Human Nutrition and Metabolism

A Ketogenic Diet Favorably Affects Serum Biomarkers for Cardiovascular

Disease in Normal-Weight Men1
Matthew J. Sharman, William J. Kraemer, Dawn M. Love, Neva G. Avery, Ana L. Gómez,
Timothy P. Scheett and Jeff S. Volek2
Human Performance Laboratory, Department of Kinesiology, University of Connecticut, Storrs, CT 06269-1110

Downloaded from https://academic.oup.com/jn/article/132/7/1879/4687418 by guest on 30 January 2021

ABSTRACT Very low-carbohydrate (ketogenic) diets are popular yet little is known regarding the effects on serum
biomarkers for cardiovascular disease (CVD). This study examined the effects of a 6-wk ketogenic diet on fasting
and postprandial serum biomarkers in 20 normal-weight, normolipidemic men. Twelve men switched from their
habitual diet (17% protein, 47% carbohydrate and 32% fat) to a ketogenic diet (30% protein, 8% carbohydrate and
61% fat) and eight control subjects consumed their habitual diet for 6 wk. Fasting blood lipids, insulin, LDL particle
size, oxidized LDL and postprandial triacylglycerol (TAG) and insulin responses to a fat-rich meal were determined
before and after treatment. There were significant decreases in fasting serum TAG (⫺33%), postprandial lipemia
after a fat-rich meal (⫺29%), and fasting serum insulin concentrations (⫺34%) after men consumed the ketogenic
diet. Fasting serum total and LDL cholesterol and oxidized LDL were unaffected and HDL cholesterol tended to
increase with the ketogenic diet (⫹11.5%; P ⫽ 0.066). In subjects with a predominance of small LDL particles
pattern B, there were significant increases in mean and peak LDL particle diameter and the percentage of LDL-1
after the ketogenic diet. There were no significant changes in blood lipids in the control group. To our knowledge
this is the first study to document the effects of a ketogenic diet on fasting and postprandial CVD biomarkers
independent of weight loss. The results suggest that a short-term ketogenic diet does not have a deleterious effect
on CVD risk profile and may improve the lipid disorders characteristic of atherogenic dyslipidemia. J. Nutr. 132:
1879 –1885, 2002.

KEY WORDS: ● triglycerides ● postprandial lipemia ● lipoprotein subclasses

Cardiovascular disease (CVD)3 is the leading cause of mor- The atherogenicity of TAG-rich lipoproteins in the postpran-
tality in most industrialized countries including the United dial state may play a greater role than fasting TAG, prompting
States (1). Diet is a major weapon used in the fight against some authors to suggest that elevated postprandial lipemia is
CVD because of its influence on the myriad of CVD risk a better predictor of CVD than fasting TAG (11,12). Abnor-
factors. Current dietary recommendations call for a low-fat mal postprandial lipemia precipitates production of highly
(⬍30% of energy), low saturated fat (⬍7% total energy), low atherogenic small LDL particles and a reduction in HDL
cholesterol (⬍300 mg/d) diet (2). However, high-carbohy- cholesterol (12), all of which contribute to the causal role for
drate diets are controversial (3,4), because they raise plasma elevated postprandial lipemia in the pathogenesis and progres-
triacylglycerols (TAG) (5) and may adversely affect LDL sion of CVD.
composition (6,7). There has been an alarming increase in the Individuals with a predominance of large buoyant LDL
popularity of diets with the common theme of reducing car- cholesterol have been classified as pattern A, whereas those
bohydrate, prompting concern regarding their safety (8). De- with a predominance of small dense LDL particles are pattern
spite the popularity of very low-carbohydrate diets, very few B (10). Individuals exhibiting higher levels of small dense LDL
scientific studies have evaluated how these diets affect CVD have a greater than 3-fold risk of CVD (13,14). This is most
risk factors (9) and no studies have examined the effect on likely a result of the longer half-life and increased susceptibil-
recently identified CVD biomarkers (i.e., LDL particle size, ity to oxidative modification (15). The fact that LDL is ex-
postprandial lipemia, oxidized LDL, etc). tremely susceptible to oxidative damage has been known for
A recent meta-analysis of prospective studies indicated that some time (16), with it now appearing that the oxidation of
elevated fasting TAG is an independent risk for CVD (10). LDL plays an important role in atherogenesis (17).
The therapeutic value of diet interventions aimed at im-
1
proving CVD risk should take into account factors other than
This study was supported by a grant from the Atkins Foundation, New York,
NY. Published in abstract form [Sharman, M. J., Volek, J. S., Gómez, A. L., Avery just fasting total cholesterol, LDL cholesterol, HDL choles-
Love, N. G., French, D. N. & Kraemer, W. J. ( 2001 ) Fasting and postprandial terol, and TAG. In this study, we evaluated the effect of a
lipoprotein responses to a ketogenic diet. Am. College Sports Med. 33: S213.] ketogenic diet on both fasting and postprandial lipoprotein
2
To whom correspondence should be addressed.
E-mail: jvolek@uconnvm.uconn.edu. metabolism including measures of postprandial lipemia, LDL
3
Abbreviations used: CVD, cardiovascular disease; TAG, triacylglycerol. size and LDL oxidation. As a first step, we studied a normal-

0022-3166/02 $3.00 © 2002 American Society for Nutritional Sciences.

Manuscript received 22 January 2002. Initial review completed 2 March 2002. Revision accepted 9 April 2002.

1879
1880 SHARMAN ET AL.

weight normolipidemic population to minimize the confound- genic group kept records each day of the experiment (7 d during
ing effects of weight loss or metabolic abnormalities on the baseline and 42 d during the ketogenic diet) and the control group
dependent variables. Based on our previous work showing a kept 7-d records during wk 1 and 6. All recorded days were analyzed
reduction in fasting TAG and postprandial lipemia after a for nutrient content (Nutritionist V, Version 2.3; N-Squared Com-
ketogenic diet rich in monounsaturated fat and supplemented puting, First Databank Division, The Hearst Corporation, San Bruno,
CA). All intervention foods and supplements were entered into the
with (n-3) PUFA (9), we hypothesized that the ketogenic diet software database and included in the analysis of nutrient intake.
used in this study would result in a similar TAG response, Additionally, dietary compliance was monitored using Ketostix re-
which would in turn result in increased HDL cholesterol and agent strips (Bayer Corporation, Elkhart, IN), which determine qual-
an increase in the average size of LDL particles. itatively the presence of acetoacetic acid in urine. Each subject
maintained a record of color changes on the reagent strips performed
MATERIALS AND METHODS daily at ⬃0800.
Fasting blood collection. Fasting blood samples were obtained on
Subjects. Twenty healthy white men free of metabolic and two separate days at wk 0, 3 and 6 after a 12-h overnight fast and
endocrine disorders volunteered to participate in this investigation. abstinence from alcohol and strenuous exercise for 24 h. Subjects
To enhance compliance to the rigorous ketogenic dietary treatment, reported to the laboratory between 0700 and 0900 h, rested quietly

