Caffeine: a double-blind, placebo-controlled study of its ... · 760 ASTRUP ET AL TABLE 1 Physical characteristics ofthe experimental subjects Subject Sex Age Height Body weight Percent

AmJCIin Nuir l990;51:759-67. Printed in USA. © 1990 American Society forClinical Nutrition 759

Caffeine: a double-blind, placebo-controlled study of itsthermogenic, metabolic, and cardiovasculareffects in healthy volunteers13

ArneAstrup, S#{248}renToubro, Stephen Cannon, Pia Hem, LezfBreum, andJoop Madsen

ABSTRACT In humans caffeine stimulates thermogenesis

by unknown mechanisms and its effect on body weight has not

been studied. The effect of placebo and 100, 200, and 400 mgoral caffeine on energy expenditure, plasma concentrations ofsubstrates and hormones, blood pressure, and heart rate wasinvestigated in a double-blind study in healthy subjects who

had a moderate habitual caffeine consumption. Caffeine in-creased energy expenditure dose dependently and the thermo-

genic response was positively correlated with the response inplasma caffeine (r = 0.52; p < 0.018), plasma lactate (r = 0.79;

p < 0.000001), and plasma triglyceride (r = 0.53; p < 0.02).Stepwise regression analysis with the thermogenic response asthe dependent variable excluded plasma caffeine and yieldedthe following equation: thermic effect (kcal/3 h) = -0.00459

x heart rate + 0.30315 X (triglyceride) + 0.531 14 X (lactate)+ 1 5.34 (r = 0.86; p 0.000 1 ). The results suggest that lactate

and triglyceride production and increased vascular smoothmuscle tone may be responsible for the major part of the

thermogenic effect of caffeine. Am J Clin Nutr 1990; 51:

759-67.

KEY WORDS Caffeine, energy expenditure, obesity, sub

strate cycles

Introduction

Caffeine, a methylxanthine derivative, is the most widely

used drug, as popular as alcohol and tobacco. It is principallypresent in coffee, tea, cocoa, chocolate, and many cola-type soft

drinks. Pharmacologically, caffeine has bronchodilatory, car-

diotonic, and diuretic effects. In addition, caffeine is contained

in several over-the-counter preparations used for slimming (1,

2). Indeed, in rodents caffeine promotes weight loss by reducing

lipid stores because of increased energy expenditure but with-

out decreasing energy intake (3, 4).In humans the use ofcaffeine as a slimming agent has never

been justified by clinical documentation ofits weight-reducingproperties. In three studies caffeine was reported to stimulate

energy expenditure and lipolysis in humans (5-7). Although

Acheson et al (5) found that a cup of coffee (4 mg caffeine/kgbody wt) consumed with a meal produced a significantlygreater thermic response than that which followed the intake

ofthe same meal with a cup ofdecaffeinated coffee, this differ-

ence can be almost totally accounted for by the thermic effect

of the caffeine. Another study also found an increased energyexpenditure by caffeinated coffee (100 mg caffeine) comparedwith decaffeinated coffee (6 mg) (6). The third study found areduced thermogenic response to caffeine (4 mg/kg ideal body

wt) in obese and postobese patients compared with lean control

subjects (7). It appears that only a few different doses of caffeinehave been studied, and other constituents of coffee thancaffeine may influence lipid metabolism (8). The thermogenicprofile and the mechanisms behind this action of caffeine are

poorly understood.The present study was undertaken to provide results from

normal adults in a double-blind, placebo-controlled dose-re-

sponse study with respect to the thermogenic, cardiovascular,and metabolic effects of caffeine and to relate the responses tothe plasma concentrations ofcaffeine and its metabolites.

Subjects and methods

Experimental design

The study was designed as a placebo-controlled, double-

blind test of caffeine (100, 200, and 400 mg), three differentdoses of ephedrine, and two similar placebo tests. Only the re-sults from the caffeine and one placebo test are reported here.The order of the tests was not entirely randomized but it was

organized by a third party (a statistician) in a sequence thatallowed testing for a carry-over effect (an effect ofthe previous

test dose on the baseline values ofthe next test). The study wasapproved by the Danish National Health Service and by theMunicipal Ethical Committee of Copenhagen.

I From the Research Department ofHuman Nutrition, Royal Veter-mary and Agricultural University, Frederiksberg, Copenhagen; the In-stitute ofMedical Physiology C, Panum Institute, University of Copen-hagen; the Department of Clinical Research, DAK-Laboratory A/S,Copenhagen; and the Departments oflnternal Medicine and Endocri-nology, Herlev Hospital, Herlev, and Hvidovre Hospital, Hvidovre,Denmark.

2 Supported by a grant from the DAK-Laboratory A/S and the lb

Berg Foundation.

