journal of dentistry 41 (2013) 949–954

Available online at www.sciencedirect.com

ScienceDirect

journal homepage: www.intl.elsevierhealth.com/journals/jden

Does orthodontic treatment harm children’s diets?

Ama Johal a,*, Feras Abed Al Jawad d, Wagner Marcenes b, Nick Croft c
a
Oral Growth & Development, Institute of Dentistry, UK
b
Dental Public Health, Institute of Dentistry, UK
c
Centre for Adult and Paediatric Gastroenterology, Institute of Cell and Molecular Sciences, Barts and The London
School of Medicine and Dentistry, Queen Mary University of London, Turner Street, London E1 1BB, UK
d
Hamad Medical Corporation, Orthodontic Section, Department of Dentistry, Doha, Qatar

article info abstract

Article history: Introduction: Despite the many courses of fixed orthodontic treatment being undertaken
Received 28 June 2013 worldwide, the question of this treatment harming children’s diets remains unanswered.
Accepted 31 August 2013 Methods: A hospital-based prospective cohort design was adopted to investigate the effects
of treatment on dietary intake and behaviour, body fat (BMI) and fat percentage in 124
patients (41.9% male) aged 11–14 (mean 13.1, SD 0.91) years, consecutively recruited to test
Keywords: and control groups. Both groups completed socio-demographic and food frequency ques-
Orthodontics tionnaires, body mass index (BMI) and body fat percentage measures at baseline and follow-
Diet up. Test patients completed follow-up pain diaries and dietary questionnaires.
Body weight Results: Both groups were comparable at baseline, with a dropout rate of 12.1%. The impact
Fat percentage change on dietary behaviour was significantly higher at 6 weeks compared to 3 months (P < 0.002).
Pain (biological factor), analgesic consumption or professional dietary advice (behaviour
modification) had no influence whilst, a high BMI status at baseline appeared to be the only
significant moderator of change in fat percentage (P < 0.05) and impacts on dietary behav-
iour (P < 0.049) at follow-up.
Conclusions: The findings show no significant detrimental effect on dietary intake or behav-
iour, BMI and fat percentage, during the first 3 months, of orthodontic treatment and may
impart a beneficial/protective effect.
# 2013 Elsevier Ltd. All rights reserved.

been undertaken to explore their impact on dietary intake and

1. Introduction
There is increasing acknowledgement for assessing patients’
expectations and experiences as being central to understand-
ing health needs, patient satisfaction with treatment and the
perceived overall quality of health systems.1,2 Fixed ortho-
dontic treatment is not without side effects: pain from teeth,3,4
ulceration and soreness5 and negative impact on daily living
and quality of life.6
Despite the knowledge of such side-effects in growing
adolescent children, no detailed investigation, to-date, has
behaviour. The relationship between oral health status and
diet is well documented since good oral health is important for
chewing and eating without causing dietary restrictions.7
Awareness of the potential negative consequences of such
side effects forms an essential part of the researchers’
commitment to produce evidence of the quality of care
orthodontists deliver. Understanding patient experiences
during treatment may help shape the process of informed
consent and provide patients with realistic expectations of
what they may experience during the course of treatment and
may improve patient compliance and outcome.8,9

* Corresponding author at: Oral Growth & Development, 5th Floor, Institute of Dentistry, Turner Street, Whitechapel, London, UK.
Tel.: +44 207 377 7686; fax: +44 207 377 7654.
E-mail address: a.s.johal@qmul.ac.uk (A. Johal).
0300-5712/$ – see front matter # 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.jdent.2013.08.025
950 journal of dentistry 41 (2013) 949–954

