dietary fiber, which has been
bohydrate consumers is not related to lower total energy reported to a have a favorable effect on BMI (34,47-49). In
intake (32,34-37,44-46). In the Nurses’ Health Study (NHS) the NHS II cohort, the difference in BMI between the
II cohort the highest quintile of carbohydrate consumption highest and lowest quintiles of total fiber intake (one BMI
had a mean BMI three units lower (ie, 26 vs 23) than women unit) was much less than the difference between the
in the lowest quintile of carbohydrate consumption, even highest and lowest quintiles of carbohydrate consump-
when adjusted for total energy intake (34). tion (three BMI units) (34). This suggests that dietary
Lower BMI in high-carbohydrate/energy consumers fiber intake does not entirely explain the generally in-
may be related to higher reported physical activity verse relationship between carbohydrate consumption
(35,46). However, in the Framingham Offspring Study and BMI (Table 1).
the lower BMI associated with high carbohydrate and
total energy consumption was not attributable to differ-
ences in physical activity (45). Furthermore, the lower CARBOHYDRATE QUALITY
BMI (1.1 units) among the highest quintile (compared to Coincident with heightened interest in low-carbohydrate
lowest quintile) of carbohydrate consumption in an anal- diets, a distinction between carbohydrates has been em-
Women
NHSb I (57) 24.7 25.1 25.2 25.3 25.4
n⫽65,173 62.9c 68.1c 70.7c 73.4c 77.9c
NHS II (33) 24.7⫾5.0d 24.6⫾5.3d 24.5⫾5.5d
n⫽91,249 69.9c 76.9c 83.1c
WHSe (35) 26.7⫾5.4d 26.3⫾5.0d 26.1⫾5.0d 25.7⫾4.8d 25.2⫾4.7d
n⫽38,446 50c 52c 53c 54c 55c
Canadian NBSSf (36) 25.2⫾10.2d 25.1⫾10.2d 24.7⫾10.2d 24.6⫾10.2d 24.3⫾10.2d
n⫽49,111 72.5c 78.4c 79.7c 81.7c 84.3c
JMETSg (59) 23.7⫾0.2h 23.9⫾0.2h 23.8⫾0.2h 24.2⫾0.2h 24.4⫾0.2h
n⫽1,354 61c 65c 67c 69c 72c
Men
HPFSi (62) 25.7 25.5 25.5 25.4 25.3
n⫽42,759 65.1c 69.7c 72.6c 75.3c 79.3c
Zutphen (64) 26.0 25.1 24.9
n⫽646 77c 82c 85c
Men/Women
ARICj (white) (63) 26.7⫾4.5d 26.7⫾4.5d 26.5⫾4.5d 26.6⫾4.5d 26.6⫾4.7d
n⫽9,529 69.1c 74.5c 77.2c 79.7c 83.4c
63.7k 55.9k 51.6k 47.2k 49.3k
ARIC (African American) (63) 29.8⫾6.2d 29.0⫾5.8d 28.9⫾6.0d 29.0⫾5.7d 28.9⫾6.3d
n⫽2,722 71.7c 77.6c 80.2c 82.6c 86.6c
76.6k 66.2k 61.0k 57.9k 52.3k
a
Not all studies reported standard error of the mean or standard deviation.
b
NHS⫽Nurses’ Health Study.
c
Mean glycemic index within each glycemic index quintile/tertile.
d
Body mass index⫾standard deviation.
e
WHS⫽Women’s Health Study.
f
NBSS⫽National Breast Screening Study.
g
JMETS⫽Japanese Multicentered Environmental Toxicants Study.
h
Body mass index⫾standard error of the mean.
i
HPFS⫽Health Professionals Follow-up Study.
j
ARIC⫽Atherosclerosis Risk in Communities Study.
k
Percentage of women study participants.