3 Address reprint requests to A Astrup, Research Department of Hu-man Nutrition, Royal Veterinary and Agricultural University, Roligh-

edsvej 25, 1958 Frederiksberg, Copenhagen, Denmark.Received March 2 1 , 1989.Accepted for publication June 7, 1989.

by guest on May 20, 2015

ajcn.nutrition.orgD

ownloaded from

http://ajcn.nutrition.org/

760 ASTRUP ET AL

TABLE 1

Physical characteristics ofthe experimental subjects

Subject Sex Age Height Body weightPercent over-

or underweightBody fatcontent

y cm kg % kg

1 F 26 172 58.9 -8.7 16.22 F 24 168 56.6 -9.4 10.1

3 F 25 171 57.6 -10.0 14.4

4 M 26 201 89.7 +1.0 9.4

5 M 25 182 82.9 +12.0 16.7

6 M 24 177 74.8 +5.0 11.0

i±SD 25±1 179±11 70.1±13.1 -1.7±8.3 13.0±2.9

Subjects

The experimental subjects were recruited from the Medical

School, University of Copenhagen. All were medical students

except one, who was a laboratory technician. Six healthy, nor-mal-weight subjects of both sexes aged between 20 and 32 ywere included. None were engaged in physical training, regular

exercise, or sports. They were all habitual coffee drinkers witha moderate coffee consumption of 150-300 mL/d. However,because heavy caffeine intake induces tolerance to at least some

of the effects of caffeine (9), subjects with a habitual intake ofmore caffeine, from coffee, tea, cola beverages, cocoa, and

chocolate, than corresponding to 1-2 daily cups of coffee

(> 100-200 mg caffeine) were not allowed to participate. Thisintake was assessed by a questionnaire on consumption of

coffee, tea, cola beverages, cocoa, and chocolate. None were

taking medicine during the study. One subject occasionallysmoked tobacco. They were instructed in a weight-mainte-nance diet containing “-250 g carbohydrate/d and 100 mmolNa/d to ensure filled glycogen stores and to avoid the influenceof a low-sodium diet on the sympathetic nervous system (10).The physical characteristics of the experimental subjects are

given in Table I.The percentage of overweight was calculated from the mdi-

vidual body weight by use of the midpoint of the medium

frame given in the 1959 Metropolitan Life Insurance Company

tables ofdesirable weights (1 1). The body fat content was esti-mated by skinfold-thickness measurements (12). This estimate

was obtained from duplicate measurements of the biceps, tri-

ceps, subscapular, and suprailiac skinfolds with a Harpendencaliper on 2 experimental days.

Test substances

The test substances were given orally as gelatin capsules, con-

taming either placebo (lactose) or caffeine (100, 200, or 400mg). All capsules had the same appearance and weight; lactosewas used as an inactive additive to fill them. The capsules wereswallowed with 300 mL tap water (20 #{176}C).

Experimental protocol

The experimental subjects abstained from food, coffee, tea,

cola beverages, cocoa, chocolate, and smoking overnight (> 12

h) before each test was started at 0830. The subjects were in-

structed to adhere strictly to the protocol, in particular regard-

ing abstinence from sweets and beverages containing caffeine,

during the study. The intake of the test substance was super-

vised and compliance was controlled by measuring plasmaconcentrations of methylxanthine metabolites before and afterintake.

Only a 150-mL glass of tap water was allowed in the morn-ing. There was a minimum of 3 d between two consecutivetests. The subjects were instructed not to have any physical ac-

tivity(bicycling,jogging, etc) in the morning. During the exper-

iments the subjects rested supine and light music was played

by a tape recorder to induce relaxation and to avoid hyperven-tilation, but sleeping was not permitted. No movements or

changes in position were allowed, in order to avoid any influ-

ence ofphysical activity on energy expenditure.

At least 60 mm before beginning the experiments, a Venfloncatheter (Viggo Products, Helsingborg, Sweden) was inserted

percutaneously into an antecubital vein for blood sampling.The catheter was kept open during the experiment by flushingwith isotonic sodium chloride solution (1 54 mmol/L) aftereach sampling. The room temperature was kept constant at 25-

27 #{176}C.All blood samples for determining substrate, metabolite,

and hormone concentrations were collected from the antecubi-tal vein. Blood was drawn -30, 0, 30, 60, 90, 120, 150, and 180mm relative to capsule intake. The subjects breathed through

a low-resistance, SCUBA one-way mouthpiece. After �‘ 10 mmofadaptation, expiratory gas was collected in Douglas bags for10 mm. Gas collection was made after each blood sampling.Before the study the experimental subjects were habituated to

the experimental procedures to prevent hyperventilation, anxi-ety, and uneasiness.

Analyses

Energy expenditure was measured by indirect calorimetry.Expiratory gas was continuously analyzed for oxygen and car-

bon dioxide with a Godart Rapox Oxygenometer (Rapox, Go-dart NV, Bilthoven, Holland) and a Beckman LB-l medicalgas analyzer. Respiratory steady state was assumed to exist

when the end-expiratory carbon dioxide fraction was constant.Expiratory gas was collected in Douglas bags and analyzed for

oxygen and carbon dioxide with gas electrodes connected to anacid-base analyzer (PHM 7 1, Radiometer A/S. Copenhagen)

and the volume was measured with a gas meter. Energy expen-diture was calculated by use ofa formula assuming a fixed pro-tein catabolism (13) because the error ofcalculating the energy


ajcn.nutrition.orgD

ownloaded from


THERMOGENIC EFFECT OF CAFFEINE 761

expenditure by omitting the exact correction from urinary ni-

trogen is negligible. The CV on resting energy expenditure at a

l-d interval and a l-wk interval was found to be �-‘3%. To ob-

tam this accuracy the gas electrodes were calibrated with stan-

dard gases ofknown composition before every sampling for gas

analysis. The standard gases were analyzed by the Scholander

microtechnique ( 14) with a measuring error on the gas fraction

of < 0.0005%, ie, an error on measurement of expired gas of

±0. 1-0.2%. The apparatus used in this investigation had a CVon repeated hourly measurements of� 3%.

