Review Articles
Low-carbohydrate nutrition and metabolism1–3
Eric C Westman, Richard D Feinman, John C Mavropoulos, Mary C Vernon, Jeff S Volek, James A Wortman,
William S Yancy, and Stephen D Phinney
ABSTRACT
The persistence of an epidemic of obesity and type 2 diabetes suggests
that new nutritional strategies are needed if the epidemic is to
be overcome. A promising nutritional approach suggested by this
thematic review is carbohydrate restriction. Recent studies show
that, under conditions of carbohydrate restriction, fuel sources shift
from glucose and fatty acids to fatty acids and ketones, and that ad
libitum–fed carbohydrate-restricted diets lead to appetite reduction,
weight loss, and improvement in surrogate markers of cardiovascular
disease. Am J Clin Nutr 2007;86:276–84.
KEY WORDS Nutrition, metabolism, macronutrients, glucose,
insulin
INTRODUCTION
The persistence of an epidemic of obesity and type 2 diabetes
suggests thatnewnutritional strategies are needed if the epidemic
is to be overcome. A historical perspective and recent research
point to some form of carbohydrate restriction as a likely candidate
for a new nutritional approach, and we present a thematic
review regarding carbohydrate restriction.
The examination of diets before modernization can remind us
of the remarkable ability of humans to adapt to their environment
and can provide a context within which to view current diets. In
contrast to current Western diets, the traditional diets of many
preagricultural peoples were relatively low in carbohydrate (1,
2). In North America, for example, the traditional diet of many
First Nations peoples of Canada before European migration comprised
fish, meat, wild plants, and berries. The change in lifestyle
of several North American aboriginal populations occurred as
recently as the late 1800s, and the numerous ensuing health
problems were extensively documented (3–5). Whereas many
aspects of lifestyle were altered with modernization, these researchers
suspected that the health problems came from the
change in nutrition—specifically, the introduction of sugar and
flour.
In a similar manner, before the discovery of insulin, the removal
of high-glycemic carbohydrates such as sugar and flour
from the diets of diabetics was found to be a successful method
of controlling glycosuria. An analysis of the pattern of food
consumption during the more recent obesity and diabetes epidemic
found that the increase in calories was almost entirely due
to an increase in carbohydrate (6). Given this context, it is reasonable
to postulate that diets low in carbohydrate may be as
healthy as, or even healthier than, the higher-carbohydrate diets
introduced into modern society only recently.
This thematic review summarizes studies involving lowcarbohydrate
diets (LCDs) published over the 4 y since the last
comprehensive reviews of the topic (7, 8). Articles were identified
by us through attendance at scientific meetings, reading of
publications, reference searching, manuscript reviews, and
weekly Medline searches from January 2002 to December 2006
with the use of the terms “diet,” “carbohydrate,” and “fat.”/SEC
THEMATIC REVIEW
Definition of low-carbohydrate diet
Much of the controversy in the study of LCDs stems from a
lack of a clear definition. The rationale of carbohydrate restriction
is that, in response to lower glucose availability, changes in
insulin and glucagon concentrations will direct the body away
from fat storage and toward fat oxidation. There is a suggestion
of a threshold effect, which has led to the clinical recommendation
of very low concentrations of carbohydrate (20–50 g/d) in
the early stages of popular diets. This typically leads to the
presence of measurable ketones in the urine and has been referred
to as a very-low-carbohydrate ketogenic diet (VLCKD) or a
low-carbohydrate ketogenic diet (LCKD). Potent metabolic effects
are seen with such diets but, beyond the threshold response,
there appears to be a continuous response to carbohydrate reduction.
The nutritional intake of 200 g carbohydrate/d has been
called an LCD, but most experts would not consider that to
provide the metabolic changes associated with an LCKD. We
suggest that LCD refers to a carbohydrate intake in the range of
50–150 g/d, which is above the level of generation of urinary
ketones for most people.