In orthodontics, the nature of evidence on how fixed 2.2. Statistical analysis

orthodontic treatment may affect dietary intake and
behaviour, body weight and body fat percentage seems to
be inadequate and deficient and warrants investigation. A
recent qualitative study developed a questionnaire to assess
changes in dietary behaviour and intake in response to fixed
appliance treatment, based on a framework analysis.10 All
patients reported varying degrees of pain, with resultant diet
change and an inability to bite and chew. If changes occur,
during this adolescent phase, it will be a significant consider-
ation for patients undergoing orthodontic treatment and may
necessitate special nutritional advice being required.
2. Materials and methods
Ethical approval and Informed consent was obtained for this
hospital-based prospective cohort study. A sample size
calculation, based on a pilot study, determined 51 subjects
were required in each group (test and control) to detect a
0.24 kg/m2 reduction in BMI (alpha of 5% and 80% power). To
allow a loss to follow up of 20%, recruitment was inflated to 62
subjects in each group. Thus, 124 subjects, aged between 11
and 14 years requiring fixed orthodontic treatment, with
similar malocclusion severity (grades 4 and 5 of the dental
health component of the Index of Treatment Need11) were
consecutively recruited from those about to commence
treatment (test) and a group assigned to the waiting list
(control) and followed up for a 3-month period.
Internal consistency of the dietary questionnaire was tested
using Cronbach’s alpha. Test–retest reliability for question-
naires, anthropometric data and VAS scores was assessed by
intra-class correlation coefficients. Paired t tests were
employed to assess anthropometric data for systematic error.
The characteristics of both groups were compared by Chi
square and independent t-tests for categorical and numeri-
cal variables, respectively. Both groups were compared with
respect to the study’s dependent variables (BMI and fat
percentage and dietary intake (FFQ)), between baseline and
follow-up periods. A one-way between group analysis of
covariance was conducted (ANCOVA). Effect sizes were
presented to assess the magnitude of differences between
groups. Independent t-tests (or Mann–Whitney U test) and
ANOVA test (or Kruskal–Wallis) were employed in the
univariate analysis, where appropriate, to select the inde-
pendent explanatory variables (significant at the 0.2 level) to
be entered into the multivariable linear regression model to
explain changes in each outcome variable and dietary
behaviour score. Based on the study’s proposed theoretical
framework, the independent variables were divided into 2
groups: non-related variables (not the focus of the current
study, which included socio-demographic variables and
treatment approach) and related variables, which included
pain levels, consumption of analgesics, BMI status at
baseline and the influence of dietary instructions given by
orthodontists.

2.1. Methods
3. Results
At baseline, subjects in both groups were asked to complete the
Food Frequency Questionnaire (FFQ12) and an assessment of A response and dropout rate of a 96.8% and 12.1% was
socio-economic status. The patients’ height and weight was achieved, respectively. The final sample size comprised
measured to calculate the Body Mass Index (BMI). Body weight
and fat percentage was measured using a digital scale (Tanita
Corp., Tokyo, Japan), based on the bioelectrical impedance
analysis principle.13 Each patient recorded pain intensity, using
a 10-cm visual analogue scale (VAS), from their teeth and from
chewing and biting, over the following 7 days. In addition, the
test group were asked whether they were influenced by dietary
advice given by their orthodontist and to complete a question-
naire to assess the impact of fixed orthodontic treatment on
dietary behaviours and on specific food items.10 Repeat
measures were undertaken after 4–6 weeks and at 3 months.
A theoretical framework based on a combination of two
proposed models (Fig. 1) was adopted to explain the changes in
the main outcome variables: dietary intake and behaviour,
BMI, fat percentage.14,15
Data analysis was conducted on the following variables:
gender; ethnicity; age; socio-economic status; pain levels; use
of analgesics, anthropometric measurements (body weight,
height, BMI and fat percentages); dietary intake (FFQ); dietary
behaviours (questionnaire) and the influence of dietary advice
from orthodontists. The FFQ was used to calculate daily energy
and macro-nutrient intake (carbohydrates, proteins and
fats12). Nutrient and energy intakes were calculated by Fig. 1 – Proposed theoretical framework for explaining the
multiplying the weight of the average portion consumed by observed changes in the outcome variables: dietary
its nutrient and energy content from the UK food tables.16 intake, body mass index (BMI) and body fat %.
 journal of dentistry 41 (2013) 949–954 951

Table 1 – Socio-demographic characteristics of the sample (n = 109).

