International Journal of Obesity (1997) 21, 860±864
ß 1997 Stockton Press All rights reserved 0307±0565/97 $12.00
Psychological and metabolic responses of
carbohydrate craving obese patients to
carbohydrate, fat and protein-rich meals
AC Toornvliet1, H Pijl1, JC Tuinenburg1, BM Elte-de Wever2, MSM Pieters5, M FroÈlich3, W Onkenhout4
and AE Meinders1
Departments of 1 General Internal Medicine, 2 Dietetics and Nutrition, 3 Clinical Chemistry, and 4 Paediatrics of the Leiden University
Hospital, and 5 Centre for Human Drug Research, Leiden, The Netherlands
RATIONALE: A defective central serotonergic neurotransmission has been suggested to result in the concomitant
occurrence of an appetite disorder and a disturbed mood. This syndrome was termed carbohydrate carving (CC)
obesity. Excessive consumption of carbohydrate-rich snacks would, through a plasma amino acid mediated mechanism, restore serotonergic neurotransmission and thereby relieve the symptoms of atypical depression.
OBJECTIVES: To test whether CC obese patients indeed exhibit symptoms of atypical depression, whether these
symptoms can be alleviated by carbohydrate-rich snacks and whether they respond differently to the snacks than
non-carbohydrate craving (NC) control subjects. Furthermore, we investigated whether differences between CC and
NC patients could be related to peripheral metabolic differences.
DESIGN: Double blinded, randomized with cross-over. Patients received three types of snacks (100/0/0, 70/29/1 and
35/3/62 energy percent carbohydrate/fat/protein respectively) on three consecutive test days. Before and after snack
administration mood and performance were tested and blood samples were obtained.
SUBJECTS: 9 CC and 17 NC obese patients, matched for sex, age and body mass index.
MEASUREMENTS: Mood states (Pro®le of Mood States and Visual Analogue Scales) and performance (BourdonWiersma cancellation test), serum glucose and insulin and plasma amino acid concentrations.
RESULTS: Before snack consumption, CC patients had slightly higher anger and fatigue scores and tended to have
lower mood scores than NC patients. The ef®ciency of performance increased in both groups after all snacks. No other
psychological effects of the snacks were registered. Psychological and metabolic responses of CC and NC patients to
the snacks were similar.
CONCLUSION: Although they may have a somewhat disturbed mood, CC obese patients do not improve their mood
states through ingestion of a carbohydrate-rich snack. It seems, from a therapeutic point of view, useless to maintain
the concept of carbohydrate craving.
Keywords: obesity; carbohydrate craving; mood; performance; amino acids
Introduction
Carbohydrate consumption increases brain 5-HT
synthesis through an insulin and amino acid mediated
metabolic pathway. Insulin, released after carbohydrate consumption, induces a decline of plasma large
neutral amino acid (LNAA) levels, through an
enhanced uptake of those amino acids into skeletal
muscle cells.1 In contrast, plasma levels of tryptophan
(Trp) are barely affected by insulin since 80±90% of
Trp is bound to albumin.2,3 Thus, insulin increases the
ratio of Trp to the other LNAAs in plasma.4±6 A
competition among all LNAAs, based on their relative
plasma concentrations, governs their transport across
the blood brain barrier. Carbohydrate consumption
therefore ultimately leads to an increase of Trp
Correspondence: Dr AC Toornvliet, Department of General
Internal Medicine, Leiden University Hospital, Building 1, C1R38, P.O. Box 9600, 2300 RC Leiden, The Netherlands.
Received 20 December 1996; revised 6 May 1997; accepted
12 May 1997
uptake into brain tissue.2,3,7 Hydroxylation of Trp is
the rate limiting step in serotonin (5-hydroxytryptamine, 5-HT) synthesis. Since tryptophan hydroxylase
is not normally saturated with its substrate, enhanced
delivery of Trp to the brain stimulates 5-HT synthesis.5
5-HT mediated neurotransmission in the brain is
involved in the regulation of many physiological,
psychological and behavioural processes.8 An impediment of 5-HT mediated neurotransmission in speci®c
brain areas was suggested to lead to a clinical syndrome termed carbohydrate craving (CC) obesity.