Through the indwelling antecubital cannula, blood was sam-pled without stasis in iced syringes. The blood was then centri-fuged for 10 mm at 3000 x g and 10 #{176}C,and nonesterified fattyacids (NEFAs) were immediately extracted and later deter-

mined as previously described (1 5). Plasma glucose and lactatewere analyzed by standard enzymatic methods (16); glycerol,

as described by Laurell and Tibbling (17); and triglyceride, as

described by Giegel et al (1 8). Blood for determining methyl-

xanthine metabolites was collected in tubes containing reduced

glutathione and EGTA. Samples were immediately centrifuged

for 10 mm at 3000 x g and 4 #{176}C,and the plasma was stored at-40 #{176}Cuntil determination of methylxanthines by high-pres-

sure liquid chromatography (HPLC) (19). Immunoreactive in-

sulin, pancreatic glucagon, and C-peptide concentrations were

measured in plasma with radioimmunoassay kits purchased

from Novo, Copenhagen.All plasma samples for determining pancreatic hormones

and methylxanthine metabolites were coded and analyzed in a

random order to avoid any systematic error attributable to the

order of analysis. Plasma sodium and potassium concentra-

tions were determined by flame spectrophotometry. CVs for

measurements were as follows: NEFAs, 3.5%; glucose, 1%; so-

dium, 1%; potassium, 2%; glycerol, 4.7%; lactate, 4%; triglycer-

ide, 4%; insulin, 9.7%; C-peptide, 1 1%; and pancreatic gluca-

gon, 42%.

Arterial blood pressure was measured in the right arm by aninflatable cuffattached to a sphygmomanometer and heart rate

was determined by palpation ofthe peripheral pulse in the ipsi-

lateral radial artery. These measurements were performed after

each blood sampling.

A Trimline apparatus (PyMaH, Copenhagen) was used tomeasure arterial blood pressure. A cuff I 2- 14 cm wide was

used. The manometer pressure was slowly and gradually re-

duced from 200 mm Hg and the first Korotkoffsound was reg-

istered as the systolic pressure. The diastolic blood pressure was

determined as the manometer pressure when the Korotkoff

sound quality changes from tapping to muffled. Subjective feel-ings ofside effects were assessed by questioning the experimen-

tal subjects after each test substance.

Statistical analt’ses

The responses to a test substance were estimated separately

for each subject as the difference between the integrated numer-

ical area ofthe response curve (by a trapezoidal approach) and

the rectangular area determined by the basal values. A two-

way analysis of variance (ANOVA) for repeated measures was

performed to test differences between experimental periods

within the same experiment and to test differences between the

responses to different doses (20). Two means were compared

by post hoc testing (20). A possible carry-over effect was evalu-

ated by comparing data obtained from the two placebo periods

by means of a paired t test. p values < 0.05 were considered

significant. Linear and step-wise regression analyses and corre-lation analysis were performed with Statgraphics software

(Graphic Software Systems, mc, Rockville, MD). All results areexpressed as i ± SEM. There were no protocol deviations ordropouts.

Results

Caffeine and metabolites

Concentrations of caffeine, theobromine, and paraxanthineare shown in Figure 1 . Analyses of plasma caffeine from four

experiments were technically unsuccessful, so all statistics in-volving these analyses are based on data from the complete 20

sets only. The low preintake concentrations of caffeine (� 5

�mol/L) verified the abstinence from caffeine in accordancewith the protocol. When the baseline values from the two pla-

cebo tests were analyzed, no difference could be found with

respect to any ofthe methylxanthine derivates. Plasma concen-

trations of caffeine increased dose dependently (p = 0.004)

whereas the increase in paraxanthine was delayed, and compar-ison ofits course after intake of200 and 400 mg caffeine pointsto saturation ofthe pathway that converts caffeine to paraxan-thine. No changes occurred in theobromine after placebo or

caffeine intake.

Energy expenditure and respiratory quotient

The changes in energy expenditure and the integrated re-

sponses after placebo and the various caffeine doses are shownin Figure 2. The integrated thermogenic response to 100, 200,

and 400 mg caffeine were 9.2 ± 5.7, 7.2 ± 6.0, and 32.4 ± 8.2kcal/h (p < 0.001) above the placebo responses. These valuesare minimum figures because the energy expenditure had not

returned to baseline after the measurements ended, 3 h after

the intake (Fig 2). The thermogenic effects of 100 and 400 mgcaffeine were significantly above the placebo effect (p < 0.05

and p < 0.001 , respectively) whereas the difference between the

effect of 200 mg caffeine and placebo was not statistically sig-nificant. Linear-regression analysis showed that there was a sig-

nificant linear relation between caffeine dose and integrated re-

sponse above baseline of plasma concentration of caffeine (y

= 0.0055x - 0.176; n = 20; r = 0.81; p = 0.000015). Linear-regression analysis also showed a significant relation between

caffeine dose and thermogenic response (y = 0.064x + 7. 1 ; r= 0.59; p = 0.006). Correlation analysis showed a significant

linear relation between plasma caffeine response and thermo-genic response(y = 0.00841x + 1 1.72; r = 0.52;p = 0.018). A

multifactor analysis ofvariance including gender as a covariatedid not show any significant influence of sex on the thermo-genic response to caffeine (p = 0.85).

