Content uploaded by Roland R Griffiths
Author content
All content in this area was uploaded by Roland R Griffiths on Nov 04, 2015
Content may be subject to copyright.
Is Caffeine a Flavoring Agent in Cola Soft Drinks?
Roland R. Griffiths, PhD; Ellen M. Vernotica, PhD
Background: Concern has been expressed about the nu-
trition and health impact of high rates of soft drink con-
sumption. Caffeine is an added ingredient in approxi-
mately 70% of soft drinks consumed in the United States.
The soft drink manufacturers’ justification to regulatory
agencies and the public for adding caffeine to soft drinks
is that caffeine is a flavoring agent.
Objective: To examine the claim that caffeine plays an
integral role in the flavor profile of soft drinks, by ex-
amining the effect of caffeine on the threshold for detec-
tion of flavor differences in cola beverages.
Design: Double-blind crossover study starting Novem-
ber 1998 and ending July 1999.
Setting: An academic research center.
Participants: Twenty-five adult regular consumers of
cola soft drinks. Based on a screening session, all were
able to detect a flavor difference between cola contain-
ing sugar and diet cola.
Intervention: A sensitive version of a forced-choice fla-
vor-detection procedure was used to evaluate the effects
of a wide range of caffeine concentrations (range, 0.05-
1.6 mg/mL) on the ability to detect flavor differences be-
tween caffeinated and caffeine-free cola beverages. Re-
peated tests permitted determination of significant
detection at each concentration in individual subjects.
Main Outcome Measures: Percentage of subjects sig-
nificantly detecting a flavor difference and mean per-
centage of trials correct at each caffeine concentration.
Results: Detection of flavor differences increased as a
function of caffeine concentration. At the 0.1-mg/mL con-
centration, which is the approximate concentration in the
majority of cola soft drink products, 2 subjects (8%) sig-
nificantly detected a flavor difference and the mean per-
centage correct (53%) was at chance levels.
Conclusions: The finding that only 8% of a group of regu-
lar cola soft drink consumers could detect the effect of
the caffeine concentration found in most cola soft drinks
is at variance with the claim made by soft drink manu-
facturers that caffeine is added to soft drinks because it
plays an integral role in the flavor profile. It is valuable
for the general public, the medical community, and regu-
latory agencies to recognize that the high rates of con-
sumption of caffeinated soft drinks more likely reflect the
mood-altering and physical dependence–producing ef-
fects of caffeine as a central nervous system–active drug
than its subtle effects as a flavoring agent.
Arch Fam Med. 2000;9:727-734
I
N 1942, the American Medical As-
sociation Council on Foods and
Nutrition stated:
From the health point of view it is de-
sirable especially to have restriction
of such use of sugar as is represented by con-
sumption of sweetened carbonated beverages
and forms of candy which are of low nutri-
tional value. The Council believes it would be
in the interest of the public health for all prac-
tical means to be taken to limit consumption
of sugar in any form in which it fails to be com-
bined with significant proportions of other food
of high nutritive quality.
1
In 1998, Americans consumed 15 bil-
lion gallons (216 billion liters) of flavored
carbonated beverages (ie, soft drinks),
which is the equivalent of 585 twelve-oz
See Editor’s Comment
at end of article
ORIGINAL CONTRIBUTION
From the Departments of
Psychiatry and Behavioral
Sciences (Drs Griffiths and
Vernotica) Neuroscience
(Dr Griffiths), Johns Hopkins
University School of Medicine,
Baltimore, Md.
(REPRINTED) ARCH FAM MED/ VOL 9, AUG 2000 WWW.ARCHFAMMED.COM
727
©2000 American Medical Association. All rights reserved.
(355-mL) cans for every man, woman, and child.
2
Con-
sumption of soft drinks in the United States has in-
creased 9-fold since the American Medical Association ex-
pressed its concern about the nutritional impact of soft
drinks, and consumption has more than doubled since
1975.
3,4
Currently, soft drinks represent the single largest
source of added sugar in the American diet
5
and there is
continuing concern that consumption of soft drinks, par-
ticularly by children and teenagers, adversely affects health
by increasing sugar and calorie consumption and by dis-
placing consumption of more healthful foods,
3,5
thereby
possibly increasing the likelihood of tooth decay,
6,7
obe-
sity,
3,8
and bone fracture.
3,9-11
In some cases, adults and chil-
dren become physiologically and psychiatrically depen-
dent on caffeinated soft drinks, experiencing a withdrawal
syndrome if they terminate regular use, and feeling com-
pelled to continue consuming soft drinks despite desires
and recommendations that they quit.
12-16
Caffeine is an added ingredient in approximately 70%
of soft drinks sold in the United States (calculated from
consumption data
2
). In response to a proposal in 1981
by the US Food and Drug Administration (Rockville, Md)
to delete caffeine from cola-type beverages,
17
soft drink
manufacturers justified adding caffeine to soft drinks on
the basis that caffeine is a flavor enhancer.
18
A similar jus-
tification for adding caffeine to citrus-flavored bever-
ages has formed the basis of a recent petition
19
by soft
drink manufacturers to the Health Protection Branch of
Health Canada, which is the Canadian food regulatory
agency (Ottawa, Ontario). The petition asserted that caf-
feine is an integral part of the overall flavor profile of soft
drinks, which have attained their popularity because they
activate a variety of flavor sensations.