Downloaded from https://academic.oup.com/jn/article/132/7/1879/4687418 by guest on 30 January 2021

we initially asked subjects if they would be willing to restrict carbo- for 10 min in the supine position, and blood was obtained from an
hydrate to ⬍10% of total energy for 6 wk. Twelve subjects volun- antecubital vein with a 20-gauge needle and vacutainers.
teered to switch from their habitual diet to a ketogenic diet for 6 wk Oral fat tolerance test. Postprandial TAG and insulin responses
(mean ⫾ SD; age: 36.7 ⫾ 11.6 y; body mass: 79.2 ⫾ 8.3 kg; fat: 20.5 to a fat challenge were assessed at wk 0 and 6 using standard
⫾ 6.2%) and the remaining eight subjects served as controls (mean procedures in our laboratory (9). Subjects arrived at the laboratory
⫾ SD; age: 35.0 ⫾ 13.0 y; body mass: 85.4 ⫾ 12.8 kg; fat: 22.2 after a 12-h overnight fast and abstinence from alcohol and strenuous
⫾ 9.0%). Subjects in the ketogenic diet group had fasting serum total exercise for 24 h. A flexible catheter was inserted into a forearm vein,
cholesterol concentrations ranging from 2.87 to 5.82 mmol/L and which was kept patent with a constant saline drip (60 mL/h). Sub-
TAG concentrations from 0.62 to 2.24 mmol/L. There were no jects rested in a seated position for 10 min and two baseline blood
significant differences in any physical characteristics or serum lipids samples were obtained separated by 10 min with a 10-mL syringe. The
concentrations between groups at the start of the study. The subjects first 3 mL of blood withdrawn was discarded to avoid dilution of the
had not lost or gained weight in the previous year, were not adhering sample and ⬃10 mL was subsequently withdrawn and processed. The
to special diets or regular consumers of nutritional supplements, and test meal then was consumed under supervision to ensure that the
habitually consumed between 19% and 43% of energy as fat (assessed entire meal was ingested within a 15-min period. The test meal was
via a 7-d food diary). All subjects were nonsmokers and were not designed to be rich in fat and low in carbohydrate similar to the
prescribed any medication known to affect serum lipoproteins. All
ketogenic diet and contained 5.44 MJ (11% carbohydrate, 2% pro-
subjects were informed of the purpose and possible risks of this
tein, 87% fat, 52 g saturated fat, 59 g monounsaturated fat, 12 g
investigation before signing an informed consent document approved
polyunsaturated fat, and 266 mg cholesterol). Postprandial blood
by an Institutional Review Board at Ball State University.
samples were obtained immediately after the meal and hourly for a
Experimental design. The study design involved normal-weight
normolipidemic men that either switched from their habitual diet total of 8 h to assess the magnitude and time course of postprandial
(32% fat) to a ketogenic diet (61% fat) or maintained their habitual lipemia. Subjects rested quietly in a seated position and consumed
diet (control group) for 6 wk. Body weight was assessed and two 12-h exactly one liter of water only during the 8-h postprandial period. All
fasting blood samples were collected at wk 0, 3 and 6. Postprandial subjects completed the entire meal and no adverse side effects (stom-
TAG and insulin responses to a fat-rich test meal were measured at ach distress, nausea, etc.) were reported.
wk 0 and 6 only in the ketogenic group. Blood analyses. Serum collected for the determination of LDL
Dietary intervention. The aim of the intervention diet was to particle size, oxidized LDL, insulin and ␤-hydroxybutyrate were im-
reduce carbohydrate intake to ⬍10% of energy. The diet was de- mediately stored at ⫺80°C. The remaining serum was analyzed for
signed so that fat comprised ⬃60% of energy with no restrictions on glucose, total cholesterol, HDL cholesterol, and TAG using auto-
the type of fat from saturated and unsaturated sources or cholesterol mated techniques (Roche Modular; Roche Diagnostics, Indianapolis,
levels. The actual diets consumed were mainly comprised of beef IN) with calculated precision values ⬍ 3%. Concentrations of LDL
(e.g., hamburger and steak), poultry (e.g., chicken and turkey), fish, and VLDL cholesterol were calculated from total cholesterol, HDL
oils, various nuts/seeds and peanut butter, moderate amounts of cholesterol, and TAG (18). Fasting serum ␤-hydroxybutyrate con-
vegetables, salads with low-carbohydrate dressing, moderate amounts centrations were enzymatically determined in duplicate using a com-
of cheese, eggs, protein powder, and water or low-carbohydrate diet mercially available kit (#310A; Sigma Diagnostics, St. Louis, MO)
drinks. Foods avoided or consumed infrequently included fruits and and spectrophotometric analysis (Spectronic 601; Milton Roy, Roch-
fruit juices, most dairy products with the exception of hard cheeses ester, NY). The intra-assay CV was 0.9%. Fasting serum and post-
and heavy cream, breads, cereals, beans, rice, desserts/sweets, or any prandial plasma insulin concentrations were determined in duplicate
other foods containing substantial amounts of carbohydrate. A por- using an ELISA kit with a sensitivity of 1.81 pmol/L (#10-1600;
tion of the food consumed during the intervention diet (⬃30 – 40% Diagnostic Systems Laboratory, Webster, TX) (19). Intra- and inter-
of total energy) was provided to subjects during weekly meetings to assay CV were 5.5% and 3.2%, respectively. Fasting oxidized LDL
review compliance with the registered dietitian. These foods included were also determined in duplicate using an enzyme-linked immu-
pumpkin seeds, roasted cheese, low-carbohydrate bars, shakes, and nosorbent assay (American Laboratory Products Company,
bake mix (Atkins Nutritionals, Hauppauge, NY) and protein powders Windham, NH) with a solid two-site enzyme immunoassay that is
(Super Whey Fuel and Fuel Plex Lite; Twin Laboratories, Hauppauge, based on the direct sandwich technique in which two monoclonal
NY). Subjects were also provided with a daily multivitamin/mineral antibodies are directed against separate antigenic determinants on
complex (Daily One Caps With Iron; Twin Laboratories). the oxidized apolipoprotein B molecule. Intra-assay CV was 7.9%.
Each subject received individual dietary instruction weekly on Absorbances were read on a multilabel counter (Wallac1420 Victor2;
how to consume meals within the specified nutrient goals and to Wallac Oy, Turku, Finland).
assess compliance. Subjects were provided with a packet outlining LDL subclasses. Separation of LDL subclasses was performed
specific lists of appropriate foods, recipes and sample meal plans that using nongradient polyacrylamide gel electrophoresis (Lipoprint LDL
were compatible with their individual preferences and the nutrient System; Quantimetrix, Redondo Beach, CA) previously described in
profile goals of the ketogenic diet. Food measuring utensils and scales detail (20,21). High-resolution 3% polyacrylamide gel tubes were
were provided to all subjects before the study to assist in the estima- used for electrophoresis. Seven bands of LDL were quantitatively
tion of portion sizes of foods and beverages. Subjects were encouraged evaluated using computer software (NIH imaging software, utilizing
to maintain adequate energy intake to maintain body weight. If body the Lipoprint LDL macro; Quantimetrix), which divides the scanned
weight changed ⬎1 kg, dietary counseling was provided. The keto- gel image at designated Rf values identified by their relative mobility.
 KETOGENIC DIETS AND BLOOD LIPIDS 1881