Test group (N = 53) Control group (N = 56) Overall (N = 109) P value
Age, mean (SD) 13.14 (0.78) 12.91 (0.94) 13.10 (0.91) 0.290

Male, n (%) 25 (47.2%) 19 (33.9%) 44 (40.4%) 0.225

Female, n (%) 28 (52.8%) 37 (66.1%) 65 (59.6%)

White, n (%) 17 (32.1%) 26 (46.4%) 43 (39.4%)

Asian, n (%) 21 (39.6%) 23 (41.1%) 44 (40.4%)
Black, n (%) 11 (20.8%) 5 (8.9%) 16 (14.7%) 0.225
Mixed, n (%) 3 (5.7%) 2 (3.6%) 5 (4.6%)
Other, n (%) 1 (1.9%) 0 (0%0 1 (0.9%)

Family composition
Which adult they live with
Living with both parents n (%) 39 (73.6%) 50 (89.3%) 89 (81.7%)
Only father, n (%) 1 (1.9%) 1 (1.8%) 22 (1.8%) 0.160
Only mother, n (%) 12 (22.6%) 5 (8.9%) 17 (15.6%)
Neither, n (%) 1 (1.9%) 0 (0) 1 (0.9%)
Parents employment
Both employed, n (%) 23 (43%) 30 (53.6%) 53 (48.6%)
Only father, n (%) 12 (22.6%) 11 (19.6%) 23 (21.1%)
Only mother, n (%) 6 (11.3%) 5 (8.9%) 11 (10.1%)
Both not employed, n (%) 12 (22.6%) 10 (17.9%) 22 (20.2%) 0.763
Crowding
Yes, n (%) 8 (15.1%) 11 (19.6%) 19 (17.4%)
No, n (%) 45 (84.9%) 45 (80.4%) 90 (82.6%) 0.532
Car ownership
Own more than 2 cars, n (%) 17 (32.1%) 24 (42.9%) 41 (37.6%) 0.507
One car only, n (%) 24 (45.3%) 21 (37.5%) 45 (41.3%)
No cars, n (%) 12 (22.6%) 11 (19.6%) 23 (21.1%)
Home ownership
Own home, n (%) 25 (47.2%) 40 (71.4%) 65 (59.6%) 0.03
Rent home, n (%) 22 (41.5%) 14 (25%) 36 (33%)
Don’t know, n (%) 6 (11.3%) 2 (3.6%) 8 (7.3%)
Access to Internet
Yes, n (%) 53 (100%) 56 (100%) 109 (100%) 0.999
No, n (%) 0 (0%) 0 (0%) 0 (0%)

Total 53 56 109

109 patients (53 test and 56 control). Sixty-five (59.6%) patients
were female, with the mean age of the sample 13.1 (SD 0.91)
years (Table 1). Asians (40.4%) and Whites (39.4%) were the
predominant ethnic backgrounds. Patients lived with both
parents (82%), who were employed (80%) and lived in non-
crowded houses (83%), where the parents owned one or more
cars (80%), lived in owned houses (60%) and all patients had
internet access. The test and control groups were comparable
at baseline.
3.1. Reliability of scales
Internal consistency for the dietary questionnaire was 0.77.
Intra-class reliability coefficients for the FFQ height, weight,
fat percentage and VAS scores were 0.96, 0.98, 1, 0.97 and 1,
respectively. Paired t tests showed no evidence of systematic
effects.
3.2. Changes in outcome variables between groups at
follow-up periods:
3.2.1. Anthropometric measurements
With respect to fat percentage change, there was significant
difference between both groups at baseline and follow-up
periods (P < 0.001), however, the corresponding size effects
were low (0.14 and 0.2, respectively). Fat percentage decreased
in the test group significantly (P < 0.001) during the first follow-
up ( 2.4%), with little subsequent change ( 0.3) and increased
in the control group (Tables 2 and 3). However, after
controlling for BMI status at baseline (normal/overweight or
obese), the difference was insignificant (P < 0.156), identifying
BMI status at baseline as a potential confounder.
3.2.2. Energy and macro-nutrient intakes (carbohydrates,
protein and fat)
Tables 2 and 3 show at follow-up, there was no significant
difference between the groups with respect to energy and
macronutrient intakes, although there was a greater reduction
in dietary intake in the test group.
3.2.3. Dietary behaviour in the test group
There was significant difference in dietary behaviour scores,
obtained from the questionnaire, in the test group, at follow-
up (P < 0.002). Two-thirds of patients (66%) agreed that pain
had caused them difficulty in eating and/or chewing during
the first period, which dropped in the second period to 29
(54.7%). Twenty-nine patients (54.7%) agreed that they ate less
snacks compared to before treatment. However, this number
952 journal of dentistry 41 (2013) 949–954