20
A comparable ra-
tionale for the role of caffeine in soft drinks is currently
provided to the general public by soft drink manufac-
turers and the National Soft Drink Association (Wash-
SUBJECTS AND METHODS
SUBJECTS
Participants were recruited to participate in a soda taste study
through notices posted on public bulletin boards through-
out a large academic medical center campus (Johns Hop-
kins Bayview, Baltimore, Md). After preliminary tele-
phone screening about their cola consumption, prospective
subjects reported to the research unit, where they com-
pleted a medical history questionnaire, their vital signs were
taken, and they were interviewed about their cola con-
sumption. Subjects were eligible for participation if they
were aged 21 to 55 years; were medically healthy; and re-
ported consuming a cola beverage at least once a week, hav-
ing a preference for a specific brand of cola (eg, Coca-Cola
Classict) based on flavor and a preference for the caffeine
content of soda (ie, a preference for caffeinated over non-
caffeinated, or vice versa) based on flavor. The study was
approved by the local institutional review board.
STUDY PROCEDURES
Of the 55 individuals who completed the preliminary tele-
phone screening, 10 were not qualified and 4 were unwill-
ing or unable to report to the research unit for further screen-
ing. Of the 41 individuals who reported for further screening,
4 were not qualified and 5 were unwilling or unable to par-
ticipate in the first flavor-detection test session. Unknown
to participants, the first session consisted of participants
attempting to detect a flavor difference between a sugar-
containing cola and a diet cola. The data from the first ses-
sion were used to eliminate individuals who were rela-
tively insensitive to flavor differences. Of the 32 individuals
who participated in the first session, 7 failed to detect a fla-
vor difference and were dropped from the study at this point.
All of the remaining 25 eligible participants completed the
subsequent 6 flavor-detection test sessions, which con-
sisted of participants attempting to detect a flavor differ-
ence between caffeinated cola and caffeine-free cola over
6 caffeine concentrations.
FLAVOR-DETECTION PROCEDURE
The procedure used for detecting flavor differences (2-
alternative, forced-choice test with warm-up; trial-by-trial
feedback; and financial payment for correct responses) is sen-
sitive and permits testing differences under conditions in
which the dimension of the flavor difference cannot be speci-
fied in advance.
45-47
Sessions, which were scheduled at the
same time each day, consisted of 25 trials during each of which
participants tasted and swallowed 2 samples (5- to 6-mL)
of 2 different cola solutions (totalling approximately 275 mL
throughout the session). The interval between trials and be-
tween samples within a trial was 30 seconds. Participants
rinsed and swallowed bottled spring water (15-30 mL on each
occasion; 750-1500 mL throughout the session) between tri-
als and between samples within a trial. Bottled spring water
(Naya; Naya Inc, Mirabel, Quebec) was chosen for the study
because it was ranked the most neutral tasting of several
bottled waters tested.
48
Cola samples and water rinse were
presented at room temperature (22°-24°C).
Within a session, the 2 samples were arbitrarily des-
ignated “sample A” and “sample B.” The first 5 trials of the
session were warm-up trials in which the participant was
informed of the contents of each sample cup (eg, “This cup
contains sample B”) immediately before tasting. Each
warm-up trial consisted of the presentation of both samples,
with the participant tasting the sample on the left first. Or-
der of samples was mixed across warm-up sessions. The
final 20 trials of the session were test trials. As before, each
trial consisted of the presentation of 2 samples, 1 of which
(REPRINTED) ARCH FAM MED/ VOL 9, AUG 2000 WWW.ARCHFAMMED.COM
728
©2000 American Medical Association. All rights reserved.
ington, DC).
21,22
The argument that the high consumer
appeal of caffeinated soft drinks is due to the important
role of caffeine in the overall flavor profile of these bev-
erages overlooks the central nervous system–mediated
stimulating, reinforcing, and physical dependence–
producing actions of caffeine that provide a compelling
alternative explanation for the high rates of consump-
tion of caffeinated soft drinks.
23-25
Based on the published literature, it is difficult to
assess the credibility of the claim that caffeine plays an
integral role in the flavor profile of soft drinks. Al-
though caffeine has been recognized as a bitter sub-
stance that can be used as a flavoring ingredient,
26,27
the
effect of caffeine on the detection of flavor differences in
soft drinks has not been reported. Since 1981, the 2 most
popular brands of cola soft drinks (Coca-Cola Classict
and Pepsi Colat), which account for more than a third
of all soft drinks consumed in the United States,
2
have
had caffeine concentrations of about 0.1 mg/mL or
lower.
28-31
Although average detection thresholds of caf-
feine in water have often exceeded this concentra-
tion,
32-36
studies have also shown that detection thresh-
olds for caffeine vary widely between individuals,
32,36-39
can be lowered with explicit taste detection training,
38
and are raised in the presence of other tastes or fla-
vors.
35,37
Furthermore, caffeine and similar bitter sub-
stances show complex interactions with other tastes. Re-
search examining simple 2-substance or 3-substance
systems has shown that the addition of a second or a third
taste can either suppress or potentiate the other tastes,
depending on the substances under study.