This syndrome is characterized by signs of atypical
depression and a concomitant urge to consume
carbohydrate-rich food items, ultimately causing
obesity.9±13 In CC obese patients, the consumption
of carbohydrates is suggested to serve as a kind of
`self-medication': it supposedly alleviates depressive
symptoms by enhancing 5-HT mediated neurotransmission in the brain.14
Wurtman et al15 de®ned CC as an appetite disorder
characterized by a strong urge to consume snacks
Effects of snacks in carbohydrate cravers
AC Toornvliet
containing virtually pure carbohydrate. By de®nition,
CC obese patients consume at least 30% of their total
daily caloric intake and more than 40% of the daily
carbohydrate intake with these snacks. It is important
to note that it concerns food items that are supposed to
contain almost no protein: as little as 4% protein
added to a snack containing pure carbohydrate was
shown to abolish the effect of carbohydrate consumption upon the plasma Trp/LNAA ratio.16
This study was designed to test the hypothesis that
CC obese patients exhibit symptoms of atypical
depression, that their mood and performance are
positively in¯uenced by carbohydrate-rich snacks
and not by protein-rich snacks, and that their psychological responses to the consumption of carbohydrate
are different from those of non-carbohydrate craving
(NC) obese patients. The metabolic response to the
intake of various macronutrients was measured to
establish whether a defect of insulin mediated amino
acid metabolism might underlie a defect of central 5HT mediated neurotransmission in the brain of CC
patients.
Methods
Subjects
Patients were recruited through a local newspaper
advertisement. All patients underwent a medical
screening and completed a three-day food record.
With the exception of obesity, their medical history,
physical examination and biochemical and haematological screening were insigni®cant. Based on the
evaluation of the food records, according to a previously reported method,17 twenty six obese patients
were enrolled. Nine of these patients were CC obese
patients according to Wurtman's de®nition, of whom
one was a male and eight were female patients. Their
average age was 42 10 y, their body mass index
(BMI) 30.2 1.8 kg/m2, and their waist-to-hip circumference ratio (WHR) 0.88 0.09. Seventeen
(one male and 16 females) patients were classi®ed
as NC obese patients. Their average age was
44 10 y, their BMI 32.8 3.5 kg/m2, and their
WHR 0.88 0.10. The two groups did not differ
statistically with respect to these parameters. All
patients had given oral informed consent and the
study protocol was approved by the Committee for
Medical Ethics of the Leiden University Hospital
before it commenced.
Study design
All tests were performed in the follicular phase of the
menstrual cycle. On each test day, the patients had a
freely chosen breakfast at home, which was consumed
before 9.00 am. At noon, they had a standard lunch at
our research unit (Department of General Internal
Medicine out-patient clinic) in order to eliminate
breakfast-induced metabolic differences within each
subject over the three days. The patients were told not
to eat between breakfast and the standard lunch. A
cannula for blood sampling was inserted into an
antecubital vein at 2.30 pm. At 3.15 pm the subjects
consumed one of three different test snacks. Mood and
performance were assessed before and at 35 and
125 min after snack consumption. Blood samples for
serum glucose and insulin analysis were drawn before
and at 30, 60, 90 and 120 min after the snack was
administered. Samples for determination of plasma
amino acid levels were drawn before and 120 min
after snack consumption. The study was double
blinded for group assignment and order of the test
snacks. Test snacks were administered in random
counterbalanced order on the three consecutive test
days. The patients were not familiar with the CC
hypothesis.
Speci®c methods
The standard lunch (450 ml of high-fat chocolate
milk) contained 1707 kJ, of which 57.2% as carbohydrate, 15.8% as protein and 27.0% as fat. The test
snacks were: (1) carbohydrate-rich (1271 kJ, 100% as
carbohydrates), (2) fat/carbohydrate-rich with the
same amount of carbohydrates (1848 kJ, 69.6% as
carbohydrates, 28.8% as fat and 0.9% as protein)
and (3) protein-rich (1267 kJ, 35.6% as carbohydrates,
3.0% as fat and 62.0% as protein) respectively. The
three types of snacks were liquid, isovolemic
( 300 ml) and had a comparable taste and appearance.