Table 2 – Changes in BMI and fat %, energy and macronutrient intakes in both groups during the first 4–6-week follow-up
period (test group n = 53 and control group n = 56). P value obtained from ANCOVA test to assess differences between
groups adjusted for baseline measurements.
Variable 4–6-Week follow-up

Test Control Effect size P value

BMI 0.03 0.25 0.020 0.147
Fat % 2.4 0.4 0.140 0.001
Energy intake 442.2 304.7 0.002 0.670
Carbohydrate intake 124.0 110.0 0.001 0.763
Protein intake 36.0 33.3 0.006 0.442
Fat intake 35.9 31.1 0.001 0.788

Table 3 – Changes in BMI and fat %, energy and macronutrient intakes in both groups during the second 3-month follow-
up period (test group n = 53 and control group n = 56). P value obtained from ANCOVA test to assess differences between
groups adjusted for baseline measurements.
Variable 4–6-Week follow-up

Test Control Effect size P value

BMI 0.01 0.36 0.020 0.122
Fat % 2.7 1.1 0.200 0.001
Energy intake 510.5 457.7 0.001 0.743
Carbohydrate intake 120.6 100.0 0.002 0.644
Protein intake 24.0 23.2 0.010 0.204
Fat intake 28.7 27.7 0.001 0.755