The respiratory quotient decreased slightly after all doses, in-

cluding after the placebo. No differences in the integrated re-

sponse could be detected between the active compounds andplacebo (data not shown). This indicates that the energy ex-pended above that observed after the placebo, ie, the expendi-ture induced by the active compounds, was caused by anequally increased carbohydrate and lipid oxidation.

Glucose and lactate

Plasma glucose concentration decreased slightly but insig-nificantly after placebo and 100 and 200 mg caffeine (Fig 3)


ajcn.nutrition.orgD

ownloaded from


Caffeine conc.(j�moI/L)

75.

Caffeine:I 400mg,�, 200mg0 100mg

#{163}Placebo

25-

0�

60 90 1�

Time after Caffeine intake (mm)

Paraxanthine(�cmoI/L)15-

12-

9-

6-

400mg

200mg

100mg

PB

400mg200mg

100mg

PB

200mg

100mg400mg

3-

0-

Theobromine(�imoI/L)

AEnergy expenditure(kcal/180 mm)

�or�1r�PB 100 200 400

Caffeine dose (mg)

FIG 2. Energy expenditure before and after oral intake of different

doses ofcaffeine or placebo. The right-hand figure shows the integratedresponses above baseline. Mean values (±SEM) ofsix subjects.

on plasma lactate (p < 0.01). Plasma lactate concentration de-

creased below baseline after placebo and 200 mg caffeinewhereas a small increase was found after 100 mg caffeine, andthere was a pronounced increase after 400 mg caffeine (Fig 3and Table 2). A positive correlation was found between caffeine

dose and plasma lactate response (Table 3).

Insulin and Cpeptide

Plasma insulin concentration decreased significantly below

baseline after placebo and 100 and 200 mg caffeine (Fig 4),

I 400mg� 200mg0 100mgA �

Glucose (mmoUL)5.2

5.0

4.8

4.6

4.4

1.0

0.9

0.8

0.7

0.6

0.5

0.4�

r � , , I � I

-30 0 30 60 90 120 150 180


FIG 3. Plasma glucose and lactate concentrations before and afterintake ofO, 100, 200, and 400 mg caffeine. Mean values (±SEM) of sixsubjects.

La�tate (mmoVL)

�--��t-�--+-i’---+ PB

10-

5-

0- . � U I �

-30 #{243}30 60 90 1�015O180


FIG 1 . Plasma concentrations of caffeine, paraxanthine, and theo-

bromine before and after oral intake of different doses of caffeine orplacebo. Mean values (±SEM) offive to six subjects.

whereas a small increase was noted between 60 and 90 mmafter intake of400 mg caffeine (p < 0.05). When the responses

from baseline were compared with placebo no difference couldbe detected (Table 2). However, there was a positive correlationbetween the caffeine dose and both plasma caffeine responseand glucose response (Table 3). Caffeine had a significant effect

762 ASTRUP ET AL


ajcn.nutrition.orgD

ownloaded from


Insulin (pmol/L)

I 400mgC-peptmde (nrnol/L) � �

0 100mg

Placebo#{163}


*p<O.O1.

t P < 0.05.

TABLE 2Integrated response from baseline over 3 h after caffeine intake*

Ca ffeine dose

0 100 200 400 Pt

mg

Systolicbloodpressure(mmHg) 3.2± 2.2 2.0± 1.2 1.5± 1.8 6.3± l.6t <0.05

Diastolic blood pressure (mm Hg) 1 .8 ± 2.0 2.7 ± 0.8 -0.2 ± 1 .2 6.3 ± I .5� <0.005

Heart rate(min’) 0.4 ± 1.8 -0.9 ± 2.2 -1.9 ± 1.7 -0.2 ± 1.2 NS

Glucose (mmol/L) -0. 1 ± 0.0 -0. 1 ± 0. 1 -0. 1 ± 0.0 0.0 ± 0. 1 NSLactate (�tmol/L) -79 ± 34 6 ± 14f -40 ± 31 98 ± 38� <0.01Glycerol (�mol/L) 1 .7 ± 3.3 1 1 . 1 ± 3.9j 14.4 ± 1 .7 II 15.0 ± 3.311 <0.05NEFAIT (j�mol/L) 2 ± 24 97 ± 66 85 ± 38 120 ± 34� NS

Triglyceride (Mmol/L) 53 ± 42 6 ± 38 10 ± 64 143 ± 38t NSInsulin (pmol/L) -2.7 ± 2.2 -2.7 ± 2.8 -5. 1 ± 2.7 0.2 ± 3.0 NSC-peptide (pmol/L) 20 ± 0 40 ± 20 0 ± 20 10 ± 20 NS

*�±SEM

t These p values indicate statistical difference between doses analyzed by a two-way analysis of variance.1�II Significantly different from placebo (by comparing i± SEM ofthe three caffeine doses with placebo by post hoc testing by use ofthe mean

square ofresiduals): jp < 0.05, §p < 0.001, ll� <0.005.#{182}Nonesterified fatty acids.

whereas no significant change was observed after 400 mg

caffeine. The changes, however, were small and no significant

difference was found between the integrated responses assessed

by ANOVA (Table 2). In contrast, a significant positive correla-

tion was found between plasma caffeine response and the inte-grated insulin response (r = 0.43; p = 0.04). Interestingly, apositive correlation was found between insulin response and

lactate response (r = 0.45; p = 0.02). No changes were found

in plasma concentrations of C peptide (Fig 4 and Table 2) or

ofpancreatic glucagon (data not shown).