35,40-43
Thus, pre-
cise prediction of how a substance such as caffeine would
affect the flavor profile of a chemically complex mixture
such as a cola soft drink that probably contains hun-
dreds of different flavors
44
is beyond the current state of
oral sensory science.
was sample A and the other sample B. The participant tasted
and swallowed the contents of both cups (left to right) and
then guessed which cup contained sample A and which cup
contained sample B. Immediately after each guess, the par-
ticipant was told whether he or she was correct or incor-
rect. In addition to their session pay of $10 per session, par-
ticipants were paid $0.25 for each correct guess. At the end
of each session, participants provided a written descrip-
tion of the basis on which they thought sample A and sample
B were different.
Participants and the research assistants who inter-
acted with them were told that the purpose of the flavor-
detection test procedure was to determine whether partici-
pants could discriminate between soft drinks containing
different commonly used taste or flavor enhancers, includ-
ing theophylline, quinine, sugar, caffeine, tannin, saccha-
rin, aspartame, and phosphoric acid. For all participants, the
first session consisted of attempting to detect a flavor dif-
ference between a sugar-containing cola vs a diet cola. Par-
ticipants who failed to correctly detect a flavor difference in
greater than or equal to 85% of 20 trials were dropped from
the study at this point. The remaining participants com-
pleted 6 additional flavor-detection test sessions. Each ses-
sion consisted of attempting to detect a flavor difference be-
tween a caffeine-containing cola vs a caffeine-free cola at a
single caffeine concentration. Six caffeine concentrations, 0.05
mg/mL, 0.1 mg/mL, 0.2 mg/mL, 0.4 mg/mL, 0.8 mg/mL, and
1.6 mg/mL (ie, 50 to 1600 ppm or 0.0002523 to 0.0082934
mol/L), were tested during the 6 sessions. The sequence of
concentrations was mixed across subjects. Sessions were gen-
erally conducted 3 to 5 times per week.
SAMPLE PREPARATION
For the first session (regular vs diet cola), samples were pre-
pared from a commercially available cola containing
high-fructose corn syrup and/or sucrose (Caffeine Free Coca-
Cola Classict) and a commercially available cola containing
aspartame (Caffeine Free Diet Coket). For each of the sub-
sequent 6 sessions, caffeinated and caffeine-free cola samples
were prepared from a single bottle of commercially available
caffeine-free cola (Caffeine Free Coca-Cola Classict). Samples
were prepared from a single bottle to eliminate confounding
by possible differences in flavor and/or carbonation be-
tween different bottles of cola. Caffeinated cola samples were
prepared by adding anhydrous caffeine to the caffeine-free
cola. Visual inspection indicated that the anhydrous caf-
feine dissolved completely within several hours of prepara-
tion. Caffeine-free samples were handled in the same way as
caffeinated samples. All samples were prepared at least 1 day
before sessions, stored in glass screw cap containers to main-
tain carbonation, and maintained at room temperature. Ali-
quots of the test samples (5-6 mL) were poured into 30-mL
medicine cups immediately before sessions.
DATA ANALYSIS
Within-Subject Analysis
For each subject in each session, the ability to detect a fla-
vor difference between the 2 cola conditions (regular vs diet
cola in session 1; caffeine vs no caffeine in sessions 2-7) was
analyzed using the binomial probability distribution. Flavor-
detection performance was considered significant if the sub-
ject was correct on 15 or more of the 20 test trials (P,.05).
Group Analysis
For analysis of the mean percentage of trials correct for regu-
lar vs diet cola and for caffeine vs no caffeine at each of the 6
caffeine concentrations, 1-tailed independent t tests were con-
ducted comparing the observed mean with a hypothesized
population mean of 10 (ie, 50% or chance performance).
(REPRINTED) ARCH FAM MED/ VOL 9, AUG 2000 WWW.ARCHFAMMED.COM
729
©2000 American Medical Association. All rights reserved.
The purpose of this study was to examine the
claim that caffeine plays an integral role in the flavor
profile of soft drinks by examining the effect of caffeine
on the threshold for detection of flavor differences in
cola beverages. Because there are large individual differ-
ences in taste sensitivity to caffeine, and to maximize
the generality of the findings to consumers of these bev-
erages, the study was conducted with volunteers who
regularly consume caffeinated colas and who reported a
preference for cola brand and for caffeine content based
on flavor. Also, to increase the generality of the findings
to consumers at large, volunteers did not receive exten-
sive training in flavor discrimination. Instead, volun-
teers were screened as being able to discriminate regu-
lar vs diet cola, and a sensitive procedure was then used
to determine the flavor-detection threshold within indi-
vidual subjects.
RESULTS
Twenty-five subjects completed the protocol (14 fe-
male, 11 male; 13 white, 2 Hispanic, 3 Asian, 7 black).
Subjects had a mean age of 29.7 years (SD, 8.3 years).
All had at least a high school diploma or equivalent, and
16 were college graduates. All were currently em-
ployed. Six were current cigarette smokers. Sixteen re-
ported consuming some alcohol weekly (mean, 3.1 drinks
per week for users). Subjects reported consuming cola
beverages a mean of 5.6 days per week (SD, 1.5 cola
beverages). Number of cola beverages consumed on a typi-
cal day was the equivalent of 2.3 (mean) 12-oz (355-
mL) servings (range, 1-6 servings; SD, 1.3 servings). Es-
timated annual cola consumption, adjusted for any
seasonal variation reported by subjects, was the equiva-
lent of 752 (mean) 12-oz (355-mL) servings (range, 176-
2184 servings; SD, 538 servings). In terms of brand fla-
vor preference, 14 preferred Coca-Cola Classict; 5, Pepsit;
4, Diet Pepsit; and 2, Diet Coket. All but 1 subject re-
ported preferring the flavor of caffeinated cola to caffeine-
free cola.