1 From the Department of Medicine, Duke University Medical Center,
Durham NC (ECW, JCM, and WSY); the Department of Biochemistry,
SUNY Downstate Medical Center, Brooklyn, NY (RDF); private practice,
Lawrence, KS (MCV); the Human Performance Laboratory, Department of
Kinesiology, University of Connecticut, Storrs, CT (JSV); the First Nations
and Inuit Health Branch, Health Canada, Vancouver, Canada (JAW); the
Center for Health Services Research, Durham Veterans Affairs Medical
Center, Durham, NC (WSY); and the Department of Medicine (Professor
Emeritus), University of California, Davis, Davis, CA (SDP).
2 Supported by a grant from the Robert C Atkins Foundation (to ECW,
MCV, and JSV) and by a Veterans Affairs Career Development Award (to
WSY).
3 Reprints not available. Address correspondence to EC Westman, Duke
University Medical Center, 4020 North Roxboro Street, Durham,NC27704.
E-mail: ewestman@duke.edu.
Received February 2, 2007.
Accepted for publication February 5, 2007.
276 Am J Clin Nutr 2007;86:276–84. Printed in USA. © 2007 American Society for Nutrition
Downloaded from www.ajcn.org by on March 25, 2009
Other macronutrients
Because an instruction only to restrict carbohydrate intake
could theoretically create a diet containing any level of daily
energy intake from protein and fat, confusion exists among researchers
and the lay public about what constitutes an LCD. As
early as 1980, LaRosa found that subjects following an LCD do
not necessarily replace the carbohydrate with either protein or
fat, but that they, rather, reduce starch and sugar intake (9). Under
such conditions, even though the absolute amounts of fat and
protein do not increase, the percentage of fat and protein will
increase. Recent research reviewed below has determined that
the reduction in calorie intake is a result of appetite and hunger
reduction. In this way, LCDs are also low-calorie diets that include
an increase in the percentage of calories from fat and
protein but not necessarily an increase in absolute amounts of fat
and protein.
General physiologic principles in carbohydrate restriction
A review outlined the way in which a marked reduction in
carbohydrate intake leads to a general change in metabolism
from a “glucocentric” (glucose) to an “adipocentric” (ketone
bodies, fatty acids) metabolism (8). The main fuel sources become
fatty acids (from dietary fat and adipose stores) and ketones
(from dietary fat, protein, and adipose stores) (Table 1).
Glucose-dependent tissues (ie, red blood cells, retina, lens, and
renal medulla) receive glucose through gluconeogenesis and glycogenolysis.
(Even if no dietary carbohydrate is consumed, it is
estimated that 200 g glucose/d can be manufactured by the liver
and kidney from dietary protein and fat.) The metabolic state
experienced by a person who is following an LCKD is often
compared with the condition of starvation. The main similarities
in metabolism between LCDs and starvation are that there is no
(or little) intake of exogenous carbohydrate and that there is a
shift from the use of glucose as fuel toward the use of fatty acids
and ketones as fuel. Under conditions of starvation, endogenous
sources (eg, muscle protein, glycogen, and fat stores) are used as
energy supplies (10). However, under conditions of LCKD intake,
exogenous sources of protein and fat provide energy, along
with endogenous glycogen and fat stores if caloric expenditure
exceeds caloric intake. Whereas the loss of lean body mass
(LBM) is typical with weight loss, under certain circumstances
when sufficient dietary protein is provided, an LCKD may preserve
LBM even during hypoenergetic conditions of weight loss
(11, 12). Under low-carbohydrate conditions, unlike those of
starvation, glucose concentrations are sustained despite the lack
of carbohydrate intake (13). The maintenance of glucose concentrations
and the lack of breakdown of endogenous protein are
important differences between starvation and very low carbohydrate
intake.
Recent studies in healthy subjects
Only in the past several years have detailed studies regarding
LCDmetabolism been performed (Table 2). In a metabolic ward
study, 8 healthy volunteers were provided a 2-d eucaloric
(weight-maintaining) diet in which 60% of energy was from
carbohydrate and 30% of energy was from fat; this diet was
followed by a 7-d eucaloric diet in which 5% of energy was from
carbohydrate and 60% of energy was from fat. Both diets were
consumed while the subjects maintained their typical sedentary
lifestyle (14). With the 5% carbohydrate diet, serum glucose
initially declined but then returned to baseline after a few days.