Mood was assessed by means of the shortened
Dutch version of the Pro®le of Mood States
(s-POMS).18 The s-POMS is a self-rating scale, consisting of 32 items, measuring ®ve non-overlapping
dimensions of mood: depression, anger, fatigue,
vigour and tension. Raw scores on each dimension
were used for statistical analyses. Subjective effects
on alertness, calmness and mood were evaluated by
using a Dutch translation of the Visual Analogue
Scales (VAS), comprising 16 items.19 Alertness was
measured by a cancellation test (Bourdon-Wiersma
test).20 The patient was requested to cancel the design
of four dots recurring among a total amount of 950
designs not identical in the number of dots. The
patients were urged to work as fast and as accurately
as possible during ®ve minutes. For each patient the
omissions (no response) and commissions (false
response) were counted and the number of errors
(omissions ‡ commissions) was calculated. The
dependent measures accuracy and ef®ciency were
used for statistical analysis. Accuracy was expressed
as: (scanned designs 7 errors)/scanned designs 100%. Ef®ciency was expressed as: (scanned designs
7errors)/total designs 6 100%.
Samples for determination of serum glucose and
insulin levels were collected in non-heparinized tubes
and centrifuged after clotting. The serum was frozen
861
Effects of snacks in carbohydrate cravers
AC Toornvliet
862
at 720 C. The samples for plasma amino acid analysis were collected in EDTA-containing tubes and
were immediately centrifuged. The plasma was then
frozen at 720 C. All samples were stored at 780 C
until assay.
Serum glucose concentrations were measured by
means of the hexokinase method (Boehringer Mannheim, Germany), using a Hitachi 747 automated
analyser. Insulin levels were measured using a radioimmunoassay (Ins-Ria-1001, Medgenix, Brussels,
Belgium) with a sensitivity of 14 pmol/L and an
interassay coef®cient of variation < 6% at various
insulin levels. Time-integrated serum glucose and
insulin responses to the various snacks were computed
as area under the response curve, using the trapezoidal
rule with 0 as baseline value. Plasma for determination of tryptophan (Trp) and other LNAA (tyrosine,
phenylalanine, leucine, isoleucine, valine) concentrations was deproteinized using an equal volume of 5%
(w/v) sulfosalicylic acid in water and analysed by
means of ion-exchange chromatography and ninhydrin derivatization on an LKB 4151 Alpha Plus
automated amino acid analyser using standard conditions.
Statistics
Statistical analyses were performed using the Statistical Package for the Social Sciences (SPSS) for
Windows version 6.1. Differences between groups
over time and effects of the type of snack were
analysed using repeated measures analysis of variance
(ANOVA). For post hoc testing, Student's t-test for
paired and unpaired samples were used. Mean differences and their 95% con®dence intervals are given. A
P-value < 0.05 was considered statistically signi®cant. Data are presented as means standard deviation, unless otherwise speci®ed.
Results
Performance
The ef®ciency and accuracy of performance did not
differ between the three occasions. Therefore the
three-day average of these values was used to compare baseline values of CC and NC obese patients
(Table 1). The ef®ciency of performance improved
signi®cantly in CC and in NC obese patients after
administration of all three snack-types. However, no
group 6 snack 6 time interaction could be revealed,
indicating that both groups did not respond differently
to the various snacks with respect to their ef®ciency
scores. The accuracy scores did not change in CC or
NC patients after any of the snack types.
Mood
Since the values of the various mood parameters
(s-POMS and VAS) before snack consumption (baseline) did not differ between the three occasions, the
average of the three values was used to assess baseline
differences between CC and NC patients. These
average values are presented in Table 1. Anger and
fatigue, both subscales of the s-POMS, were slightly
but signi®cantly higher in CC patients than in NC
patients. The mood score, as assessed by means of the
VAS only tended to be lower in CC patients. Further
analysis of the mood parameters is based on the raw
data, not on the average values. None of the mood
parameters changed signi®cantly in CC or NC obese
patients after any of the snack types.