The
Figure shows the percentage of subjects sig-
nificantly detecting flavor differences (top) and the mean
percentage of trials correct (bottom) for each of the ex-
perimental conditions in the 25 subjects. The figure shows
that all subjects correctly detected a flavor difference be-
tween diet and regular cola, with a mean percentage cor-
rect of 98%. The high detection accuracy in this condi-
tion reflects the fact that participants who were unable
to discriminate diet vs regular cola at greater than or equal
to 85% were dropped from the study. The figure also
shows that both the percentage of subjects significantly
detecting a flavor difference and the mean percentage tri-
als correct increased as a function of caffeine concentra-
tion across the 6 concentrations. No subject was able to
significantly detect a difference ($75%) at the lowest caf-
feine concentration (0.05 mg/mL) and the mean per-
centage correct was at chance levels (52%). At the 0.1
mg/mL concentration, 2 subjects (8%) significantly de-
tected a flavor difference and the mean percentage cor-
rect remained at chance levels (53%). Increasing con-
centrations of caffeine (0.2 mg/mL, 0.4 mg/mL, 0.8 mg/
mL, and 1.6 mg/mL) were significantly detected by 56%,
96%, 100%, and 100% of subjects, respectively, with the
mean percentages correct being 74%, 95%, 99%, and
100%, respectively.
At the end of each session, participants provided a
written description of the basis on which they thought
the samples were different. Those subjects who cor-
rectly detected a flavor difference between caffeine vs
100
60
80
40
20
0
Subjects Significantly Detecting
a Flavor Difference, %
∗
∗
∗
∗
Diet
vs
Classic
100
60
70
80
90
40
50
20
10
30
0
0.05 0.1 0.2 0.4 0.8 1.6
Caffeine vs No Caffeine, mg/mL
Mean Trials Correct, %
0.05 0.1 0.2 0.4 0.8 1.6
Effects of caffeine concentration on the ability to detect a flavor
difference in a cola soft drink. The first bar shows the results of the
screening session in which all subjects significantly detected a flavor
difference between a sugar-containing cola vs a diet cola. The other bars
show the results of sessions testing the detection of flavor differences
between caffeinated cola and caffeine-free cola across 6 caffeine
concentrations. Top, The percentage of subjects significantly detecting a
flavor difference. Flavor detection was significant in a given subject if
the subject was correct on 15 or more of the 20 test trials (
P,
.05).
Bottom, Mean percentage trials correct. Bars show means; brackets
show +1 SD; asterisks indicate which conditions were significantly
different from chance (
P,
.05). The approximate caffeine concentration
in the 2 most widely consumed cola soft drink products is 0.1 mg/mL.
(REPRINTED) ARCH FAM MED/ VOL 9, AUG 2000 WWW.ARCHFAMMED.COM
730
©2000 American Medical Association. All rights reserved.
no caffeine at the lower concentrations (0.1 mg/mL and
0.2 mg/mL) most often described the caffeine samples
as being “bitter” and/or as having an “aftertaste.” Higher
caffeine concentrations were often further described as
being extremely unpleasant (eg, “like medicine” or
“nasty”).
COMMENT
This study shows that in a group of regular consumers
of cola soft drinks, the effect of caffeine on the thresh-
old for detection of flavor differences in a cola beverage
is an orderly increasing function of caffeine concentra-
tion. At the caffeine concentration of 0.1 mg/mL, only 2
subjects (8%) could detect a flavor difference, and the
overall data for the group was nonsignificant. This caf-
feine concentration is similar to or somewhat higher
than that in the 2 most popular brands of cola soft
drinks (Coca-Cola Classict and Pepsit),
28-31
which
account for more than half of cola soft drinks and more
than one third of all soft drinks consumed in the United
States.
2
At 0.2 mg/mL, which is the highest caffeine
concentration permitted in soft drinks by the Food and
Drug Administration,
49
56% of the subjects were able to
detect a flavor difference. At this concentration, subjects
usually described the caffeine-containing sample as bit-
ter or as having an aftertaste. There is only one com-
mercially marketed cola to have this level of caffeine
(Jolt Cola; Wet Planet Beverages, Rochester, NY),
whose marketing and advertising strategy is based not
on superior flavor but on having “twice the caffeine.”
Overall, the results of the study are at variance with the
claim made by soft drink manufacturers that caffeine is
added to soft drinks because it plays an integral role in
the flavor profile.
The subject population used in our study was cho-
sen as most appropriate for examining the effect of caf-
feine on the flavor of cola soft drinks among usual con-
sumers of those beverages. All subjects were regular
cola drinkers who claimed that they already had a pref-
erence for either caffeinated cola to caffeine-free cola, or
vice versa. All subjects also demonstrated flavor sensi-
tivity by significantly discriminating regular cola from
diet cola. In addition to the practice obtained in the
screening session, the flavor-detection procedure
involved explicit flavor-detection training during the
warm-up and test trials in which trial-by-trial feedback
with a financial contingency was provided. It is possible
that more extensive flavor-detection training would
have increased the sensitivity of subjects; however, such
a demonstration would not be relevant to the question
of whether a naturally occurring preference for cola
drinks depends on the role of caffeine. Finally, although
the study was conducted in adults, it is likely that simi-
lar results would have been obtained in children, many
of whom are also regular consumers of soft drinks,
3
since previous research has shown that caffeine detec-
tion thresholds are not lower in children than adults.