Whereas fasting insulin did not differ between the 2 diets, the
24-h area under the curve (AUC) for insulin was 50% lower
TABLE 1
Fuel sources in a low-carbohydrate ketogenic diet
Fatty acids (70% of caloric requirements)
Dietary fat
Lipolysis
Adipose stores
Ketone bodies (20% of caloric requirements)
Dietary fat and protein
Lipolysis and ketogenesis
Adipose stores
Glucose (10% of caloric requirements)
Gluconeogenesis
Dietary protein and fat (glycerol)
Glycogenolysis
TABLE 2
Studies of low-carbohydrate ketogenic diet metabolism1
Study Duration Subjects
Macronutrients
CHO Pro Fat Energy RQ Insulin Glucagon Glucose Fatty acids -Hydroxybutyrate
d n % of daily intake kcal/d U/mL pg/mL mmol/L mmol/L mmol/L
Bisschop et al, 2000 (16) 11 6 2 15 85 —2 0.73 3.7 65 4.7 0.78
11 6 44 15 41 — 0.81 8.4 57 5.2 0.36
11 6 85 15 15 — 0.86 8.4 60 5.4 0.36
Allick et al, 2004 (22)3 14 5 0 11 89 3500 0.73 10 79 6.8 0.79
89 11 0 3500 0.79 12 73 8.2 0.70
Harber et al, 2005 (14) 7 8 5 35 60 — 7.0 5.0 0.45 0.5
2 60 10 30 — 7.0 5.0 0.3 0.05
Boden et al, 2005 (13)3 7 10 4 28 68 2164 6.7 89 6.3 0.65
39 17 44 3190 9.2 78 7.3 0.13
Noakes et al, 2006 (21) 844 24 12 31 54 7.1 5.3 0.1
21 49 21 28 7.4 5.3 0.1
22 66 20 13 9.9 5.2 0.1
1 CHO, carbohydrate; Pro, protein; RQ, respiratory quotient. Insulin, glucagon, glucose, fatty acids, and -hydroxybutyrate were fasting serum samples.
2 Weight-maintaining diet; the same caloric content in each group (all such).
3 Patients had type 2 diabetes.
4 8 wk of isocaloric weight loss and then 4 wk of weight maintenance.
LOW-CARBOHYDRATE NUTRITION 277
Downloaded from www.ajcn.org by on March 25, 2009
with the 5% carbohydrate diet than with the 60% carbohydrate
diet. After 1–2 d of the 5% carbohydrate diet and persisting
through the 7-d period, serum-hydroxybutyrate increased from
0.1 to 0.4 mmol/L and free fatty acids increased from 0.2 to
0.4 mmol/L. In addition, muscle glycogen (measured by muscle
biopsy) was reduced by 20% after 9 d.
Glucose kinetics were assessed by stable-isotope techniques
while resting metabolic rates were calculated from oxygen consumption
(V˙ O2) and carbon dioxide production (V˙ CO2) was measured
by using a metabolic cart. By day 2 of the 5% carbohydrate
diet, both the glucose rate of appearance and rate of disappearance
decreased by 20%, and they remained suppressed on day 7.
In addition, postabsorptive carbohydrate oxidation decreased
progressively over the 7-d duration, and this decline was greater
than the decline in glucose uptake. This means that the rate of
nonoxidative glucose disposal (ie, carbohydrate storage) increased
in the postabsorptive state with the5%carbohydrate diet.
These changes suggest that there is a shift from the use of glucose
to the use of ketones and free fatty acids as metabolic fuels, and
that glycogen formation increases from baseline.
Another set of studies was performed to evaluate the metabolic
effects of diets consisting of 0–2% carbohydrate, 11–15% protein,
and 83–88% fat in healthy volunteers (15–19). (These experimental
diets contained a higher percentage of fat than is
typically observed in an ad libitumLCD,and thus they were more
characteristic of an ancestral Inuit diet or the ketogenic diet for
epilepsy.) Nonetheless, these studies elucidate many metabolic
aspects of carbohydrate restriction. Serum glucose, insulin, and
C-peptide concentrations with the 2% carbohydrate diet were
lower than those with the 85% carbohydrate control diet. After
11 d of the 2% carbohydrate diet, gluconeogenesis was 15%
higher and glycogenolysis was 55% lower than that after 11 d of
the 85% carbohydrate diet (15). In a related study by the same
group, weight-maintaining diets containing either 89% carbohydrate,
11% protein, and 0% fat or 0% carbohydrate, 11% protein,
and89%fat were compared over a 15-d period (18). In that study,
gluconeogenic rates did not differ significantly between the diets.