Metabolic
The baseline serum glucose and insulin concentrations, plasma tryptophan and large neutral amino acid
concentrations and Trp/LNAA ratio did not differ
Table 1 Three-day average of various determinants of mood and performance and of the metabolic parameters as established
before snack consumption in carbohydrate craving obese (CC, n ˆ 9) and non-carbohydrate craving obese (NC, n ˆ 17) patients
Determinant
Performance
Accuracy (0±100%)
Ef®ciency (0±100%)
Mood (s-POMS)
Anger (7±35)
Depression (8±40)
Fatigue (6±30)
Tension (6±30)
Vigour (5±25)
Mood (VAS)
Alertness (0±100)
Calmness (0±100)
Mood (0±100)
Metabolic
Glucose (mmol/L)
Insulin (pmol/L)
Trp (mmol/L)
LNAA (mmol/L)
Trp/LNAA (1073)
CC
NC
Mean diff. (95% CI)
P-value
98.6 0.7
65.1 8.5
98.7 0.8
61.4 9.9
70.1 (70.8 to 0.5)
3.7 (74.4 to 11.8)
0.68
0.36
7.6 1.0
8.6 1.2
8.4 2.4
7.1 2.7
16.2 2.9
7.0 0.1
8.2 0.4
6.6 0.9
7.1 1.3
18.1 3.2
0.6
0.4
1.8
0.0
71.9
0.017
0.16
0.012
1.0
0.15
71.9 18.6
77.1 12.5
76.0 16.3
79.5 10.6
79.8 11.9
85.1 9.2
77.6 (719.6 to 4.4)
72.7 (713.2 to 7.8)
79.1 (719.6 to 1.4)
0.19
0.59
0.079
4.5 0.3
83.2 29.1
43 7
482 53
89 12
4.6 0.4
111.8 58.6
46 9
468 63
98 14
70.1 (70.4 to 0.3)
728.7 (714.5 to 71.8)
74 (711 to 4)
14 (737 to 65)
710 (721 to 2)
0.72
0.18
0.33
0.58
0.10
Mean diff. ˆ mean difference.
95% CI, 95% con®dence interval.
Bracketed values for s-POMS and VAS parameters indicate range of possible values.
(0.1 to 1.1)
(70.2 to 1.0)
(0.4 to 3.2)
(71.6 to 1.6)
(74.6 to 0.8)
Effects of snacks in carbohydrate cravers
AC Toornvliet
Table 2 Time-integrated serum glucose and insulin responses to the three snack-types in carbohydrate craving obese (CC, n ˆ 9) and
non-carbohydrate craving obese (NC, n ˆ 17)
Glucose (mmol/L/120 min)
Carbohydrate
Fat
Protein
Insulin (103 pmol/L/120 min)
Carbohydrate
Fat
Protein
CC
NC
Mean diff. (95% CI)
P-value
953.8 128.0
985.0 102.7
629.5 44.1
1008.7 157.2
1029.1 177.7
635.1 36.6
754.9 (7183.7 to 73.9)
744.1 (7180.4 to 92.2)
75.6 (739.7 to 28.5)
0.38
0.5
0.73
78.5 24.2
70.8 24.3
54.9 18.1
96.8 54.7
105.1 71.7
65.9 50.6
718.3 (759.0 to 22.5)
734.4 (786.8 to 18.0)
711.0 (748.0 to 26.1)
0.35
0.18
0.54
Mean diff. ˆ mean difference.
95% CI, 95% con®dence interval.
Figure 1 Plasma ratios of tryptophan (Trp) to the other large
neutral amino acids (LNAA) before and 120 min after the consumption of three types of snacks in carbohydrate craving (n ˆ 9)
and non-carbohydrate craving (n ˆ 17) obese patients. *Indicates
a statistically signi®cant change from before snack consumption.
No differences between the two groups were observed. Data are
means standard error of the mean.
between the three occasions. Therefore, the average of
the three values was used to compare CC and NC
patients. The results are shown in Table 1. None of
these measures exhibited a statistically signi®cant
difference between CC and NC patients. Further
analyses are based on the data of the separate study
days, not on the three-day averages.
The serum glucose and insulin responses to the
snacks did not differ between the two groups (Table
2). In CC as well as in NC patients, the plasma Trp
concentration increased signi®cantly after the protein
snack (P < 0.0005), whereas it did not change after
the carbohydrate and fat snack (both P ˆ 0.17). The
summed concentrations of the other LNAAs
decreased signi®cantly after carbohydrate and fat
consumption and increased after protein consumption
(all P < 0.0005). Consequently, the Trp/LNAA ratio
decreased signi®cantly after protein consumption and
increased after the carbohydrate and the fat snack
(Figure 1). No differences could be revealed between
CC and NC obese patients.