36
The failure to detect a flavor difference at the caf-
feine concentration found in popular cola beverages (0.1
mg/mL) cannot be attributed to using an insensitive flavor-
detection procedure. The study used a 2-alternative forced
choice test with a warm-up procedure that is among the
most sensitive of flavor-detection methods developed, and
is appropriate for testing differences under conditions in
which the dimensions of flavor difference cannot be speci-
fied in advance.
45-47
The addition of trial-by-trial feed-
back with a financial payment for correct responses could
be expected to further increase the sensitivity of the pro-
cedure.
50
Other testing parameters (eg, sample volume,
temperature, intersample interval, rinsing between
samples, and sample swallowing requirement) are simi-
lar to those that have been used in previous research dem-
onstrating low caffeine detection thresholds in aqueous
solutions in certain individuals.
32,33,36
Although the sample size in our study was only 25,
it is sufficient to provide meaningful information about
the detectability of caffeine in cola beverages. To our
knowledge, almost all of the caffeine taste detection
studies published to date used sample sizes smaller than
25 per group.
32-35,37-39
In our study, the number of sub-
jects detecting a difference and the mean percentage of
trials correct were monotonically increasing functions
of caffeine concentration, which demonstrates that the
procedure used was sensitive to the manipulation of
caffeine concentration. At the concentration of 0.1
mg/mL, which is similar to that delievered in most cola
soft drinks, only 2 of 25 subjects detected a flavor dif-
ference, and the mean percentage of trials correct was
only 53% (SD, 14.7%), which is very close to 50%, or
chance levels. From these data, it seems that increasing
the sample size would not have affected the conclusion
that caffeine does not play an important role in the fla-
vor of cola beverages for the large majority of regular
cola consumers.
It is interesting that the subjects in our study claimed
to have a preference for cola based partly on their belief
that caffeinated cola tastes different from noncaffeinated
cola. Furthermore, all but 1 of the subjects preferred caf-
feinated cola to caffeine-free cola, which would seem to
be consistent with manufacturers’ claims about the role
of caffeine. It would be informative to know whether sub-
jects could, in fact, detect differences in flavor under double-
blind conditions. We did not attempt to assess the claim
that caffeine contributes to the flavor profile of cola by di-
rectly comparing commercially available caffeine-
containing cola with caffeine-free cola because the con-
tents of those products are trade secrets and may differ on
dimensions other than just caffeine. Related to this, a pos-
sible critique of our study is that we tested the role of caf-
(REPRINTED) ARCH FAM MED/ VOL 9, AUG 2000 WWW.ARCHFAMMED.COM
731
©2000 American Medical Association. All rights reserved.
feine by adding caffeine to a commercially available caffeine-
free beverage (Caffeine Free Coca-Cola Classict). It is
possible that other changes in the formula of this bever-
age render it less sensitive to the effects of caffeine. Be-
cause the results of the present study have important im-
plications for the justification of adding caffeine to noncola
soft drinks (eg, Mountain Dew [PepsiCo Inc, Purchase,
NY], Sunkist [Sunkist Growers Inc, Sherman Oaks, Calif]),
it would also be informative to replicate our design using
a broader range of flavored soft drink vehicles.
Caffeine’s effects on the central nervous system pro-
vide a compelling alternative explanation for the strong
preference for caffeine-containing over caffeine-free cola
drinks claimed by participants in this study, and for the
high rates of consumption of these caffeinated soft drinks.
Caffeine is a stimulant drug that produces mild positive
mood changes (eg, increases in feelings of energy and alert-
ness)
51-53
and is self-administered
54-56
at doses delivered
in soft drinks. With daily consumption of caffeine at a
dose equivalent to that of 3 servings of cola soft drink
(100 mg), caffeine has been shown to produce physical
dependence characterized by lethargy and headache on
cessation of intake.
57,58
Suppression of low-grade with-
drawal symptoms after overnight abstinence has been sug-
gested as a principal mechanism by which caffeine en-
genders daily self-administration,
24,59-61
including
preference for flavors paired with caffeine.
24,25
In some
cases, people become psychiatrically dependent on caf-
feinated soft drinks, feeling compelled to continue con-
suming soft drinks despite desires and recommenda-
tions that they quit.
12,14,16
In addition to caffeine producing physical depen-
dence and playing a prominent role in maintaining bev-
erage consumption, various other caffeine-related health
concerns have been recently expressed,
62
including the
displacement of nutritionally healthful foods by caffein-
ated sugar-containing soft drinks,
3,5
and possible ad-
verse effects of caffeine in anxiety and sleep disorders
63
and in pregnancy.
64
In response to potential health con-
cerns and the aggressive marketing of caffeinated soft
drinks, in 1997 the American Medical Association adopted
a resolution to work with the Food and Drug Adminis-
tration to ensure that, when caffeine is added to a prod-
uct, the amount of caffeine is displayed prominently on
the label.
65
The Food and Drug Administration is cur-
rently considering a petition from the Center for Sci-
ence in the Public Interest (Washington, DC), which also
calls for labeling the caffeine dose.