Glycerol and nonesterifiedfatty acids

Plasma glycerol concentration increased slightly after pla-

cebo as a result oflipolysis induced by fasting (Fig 5 and Table

TABLE 3Correlation coefficients between caffeine dose and serum

concentrations and integrated responses ofthermogenesis, substrates,metabolites, and hormones

Caffeine dose(n=24)

Serum caffeineconcentration

(n=l8)

Thermogenesis 059* 0.52t

Glucose 0.56* 0.43tLactate 0.52t 0.29Glycerol O.48t 0.15NEFA 0.35 0.14Triglyceride 0.49t #{216}#{149}59*Insulin 0.25 0.43tC-peptide 0. 19 0.21Glucagon -0. 13 -0.15Systolic blood pressure 0.42t 0.37Diastolic blood pressure 0.27 0.05Heartrate -0.10 0.11

2). By contrast, caffeine had a pronounced impact on plasmaglycerol concentration, the most potent stimuli being 200- and400-mg doses. There was also a positive correlation between

the caffeine dose and the glycerol response (Table 3). In theplacebo test nonesterified fatty acids (NEFA) did not change

significantly above baseline (Fig 5 and Table 2). All doses ofcaffeine had a pronounced effect on NEFA but a considerable

variation between subjects was observed.

The plasma triglyceride concentration increased slightly butsignificantly after placebo and 100 and 200 mg caffeine (p

< 0.05)(Fig 5 and Table 2). After 400 mg caffeine a more markedincrease was found. After this dose triglyceride increased progres-

90

70

60


FIG 4. Plasma concentration of insulin and C peptide before andafter oral intake of 0, 100, 200, and 400 mg caffeine. Mean values(±SEM) ofsix subjects.


ajcn.nutrition.orgD

ownloaded from


G1�erol (1amoVL)

764

20�

150

10�

700

ASTRUP ET AL

caffeine and the changes did not differ from those observed af-

ter placebo (Fig 6). In contrast, 400 mg caffeine increased sys-tolic blood pressure, with a peak of 10 mm Hg 60 mm after theintake and with an average increase of6.3 mm Hg (Table 2). A

similar increase was observed in diastolic blood pressure. Therewas a positive correlation between caffeine dose and integratedincrease in systolic blood pressure (p = 0.04) whereas the rela-

tion between serum caffeine concentration and systolic bloodpressure was not statistically significant (p < 0.09). Analysis

showed a linear, positive correlation between the integratedresponses of systolic and diastolic blood pressures (r = 0.86;

n = 24; p < 0.00001).The heart rate increased, on average, 4 beats/mm after pla-

cebo (p < 0.05). After all three doses of caffeine a diphasic re-

sponse was observed (Fig 6). Initially, heart rate decreased by3-4 beats/mm 30-90 mm after the intake and subsequentlyincreased again at 90-1 80 mm. Only after 400 mg caffeine was

the final increase significantly above baseline.Nones�edfatty acids(1Am0VL)

. 400mg

� 200mg

0 100mg

#{163}Placebo

Triglyceride (mmol/L)

600

500

400

300

200

0.8

0.7

0.6

FIG 5. Plasma concentrations ofglycerol, nonesterified fatty acids,and triglyceride before and after intake of different doses of caffeine.Mean values ofsix subjects.

Systolic and Diastolicblood pressure (mmHg)

A 200mg

0 100mg

A Placebo

12�

110�

100

90.

80-

70

60

64

62

60

58

56

30 60 90 1:

�1ime after Caffeine intake (mm)

Side effects

Few (0- 1) side effects were reported after placebo and 100and 200 mg caffeine. By contrast, after 400 mg caffeine signifi-cantly more subjects reported side effects compared with pla-cebo (p < 0.01): four reported palpitation; three, anxiety; three,headache; two, restlessness; and one, dizziness.

Analysis offactors determining the thermogenic response

When all single factors were tested separately against the in-

tegrated thermic response to caffeine, only three significant re-

sively until the end of the study. There was also a highly signifi-

cantly positive correlation between both caffeine dose and plasma

caffeine response and the triglyceride response (Table 3).

Arterial bloodpressure and heart rate

Only small and insignificant changes could be detected insystolic and diastolic blood pressure after 100 and 200 mg

-� o �6�9b120


1�01�

FIG 6. Systolic and diastolic blood pressure and heart rate beforeand after intake ofdifferent doses ofcaffeine. Mean values (±SEM) of

six subjects.


ajcn.nutrition.orgD

ownloaded from


Thermogenic responseto caffeine (kcall3h)

.

.

S

60-

40-

20-

0-

.20-

S

S

r = 0.79p < 0.000001


.

. �

/ I #{149}.