When a hyperinsulinemic euglycemic clamp technique (200
pmol/L) was used, insulin-mediated suppression of glucose production
and stimulation of glucose disposal did not differ significantly
between the diets. The investigators wrote, “After 14 d on
this [0% carbohydrate, 11% protein, and 89% fat] diet, 3 h of
hyperinsulinemia were not sufficient to suppress fat oxidation
and increase glucose oxidation. Because fatty acid use and oxidation
[are] impaired in patients with type 2 diabetes, the 14-d
high-fat diet seemed to have reversed this defect by allowing
adaptation of fuel selection toward fatty acids as the main energy
substrates and maintaining glucose oxidation at a minimum.”
Other findings support the metabolic differences between
“starvation” and “carbohydrate deprivation.” After 7 d of a 2%
carbohydrate, 15% protein, and 83% fat weight-maintenance
(33 kcal/kg) diet, 24-h nitrogen excretion was higher, without a
change in postabsorptive hepatic or whole-body protein metabolism,
than it was with the 2 comparison diets composed of 0%
fat and 41% fat (15). A previous study found that this rise in
nitrogen excretion after carbohydrate withdrawal is short-lived,
however, as both nitrogen balance and LBM retention were observed
after a 1–2-wk adaptation to a 0% cholesterol, 15% protein,
and 85% fat diet (20).
The 2% carbohydrate, 15% protein, and 83% fat weightmaintenance
diet also resulted in lower absorptive and postabsorptive
plasma insulin concentrations than did the 0% fat and
41% fat diets (15). Postabsorptive rates of appearance of leucine
and of leucine oxidation—measures of proteolysis— did not differ
significantly among the 3 diets. In addition, dietary carbohydrate
did not affect the synthesis rates of fibrinogen and albumin.
However, this study was limited in that the experimental manipulation
did not provide sufficient potassium or sodium intake, nor
did it allow time for keto-adaptation to reflect the conditions of
chronic, very low carbohydrate consumption. Both mineral nutriture
and time for adaptation have been addressed in earlier
eucaloric, very-low-carbohydrate feeding studies (20).
Another outpatient feeding study randomly assigned 83 subjects
to 1 of 3 diets ranging in carbohydrate content from 12% to
66%for 8-wk weight-loss and 4-wk weight-maintenance periods
(21). During the weight-maintenance period, the 3 diets contained
an estimated66%carbohydrate,20%protein, and13%fat;
49% carbohydrate, 21% protein, and 28% fat; or 12% carbohydrate,
31% protein, and 54% fat. The 12% carbohydrate diet led
to the greatest reduction in fasting insulin concentrations,
whereas fasting glucose concentrations did not differ significantly
among the groups. That study confirmed that the postprandial
rise of glucose and insulin after typical meals does not
occur after a 12% carbohydrate meal when given to a person who
has been adapted to a 12% carbohydrate diet.
Recent studies in diabetic subjects
Another metabolic ward study examined the effects of an ad
libitum LCKD in obese persons with type 2 diabetes (13). Ten
subjects were monitored while eating their usual diet for 7 d and
then a VLCD for 14 d. Carbohydrate intake was reduced to21
g/d, but patients could eat as much protein and fat as they wanted
and as often as they wanted. The final diet consumed was
weighed and estimated to contain a daily average of 21 g carbohydrate,
151 g protein, and 164 g fat, representing a spontaneous
reduction in caloric intake of 947 kcal, which resulted in a mean
weight loss of 2 kg over 14 d (P 0.042). During the lowcarbohydrate-
diet period, mean fasting glucose decreased from
7.5 mmol/L (135 mg/dL) on day 8 to 6.3 mmol/L (113 mg/dL) on
day 22 (P 0.025), glycated hemoglobin decreased from 7.3%
to 6.8% (P0.006), and 24-h glucose and insulin concentrations
decreased significantly. This reduction in glucose concentrations
required a decrease in diabetes medication in 5 of the 10 patients.