66
In conclusion, the findings of our study suggest that
despite claims to the contrary, caffeine plays a relatively
minor role as a flavoring agent in cola soft drinks for the
majority of regular consumers. We cannot, of course, rule
out the possibility that some alteration in the testing pro-
tocol or parameters might permit development of a pro-
cedure uniquely sensitive to demonstrating, among nor-
mal users, a prominent role of caffeine as a flavoring agent
in soft drinks. In the meantime, we believe that it is worth-
while for the general public, the medical community, and
regulatory agencies to recognize the possibility that the
high consumption rates of caffeine-containing soft drinks
are more likely to reflect the mood-altering and physi-
cal dependence–producing effects of caffeine as a cen-
tral nervous system–active drug than its subtle effects as
a flavoring agent.
Accepted for publication May 17, 2000.
This study was supported in part by grant R01 DA03890
from the US Public Health Service (Rockville, Md).
A preliminary version of this article was presented to
The College on Problems of Drug Dependence, Acapulco,
Mexico, June 15, 1999.
We thank Joseph Harrison, for technical assistance in
conducting the study, Linda Felch, for consultation on sta-
tistical analysis, and Eric Strain, MD, Una McCann, MD,
and George Bigelow, PhD, for their helpful comments on the
manuscript.
Corresponding author: Roland Griffiths, PhD, Depart-
ment of Psychiatry, Behavioral Pharmacology Research
Unit, Johns Hopkins University School of Medicine, 5510
Nathan Shock Dr, Baltimore, MD 21224 (e-mail: rgriff
@jhmi.edu).
REFERENCES
1. American Medical Association, Council on Foods and Nutrition. Some
nutritional aspects of sugar, candy, and sweetened carbonated bev-
erages.
JAMA
. 1942;120:763-765.
2. Beverage Digest Company.
Beverage Digest Fact Book 1999
. Bedford
Hills, NY: Beverage Digest Co; 1999.
3. Jacobson MF.
Liquid Candy: How Soft Drinks Are Harming America’s
Health
. Washington, DC: Center for Science in the Public Interest; 1998.
4. Gilbert RM. Caffeine consumption. In: Spiller GA, ed.
Methylxanthine
Beverages and Foods: Chemistry, Consumption, and Health Effects
.
New York, NY: Liss; 1984:185-213.
5. Bowman SA. Diets of individuals based on energy intakes from added
sugars.
Fam Econ Nutr Rev
. 1999;12:31-38.
6. Ismail AI, Burt BA, Ekland SA. The cariogenicity of soft drinks in the
United States.
J Am Dent Assoc
. 1984;109:241-245.
7. Glynn K. The role of food and beverages in causing and preventing
dental decay. The Canadian Soft Drink Association Web site. Avail-
able at: http://www.softdrink.ca/prw94en.htm#10.0. Accessed De-
cember 3, 1999.
8. Troiano RP, Briefel RR, Carroll MD, Bialostosky K. Energy and fat in-
take of children and adolescents in the United States: data from the
National Health and Nutrition Examination Surveys.
Am J Clin Nutr
.
In press.
9. Wyshak G, Frisch RE. Carbonated beverages, dietary calcium, the di-
etary calcium/phosphorous ratio, and bone fractures in girls and boys.
J Adolesc Health
. 1994;15:210-215.
10. Wyshak G, Frisch RE, Albright TE, Albright NL, Schiff I, Witschi J. Non-
alcoholic carbonated beverage consumption and bone fractures among
women former college athletes.
J Orthop Res
. 1989;7:91-99.
11. Wyshak G. Teen-aged girls, carbonated beverage consumption, and
bone fractures.
Arch Pediatr Adolesc Med
. 2000;154:610-613.
(REPRINTED) ARCH FAM MED/ VOL 9, AUG 2000 WWW.ARCHFAMMED.COM
732
©2000 American Medical Association. All rights reserved.
12.
Strain EC, Mumford GK, Silverman K, Griffiths RR. Caffeine depen-
dence syndrome: evidence from case histories and experimental evalu-
ations.
JAMA
. 1994;272:1043-1048.
13.
Goldstein A, Wallace ME. Caffeine dependence in school children?
Exp
Clin Psychopharm
. 1997;5:388-392.
14.
Hughes JR, Oliveto AH, Liguori A, Carpenter J, Howard T. Endorse-
ment of
DSM-IV
dependence criteria among caffeine users.
Drug Al-
cohol Depend
. 1998;52:99-107.
15.
Bernstein GA, Carroll ME, Dean NW, Crosby RD, Perwien AR, Beno-
witz NL. Caffeine withdrawal in normal school-age children.
J Am Acad
Child Adolesc Psychiatry
. 1998;37:858-865.
16.
Bernstein GA, Carroll ME, Thuras P, Cosgrove K, Monroe M, Winters
K. Caffeine dependence in adolescents. In:
Scientific Proceedings 45th
Annual Meeting of the American Academy of Child and Adolescent Psy-
chiatry.
Washington, DC: American Academy of Child and Adolescent
Psychiatry; 1998:106.
17.
Caffeine: deletion of GRAS status, proposed declaration that no prior
sanction exists, and use on an interim basis pending additional study,
Federal Register
69817-69838 (1980).
18.
PepsiCo, Inc. The physical or technical effect of caffeine in cola bev-
erages, July 20, 1981. In: Volume III, Appendix XII of Comments of
the National Soft Drink Association submitted to the Department of
Health and Human Services Food and Drug Administration in Re-
sponse to the Proposal to Delete Caffeine in Cola-Type Beverages From
the List of Substances Generally Recognized as Safe and to Issue an
Interim Food Additive Regulation Governing Its Future Use
.