the lowest (four servings per week) and highest (30 serv- DIETARY FIBER
ings per week) quintile of refined-grain intake. In the one The inverse association between carbohydrate intake and
male cohort refined-grain intake was unrelated to BMI BMI may be due in part to dietary fiber (33,34,44,47-
(44) (Table 5). Data on combined men and women are 49,78,85-87). Cereal fiber, in particular, appears to play a
confounded due to the decreasing percentage of women beneficial role in weight control (33,85,87-92). In the NHS
across higher quartiles/quintiles of refined-grain intake I and II cohorts, the highest quintile of cereal fiber con-
(Table 5) and, hence, should be interpreted with caution. sumption had a mean BMI 0.7 (NHS I) and 1.5 (NHS II)
Despite this bias toward finding a positive relationship, units less than that of the lowest quintile of cereal fiber
data from the Framingham Offspring Study revealed no consumption (33,85). Breakfast cereal consumption is
trend between refined-grain intake and BMI (45). predictive of lower BMI (81,88-91). The number of Amer-
In the Physician’s Health Study (82), intake of refined- icans who skipped breakfast between 1965 and 1991 rose
grain (as well as whole-grain) breakfast cereal was asso- from 14% to 25% (88). Since 30% of the US Department of
ciated with lower BMI and was inversely associated with Agriculture’s recommended three or more whole-grain
body weight gain over 8 years. Similarly, in the Women’s servings per day are delivered in the form of breakfast
Health Study (35), servings/day of refined grain was in- foods, reversing the trend of skipping breakfast may also
versely related to BMI. Thus, current data from cohort increase the low percentage of Americans who comply
studies are not sufficiently consistent to conclude that with the US Department of Agriculture’s recommenda-
refined grain intake has a deleterious affect on BMI. tion (92).
Women
NHSb I (31) 25.2⫾4.6c 25.2⫾4.6c 25.1⫾4.6c 24.9⫾4.6c 24.7⫾4.5c
n⫽71,919 107d 132d 144d 156d 181d
NHS II (33) 25.7⫾5.8c 24.5⫾5.1c 23.6⫾4.8c
n⫽91,249 133d 171d 217d
WHSe (35) 26.7⫾5.4c 26.3⫾5.0c 26.1⫾5.0c 25.7⫾4.8c 25.2⫾4.7c
n⫽38,446 92d 106d 117d 127d 143d
PLCOf (39) 27.4⫾5.6c 27.2⫾5.6c 26.5⫾5.2c
n⫽18,341 89.7d 117.1d 142.8d
Canadian NBSSg (36) 25.2⫾10.2c 25.1⫾10.2c 24.7⫾10.2c 24.6⫾10.2c 24.3⫾10.2c
n⫽49,111 98.6d 129.5d 147.2d 164.5d 196.0d
JMETSh (59) 24.2⫾0.3i 23.8⫾0.2i 24.0⫾0.2i 24.2⫾0.2i 23.8⫾0.3i
n⫽1,354 69d 80d 87d 95d 107d
SMCj (69) 24.9 24.7 24.7 24.6 24.6
n⫽61,433 157d 167d 179d 190d 207d
Health ABCk (68) 27.4⫾5.4b 26.7⫾5.4 27.8⫾5.6 26.2⫾4.9 27.1⫾5.7
n⫽1,169 88.5d 108.7d 118.2d 128.0d 148.3d
Men
HPFSl (31) 26.1⫾3.4c 25.9⫾3.3c 25.6⫾3.2c 25.3⫾3.2 24.8⫾3.2c
n⫽39,926 131d 163d 181d 198d 231d
PLCO (39) 28.1⫾4.4c 27.6⫾4.1c 26.8⫾3.9c
n⫽20,172 106.6d 144.7d 182.8d
Health ABC (68) 26.7⫾4.2c 26.8⫾4.0c 26.8⫾3.4c 26.6⫾4.2c 26.7⫾3.4c
n⫽1,079 107.4d 132.7d 143.9d 156.8d 185.3d
a
Not all studies reported standard error of the mean or standard deviation.
b
NHS⫽Nurses’ Health Study.
c
Body mass index⫾standard deviation.
d
Mean glycemic load within the quintile.
e
WHS⫽Women’s Health Study.
f
PLCO⫽Prostate, Lung, Colorectal, and Ovarian Screening Study.
g
NBSS⫽National Breast Screening Study.
h
JMETS⫽Japanese Multicentered Environmental Toxicants Study.
i
Body mass index⫾standard error of the mean.
j
SMC⫽Swedish Mammography Cohort.
k
Health ABC⫽Aging and Body Composition.
l
HPFS⫽Health Professionals Follow-up Study.