.- I j � I ‘ ‘ ‘ I ‘ ‘ ‘ � I.40 .20 0 20 40 60

Plasma lactate response (�mol/L)

FIG 7. Correlation analysis between the integrated plasma lactateresponse above baseline and the thermogenic response to caffeine.

lations were found: the serum caffeine concentration (r = 0.52;

p < 0.01 8), the plasma lactate response (r = 0.79; p < 0.000001;

Fig 7), and the plasma triglyceride response(r = 0.53; p < 0.02).

However, when all plasma substrate and hormone responses to

caffeine, age, body weight, lean body mass, blood pressures,and heart rate as independent variables were included in a step-wise-selection regression analysis against the thermogenic re-

sponse as the dependent variable, the serum caffeine response

was excluded and the heart rate increase was included, yieldingthe following equation (Table 4):

Thermic effect (kcal/3 h) = -0.00459 X (pulse) + 0.30315

x (triglyceride) + 0.531 14 X (lactate) + 15.34

Discussion

(r = 0.86; radjusted = 0.82; p = 0.0001).

These results confirm previous reports of a thermogenic

effect ofcaffeine in humans who have a moderate daily caffeine

consumption (5-7). In addition, they provide the evidence thatcaffeine stimulates thermogenesis in a linear, dose-dependentway. Caffeine, but not its metabolites, is responsible for the

thermogenic effect as well as for the other cardiovascular and

metabolic actions determined because the majority of effects

were positively correlated to either caffeine dose or plasma

caffeine response, or both (Table 3). By contrast, no significantcorrelation was found between the plasma responses of para-

xanthine, theobromine, or theophylline nor any of the vari-ables listed in Table 3.

The results also point to possible mechanisms responsible for

the thermogenic effect ofcaffeine. A highly significant correla-tion was found between the plasma response of lactate and the

thermogenic response (r = 0.79; p = 0.000001; Fig 7). Also, theintegrated increase in plasma triglyceride was positively corre-lated with the thermogenic response (r = 0.59; p = 0.009). To

elucidate the possible contribution of these changes and othervariables to the thermogenic effect, a stepwise regression analy-sis was performed with the thermogenic response as the depen-dent variable and all integrated changes in plasma substrates,hormones, blood pressure, and heart rate as independent vari-ables. The analysis demonstrated that the changes in lactate,

triglyceride, and heart rate were significant predictors of thethermogenic response (Table 4). The squared coefficient of cor-

relation (r2 = 0.822 0.67) implies that 67% of the variationin thermic response may be accounted for by these three vari-

ables.Although initially well correlated with the thermogenic re-

sponse, the plasma caffeine response was excluded during the

analysis. This points to caffeine as exerting its thermogenic

effect not directly but indirectly through energy-consuming

processes related to the three variables.

It is likely that substrate cycles involving lactate and triglyc-

eride are responsible for the major part ofthe thermogenic re-sponse. Clearly, the changes in plasma concentrations are not

accurate measures ofthe fluxes through the energy-consumingsubstrate cycles. Thus determination of the fluxes by isotopesmay produce results that will better explain the variation in

thermogenic response to caffeine.

The relative importance of the mechanisms by which

caffeine exerts its various effects is not fully clarified. Caffeineprobably exerts most ofits effects through antagonism of aden-

osine receptors although phosphodiesterase inhibition and cal-

cium mobilization may also be important. The effects on car-

bohydrate metabolism are generally considered to be small andinconsistent (2 1), and the increases in plasma glucose and insu-

lin after caffeine in the study by Acheson et al (5) were notstatistically significant (seeTable 3 in ref5). By contrast, a smallbut significant increase was found in blood glucose after a dose

of250 mg caffeine compared with water (22). Also, in the pres-

ent study the impact ofcaffeine on plasma glucose, insulin, andC peptide was very modest (Figs 3 and 4 and Table 2). How-ever, a significant positive correlation was found between theincrease in plasma caffeine and both plasma glucose and insu-lin (Table 3). Furthermore, a pronounced increase was ob-

served in plasma lactate after 400 mg caffeine (Fig 3), an effectthat appeared to be dose-dependent (Table 3). The increase inlactate did not correlate with plasma caffeine (r = 0.29; NS) but

with plasma insulin (r = 0.45; p = 0.02). These findings and

TABLE 4

Results ofmodel fitting by use ofstepwise regression analysis of thethermogenic response to caffeine as the dependent variable andsubstrate, hormone, and other measures (see text) as the independentvariables

k SEE i p

Constant 15.341 2.0092 7.6356 0.0000

Lactate 0.531 0.1176 4.5153 0.0004

Triglyceride 0.303 0.0894 3.3878 0.0038Heart rate -0.005 0.0022 -2.0422 0.0580


ajcn.nutrition.orgD

ownloaded from


766 ASTRUP ET AL

the association between the lactate increase and thermogenic

response suggest the involvement of the thermogenic Cori cy-cle, ie, the conversion ofglycogen and glucose to lactate in mus-

cle and adipose tissue (23) and subsequent hepatic gluconeo-

genesis and glycogenesis (24). This substrate cycle and other

hepatic thermogenic processes triggered by lactate (25) may ex-plain the thermogenic contribution of lactate to the effect of

caffeine. Svedmyr (26) also found that the increase in plasmalactate was closely correlated with the thermogenic effect of epi-