This is based on differences in particle size (smaller particles migrate Fasting serum metabolic and insulin responses. Com-
further) and calculates the area under the curve for each fraction. We pared with baseline in the men who followed the ketogenic
report the relative percentage of LDL cholesterol in each band and diet, serum ␤-hydroxybutyrate concentrations were signifi-
the mean and peak particle diameter. The peak particle diameter for cantly increased at wk 3 (⫹427%) and remained significantly
phenotype A is generally ⬎25.5 nm, in contrast the major peak for
phenotype B is usually ⬍25.5 nm (22). The determination of a elevated at wk 6 (⫹250%; Table 2). All subjects following the
sample being characterized as either phenotype A or B is based on ketogenic diet had ␤-hydroxybutyrate concentrations ⬎ 0.20
LDL migration rates and is described in detail in Hoefner et al. (20). mmol/L, indicating compliance with the ketogenic diet. Se-
Insulin sensitivity. Because high-fat diets have been associated rum insulin concentrations were significantly reduced at 3 and
with insulin resistance, we estimated insulin sensitivity using the 6 wk (⫺34.2%) in the ketogenic group but unchanged in the
homeostasis model analysis using fasting glucose and insulin concen- control group (Table 2). From 0 to 6 wk, the estimation of
trations (23). Assuming that normal-weight subjects aged ⬍35 y have insulin resistance was not affected in the ketogenic group
a insulin resistance of 1, the values for a subject can be assessed from (0.75 ⫾ 0.62 3 0.52 ⫾ 0.35) or the control group (0.52
the insulin and glucose concentrations by the formula: insulin resis-
tance (near approximation) ⫽ insulin/(22.5e⫺ln glucose). ⫾ 0.41 3 0.51 ⫾ 0.41).
Statistical analyses. Means and SD were calculated for all vari- Fasting serum lipid responses. There were significant
increases in total and LDL cholesterol that returned to values

Downloaded from https://academic.oup.com/jn/article/132/7/1879/4687418 by guest on 30 January 2021

ables using conventional methods. Two fasting samples were obtained
for each blood variable and the mean of these two values used for that were not significantly different than wk 0 values at the
statistical analysis. A two-way repeated-measures ANOVA was used end of the ketogenic diet period (Table 2). Individual re-
to evaluate changes over time in the ketogenic and control groups. sponses in total cholesterol were variable, decreasing in five
When a significant F value was achieved, Fisher’s least significant subjects (range: ⫺2 to ⫺17%) and increasing in seven subjects
difference test was used to locate the pair-wise differences between (range: 1 to 60%). LDL cholesterol decreased in four subjects
means. The total area (serum concentration ⫻ time) under the line
connecting the postprandial TAG and insulin responses was calcu-
(range: ⫺4 to ⫺11%) and increased in seven subjects (range:
lated using the trapezoidal method. The change in body weight was 2 to 70%; Fig. 1). Compared with wk 0 in the ketogenic group,
used as a covariate in all analyses. Relationships between variables there was a significant increase in HDL cholesterol at wk 3 and
were examined using Pearson’s product-moment correlation coeffi- a trend (P ⫽ 0.066) at wk 6. Serum HDL cholesterol decreased
cient. A criterion ␣-level of P ⱕ 0.05 was used for all statistical in three subjects (range: ⫺6 to ⫺20%) and increased in nine
comparisons. subjects (range: 1 to 71%; Fig. 1). There was a significant
reduction in VLDL cholesterol at wk 3 and 6 of the ketogenic
diet. Serum TAG were significantly reduced at wk 3 (⫺30.9%)
RESULTS and wk 6 (⫺33.0%). All but one subject had a decrease in
Body mass and dietary intake. All dietary nutrients were fasting TAG (range: ⫺1 to ⫺50%). There were no significant
significantly different when men consumed the ketogenic diet changes in serum lipids in the control group.
compared with their habitual diet with the exception of di- LDL subclass profiles. There was a significant increase in
etary energy and alcohol consumption (Table 1). Dietary peak particle diameter and the LDL-1 subclass distribution
protein, fat and cholesterol were significantly greater and after the ketogenic diet was consumed (Table 3). We also
dietary carbohydrate was significantly lower (8% of total en- analyzed subjects classified as pattern A or pattern B at the
ergy) during the ketogenic diet period. There were no signif- start of the study. Before the diet intervention, seven subjects
icant changes in dietary nutrient intake in the control group were classified as pattern A and five subjects were pattern B.
from 0 to 6 wk. There was a small but significant decrease in Three of the five pattern B subjects switched to pattern A after
body mass in the ketogenic group (⫺2.2 ⫾ 1.7 kg) but not the the ketogenic diet; all seven pattern A subjects remained
control group (⫹0.4 ⫾ 4.1 kg). pattern A after the ketogenic diet. At the start of the study,