During the euglycemic hyperinsulinemic clamp procedure,
the mean glucose infusion rate needed to maintain euglycemia
increased by 30%, from 12.9 mol kg body wt1 min1 during
the usual diet to 16.8 mol kg body wt1 min1 after the
LCD (P 0.03). After adjustment for the differences in clamp
insulin concentrations on the 2 diets, glucose infusion rates
increased by 75%. Mean insulin-stimulated rates of glucose
disappearance (ie, insulin sensitivity), after adjustment, increased
from 0.01 mol kg1 min1/pmol L1 to 0.03
mol kg1 min1/pmol L1. So, the increase in glucose disappearance
effectively explained all of the increase in the glucose
infusion rates, whereas endogenous glucose production did
not change significantly. Therefore, insulin sensitivity improved
largely because of an increase in peripheral glucose uptake.
In another study involving persons with type 2 diabetes, a
3500-kcal weight-maintenance diet of 89% carbohydrate, 11%
protein, and 0% fat was compared with a 3500-kcal diet of 0%
278 WESTMAN ET AL
Downloaded from www.ajcn.org by on March 25, 2009
carbohydrate, 11% protein, and 89% fat in 7 persons. This research
group used stable-isotope and euglycemic hyperinsulinemic
clamp techniques to partition glucose disposal into oxidative
and nonoxidative disposal (22). In findings similar to those of
Harber et al (5), oxidative glucose disposal decreased 90%,
whereas nonoxidative glucose disposal increased. The main
finding was that, “in subjects with mild type 2 diabetes, maximal
eucaloric variation in carbohydrate to fat ratio modulates plasma
glucose concentration exclusively by alterations in hepatic glucose
production.” Allick et al concluded that it was remarkable
that, in the context of diabetes risk, 2 aspects of glucose homeostasis
actually improved after consumption of the high-fat,
low-carbohydrate diet: basal endogenous glucose production decreased,
and insulin-stimulated nonoxidative glucose disposal
increased.
The role of ketogenesis
Ketone concentrations after LCD intake have now been measured
in several studies. In a study of persons with type 2 diabetes,
urinary ketone body excretion increased from a mean of
0.10 mmol/d at the end of the usual diet to a peak of 2.75 mmol/d
after 1 wk of the LCKD (P0.001); it then decreased gradually
for a week but remained above baseline (13). The mean plasma
total-body ketone concentrations were 130 mol/L (91.5%
-hydroxybutyrate) at the end of the usual diet and increased
5-fold to 653 mol/L (97.4% -hydroxybutyrate) at the end
of the LCKD (P 0.001). Two longer-term studies, in
persons without diabetes, that measured fasting blood
-hydroxybutyrate concentrations over 10 wk found that,
whereas the concentrations increased over the first 2–4 wk, they
then decreased and, after 10–12 wk, remained only slightly
higher than those of dieters following other diets (21, 23).
Effects on appetite and satiety factors
Several studies confirm that there is a spontaneous reduction
in caloric intake when carbohydrate intake only is restricted to
5–10% of caloric intake (24). In the most controlled study to date,
an LCD led to hunger levels similar to those of a low-fat diet,
even though the daily caloric intake with the LCD was 1000 kcal
lower (13). Another study used the Eating Inventory, a validated
questionnaire assessing hunger and cognitive restraint, and
found that hunger was reduced by50%when measured after 1wk
of an LCD (25). Another study examining a 20-g carbohydrate
diet found that fasting serum leptin was reduced by 50% and
fasting serum neuropeptide Y was reduced by 15% (26). It may
also be that the mere lowering of serum insulin concentrations, as
is seen with LCDs, may lead to a reduction in appetite. In support
of this idea, several studies have found that insulin increases food
intake, that foods with high insulin responses are less satiating,
and that suppression of insulin with octreotide leads to weight
loss (27–29).