nephrine in humans.The increases in plasma glycerol and NEFAs are due to a

well-described lipolytic action ofcaffeine and other methylxan-

thine derivates (5, 7). By contrast, the increase in triglyceride is

less well recognized (2 1). In the present study this effect was

dose-dependent and correlated with the plasma caffeine con-centration (Fig 5 and Tables 2 and 3). The increase in plasma

triglyceride was positively correlated with the thermogenic re-

sponse (r = 0.53) and this variable was not excluded by thestepwise regression analysis(Table 4). Thus lipolysis in adiposetissue and subsequent hepatic reesterfication, ie, formation of

triglyceride, may explain why the increase in triglyceride corre-lates with the thermic effect of caffeine. This is supported bythe findings that inhibition oflipolysis by nicotinic acid reducesthe thermogenic response to norepinephrine (27) and that rais-ing the NEFA concentrations by adding triglyceride exoge-nously has a pronounced thermic effect and increases hepatic

blood flow (28).Except for 400 mg caffeine, only a weak influence of caffeine

was detected on blood pressure and heart rate (Fig 6 and Tables

2 and 3). The cardiovascular effects of caffeine are known in

detail because of the studies by Robertson et al (9, 10) andWhitsett et al (29). When 250 mg oral caffeine was given to

people who do not drink coffee, systolic blood pressure in-creased by 10 mm Hg whereas heart rate showed a diphasicdecrease after the first hour followed by an increase above base-line after 2 h (10). However, in a subsequent study Robertsonet al (9) found that during chronic caffeine intake an essentiallycomplete tolerance developed to these effects after 1-4 d of con-sumption of750-mg caffeine/d.

Because a modest daily caffeine intake was allowed in thepresent study, our subjects may have developed partial toler-ance. This could explain why only the highest dose of caffeinehad a substantial effect on blood pressure (Fig 6). The diphasicresponse in heart rate was found after 400 mg caffeine (Fig 6).

That the response in heart rate inversely contributes to the ther-

mic effect of caffeine is not immediately intelligible. Caffeine

significantly increases cardiac contractility and hence strokevolume and cardiac output (30). In addition, caffeine increasesperipheral vascular resistance (29). Consequently, the decrease

in heart rate is likely mediated by reflexes secondary to the in-crease in blood pressure, and the magnitude of the decreasemay reflect the energy expenditure associated with increasedcardiac contractility and vasoconstriction caused by increased

smooth muscle tone.Caffeine significantly increases plasma concentrations of

norepinephrine and epinephrine (9, 10). Conceivably, the ther-mogenic effect may be mediated by �3-adrenergic stimulation.However, this possibility was ruled out by Jung et al (7), whofound that a L�-adrenergic blockade did not reduce the thermo-genic or lipolytic effect of caffeine. In addition, Robertson et

al (10) reported that the cardiovascular effects of caffeine are

normal in patients who lack a functioning autonomic nervous

system.A significant thermogenic effect was found even after the

lowest dose ofcaffeine (100 mg) despite the fact that our experi-mental subjects had a habitual caffeine intake of 100-200 mg/

d, which was confirmed by the finding ofa fasting plasma con-centration ofcaffeine of2-6 �moI/L (Fig I). The magnitude ofthe thermogenic response was clearly underestimated becausethe energy expenditure had not returned to baseline levelswhen the measurements were ended 3 h after the intake (Fig2). Although a certain degree of tolerance to the thermogeniceffect of caffeine may develop, these results suggest that a sub-stantial effect remains during a moderate daily caffeine con-

sumption. In a very recent study Dulloo et al (3 1) found a 3-4% increase in energy expenditure after oral administration of100 mg caffeine in subjects who consumed 250-500 mg/d,

which is twice the daily intake of our subjects. Also, physical

activity may enhance the thermogenic response to caffeine(32), which may be important for weight reduction. The ther-mogenic effect ofcaffeine may reduce body fat stores in coffee

drinkers if energy balance is not maintained by an increasedenergy intake. Thus the effect of caffeine in the treatment ofobesity remains to be determined. 13

References

1. Dulloo AG, Miller DS. The thermogenic properties of ephedrine/methylxanthine mixtures: human studies. Int J Obes 1986; 10:467-81.

2. Malchow-M#{248}ller A, Larsen S, Hey H, Stokholm KH, Juhl E,Quaade F. Ephedrine as an anorectic: the story of the “Elsinorepill.” IntJObes l98l;5:l83-7.

3. Bukowiecki U, Lupien J, Foll#{234}aN, Jahjah L. Effects of sucrose,caffeine, and cola beverages on obesity, cold resistance, and adi-pose tissue cellularity. Am J Physiol 1983;244:R500-7.

4. Cheung WT, Lee CM, Ng TB. Potentiation of the antilipolyticeffect of 2-chloroadenosine after chronic caffeine treatment. Phar-macology l988;36:33 1-9.

5. Acheson IU, Zahorska-Markiewicz B, Pittet P, Anantharaman K,J#{233}quierE. Caffeine and coffee: their influence on metabolic rateand substrate utilization in normal weight and obese individuals.Am J Clin Nutr 1980; 33:989-97.

6. Hollands MA, Arch iRS, Cawthrone MA. A simple apparatus forcomparative measurements of energy expenditure in human sub-

jects: the thermic effect of caffeine. Am J Clin Nutr l981;34:2291-4.