TABLE 1
Daily intake of dietary energy and nutrients in men who switched from their habitual diet to a ketogenic diet for 6 wk
and in a control group who continued to consume their habitual diet for 6 wk1

Ketogenic diet Ketogenic diet Habitual diet Habitual diet

(wk 0) (wk 6) (wk 0) (wk 6)

Energy, kJ 10627 ⫾ 2469 9770 ⫾ 1569 8489 ⫾ 1962 7594 ⫾ 816

Protein, g 113 ⫾ 40b 176 ⫾ 45a 82 ⫾ 16b 70 ⫾ 10b
Protein, %2 17 ⫾ 4b 30 ⫾ 5a 16 ⫾ 2b 15 ⫾ 1b
Carbohydrate, g 306 ⫾ 100a 46 ⫾ 10b 287 ⫾ 79a 271 ⫾ 47a
Carbohydrate, % 48 ⫾ 10a 8⫾ 3b 55 ⫾ 5a 59 ⫾ 7a
Total fat, g 91 ⫾ 31b 157 ⫾ 27a 65 ⫾ 18b 50 ⫾ 14b
Total fat, % 32 ⫾ 8b 61 ⫾ 4a 29 ⫾ 5b 25 ⫾ 8b
Saturated fat, g 31 ⫾ 12b 56 ⫾ 11a 21 ⫾ 6b 16 ⫾ 6b
Saturated fat, % 14 ⫾ 4b 25 ⫾ 2a 14 ⫾ 4b 12 ⫾ 4b
Monounsaturated fat, g 27 ⫾ 11b 57 ⫾ 12a 14 ⫾ 5c 12 ⫾ 8c
Monounsaturated fat, % 12 ⫾ 4b 25 ⫾ 3a 9⫾ 2b 9 ⫾ 5b
Polyunsaturated fat, g 12 ⫾ 6b 24 ⫾ 5a 9⫾ 3b 6 ⫾ 3b
Polyunsaturated fat, % 6⫾ 2b 11 ⫾ 2a 6⫾ 1b 4 ⫾ 2b
Cholesterol, mg 332 ⫾ 126b 741 ⫾ 254a 130 ⫾ 14c 118 ⫾ 7c
Alcohol, % 3⫾ 3 1⫾ 2 0⫾ 0 0⫾ 1

1 Values are means ⫾ SD. Ketogenic diet group, n ⫽ 12. Control group, n ⫽ 8. Means in a row with different superscripts differ (P ⱕ 0.05).
2 Percentage of total energy intake.
1882 SHARMAN ET AL.

TABLE 2
Fasting blood lipid, metabolic and insulin responses in men who switched from their habitual diet to a ketogenic diet for 6 wk and
in a control group who continued to consume their habitual diet for 6 wk1

Ketogenic group (n ⫽ 12) Control group (n ⫽ 8)

Wk 0 Wk 3 Wk 6 Percent ⌬2 Wk 0 Wk 6 Percent ⌬

TC, mmol/L 4.27 ⫾ 0.8b 4.78 ⫾ 0.9a 4.47 ⫾ 0.81b 4.7% 4.24 ⫾ 1.0b 4.10 ⫾ 1.2b ⫺3.3%
TAG, mmol/L 1.09 ⫾ 0.5a 0.75 ⫾ 0.3b 0.73 ⫾ 0.3b ⫺33.0% 1.14 ⫾ 0.3a 1.08 ⫾ 0.7a ⫺5.3%
HDL-C, mmol/L 1.22 ⫾ 0.2b 1.43 ⫾ 0.3a 1.36 ⫾ 0.4b 11.5% 1.16 ⫾ 0.2b 1.16 ⫾ 0.5b 0.0%
LDL-C, mmol/L 2.87 ⫾ 0.8b 3.22 ⫾ 0.9a 2.99 ⫾ 0.8b 4.2% 2.89 ⫾ 0.9b 2.74 ⫾ 1.1b ⫺5.2%
VLDL-C, mmol/L 0.17 ⫾ 0.1a 0.12 ⫾ 0.0b 0.12 ⫾ 0.0b ⫺29.4% 0.18 ⫾ 0.0a 0.20 ⫾ 0.1a 11.1%
TC/HDL 3.60 ⫾ 0.9 3.45 ⫾ 0.88 3.45 ⫾ 0.9 ⫺4.2% 3.67 ⫾ 0.7 3.59 ⫾ 0.8 ⫺2.2%
Insulin, pmol/L 23.7 ⫾ 16.3a 19.1 ⫾ 12.2b 15.6 ⫾ 8.9b ⫺34.2% 21.5 ⫾ 6.7a 24.3 ⫾ 9.9a 13.0%

Downloaded from https://academic.oup.com/jn/article/132/7/1879/4687418 by guest on 30 January 2021

Glucose, mmol/L 5.00 ⫾ 0.4 4.84 ⫾ 0.4 4.99 ⫾ 0.3 ⫺0.2% 5.00 ⫾ 0.3 5.09 ⫾ 0.3 1.8%
␤-HBA, mmol/L 0.08 ⫾ 0.1b 0.40 ⫾ 0.3a 0.28 ⫾ 0.09a 250.0% 0.09 ⫾ 0.1b 0.10 ⫾ 0.1b 11.1%

1 Values are means ⫾ SD. Data were analyzed with a two-way ANOVA using body weight as a covariate. Ketogenic Diet group, n ⫽ 12, Control
group, n ⫽ 8. Means in a row with different superscripts differ (P ⱕ 0.05). TC, total cholesterol; TAG, triacylglycerol; ␤-HBA, ␤-hydroxybutyrate.
2 Percent change from wk 0 to wk 6.