In summary, new metabolic studies of very-low-carbohydrate
conditions have found that serum glucose homeostasis is maintained
and serum ketone concentrations are increased. Muscle
glycogen is reduced but still present. With the exception of one
study (20), these metabolic studies are limited by the short study
duration (typically, 7–14 d), which probably does not allow sufficient
time for full adaptation to low-carbohydrate conditions.
Unanswered questions on low-carbohydrate metabolism
The study of LCD metabolism has been used to illustrate
metabolic pathways in medical school curricula and has also
highlighted some of the gaps in our current understanding of
biochemistry and metabolism (30). Whereas a spontaneous reduction
in caloric intake is a major effect of LCDs, there are many
reports indicating a so-called “metabolic advantage”—that is, a
greater amount of weight lost per calorie consumed. This is a
controversial idea, and there are perceptions that variable weight
loss with isocaloric diets would somehow violate the laws of
thermodynamics and that there must be some experimental error.
It is widely held that only caloric intake is important (as expressed
by the statement, “A calorie is a calorie.”). Variable
energy efficiency, however, is known in many contexts: hormonal
imbalance (31, 32), studies of weight regain (33, 34), and,
most strikingly, in knockout experiments in animals (35–37).
The fat-specific insulin receptor knock-out mouse weighs only
60% as much as the wild-type mouse, even though the 2 types
of mice eat the same amount of food (35). Most of the time, of
course, a calorie is a calorie, and we do not maintain that, in
carbohydrate restriction, metabolic advantage always occurs, but
only 1) that it can occur (11), 2) that it is not excluded by a correct
thermodynamic analysis, and 3) that, because of the importance
of obesity, it is sensible to try to identify the conditions under
which it can occur and to maximize the effect. The thermodynamic
analysis leads to the conclusion that variable efficiency is
the expected outcome from physical principles, and therefore,
when a calorie is a calorie, it is not explained by thermodynamics
but rather by the unique characteristics of living systems. In other
words, it is energy balance that needs to be explained. The mechanisms
that explain metabolic advantage emphasize the inefficiencies
introduced by substrate cycling and the requirements for
increased gluconeogenesis (38, 39). In addition, that thermogenesis
varies for different macronutrients is widely accepted, but it
is somehow expected to be ignored in weight-loss experiments,
even though the levels are in the range of 20% for protein compared
with 5% for carbohydrate. Moreover, discussion generally
centers on equilibrium thermodynamics, but living systems are
maintained far from equilibrium, and nonequilibrium thermodynamics,
which emphasizes kinetic fluxes as well as thermodynamic
forces, is more relevant. Most simply, the argument that “a
calorie is a calorie” rests on the fact that free energy and other
thermodynamic variables are state variables—that is, that they
are independent of mechanism or path. In fact, the change in a
weight-loss experiment is extremely far from the total change
embodied in free energy values, and the change that is measured
(technically, the partial derivative of the energy with respect to
reaction) is not path-independent and is notably influenced by the
activity of enzymes (40, 41).
Studies suggested that xylulose-5-phospate (Xu-5-P) is a signal
for the coordinated control of glucose metabolism and lipogenesis
(42). Xu-5-P is generated from glucose metabolism in the
hexose monophosphate pathway, which activates phosphofructokinase
and promotes the transcription carbohydrate-responsive
element–binding protein (ChREBP), thereby increasing the enzymes
of lipogenesis, the hexose monophosphate shunt, and glycolysis,
all of which are required for lipogenesis. The control of both
glycolysis and lipogenesis by one transcription factor shows the
close relation between these pathways.
LOW-CARBOHYDRATE NUTRITION 279
Downloaded from www.ajcn.org by on March 25, 2009
It is also likely that the regulation of hepatic glucose output is
substantially altered after adaptation to anLCKD(keto-adaptation).
For example, one study compared a very-low-energy (624 kcal),
low-carbohydrate (20% of daily energy intake) diet to a baseline
isoenergetic (30 kcal/kg), high-carbohydrate (55%) diet in obese
subjects with type 2 diabetes (43). After 3 wk of adaptation, the
very-low-energy, LCD diet resulted in significantly less hepatic
glucose output, and, across all subjects and diets, basal hepatic glucose
output was negatively correlated with plasma ketones (r
0.71, P 0.05).