decreased to 21 (39.6%) in the second period. Almost half of the
patients (49.1%) agreed that they had to cut their food into
pieces or cooked in a different way during the first period and
41.5% in the second period. In both follow-up periods, the
majority of patients agreed that they ate less sticky and hard
food types as a result of their braces (79.3%) or because they
were advised by their orthodontist. Two-thirds of patients
(64.2%) in the first period and more than half (56.6%) in the
second period disagreed that the braces resulted in them
eating less healthy.
3.2.4. Pain levels in the test group
Pain from the teeth and on biting and chewing declined
significantly on days 3 and 2 in the first and second follow-up
periods, respectively when compared to baseline (P < 0.001).
This decline continued, indicating that there was an adapta-
tion to pain.
3.3. Multivariable analysis of changes in fat percentage
and dietary behaviour scores with independent variables
Uni-variate analysis demonstrated that pain from chewing
and biting (P < 0.106) and BMI status at baseline (P < 0.107)
were the only statistically significant independent variables
that were entered into the final multiple regression model,
explaining 12% of the variance in change of fat percentage
(R2 = 0.12). The overall significance of the model was P < 0.04.
BMI at baseline had a stronger contribution than pain with
beta coefficient of 0.27 and 0.19, respectively.
Uni-variate analysis showed that pain from chewing and
biting, analgesics consumption, dietary instructions and BMI
status at baseline were the only independent variables that
were statistically significant at the 0.2 level (P < 0.049,
P < 0.049, P < 0.062 and P < 0.025, respectively) and were
entered into the final multiple regression model. This
explained 47% of the variance in dietary behaviour scores
(R2 = 0.47), of which BMI status at baseline remained signifi-
cant (P = 0.049) explaining 26% of the variance.
Thus, it appears BMI status at baseline was an important
predictor for change in fat percentage and dietary behaviour
scores, with overweight/obese patients showing greatest
reduction in fat percentage and more impact on their dietary
behaviour due to treatment.
4. Discussion
Whilst fixed appliance therapy is known to achieve optimal
dental correction of malocclusion, further patient-centred
research is needed to aid in our understanding of its impact on
dietary intake and behaviour, body fat and fat percentage.
Research in this area to-date is scarce and limited by the
recruitment of ill-defined samples; unclear methodological
design; a lack of control groups and invalid dietary assessment
techniques.17,18 The present research addressed these short-
comings through a combination of qualitative10 and quantita-
tive research methods.
4.1. Changes in BMI and fat percentage
Bioelectrical impedance analysis provided a more valid
method of assessing shifts in fat distribution, than BMI
alone.19 Both BMI and fat percentage decreased during follow
up in the test group and increased in the control group, with a
significant difference in fat percentage changes (P < 0.001).
The latter was not a surprising finding, reflecting normal
physiological changes that take place during adolescence.20
Baseline BMI was the strongest predictor of change in fat
percentage (P < 0.05), with patients who were overweight or
obese more likely to lose fat than ‘normal’ weight patients.
 journal of dentistry 41 (2013) 949–954
Indeed, this finding is consistent with studies that introduced
intervention programmes to prevent obesity in adoles-
cents12,21 and could be linked indirectly to the findings of
the qualitative study that showed patients thought their
eating habits were healthier.10
4.2. Changes in energy and macro-nutrient intake
Patients in the test group were observed to reduce energy and
macro-nutrient (carbohydrates, protein and fat) intake during
the first follow-up period compared to the control group, after
which patients started to resume normal physiological
behaviour. This is consistent with the medical literature, in
which lowering dietary intake will result in a series of
physiological responses, of which the principal response is
a reduction in body weight and reduced muscle mass and fat
stores.22 However, this change was not statistically significant
and such changes could be attributed to other factors: The
Food Frequency Questionnaire (FFQ) provides a validated
measure to assess dietary intake in adolescents12 but does not
retrieve unique details of the individual’s diet and respon-
dents can misreport their intake.23 In children, an additional
difficulty is their cognitive ability to remember their diets.24
This might explain the decrease in energy and macro-nutrient
intake in patients in the control group at 4–6 weeks compared
to their intake at baseline, although the decrease was less than
that observed in the test group. Adolescent dietary behaviour
is characterized by having irregular meals, snacking, peer
influence and overweight subjects under-reporting their
intake.25 Seasonal variation might also influence dietary
intake, however this was minimized by recruiting patients
to both groups concurrently, during both the winter and spring
seasons. Despite the aforementioned limitations of FFQ in
estimating energy and macro-nutrient intake, it was the most
appropriate, convenient and practical method to be used in
the sample being recruited when compared to other self-
reported methods such as weighed or estimated records.
Energy and macro-nutrient intake changes estimated using
the FFQ in the present study appeared to follow the same trend
of changes observed with the objective measures (BMI and fat
percentage) being applied, although the correlations were
weak.
4.3. Interaction of the study’s independent variables with
outcome variables based on the proposed theoretical
framework (Fig. 1)
Socio-economic status and material resources can affect food
choice at both a society and an individual level, the higher the
social class and income, the healthier the diet.26,27 However,
no such influence was identified, with BMI status at baseline
the only explanatory environmental variable that moderated
changes in fat percentage and dietary behaviour scores.