pattern B subjects had significantly smaller mean and peak subjects. There were no changes from 0 to 6 wk in the
LDL particle diameters, a significantly greater percentage of concentrations of oxidized LDL in either the ketogenic group
LDL-3 and LDL-4, and a significantly smaller percentage of (44.38 ⫾ 33.7 U/L 3 46.45 ⫾ 15.6 U/L) or control group
LDL-1 compared with pattern A subjects. There were no (36.49 ⫾ 10.9 U/L 3 39.56 ⫾ 16.9 U/L).
significant changes in the percentage of any LDL subclasses or Postprandial TAG and insulin responses. Postprandial
the mean and peak particle size in pattern A subjects. There TAG concentrations peaked 3 h after the meal and started to
was a significant increase in peak LDL particle diameter from decline toward fasting values ⬃5 h after the meal (Fig. 3).
25.28 nm to 26.16 nm after the ketogenic diet in pattern B Compared with wk 0, peak postprandial TAG concentrations
subjects (Fig. 2), and also a significant increase in mean LDL were significantly lower (⫺24%) after the ketogenic diet (2.57
particle diameter. There was a significant increase in the ⫾ 1.4 to 1.96 ⫾ 0.7 mmol/L). The area under the postprandial
percentage of LDL-1 and a significant decrease in the percent- TAG curve was also significantly lower (⫺29%) after the
ages of LDL-3 and LDL-4 after the ketogenic diet in pattern B ketogenic diet (17.47 ⫾ 9.3 to 12.39 ⫾ 4.2 mmol/L ⫻ h).
Postprandial insulin concentrations peaked immediately after
the meal at wk 0 and 1 h after the meal at wk 6. Compared
with wk 0, the area under the postprandial insulin curve was
unaffected at wk 6 (339 ⫾ 168 to 283 ⫾ 140 pmol/L ⫻ h).

DISCUSSION
The primary objective of this study was to examine how
healthy normolipidemic, normal-weight men respond to a
ketogenic diet in terms of fasting and postprandial CVD bi-
omarkers. Ketogenic diets have been criticized on the grounds
they jeopardize health (8); however, very few studies have
directly evaluated the effects of a ketogenic diet on fasting and
postprandial risk factors for CVD. Subjects consumed a diet
that consisted of 8% carbohydrate (⬍50 g/d), 61% fat, and
30% protein. Adaptation to this ketogenic diet resulted in
significant reductions in fasting TAG (⫺33%), postprandial
lipemia after a fat-rich meal (⫺29%), and fasting insulin
concentrations (⫺34%). There were significant increases in
LDL particle size, and no change in the oxidative LDL con-
centrations. There was a significant increase in HDL choles-
terol at wk 3 after the ketogenic diet. Collectively, the re-
sponses in serum lipids, insulin and lipid subclasses to the
ketogenic diet were favorable in terms of overall CVD risk
profile.
Only a few studies have examined the effects of a diet with
very low amounts of carbohydrate on blood lipids (9,24). Our
laboratory recently examined the effects of a ketogenic diet
rich in monounsaturated fat and supplemented with (n-3)
FIGURE 1 Individual responses of men (n ⫽ 12) in LDL-choles- PUFA on blood lipids in normolipidemic men (9). Fasting
terol (C; upper graph) and HDL-C (lower graph) after consuming a TAG, total cholesterol, LDL cholesterol, and HDL cholesterol
ketogenic diet for 6 wk in normolipidemic, normal-weight men. changed ⫺55%, ⫹2%, ⫹10%, and ⫹10%, respectively (9).
 KETOGENIC DIETS AND BLOOD LIPIDS 1883

TABLE 3
Serum LDL subclass responses in 12 men who consumed a ketogenic diet who started as either pattern A or pattern B1,2

% LDL-1, % LDL-2, % LDL-3, % LDL-4, Peak LDL size, Mean LDL

27.7 nm 26.1 nm 24.5 nm 23.0 nm nm size, nm

Ketogenic group (n ⫽ 12)

Wk 0 13.9 ⫾ 6.4b 16.6 ⫾ 5.0 6.1 ⫾ 5.2 2.0 ⫾ 3.9 26.19 ⫾ 10.7b 26.28 ⫾ 7.8
Wk 3 18.4 ⫾ 6.8a 19.2 ⫾ 5.5 5.5 ⫾ 3.7 0.8 ⫾ 1.0 26.76 ⫾ 7.8a 26.54 ⫾ 4.3
Wk 6 18.2 ⫾ 6.2a 19.4 ⫾ 5.7 4.7 ⫾ 3.6 0.6 ⫾ 1.3 26.71 ⫾ 7.8b 26.57 ⫾ 4.5
Pattern A subjects (n ⫽ 7)
Wk 0 18.2 ⫾ 4.1 17.0 ⫾ 4.9 2.0 ⫾ 1.6 0.1 ⫾ 0.2 26.83 ⫾ 5.1 26.80 ⫾ 1.6
Wk 3 21.8 ⫾ 6.4 19.1 ⫾ 5.5 4.2 ⫾ 2.7 0.4 ⫾ 0.5 27.04 ⫾ 7.8 26.70 ⫾ 3.0
Wk 6 22.2 ⫾ 4.1 19.0 ⫾ 7.1 2.8 ⫾ 1.8 0.1 ⫾ 0.3 27.10 ⫾ 7.2 26.81 ⫾ 2.5
Pattern B subjects (n ⫽ 5)
Wk 0 7.9 ⫾ 3.1b 16.0 ⫾ 6.1 11.7 ⫾ 2.3b 4.7 ⫾ 5.5a 25.28 ⫾ 9.9b 25.54 ⫾ 7.0b

Downloaded from https://academic.oup.com/jn/article/132/7/1879/4687418 by guest on 30 January 2021

Wk 3 13.7 ⫾ 4.2a 19.2 ⫾ 6.1 7.2 ⫾ 4.6a 1.3 ⫾ 1.2b 26.37 ⫾ 6.6a 26.32 ⫾ 5.3a
Wk 6 12.6 ⫾ 3.8a 19.9 ⫾ 3.7 7.3 ⫾ 4.0a 1.4 ⫾ 1.9b 26.16 ⫾ 5.1a 26.22 ⫾ 4.6a

1 Values are means ⫾ SD. Means in a column within a group with different superscripts differ (P ⱕ 0.05).
2 Individuals with pattern A have a predominance of large LDL particles and those with pattern B have a predominance of smaller LDL particles.