Insulin resistance is reduced with an LCKD, possibly by a
reduction in the availability of dietary glucose, which causes
hyperinsulinemia (44, 45). A consideration of the physiology of
very-low-carbohydrate dieting leads to a different perspective on
insulin resistance. That is, rather than treating insulin resistance
by increasing glucose disposal through an increase in nonstorage
cellular influx (eg, by increasing either the insulin dose or its
effect), it could be treated by reducing glucose availability to
insulin-resistant tissue (eg, by reducing carbohydrate intake or
absorption and basal hepatic glucose output), which would reduce
the nonstorage cellular influx. Reductions in dietary carbohydrate
should be used as a strategy to treat insulin resistance.
Low-carbohydrate diets and exercise
Over the past several years, 2 reviews focused on LCKD and
exercise have been published. One of these reviews concluded
that submaximal endurance performance can be sustained despite
the virtual exclusion of carbohydrate from the human diet
(46). The other review addressed the intramuscular enzyme adaptation
that occurs with these diets (47).
Several important issues arise in the consideration of LCKD
studies in general and of exercise studies in particular: 1) the time
allowed for keto-adaptation, 2) the use of electrolyte supplementation,
and 3) the amount of protein intake. To try to examine the
first issue, we can consider the multiple studies comparing lowcarbohydrate
with high-carbohydrate diets to test the hypothesis
that “carbohydrate loading” can enhance physical performance.
None of the studies that support this hypothesis maintained the
LCD for 2 wk (48), and most maintained the LCDs for 7 d
(49). No studies have carefully examined the process or duration
of keto-adaptation, but clinical observation suggests that it probably
takes from 2 to 4 wk for keto-adaptation to occur.
The second issue has to do with the maintenance of adequate
mineral supplementation as long as the ketogenic state is maintained.
One group of investigators provided supplements containing
3–5 g sodium/d and 2–3 g potassium/d and found that
circulatory competence during submaximal exercise was sustained.
These supplements also allowed the subjects to achieve
nitrogen balance, which had not been achieved in studies that did
not use supplements (20).
The third issue affecting physical performance is adequate
protein intake. It is generally accepted that the preservation of
LBM and of physical performance during any degree of energy
restriction occurs when protein is in the range of 1.2 to 1.7 g kg
reference body wt1 d1. The use of the mid-range value of 1.5
g kg1 d1 for adults with reference weights ranging from 60
to 80 kg, this translates into total daily protein intakes of 90 to 120
g/d. When adequate protein intake is expressed in the context of
total daily energy expenditures of 2000 to 3000 kcal/d,15% of
daily energy expenditure should be provided as protein.
Further research on exercising under conditions of LCDs is
needed. These studies may be optimized by careful attention to
the time needed for keto-adaptation, to mineral supplementation,
and to the daily protein dose. Therapeutic use of ketogenic diets
should not limit most forms of physical activity, with the caveat
that anaerobic performance (ie, weight lifting or sprinting) may
be limited by lower-muscle glycogen concentrations.
Outpatient clinical trials for obesity
The efficacy of an LCKD for weight loss has now been established
in 6 outpatient randomized controlled trials (23, 50–55).
All of these trials used the most widely recommended diet at that
time, a 30%-fat, reduced-calorie diet, as the comparison diet
(Table 3). There were differences in the intensity of the interventions
in these outpatient studies. For example, the amount of
behavioral support ranged from simply providing a popular diet
book along with minimal education to providing biweekly group
sessions with extensive handouts and close monitoring (51, 55).
Across these studies, there appeared to be better adherence and
greater weight loss as the intensity of the intervention increased.