July 29,
1981. FDA Docket No. 80N-0418.
19.
Government of Canada, Department of Health. Regulations Amend-
ing the Food and Drug Regulations (1079),
Canada Gazette.
Part I. Janu-
ary 3, 1998.
20.
Pepsi-Cola Canada Beverages. Use of Caffeine as a Flavoring Agent in
Non-Cola Carbonated Soft Drinks, Volume 1: Food Additive Submis-
sion for the Health Protection Branch. Ottawa, Ontario: Health Canada;
1996. Health Canada submission ADDFS96100901.
21.
The Coca-Cola Company.
Top Ten Answers to the Ten Most Fre-
quently Asked Questions About Our Soft Drinks
. Atlanta, Ga: Coca-
Cola Co; 1997.
22.
National Soft Drink Association. What’s in Soft Drinks. Available at:
http://www.nsda.org/softdrinks/History/whatsin.html. Accessed No-
vember 17, 1999.
23.
Griffiths RR, Mumford GK. Caffeine–a drug of abuse? In: Bloom FE,
Kupfer DJ, eds.
Psychopharmacology: The Fourth Generation of
Progress
. New York, NY: Raven Press Ltd; 1995:1699-1713.
24.
Rogers PJ, Richardson NJ, Elliman NA. Overnight caffeine absti-
nence and negative reinforcement of preference for caffeine-
containing drinks.
Psychopharmacology
. 1995;120:457-462.
25.
Yeomans MR, Spetch H, Rogers PJ. Conditioned flavour preference
negatively reinforced by caffeine in human volunteers.
Psychophar-
macology
. 1998;137:401-409.
26.
Hall RL, Oser BL. Recent progress in the consideration of flavoring
ingredients under the food additives amendment, III. GRAS sub-
stances.
Food Technol
. 1965;253:151-197.
27.
Brieskorn CH. Physiological and therapeutical aspects of bitter com-
pounds. In: Rouseff RL, ed.
Bitterness in Foods and Beverages
. New
York, NY: Elsevier; 1990:15-33.
28.
Consumers Union. Caffeine: how to consume less.
Consumer Re-
port.
1981;46:597-599.
29.
Stavric B, Klassen R. Caffeine content in colas from New York State
and Ontario.
J Food Safety
. 1987;8:179-185.
30.
Galasko GTF, Furman KI, Alberts E. The caffeine contents of non-
alcoholic beverages.
Food Chem Toxicol
. 1989;27:49-51.
31.
Mumford GK, Benowitz NL, Evans SM, et al. Absorption rate of meth-
ylxanthines following capsules, cola and chocolate.
Eur J Clin Phar-
macol
. 1996;51:319-325.
32.
Mela DJ, Mattes RD, Tanimura S, Garcı´a-Medina MR. Relationship be-
tween ingestion and gustatory perception of caffeine.
Pharmacol Bio-
chem Behav
. 1992;43:513-521.
33.
Mattes RD. Influences on acceptance of bitter foods and beverages.
Physiol Behav
. 1994;56:1229-1236.
34.
Schiffman SS, Gatlin LA, Frey AE, Heiman SA, Stagner WC, Cooper
DC. Taste perception of bitter compounds in young and elderly per-
sons: relation to lipophilicity of bitter compounds.
Neurobiol Aging
.
1994;15:743-750.
35.
Schiffman SS, Gatlin LA, Sattely-Miller EA, et al. The effect of sweet-
eners on bitter taste in young and elderly subjects.
Brain Res Bull
. 1994;
35:189-204.
36.
James CE, Laing DG, Oram N. A comparison of the ability of 8-9-year-
old children and adults to detect taste stimuli.
Physiol Behav
. 1997;
62:193-197.
37.
Mackey AO, Valassi K. The discernment of primary tastes in the pres-
ence of different food textures.
Food Technol
. 1956;10:238-240.
38.
Pangborn RM. Influence of hunger and sweetness preferences on taste
thresholds.
Am J Clin Nutr
. 1959;7:280-287.
39.
Hall MJ, Bartoshuk LM, Cain WS, Stevens JC. PTC taste blindness and
the taste of caffeine.
Nature
. 1975;253:442-443.
40.
Pangborn RM. Taste interrelationships.
Food Res
. 1960;25:245-256.
41.
Schiffman SS, Diaz C, Beeker TG. Caffeine intensifies taste of certain
sweeteners: role of adenosine receptor.
Pharmacol Biochem Behav
.
1986;24:429-432.
42.
Calvin˜o AM, Garcı´a-Medina MR, Cometto-Mun˜iz JE, Rodrı´guez MB.
Perception of sweetness and bitterness in different vehicles.
Percept
Psychophys
. 1993;54:751-758.
43.
Breslin PAS, Beauchamp GK. Salt enhances flavour by suppressing
bitterness.
Nature
. 1997;387:563.
44.
Pendergrast M. The sacred formula [appendix]. In:
For God, Country and
Coca-Cola
. New York, NY: Charles Scribner’s Sons; 1993:421-425.
45.
Ennis DM. Relative power of difference testing methods in sensory
evaluation.
Food Technol
. April 1990:114-118.
46.