Larosa et al. (24) examined the effects of a hypocaloric keto- There was a significant decrease in postprandial lipemia
genic diet on blood lipids in moderately overweight normo- after the fat-rich meal (⫺29%), which was significant but
lipidemic subjects. Fasting TAG, total cholesterol, LDL cho- somewhat lower than the decrease (⫺50%) we observed in
lesterol and HDL cholesterol changed ⫺33%, ⫹6%, ⫹18% response to a ketogenic diet rich in monounsaturated fat and
and ⫺6%, respectively. Corresponding changes in serum lipids supplemented with (n-3) polyunsaturated fatty acids (9). In
in this study for fasting TAG, total cholesterol, LDL choles- contrast to our results, Miller et al. (26) reported that a low-fat
terol, and HDL cholesterol were ⫺33%, ⫹4%, ⫹4%, and (19% of total energy)/high-carbohydrate (64% of total energy)
⫹11%, respectively. Confounding variables in these studies diet significantly reduced postprandial lipemia compared with
include varying degrees of weight loss (⫺2.2 to ⫺7.7 kg) and a diet higher in fat (41% of total energy) in normolipidemic
slight differences in the type of fat consumed. Nevertheless, men. The high-fat diet in the study by Miller et al. (26) still
these studies collectively indicate that carbohydrate restriction contained significant amounts of carbohydrate (42% of total
result in significant decreases in serum TAG, small increases in energy), which likely explains the conflicting results with our
total and LDL cholesterol, and moderate increases in HDL postprandial TAG response in men that consumed a very low
cholesterol in normolipidemic individuals. The small but sig- carbohydrate (8% of total energy) diet.
nificant weight loss (⫺2.2 kg) could have partially explained The significant reduction in fasting TAG was probably due
the HDL and TAG responses in this study. A meta-analysis by to the combination of a reduced VLDL production rate, which
Dattilo and Kris-Etherton (25), showed that for every kilo- has been shown to increase on a high-carbohydrate diet (27),
gram decrease in body weight during weight loss, HDL-C and an increase in TAG removal because high-fat diets (46 –
increases 0.009 and TAG decreases 0.015 mmol/L. Using 65% of total energy) significantly increase postheparin plasma
these estimates, the ⫺2.2 kg weight loss would have been LPL activity and skeletal muscle LPL activity in humans
predicted to increase HDL by 0.198 mmol/L and decrease (28 –30). A greater VLDL-TAG pool size would also compete
TAG by 0.033 mmol/L, which amounts to only 14% and 9%
of the observed changes in these parameters. Thus, dietary
composition most likely contributed to the changes in blood
lipids in this study.

FIGURE 2 Peak LDL particle size responses to a ketogenic diet in FIGURE 3 Postprandial serum triacylglycerol (TAG) responses
normolipidemic, normal-weight men classified as pattern A (n ⫽ 7) or (mean ⫾ SD; n ⫽ 12) to a fat-rich meal before (wk 0) and after (wk 6) a
pattern B (n ⫽ 5) at the start of the diet. Individuals with pattern A have ketogenic diet (n ⫽ 12) in normolipidemic, normal-weight men. The area
a predominance of large LDL particles and those with pattern B have a under the postprandial TAG curve was significantly (P ⱕ 0.05) lower
predominance of smaller LDL particles. The peak particle diameter for (⫺29%) after the ketogenic diet (17.47 ⫾ 9.3 to 12.39 ⫾ 4.2 mmol/L
pattern B is usually ⬍25.5 nm. *P ⱕ 0.05 from corresponding wk 0 ⫻ h). *P ⱕ 0.05 from corresponding wk 0 value. #P ⱕ 0.01 from
value. corresponding wk 0 value.
1884 SHARMAN ET AL.

with TAG from intestinal origin for removal during the post- Numerous studies now suggest that high-carbohydrate diets
prandial period. Thus, elevated fasting TAG (primarily VLDL- can raise TAG levels, create small, dense LDL particles, and
TAG) is associated with enhanced postprandial TAG (pri- reduce HDL cholesterol (i.e., atherogenic dyslipidemia)—a
marily chylomicron-TAG) due to competition for removal combination along with insulin resistance, that has been
(31). It follows then that a reduction in fasting TAG should be termed syndrome X (42,43). Syndrome X is postulated to be
directly related to a reduction in TAG responses to a fat-rich resistance to insulin-mediated glucose disposal by muscle (44),
meal, which was the case in this study (r ⫽ 0.59; P ⬍ 0.05). 30% of adult males and 10% to 15% of postmenopausal
Although the majority of studies have reported significant women have this particular syndrome X profile, which is
correlations between changes in fasting and postprandial TAG associated with several-fold increase in heart disease risk.
(9), a recent study demonstrated that a dietary regimen that Replacing saturated fat with carbohydrate appears to accentu-
lowered fasting TAG did not result in a reduction in postpran- ate insulin concentrations and the atherogenic dyslipidemia
dial TAG (32), emphasizing the importance of measuring associated with syndrome X (44,45). The ketogenic diet in this
postprandial TAG to assess overall CVD risk. study resulted in favorable responses in fasting TAG, postpran-
Dietary cholesterol intake increased ⬎100% (332–741 dial lipemia, HDL-C, LDL particle size, and insulin levels in
mg/d) when subjects switched to the ketogenic diet, which healthy normolipidemic men. Although the duration of the

Downloaded from https://academic.oup.com/jn/article/132/7/1879/4687418 by guest on 30 January 2021