Discussion
The results of this study show that, to a limited extent,
CC obese patients exhibit more signs of a disturbance
of mood than NC obese patients. Consumption of
snacks containing (virtually) no protein appears not to
improve mood in CC patients, in spite of a snackinduced increase of the plasma Trp/LNAA ratio
suf®cient to enhance brain 5-HT synthesis.21 The
protein-rich snack, decreasing the Trp/LNAA ratio,
did not alter the mood of CC patients in any direction
either. The ef®ciency of performance increased in
both groups after all snacks and is therefore nonspeci®c in nature and should not be attributed to a
5-HT mediated behavioural effect of the snacks. We
did not observe any differences between CC and NC
patients as far as their psychologic and metabolic
responses to the different snacks were concerned.
Thus, although CC obese patients might have a somewhat more disturbed mood than NC control subjects,
the results of this study do not support the concept that
carbohydrate craving serves as a kind of `nutritional
medication' to alleviate mood disturbances through a
5-HT mediated mechanism.
The ®nding that CC patients appear to have a
disturbed mood compared to NC patients is in keeping
with the notion that central 5-HT mediated mechanisms could be involved in the pathogenesis of this
appetite disorder. As 5-HT mediated neurotransmission plays a role in the modulation of both carbohydrate consumption and mood, it is obvious to believe
that a concomitant disturbance of these two clinical
characteristics is related to a defect of central 5-HT
pathways. In fact, disturbances of mood in patients
with an eating disorder and disturbances of
appetite and weight regulation in patients with an
affective disorder have been shown to occur frequently (both in about 70% of the patients).10 The
results of our metabolic studies indicate that the
serotonergic defect, if present, is not caused by a
disturbance of insulin mediated amino acid metabolism.
In a previous study, CC and NC obese patients
responded similarly to treatment with a serotonergic
drug (d-fen¯uramine) with respect to eating behaviour
and weight loss.17 The pharmacological effects of this
drug, just like the effects of consumption of carbohydrates, are identical in CC and NC patients. Thus, it
appears that we presently do not have the clinical
ability to counteract the presumed neurophysiological
disorder in these patients.
863
Effects of snacks in carbohydrate cravers
AC Toornvliet
864
The question arises why we did not observe a
positive effect of high-carbohydrate snacks on mood
in CC obese patients or a negative effect in NC obese
patients, whereas Lieberman and colleagues, in the
only other study on this issue, did.14 One of the
differences with the previous study is that we randomly offered three different types of snacks with a
similar taste and appearance. Lieberman and colleagues offered their subjects only a carbohydrate lunch
in the form of cookies. It is not clear whether their
subjects were aware of expected effects of the lunch.
Furthermore, the phase of the menstrual cycle, during
which classi®cation and testing in women are performed, probably in¯uences the results. Our subjects
completed their food records and they were tested
during the follicular phase of the cycle, since both
food intake and mood are in¯uenced by the phase of
the cycle.22,23 Lieberman and colleagues do not report
on this. If they selected patients in the luteal phase of
the cycle, the patients may just have exhibited a
premenstrual eating pattern. If testing was performed
during the luteal phase, premenstrual symptoms may
have in¯uenced the subjects' mood states as well.
Conclusions
Finally, it should be kept in mind that the existence of
CC patients has never been established beyond any
doubt. When reviewing the results of our studies, it is
still possible that some central serotonergic impediment is responsible for the eating behaviour and the
somewhat disturbed mood in CC patients. However, if
the serotonergic neurotransmission is de®cient in CC
patients, it cannot be corrected by either serotonergic
drug treatment or a carbohydrate-rich snack. The
arguments for this conclusion are twofold. Firstly,
CC patients do not lose more weight during serotonergic drug treatment.17 Secondly, CC patients do not
exhibit the expected mood changes following highcarbohydrate and high-protein snacks. Therefore, it
seems, from a therapeutic point of view, useless to
maintain the concept of carbohydrate craving for
subdividing obese patients.
References
1 Smith K, Rennie MJ. Protein turnover and amino acid metabolism in human skeletal muscle. Baillieres Clin Endocrinol
Metab 1990; 4: 461±498.
2 Yuwiler A, Oldendorf WH, Geller E, Braun L. Effect of
albumin binding and amino acid competition on tryptophan
uptake into brain. J Neurochem 1977; 28: 1015±1023.
3 Pardridge WM. The role of blood-brain barrier transport of
tryptophan and other neutral amino acids in the regulation of
substrate-limited pathways of brain amino acid metabolism.