O’Mahony M, Thieme U, Goldstein LR. The warm-up effect as a means
of increasing the discriminability of sensory difference tests.
J Food
Sci
. 1988;53:1848-1850.
47.
Thieme U, O’Mahony M. Modifications to sensory difference test pro-
tocols: the warmed up paired comparison, the single standard duo-
trio and the A-not-A test modified for response bias.
J Sens Stud
. 1990;
5:159-176.
48.
Kemp SE, Beauchamp GK. Flavor modification by sodium chloride and
monosodium glutamate.
J Food Sci
. 1994;59:682-686.
49.
Food and Drug Administration. Caffeine. Washington, DC: US Gov-
ernment Printing Office,
Code of Federal Regulations
Title 21, Food
and Drugs, Part 182.11180; 1998.
50.
Breslin PAS, Beauchamp GK, Pugh EN Jr. Monogeusia for fructose,
glucose, sucrose, and maltose.
Percept Psychophys
. 1996;58:327-
341.
51.
Lieberman HR, Wurtman RJ, Emde GG, Coviella ILG. The effects of
caffeine and aspirin on mood and performance.
J Clin Psychophar-
macol
. 1987;7:315-320.
52.
Griffiths RR, Evans SM, Heishman SJ, et al. Low-dose caffeine dis-
crimination in humans.
J Pharmacol Exp Ther
. 1990;252:970-978.
53.
Silverman K, Griffiths RR. Low-dose caffeine discrimination and self-
reported mood effects in normal volunteers.
J Exp Anal Behav
. 1992;
57:91-107.
54.
Silverman K, Mumford GK, Griffiths RR. Enhancing caffeine reinforce-
ment by behavioral requirements following drug ingestion.
Psycho-
pharmacol
. 1994;114:424-432.
55.
Hughes JR, Oliveto AH, Bickel WK, Higgins ST, Badger GJ. The ability
of low doses of caffeine to serve as reinforcers in humans: a replica-
tion.
Exp Clin Psychopharmacol
. 1995;3:358-363.
56.
Liguori A, Hughes JR, Oliveto AH. Caffeine self-administration in hu-
mans: 1. Efficacy of cola vehicle.
Exp Clin Psychopharmacol
. 1997;5:
286-294.
(REPRINTED) ARCH FAM MED/ VOL 9, AUG 2000 WWW.ARCHFAMMED.COM
733
©2000 American Medical Association. All rights reserved.
57.
Griffiths RR, Evans SM, Heishman SJ, et al. Low-dose caffeine physi-
cal dependence in humans.
J Pharmacol Exp Ther
. 1990;255:1123-
1132.
58.
Evans SM, Griffiths RR. Caffeine withdrawal: a parametric analysis of
caffeine dosing conditions.
J Pharmacol Exp Ther
. 1999;289:285-
294.
59.
Hughes JR, Oliveto AH, Bickel WK, Higgins ST, Badger GJ. Caffeine
self-administration and withdrawal: incidence, individual differences
and interrelationships.
Drug Alcohol Depend
.1993;32:239-246.
60.
Schuh KJ, Griffiths RR. Caffeine reinforcement: the role of with-
drawal.
Psychopharmacology
. 1997;130:320-326.
61.
Garrett BE, Griffiths RR. Physical dependence increases the relative
reinforcing effects of caffeine versus placebo.
Psychopharmacology
.
1998;139:195-202.
62. James JE.
Understanding Caffeine: A Biobehavioral Analysis
. Thou-
sand Oaks, Calif: Sage Publications; 1997.
63. Strain EC, Griffiths RR. Caffeine. In: Tasman A, Kay J, Lieberman JA, eds.
Psychiatry
. Vol 1. Philadelphia, Pa: WB Saunders Co; 1997:779-794.
64. Eskenazi B. Caffeine – filtering the facts.
N Engl J Med
. 1999;341:
1688-1689.
65. American Medical Association. Caffeine Drinks (Resolution 523).
Adopted by the American Medical Association House of Delegates at
the June 1997 annual meeting.
66. Center for Science in the Public Interest. Petition for Amendment of
Food-Labeling Regulations to Require Quantitative Labeling of Caf-
feine Content and Request for Review of Health Effects of Caffeine:
Petition to the Food and Drug Administration. Washington, DC: Cen-
ter for Science in the Public Interest; July 31, 1997. FDA Docket No.
97P-0329/CP1.
Editor’s Comment
Editor’s Comment
T
hus, caffeine may affect the flavor in soft drinks in 2 circum-
stances: a small minority of people may detect caffeine in con-
centrations found in common soft drinks, or only at high doses
not usually found in soft drinks. While this study was not definitive proof—
the numbers of subjects was modest and perhaps caffeine does taste dif-
ferent when combined with other ingredients not tested in this study—I
believe it is basically true. In a less controlled setting, we have done taste
tests in our home and found our children could not reliably tell the dif-
ference between caffeinated and uncaffeinated soft drinks.
Soft drinks are generally of low or no nutritional value. If the caffeine is
primarily in soft drinks for its mood-altering affects, encouraging people
to drink more nutrition-poor liquid and potentially causing insomnia and
withdrawal, we as family physicians should oppose this, particularly for
children and adolescents.
Marjorie A. Bowman, MD, MPA
(REPRINTED) ARCH FAM MED/ VOL 9, AUG 2000 WWW.ARCHFAMMED.COM
734
©2000 American Medical Association. All rights reserved.