would be predicted to result in significant increases in total diet was short (6 wk), these data suggest that a ketogenic
cholesterol and LDL cholesterol, although these were not diet does not have an adverse effect on accepted biochem-
significantly elevated after 6 wk of the ketogenic diet. There is ical risk factors for CVD and improves those associated with
great variability in the responsiveness of blood cholesterol syndrome X.
after increases in dietary cholesterol, which may be due to
differences in hormonal factors, obesity, and genetic predispo- LITERATURE CITED
sition (33). There were changes in the distribution of the LDL
1. Ordovas, J. M. (1999) The genetics of serum lipid responsiveness to
subfractions that would be considered favorable in terms of dietary interventions. Proc. Nutr. Soc. 58: 171–187.
CVD. We observed general increases in the mean and peak 2. National Cholesterol Education Program Expert Panel on Detection, Eval-
LDL particle sizes during the ketogenic diet, which were more uation, and Treatment of High Blood Cholesterol in Adults (Adult Treatment Panel
III). (2001) National Heart, Lung and Blood Institute and National Institutes of
pronounced in subjects that exhibited a pattern B distribution Health, Bethesda, MD. (NIH Publication No. 01-3670) May 2001.
at the start of the study. Individuals with pattern B exhibit a 3. Katan, M. B., Grundy, S. M. & Willett, W. C. (1997) Should a low-fat,
predominance of small dense lipoproteins and this distribution high-carbohydrate diet be recommended for everyone? Beyond low-fat diets.
is associated with increased risk of CVD (13,14) and was N. Engl. J. Med. 337: 563–567.
4. Schneeman, B. O. (2001) Carbohydrate: friend or foe? Summary of
recently shown to be the best discriminate factor for the research needs. J. Nutr. 131: 2764S–2765S.
presence of CVD even when adjusting for other risk factors 5. Parks, E. J. & Hellerstein, M. K. (2000) Carbohydrate-induced hyper-
(21). Although the characterization of pattern B is likely to triacylglycerolemia: historical perspective and review of biological mechanisms.
Am. J. Clin. Nutr. 71: 412– 433.
have a genetic origin (22), changes in diet are known to 6. Dreon, D. M., Fernstrom, H. A., Miller, B. & Krauss, R. M. (1994)
influence the distribution of LDL subclasses. For example, Low-density lipoprotein subclass patterns and lipoprotein response to a reduced-
switching to a fat-rich diet (46% vs. 24% of total energy) was fat diet in men. FASEB J. 8: 121–126.
7. Dreon, D. M., Fernstrom, H. A., Williams, P. T. & Krauss, R. M. (1999) A
shown to increase mean particle diameter and large LDL-1 very low-fat diet is not associated with improved lipoprotein profiles in men with
mass and decrease small dense LDL-III cholesterol (28), while a predominance of large, low-density lipoproteins. Am. J. Clin. Nutr. 69: 411– 418.
reductions in dietary fat have the opposite effect (6,7). Despite 8. Blackburn, G. L., Phillips, J. C. & Morreale, S. (2001) Physician’s guide
to popular low-carbohydrate weight-loss diets. Cleve Clin. J. Med. 68: 761–773.
the changes in LDL size, we did not observe any significant 9. Volek, J. S., Gomez, A. L. & Kraemer, W. J. (2000) Fasting lipoprotein
changes in oxidized LDL concentrations. Collectively these and postprandial triacylglycerol responses to a low-carbohydrate diet supple-
studies indicate that when dietary fat is reduced, the distribu- mented with n-3 fatty acids. J. Am. Coll. Nutr. 19: 383–391.
10. Austin, M. A., Hokanson, J. E. & Edwards, K. L. (1998) Hypertriglyc-
tion of LDL moves toward a smaller more dense particle and eridemia as a cardiovascular risk factor. Am. J. Cardiol. 81: 7B–12B.
when dietary fat is increased the distribution of LDL moves 11. Patsch, J. R., Miesenbock, G., Hopferwieser, T., Muhlberger, V., Knapp,
toward a larger less dense particle. The reason some individ- E., Dunn, J. K., Gotto, A. M., Jr. & Patsch, W. (1992) Relation of triglyceride
uals are more stable in their LDL subclass distribution in metabolism and coronary artery disease: studies in the postprandial state. Arte-
rioscler. Thromb. 12: 1336 –1345.
response to changes in diet is unknown but is likely to reflect 12. Roche, H. M. & Gibney, M. J. (2000) The impact of postprandial
complex interactions between metabolic and genetic traits lipemia in accelerating atherothrombosis. J. Cardiovasc. Risk 7: 317–324.
that are influenced to varying extents depending on the level 13. Austin, M. A., Breslow, J. L., Hennekens, C. H., Buring, J. E., Willett, W. C.
& Krauss, R. M. (1988) Low-density lipoprotein subclass patterns and risk of
of dietary fat (1,7). myocardial infarction. J. Am. Med. Assoc. 260: 1917–1921.
We observed a significant decrease in fasting and postpran- 14. Griffin, B. A., Freeman, D. J., Tait, G. W., Thomson, J., Caslake, M. J.,
dial insulin responses after the ketogenic diet. Decreases in Packard, C. J. & Shepherd, J. (1994) Role of plasma triglyceride in the
regulation of plasma low density lipoprotein (LDL) subfractions: relative contribu-
resting insulin concentrations have been reported in response tion of small, dense LDL to coronary heart disease risk. Atherosclerosis 106:
to 3– 4 d of a low-carbohydrate diet high in fat (34 –38). The 241–253.
mechanism for such a response probably resides in the greater 15. de Graaf, J., Hak-Lemmers, H. L., Hectors, M. P., Demacker, P. N.,
Hendriks, J. C. & Stalenhoef, A. F. (1991) Enhanced susceptibility to in vitro
reliance on fat oxidation induced by dietary carbohydrate oxidation of the dense low density lipoprotein subfraction in healthy subjects.
restriction (39) and subsequent reduced requirement for insu- Arterioscler. Thromb. 11: 298 –306.
lin to assist in glucose uptake. To our knowledge, the reduced 16. Schuh, J., Fairclough, G. F., Jr. & Haschemeyer, R. H. (1978) Oxygen-
mediated heterogeneity of apo-low-density lipoprotein. Proc. Natl. Acad. Sci.
postprandial insulin response to a fat-rich meal observed after USA 75: 3173–3177.
a ketogenic diet has not been reported in the literature. 17. Steinberg, D. (1997) Low density lipoprotein oxidation and its patho-
According to our estimate of insulin resistance using fasting biological significance. J. Biol. Chem. 272: 20963–20966.
levels of glucose and insulin, subjects in this study were not 18. Friedewald, W. T., Levy, R. I. & Fredrickson, D. S. (1972) Estimation of
the concentration of low-density lipoprotein cholesterol in plasma, without use of
insulin resistant and there was no adverse effect of the keto- the preparative ultracentrifuge. Clin. Chem. 18: 499 –502.
genic diet on insulin sensitivity. This is in agreement with 19. Volek, J. S., Gomez, A. L., Love, D. M., Avery, N. G., Sharman, M. J. &
other studies showing no adverse effects on glucose metabo- Kraemer, W. J. (2001) Effects of a high-fat diet on postabsorptive and post-
prandial testosterone responses to a fat-rich meal. Metabolism 50: 1351–1355.
lism or insulin resistance after ketogenic diets using the insulin 20. Hoefner, D. M., Hodel, S. D., O’Brien, J. F., Branum, E. L., Sun, D.,
clamp technique (40,41). Meissner, I. & McConnell, J. P. (2001) Development of a rapid, quantitative
 KETOGENIC DIETS AND BLOOD LIPIDS 1885