J Neural Transm Suppl 1979; 15: 43±54.
4 Wurtman RJ, Hefti F, Melamed E. Precursor control of
neurotransmitter synthesis. Pharmacol Rev 1980; 32:
315±335.
5 Fernstrom JD. Dietary effects on brain serotonin synthesis:
relationship to appetite regulation. Am J Clin Nutr 1985; 42:
1072±1082.
6 Pijl H, Koppeschaar HP, Cohen AF, Iestra JA, Schoemaker
HC, FroÈlich M, Onkenhout W, Meinders AE. Evidence for
brain serotonin-mediated control of carbohydrate consumption
in normal weight and obese humans. Int J Obes Relat Metab
Disord 1993; 17: 513±520.
7 Perez Cruet J, Chase TN, Murphy DL. Dietary regulation of
brain tryptophan metabolism by plasma ratio of free tryptophan and neutral amino acids in humans. Nature 1974; 248:
693±695.
8 Pijl H, Meinders AE. Brain serotonin and food selection:
history and current perceptions. J Serotonin Res 1994; 1:
21±45.
9 Wurtman RJ, Wurtman JJ. Carbohydrates and depression. Sci
Am 1989; 260: 68±75.
10 Rosenthal NE, Heffernan MM. Bulimia, carbohydrate craving,
and depression: a central connection? In: Wurtman RJ, Wurtman JJ (eds). Nutrition and the Brain. Raven Press: New York,
1986, pp 139±166.
11 Wurtman JJ. Carbohydrate craving: Relationship between
carbohydrate intake and disorders of mood. Drugs Suppl
1990; 39/3: 49±52.
12 Wurtman RJ, O'Rourke D, Wurtman JJ. Nutrition imbalances
in depressive disorders. Possible brain mechanisms. Ann N Y
Acad Sci 1989; 575: 75±85.
13 Wurtman RJ, Wurtman JJ. Carbohydrate craving, obesity and
brain serotonin. Appetite Suppl 1986; 7: 99±103.
14 Lieberman HR, Wurtman JJ, Chew B. Changes in mood after
carbohydrate consumption among obese individuals. Am J
Clin Nutr 1986; 44: 772±778.
15 Wurtman J, Wurtman R, Berry E, Gleason R, Goldberg H,
McDermott J, Kahne M, Tsay R. Dexfen¯uramine, ¯uoxetine,
and weight loss among female carbohydrate cravers. Neuropsychopharmacology 1993; 9: 201±210.
16 Teff KL, Young SN, Blundell JE. The effect of protein or
carbohydrate breakfasts on subsequent plasma amino acid
levels, satiety and nutrient selection in normal males. Pharmacol Biochem Behav 1989; 34: 829±837.
17 Toornvliet AC, Pijl H, Hopman E, Elte-De Wever BM,
Meinders AE. Serotoninergic drug-induced weight loss in
carbohydrate craving obese patients. Int J Obes Relat Metab
Disord 1996; 20: 917±920.
18 Wald FDM, Mellenbergh GJ. De verkorte versie van de
Nederlandse vertaling van de Pro®le of Mood States
(POMS). Ned Tijdschr Psychol 1990; 45: 86±90.
19 Bond A, Lader M. The use of analogue scales in rating
subjective feelings. Br J Med Psychol 1974; 47: 211±218.
20 De Zeeuw J. Bourdontest (DBT). Anonymous. Algemene
psychodiagnostiek I: Testmethoden. 6th edn. Swets & Zeitlinger B.V.: Lisse, 1986, pp 201±207.
21 Ashley DV. Dietary control of brain 5-hydroxytryptamine
synthesis: implications in the etiology of obesity. Int J Vitam
Nutr Res Suppl 1986; 29: 27±40.
22 Lissner L, Stevens J, Levitsky DA, Rasmussen KM, Strupp
BJ. Variation in energy intake during the menstrual cycle:
implications for food-intake research. Am J Clin Nutr 1988;
48: 956±962.
23 Gallant SJ, Hamilton JA, Popiel DA, Morokoff PJ, Chakraborty PK. Daily moods and symptoms: effects of awareness of
study focus, gender, menstrual-cycle phase, and day of the
week. Health Psychol 1991; 10: 180±189.