Women
NHSb I (76) 24.9⫾5c 24.9⫾5c 24.5⫾4c
n⫽74,091
NHS I (Diabetic) (78) 30.9 29.8 28.4
n⫽902
NHS II (46) 24.9⫾0.5d 24.3⫾0.4d 23.9⫾0.5d
n⫽470 5.9 to 10.8 g/de 20.8 to 23.5 g/de 39.8 to 49.2 g/de
IWHSf (79) 27.2 27.0 27.0 26.6 26.9
n⫽34,492 1.5 s/wke 6.0 s/wke 8.5 s/wke 13.0 s/wke 22.5 s/wke
Men
HPFSg (44) 25.7 25.6 24.9
n⫽42,540 0.2 s/de 1.1 s/de 3.4 s/de
Men/Women
FOSh (45) 27.0 26.9 27.3 26.7 26.6
n⫽2,941 0.9 s/wke 3.5 s/wke 6.4 s/wke 9.5 s/wke 20.5 s/wke
48i 52i 59i 58i 55i
Finnish (74) 26.7⫾4.3c 26.5⫾4.0c 26.3⫾3.8c 26.4⫾3.7c
n⫽4,316 79 g/de 136 g/de 198 g/de 303 g/de
56i 48i 43i 41i
Boston (83) 26.4 25.5 25.3 25.2
n⫽535 0.31 s/de 0.86 s/de 1.49 s/de 2.90 s/de
69i 67i 71i 59i
TLGSj (84) 26.4⫾4.8c 25.8⫾5.1c 25.1⫾4.8c 24.7⫾4.9c
n⫽827 ⱕ10 g/de 10 to ⬍71 g/de 71 to ⬍143 g/de ⱖ143 g/de
58i 61i 54i 48i
a
Not all studies reported standard error of the mean or standard deviation.
b
NHS⫽Nurses’ Health Study.
c
Body mass index⫾standard deviation.
d
Body mass index⫾standard error of the mean.
e
Mean or range of whole-grain intake, in g/d, servings/d (s/d), or servings/wk (s/wk).
f
IWHS⫽Iowa Women’s Health Study.
g
HPFS⫽Health Professionals Follow-up Study.
h
FOS⫽Framingham Offspring Study.
i
Percentage of women study participants.
j
TLGS⫽Tehran Lipids and Glucose Study.
a low-fat diet consistently resulted in spontaneous of premenopausal women consuming an ad libitum, 20%
weight loss. fat diet, systematic adjustments in energy intake de-
In a number of ad libitum, low-fat diet interventions in signed to prevent weight loss were unsuccessful (101).
which weight loss was one of several outcome measures, During a 20-week period, despite increases in energy
weight loss was not encouraged (96-99). Some of these intake to maintain weight throughout the study, women
studies either discouraged weight loss or modified caloric lost an average of 2.8% of body weight and 11.3% of fat
intake in an effort to prevent weight loss (without suc- weight. Thus, weight loss on a low-fat diet, even when not
cess) (100,101). In the Carbohydrate Ratio Manipulation encouraged, may reflect imprecision in food-intake con-
in European National Diets Study, subjects consuming trol when the energy density of the diet is altered (97,101-
an ad libitum, low-fat, complex-carbohydrate diet lost 4.5 104).