method for LDL subfractionation with use of the Quantimetrix Lipoprint LDL dence for extreme interindividual variation in dietary cholesterol absorption in
System. Clin. Chem. 47: 266 –274. humans. J. Lipid Res. 39: 2415–2422.
21. Rajman, I., Kendall, M. J., Cramb, R., Holder, R. L., Salih, M. & Gammage, 34. Fery, F., Bourdoux, P., Christophe, J. & Balasse, E. O. (1982) Hor-
M. D. (1996) Investigation of low density lipoprotein subfractions as a coronary monal and metabolic changes induced by an isocaloric isoproteinic ketogenic
risk factor in normotriglyceridaemic men. Atherosclerosis 125: 231–242. diet in healthy subjects. Diabetes Metab. 8: 299 –305.
22. Austin, M. A., King, M. C., Vranizan, K. M. & Krauss, R. M. (1990) 35. Galbo, H., Holst, J. J. & Christensen, N. J. (1979) The effect of different
Atherogenic lipoprotein phenotype: a proposed genetic marker for coronary heart diets and of insulin on the hormonal response to prolonged exercise. Acta
disease risk. Circulation 82: 495–506. Physiol. Scand. 107: 19 –32.
23. Matthews, D. R., Hosker, J. P., Rudenski, A. S., Naylor, B. A., Treacher, 36. Johannessen, A., Hagen, C. & Galbo, H. (1981) Prolactin, growth
D. F. & Turner, R. C. (1985) Homeostasis model assessment: insulin resistance hormone, thyrotropin, 3,5,3⬘-triiodothyronine, and thyroxine responses to exer-
and ␤-cell function from fasting plasma glucose and insulin concentrations in cise after fat- and carbohydrate-enriched diet. J. Clin. Endocrinol. Metab. 52:
man. Diabetologia 28: 412– 419. 56 – 61.
24. Larosa, J. C., Fry, A. G., Muesing, R. & Rosing, D. R. (1980) Effects of 37. Langfort, J., Pilis, W., Zarzeczny, R., Nazar, K. & Kaciuba-Uscilko, H.
high-protein, low-carbohydrate dieting on plasma lipoproteins and body weight. (1996) Effect of low-carbohydrate-ketogenic diet on metabolic and hormonal
J. Am. Diet. Assoc. 77: 264 –270. responses to graded exercise in men. J. Physiol. Pharmacol. 47: 361–371.
25. Dattilo, A. M. & Kris-Etherton, P. M. (1992) Effects of weight reduction 38. Langfort, J., Zarzeczny, R., Pilis, W., Nazar, K. & Kaciuba-Uscitko, H.
on blood lipids and lipoproteins: a meta-analysis. Am. J. Clin. Nutr. 56: 320 –328. (1997) The effect of a low-carbohydrate diet on performance, hormonal and
26. Miller, M., Teter, B., Dolinar, C. & Georgopoulos, A. (1998) An NCEP II
metabolic responses to a 30-s bout of supramaximal exercise. Eur. J. Appl.
diet reduces postprandial triacylglycerol in normocholesterolemic adults. J. Nutr.
Physiol. Occup. Physiol. 76: 128 –133.

Downloaded from https://academic.oup.com/jn/article/132/7/1879/4687418 by guest on 30 January 2021

128: 582–586.
39. Phinney, S. D., Bistrian, B. R., Evans, W. J., Gervino, E. & Blackburn, G. L.
27. Chen, Y. D., Coulston, A. M., Zhou, M. Y., Hollenbeck, C. B. & Reaven,
(1983) The human metabolic response to chronic ketosis without caloric restric-
G. M. (1995) Why do low-fat high-carbohydrate diets accentuate postprandial
tion: preservation of submaximal exercise capability with reduced carbohydrate
lipemia in patients with NIDDM? Diabetes Care 18: 10 –16.
28. Campos, H., Dreon, D. M. & Krauss, R. M. (1995) Associations of oxidation. Metabolism 32: 769 –776.
hepatic and lipoprotein lipase activities with changes in dietary composition and 40. Bisschop, P. H., de Metz, J., Ackermans, M. T., Endert, E., Pijl, H.,
low density lipoprotein subclasses. J. Lipid Res. 36: 462– 472. Kuipers, F., Meijer, A. J., Sauerwein, H. P. & Romijn, J. A. (2001) Dietary fat
29. Jackson, R. L., Yates, M. T., McNerney, C. A. & Kashyap, M. L. (1990) content alters insulin-mediated glucose metabolism in healthy men. Am. J. Clin.
Relationship between post-heparin plasma lipases, triglycerides and high density Nutr. 73: 554 –559.
lipoproteins in normal subjects. Horm. Metab. Res. 22: 289 –294. 41. Cutler, D. L., Gray, C. G., Park, S. W., Hickman, M. G., Bell, J. M. &
30. Kiens, B., Essen-Gustavsson, B., Gad, P. & Lithell, H. (1987) Lipopro- Kolterman, O. G. (1995) Low-carbohydrate diet alters intracellular glucose
tein lipase activity and intramuscular triglyceride stores after long-term high-fat metabolism but not overall glucose disposal in exercise-trained subjects. Metab-
and high-carbohydrate diets in physically trained men. Clin. Physiol. 7: 1–9. olism 44: 1264 –1270.
31. Wilson, D. E., Chan, I. F., Buchi, K. N. & Horton, S. C. (1985) Post- 42. Taubes, G. (2001) Nutrition: the soft science of dietary fat. Science
challenge plasma lipoprotein retinoids: chylomicron remnants in endogenous 291: 2536 –2545.
hypertriglyceridemia. Metabolism 34: 551–558. 43. Zammit, V. A., Waterman, I. J., Topping, D. & McKay, G. (2001) Insulin
32. Zoppo, A., Maggi, F. M. & Catapano, A. L. (1999) A successful dietary stimulation of hepatic triacylglycerol secretion and the etiology of insulin resis-
treatment fails to normalize plasma triglyceride postprandial response in type IV tance. J. Nutr. 131: 2074 –2077.
patients. Atherosclerosis 146: 19 –23. 44. Reaven, G. M. (2000) Diet and syndrome X. Curr. Atheroscler. Rep. 2:
33. Sehayek, E., Nath, C., Heinemann, T., McGee, M., Seidman, C. E., 503–507.
Samuel, P. & Breslow, J. L. (1998) U-shape relationship between change in 45. Krauss, R. M. (2001) Atherogenic lipoprotein phenotype and diet-gene
dietary cholesterol absorption and plasma lipoprotein responsiveness and evi- interactions. J. Nutr. 131: 340S–343S.