Furthermore, this study does not support the assumption
that pain (biological factor) or analgesic consumption (behav-
iour modification) during treatment was a strong predictor of
changes in fat percentage and dietary behaviour scores during
the study. An explanation of this could be that the highest
intensity of pain occurred during the first few days, after
which it started to decline and patients resumed their normal
953
daily dietary behaviour. In addition, dietary advice (behaviour
modification) given to the patient did not show that this
variable influenced changes in fat percentage and dietary
behaviour scores significantly in the final regression analysis.
This might be due to the fact that patients shifted to other food
items of softer consistency, or changed the method and the
manner of eating and/or the preparation of food. Finally,
according to Wilson and Cleary,15 personality factors such as
sense of coherence, health locus control and self-esteem
might have mediated the effects of the proposed independent
variables on the study’s outcome variables, although their
effects were not tested in the present study.
Fixed appliance treatment does not significantly affect
energy or macro-nutrient intake, body mass index or body fat
percentage. As such orthodontic treatment does not harm a
child’s diet.
Acknowledgements
This research received no specific grant from any funding
agency in the public, commercial, or not-for-profit sectors. The
authors declare no conflicts of interest with respect to the
authorship and/or publication of this article.
references
1. Locker D. Oral health and quality of life. Oral Health and
Preventative Dentistry 2004;2:247–53.
2. Newsome PR, McGrath C. Patient-centred measures in
dental practice: 1. An overview. Dental Update 2006;33:
596–8. 600.
3. Firestone A, Scheurer P, Burgen W. Patient’s anticipation
of pain and pain-related side effects and their perception
of pain as a result of orthodontic treatment with
fixed appliances. European Journal of Orthodontics 1999;21:
387–96.
4. Bartlet B, Firestone A, Vig K, Beck M, Marucha P. The
influence of a structerd telephone call on orthodontic pain
and anxiety. American Journal of Orthodontics and Dentofacial
Orthopedics 2005;128:435–41.
5. Sinclair PM, Cannito MF, Goats LJ, Solomos LF, Alexander
CM. Patients responses to lingual appliances. Journal of
Clinical Orthodontics 1986;20:396–404.
6. Mandall NA, Vine S, Hulland R, Worthington H. The impact
of fixed orthodontic appliances on daily life. Community
Dental Health 2006;23:69–74.
7. Acs G, Lodolini G, Kaminsky S, Cisneros G. Effect of nursing
caries on body weight in a pediatric population. Pediatric
Dentistry 1992;149:302–5.
8. Sergl H, Klages U, Zentner A. Functional and social
discomfort during orthodontic treatment – effects on
compliance and prediction of patients’ adaptation by
personality variables. European Journal of Orthodontics
2000;22:307–15.
9. Chen M, Wang DW, Wu LP. Fixed orthodontic
appliance therapy and its impact on oral health-related
quality of life in Chinese patients. Angle Orthodontist
2010;80:49–53.
10. Al Jawad F, Croft N, Cunningham S, Johal A. A qualitative
study of the effects of fixed orthodontic treatment on
dietary intake and behaviours. European Journal of
Orthodontics 2011. May 13 [Epub ahead of print].
954 journal of dentistry 41 (2013) 949–954
11. Brook PH, Shaw WC. The development of an index for
orthodontic treatment priority. European Journal of
Orthodontics 1980;11:309–32.
12. Robinson S, Skelton R, Barker M, Wilman C. Assessing the
diet of adolescent girls in the UK. Public Health and Nutrition
1999;2:571–7.
13. Jebb S, Cole T, Doman D, Murgatroyd P, Prentice A.
Evaluation of the novel Tanita body-fat analyser to measure
body composition by comparison with a four-compartment
model. British Journal of Nutrition 2000;83:115–22.
14. Khan MA. Evaluation of food selection patterns and
preferences. Critical Review of Food Science and Nutrition
1981;15:129–53.
15. Wilson IB, Cleary PD. Linking clinical variables with
health-related quality of life. A conceptual model of
patient outcomes. Journal of American Medical Association
1995;273:59–65.
16. Food Standards Agency. McCance and Widdowson’s The
composition of foods. Sixth summary edition. Cambridge:
Royal Society of Chemistry; 2002.
17. Cheraskin E, Ringsdorf W. Biology of the orthodontic
patient: II. Lingual Vitamine C test scores. Angle Orthodontist
1969;39:324–5.
18. Riordan D. Effects of orthodontic treatment on nutrient
intake. American Journal of Orthodontics and Dentofacial
Orthopedics 1997;111:554–61.
19. Reilly JJ, Dorosty AR, Emmett PM. Identification of the obese
child: adequacy of the body mass index for clinical practice
and epidemiology. International Journal of Obesity
2000;24:1623–7.
20. Rogol AD, Clark PA, Roemmich JN. Growth and pubertal
development in children and adolescents: effects of diet and
physical activity. American Journal of Clinical Nutrition
2000;72:521S–8S.
21. Ebbeling CB, Feldman HA, Osganian SK, Chomitz VR,
Ellenbogen SJ, Ludwig DS. Effects of decreasing sugar-
sweetened beverage consumption on body weight in
adolescents: a randomized. controlled pilot study. Pediatrics
2006;117:673–80.
22. Shetty P. Adaptation to low energy intakes: the
responses and limits to low intakes in infants, children
and adults. European Journal of Clinical Nutrition 1999;53:
14–33.
23. Kristal AR, Peters U, Potter JD. Is it time to abandon the food
frequency questionnaire? Cancer Epidemiology Biomarkers and
Prevention 2005;14:2826–8.
24. Rockett HR, Colditz GA. Assessing diets of children and
adolescents. American Journal of Clinical Nutrition
1997;65:1116S–22S.
25. Samuelson G. Dietary habits and nutritional status in
adolescents over Europe. An overview of current studies in
the Nordic countries. European Journal of Clinical Nutrition
2000;54:S21–8.
26. Friel S, Kelleher CC, Nolan G, Harrington J. Social diversity of
Irish adults nutritional intake. European Journal of Clinical
Nutrition 2003;57:865–75.
27. Giskes K, Lenthe Fv F, Brug HJ, Mackenbach J. Dietary
intakes of adults in the Netherlands by childhood and
adulthood socioeconomic position. European Journal of
Clinical Nutrition 2004;58:871–80.