kg after 6 months despite no encouragement to actively
reduce caloric intake (96). In view of the concern about
the potential contribution of simple carbohydrates to obe- WEIGHT MAINTENANCE
sity, it is important to note that increasing consumption Long-term weight-loss maintenance is poor (105). Be-
of simple carbohydrates over the 6-month trial did not cause ad libitum, low-fat diets lead to only modest weight
lead to weight gain (96). loss, the greatest potential advantage for advocating a
Ad libitum, low-fat diets may result in weight loss low-fat, high-carbohydrate diet is in prevention of weight
despite efforts to prevent this from occurring. In a study gain (95,106,107). In the WHI (95), 48,835 women were
Women
NHSb I (76) 24.6⫾4c 24.9⫾5c 25.2⫾5c
n⫽74,091
IWHSd (79) 26.9 26.8 26.9 27.0 27.1
n⫽34,492 4.0 s/wke 8.0 s/wke 12.0 s/wke 18.0 s/wke 30.0 s/wke
Men
HPFSf (44) 25.4 25.5 25.4
n⫽42,540 0.6 s/de 1.7 s/de 4.3 s/de
Men/Women
FOSg (45) 26.9 26.8 26.9 27.0 26.8
n⫽2,941 6.9 s/wke 11.8 s/wke 16.7 s/wke 23.6 s/wke 38.9 s/wke
61h 58h 57h 52h 46h
Boston (83) 25.4 25.4 25.5 26.2
n⫽535 1.6 s/de 2.9 s/de 4.1 s/de 6.1 s/de
77h 78h 65h 45h
TLGSi (84) 24.9⫾4.2c 25.5⫾4.7c 26.2⫾5.2c 26.9⫾5.3c
n⫽827 ⬍125 g/de 125 to 203 to ⱖ281 g/de
⬍203 g/de ⬍281 g/de
62h 64h 53h 45h
a
Not all studies reported standard error of the mean or standard deviation.
b
NHS⫽Nurses’ Health Study.
c
Body mass index⫾standard deviation.
d
IWHS⫽Iowa Women’s Health Study.
e
Mean or range of refined grain intake, in g/d, servings/d (s/d), or servings/wk (s/wk).
f
HPFS⫽Health Professionals Follow-up Study.
g
FOS⫽Framingham Offspring Study.
h
Percentage of women study participants.
i
TGLS⫽Tehran Lipids and Glucose Study.
randomized to assess the long-term benefits and risks of (111). In addition, an ad libitum, low-fat, high-carbohy-
a low-fat dietary pattern on breast and colorectal cancers drate diet was more effective than a fixed-energy intake
and cardiovascular disease. The intervention did not en- for maintaining weight after a major weight loss (112).
courage weight loss or energy restriction. Average weight Two years after a 12- to 13-kg weight loss, the ad libitum,
decreased by 2.2 kg during the first year, which was low-fat, high-carbohydrate group maintained three times
significantly more than that of the control group. Al- as much weight loss as the fixed-energy group (8.0 kg vs
though the 2.2-kg difference in weight between interven- 2.5 kg), and more subjects maintained a weight loss of ⱖ5
tion and control groups diminished during the 7.5-year kg (65% vs 40%).
follow-up, the difference in weight remained significant Ad libitum, low-fat diets may help those with unre-
during the entire follow-up period. strained eating behavior maintain body weight (113).
Because weight gain during adulthood is common in Whereas individuals who consume hypocaloric diets gen-
the United States, an ad libitum, low-fat diet could assist erally experience increased hunger, subjects consuming
weight control by limiting weight gain. Behavioral Risk ad libitum, low-fat diets do not (26,96,100). This may be
Factor Surveillance System surveys (108) indicated adult attributable in part to lower energy density of low-fat,
weight gain averaged about 1 lb/year during the 1990s high-carbohydrate diets (114,115). A low-fat, high-carbo-
(ie, during which time much of the WHI was conducted). hydrate diet may also increase sensitivity to leptin and
Data from the WHI suggest that an ad libitum, low-fat avoid the increase in ghrelin caused by energy restriction
diet might attenuate this weight gain (95). (116). Weigle and colleagues (116) reported that 12 weeks
Low-fat, high-carbohydrate diets seem to be most com- of an ad libitum, low-fat, high-carbohydrate diet resulted
mon among people who successfully maintain weight loss in a spontaneous decrease in energy intake without an
(109). Lower fat intake was a predictor of both initial and increase in 24-hour area-under-curve for ghrelin. Al-
sustained weight loss in persons with obesity consuming though leptin 24-hour area-under-curve was decreased,
an ad libitum diet (110), and a low-fat, high-fiber diet there was an increase in the percentage change between
predicted long-term (3 years) weight reduction in initially nadir and peak 24-hour leptin levels, which was strongly
overweight persons with impaired glucose tolerance correlated to the absolute change in both weight and fat