TABLE 3
Randomized outpatient trials of a low-carbohydrate ketogenic diet for obesity: estimated dietary intake and effect on weight and fasting serum lipids1
Reference Duration Subjects
Low-fat diet Low-carbohydrate diet
Macronutrients
Energy Weight LDL TG HDL
Macronutrients
CHO Fat Pro CHO Fat Pro Energy Weight LDL TG HDL
n % of energy kcal/d kg % % % % of energy kcal/d kg % % %
Brehm et al, 2003 (54) 6 mo 42 53 29 18 1245 3.9 5 2 8 23 46 30 1302 8.5 0 23 13
Foster et al, 2003 (51) 6 mo 63 —2 — — — 5.3 3 13 4 — — — — 9.7 4 21 20
12 mo 4.5 6 1 3 7.3 1 28 18
Meckling et al, 2004 (23) 10 wk 40 62 20 18 1447 6.8 32 25 15 15 56 26 1528 7.0 0 29 12
Samaha et al, 2003 (52) 6 mo 132 51 33 16 1576 1.9 3 4 2 37 41 22 1630 5.8 4 20 NC
12 mo 3.1 3 2 12 5.1 6 29 2
Sondike et al, 2003 (50) 3 mo 30 56 12 32 1100 4.1 17 6 2 8 60 32 1830 9.9 4 48 4
Yancy et al, 2004 (55) 6 mo 119 51 18 31 1588 6.5 3 15 1 10 60 30 1472 12.0 2 42 13
1 CHO, estimated daily carbohydrate intake; Pro, estimated daily protein intake; Energy, estimated daily calorie intake; weight, total body weight; TG,
triacylglycerols; NC, no change.
2 Not measured (all such).
280 WESTMAN ET AL
Downloaded from www.ajcn.org by on March 25, 2009
One study (23) used a tapering of carbohydrate, whereas all other
studies used a sudden reduction in carbohydrate.
Several of these studies collected detailed outpatient nutritional
intake information (23, 54). Whereas instruction in an
LCD does not mention calories, the restriction of dietary carbohydrate
leads to a reduction in caloric intake from baseline. The
ad libitum intake can vary from person to person, but, in many
cases, the protein and fat intakes, in absolute terms, are not much
higher than those of a typical American diet, because the total
caloric intake is lower. As such, the LCD is not necessarily a
high-protein diet or a high-fat diet. In addition, whereas the diet
typically contains high amounts of saturated fat, it also contains
high amounts of monounsaturated and polyunsaturated fats.
Cardiovascular disease risk factors
As shown in Table 3, the outpatient obesity studies found a
consistent reduction in fasting serum triacylglycerols and a fairly
consistent increase in HDL cholesterol, but little change in total
or LDL cholesterol, in the LCD groups. Two of these studies
published examinations of the fasting serum lipid concentrations
by using a lipid subfraction technique, and both found an average
change in LDL-cholesterol type from small LDL to large LDL
cholesterol, which corresponded with a decrease in LDL particle
concentration for subjects following theLCKDfor 6mo(56, 57).
Two studies assessed the effectiveness and adherence rates of
several popular diet plans with minimal behavioral counseling
(58, 59). Several outpatient diet studies have shown reductions in
CVD risk factors after an 8–12-wk LCKD, during weight loss,
and during weight maintenance (21, 60–62).
Clinical practice summaries
Several retrospective and prospective clinical series on the
potential effectiveness of anLCDin the clinical setting have been
published. A group in Kuwait published 2 case series involving
185 patients following a diet with 20–40 g carbohydrate/d; those
investigators found reductions in weight, cholesterol, and LDL
cholesterol and an increase in HDL cholesterol (63, 64).Agroup
in Bahrain conducted a pilot study of 13 obese patients and found
positive effects (65). Two clinical studies from the United States
used LCDs in conjunction with statin therapy (66, 67). Five other
clinical series involving 229 patients suggested that the LCD has
a potent effect on obesity and type 2 diabetes (68 –74). Another
clinical series involved 37 adolescents who were instructed to
follow either an LCD (30 g/d; n 27) or a low-calorie diet
(n10) for 2 mo. Compliance and weight loss were better in the
LCD group (75).
Whereas these studies are limited because of their clinical
measures and possible selection bias, they show that many clinicians
already find that it is feasible to implement the LCD in
clinical practice. All of the case series studies showed improvements
in clinical measures, which indicates that these outcomes
can be obtained in the outpatient practice setting in at least a
subset of adherent patients.
SEE ATTACHMENT FOR FULL ARTICLE
7587-Lowcarbohydratenutritionandmetabolism13.pdf (133 KB)