7. Jung RT, Shetty PS, James WPT, Barrand MA, Callingham BA.Caffeine: its effect on catecholamines and metabolism in lean andobese subjects. Clin Sci 198 l;60:527-35.

8. B#{248}naaK, Arnesen E, Thelle DS, F#{248}rdeOH. Coffee and cholesterol:is it all in the brewing? The Troms� study. Br Med J l988;297:

1103-4.9. Robertson D, Wade D, Workman R, Woosley RL. Tolerance to

the humoral and hemodynamic effects of caffeine in man. J ClinInvest l98l;67:llll-7.

10. Robertson D, Fr#{246}lichJC, Carr RK, et al. Effects of caffeine onplasma renin activity, catecholamines and blood pressure. N EnglJMed l978;298:18l-6.

1 1 . Metropolitan Life Insurance Company. New weight standards formen and women. Stat Bull Metropol Life Insur Co, 1959;40: 1-4.

12. Durnin JVGA, Womersley J. Body fat assessed from total bodydensity and its estimation from skinfold thickness. Br J Nutr1974; 32:77-97.

13. Garby L, Astrup A. The relationship between the respiratory quo-


ajcn.nutrition.orgD

ownloaded from



tient and the energy equivalent ofoxygen during simultaneous glu-cose and lipid oxidation and lipogenesis. Acts Physiol Scand

1987; 129:443-4.14. Scholander PF. Analyser for accurate estimation of respiratory

gases in one-halfcubic centimeter samples. J Biol Chem 1947; 167:

235-9.

15. Laurell 5, Tibbling G. Colorimetric microdetermination of free

fatty acids in plasma. Clin Chim Acta 1967; 16:57-62.16. Bergmayer HV, ed. Methodsofenzymatic analysis. New York: Ac-

ademicPress, 1974.

17. Laurell 5, Tibbling G. An enzymatic fluorometric micromethod

for the determination of glycerol. Clin Chim Acta 1966; 13:3 17-

22.

18. Giegel JL, Ham AB, Clema W. Manual and semi-automated pro-cedures for measurement of triglyceride in serum. Clin Cheml975;2l: 1575-8 1.

19. Willems HJJ, Van der Horst A, de Goede PNFC, HaakmeesterGJ.

Determination of some anticonvulsants, antiarrhytmics, benzodi-azepines, xanthines, paracetamol and chloramphenicol by re-versed phase HPLC. Pharm Weekbl [Sci] l985;7: 150-5.

20. Brown BW, Hollander M. Statistics, a biomedical introduction.New York: John Wiley & Son, 1977.

21. Fredholm BB. Gastrointestinal and metabolic effects of methyl-xanthines. In: Dews PB, ed. The methylxanthine beverages and

foods: chemistry, consumption, and health effects. New York: AR

Lisa Press, 1984:331-54.22. Cheraskin E, Ringsdorf WM, Setyaadmadja ATSH, Barrett RA.

Effect ofcaffeine versus placebo supplementation on blood-glucoseconcentration. Lancet l967;2: 1299-1300.

23. Newby FD, Sykes MN, DiGirolamo M. Regional differences in

adipocyte lactate production from glucose. Am J Physiol

1988;255:E7 16-22.

24. Kusaka M, Ui M. Activation ofthe Con cycle by epinephrine. Am

J Physiol l977;232:El45-55.25. Lardy H, Su CY, Kneer N, Wielgus S. Dehydroepiandrosterone

induces enzymes that permit thermogenesis and decrease meta-bolic efficiency. In: Lardy H, Stratman F, eds. Hormones, thermo-

genesis and obesity. New York: Elsevier, 1989:415-26.26. Svedmyr N. Studies on the relationships between some metabolic

effects of thyroid hormones and catecholamines in animals andman. Acta Physiol Scand l966;68(suppl 274): 1-46.

27. Havel Ri, Carlson LA, Ekelund L-G, Holmgren A. Studies on the

relation between mobilization offree fatty acids and energy metab-

olism in man: effects ofnorepinephrine and nicotinic acid. Metab-

olism l964;l3:l402-l2.

28. Astrup A, B#{252}lowJ, Christensen NJ. The effect of nonesterifiedlong-chain fatty acids on blood flow and thermogenesis in brownadipose tissue in the young dog. Acta Physiol Scand 1985; 124:

8 1-5.

29. Whitsett TL, Manion CV, Christensen HD. Cardiovascular effectsofcoffee and caffeine. Am J Cardiol 1984;53:9l8-22.

30. Gould L, Venkataraman K, Goswami M, Gomprecht RF. The car-diac effects ofcaffeine. Angiology 1973;24:455-63.

31. Dulloo AG, Geissler GA, Horton T, Collins A, Miller DS. Normalcaffeine consumption: influence on thermogenic and daily energyexpenditure in lean and postobese human volunteers. Am J Clin

Nutr l989;49:44-50.32. LeBlanc J, Jobin M, C#{244}t#{233}J, Samson P, Labrie A. Enhanced meta-

bolic response to caffeine in exercise-trained human subjects. JAppI Physiol 1985; 59:832-7.


ajcn.nutrition.orgD

ownloaded from


Documents

Caffeine: a double-blind, placebo-controlled study of its ... · 760 ASTRUP ET AL TABLE 1 Physical characteristics ofthe experimental subjects Subject Sex Age Height Body weight Percent