Content uploaded by Andrew Gray
Author content
All content in this area was uploaded by Andrew Gray on Jun 29, 2016
Content may be subject to copyright.
Public Health Nutrition: 15(12), 2272–2279 doi:10.1017/S1368980012000882
Eating in response to hunger and satiety signals is related to
BMI in a nationwide sample of 1601 mid-age New Zealand
women
Clara EL Madden
1
, Sook Ling Leong
1
, Andrew Gray
2
and Caroline C Horwath
1,
*
1
Department of Human Nutrition, University of Otago, Dunedin 9054, New Zealand:
2
Department of Preventive
and Social Medicine, University of Otago, Dunedin, New Zealand
Submitted 30 May 2011: Final revision received 6 February 2012: Accepted 10 February 2012: First published online 23 March 2012
Abstract
Objective: To examine the association between eating in response to hunger and
satiety signals (intuitive eating) and BMI. A second objective was to determine
whether the hypothesized higher BMI in less intuitive eaters could be explained
by the intake of specific foods, speed of eating or binge eating.
Design: Cross-sectional survey. Participants were randomly selected from a
nationally representative sampling frame. Eating in response to hunger and satiety
signals (termed ‘intuitive eating’), self-reported height and weight, frequency of
binge eating, speed of eating and usual intakes of fruits, vegetables and selected
high-fat and/or high-sugar foods were measured.
Setting: Nationwide study, New Zealand.
Subjects: Women (n2500) aged 40–50 years randomly selected from New Zealand
electoral rolls, including Ma
¯ori rolls (66 % response rate; n1601).
Results: Intuitive Eating Scale (IES) scores were significantly associated with BMI
in an inverse direction, after adjusting for potential confounding variables. When
controlling for confounding variables, as well as potential mediators, the inverse
association between intuitive eating (potential range of IES score: 21–105) and
BMI was only slightly attenuated and remained statistically significant (5?1%
decrease in BMI for every 10-unit increase in intuitive eating; 95 % CI 4?2, 6?1%;
P,0?0 0 1). The relationship between intuitive eating and BMI was partially
mediated by frequency of binge eating.
Conclusions: Eating in response to hunger and satiety signals is strongly asso-
ciated with lower BMI in mid-age women. The direction of causality needs to be
investigated in longitudinal studies and randomized controlled trials.
Keywords
Hunger and satiety
Intuitive eating
Mid-age
Women
BMI
Obesity
Traditional obesity treatments based on energy-restrictive
diets show poor long-term success and weight regain
is common
(1)
. There is evidence from both prospective
studies and randomized trials that dieting or restrictive
eating can promote increased food preoccupation
(2,3)
,loss
of control
(2,4)
and overeating
(2,3)
. Some studies suggest that
dieting or weight-loss attempts may be associated with
subsequent weight gain
(5,6)
.
In response to the poor long-term outcomes of dieting
treatments, eating in response to hunger and satiety signals
(termed ‘intuitive eating’) has been promoted as an alter-
native to deliberate dietary restriction and the feeling of
deprivation that often accompanies it. Intuitive eating has
been defined as ‘trust in and connection with physiological
hunger and satiety cues and eating in response to these
cues’
(7)
. Intuitive eating is a key recommendation of the
non-dieting
(8)
andHealthAtEverySizeapproaches
(9)
.
These approaches also advocate a shift in focus away from
body weight to the improvement of health behaviours and
psychological well-being. Important guiding principles of
the Health At Every Size approach also include: accepting
and respecting the diversity of body shapes and sizes;
promoting eating in a manner that balances individual
nutritional needs, hunger, satiety, appetite and pleasure;
promoting enjoyable, life-enhancing activity, rather than
exercise focused on achieving weight loss; and promoting
all aspects of health and well-being for people of all sizes
(9)
.
Intuitive eating has been shown to be negatively related
to eating disorder symptomatology, body dissatisfaction
and internalization of the thin ideal, and positively related
to well-being
(10)
.
Recent randomized trials in treatment-seeking obese
women have shown that training in intuitive eating (eating
in response to hunger and satiety signals) can prevent
*Corresponding author: Email caroline.matthaei@otago.ac.nz rThe Authors 2012
weight gain over 2 years
(9,11)
. Among college women,
intuitive eating has been associated with lower BMI
(10,12)
,
lower TAG levels, higher levels of HDL cholesterol and
decreased cardiovascular risk
(12)
.
To our knowledge, only two previous studies have
examined the relationship between intuitive eating and
BMI in adult women
(7,10)
, and no such studies have
been undertaken in representative nationwide samples.
Mid-age is a time when women are at particular risk of
weight gain and obesity
(13)
. Perhaps in response to this,
mid-aged women have been reported to rely to a lesser
extent than younger women on their internal hunger and
satiety cues to guide eating and to give themselves less
permission to eat when hungry
(7)
. Thus, the aim of the
present study was to test the hypothesis that lower levels
of intuitive eating are associated with higher BMI in a
representative sample of 40–50-year-old women. The
study also aimed to determine whether the hypothesized
higher BMI in less intuitive eaters could be explained by
the intake of specific foods, speed of eating or frequency
of binge eating (Fig. 1).
Experimental methods
Design and sample
A nationwide, self-administered mail survey of 2500
New Zealand women aged 40–50 years, randomly selected
from the New Zealand electoral rolls (listings of New
Zealand residents eligible to vote in elections including
both the General electoral rolls and the Ma
¯ori electoral
rolls), was conducted in May 2009. The New Zealand
electoral rolls contain up-to-date mailing details from
approximately 97 % of the estimated eligible population
of people aged 40–49 living in New Zealand
(14)
. The
twenty-one-page questionnaire examined factors poten-
tially influencing eating behaviour and weight, and took
30–40 min to complete. Prior to the main survey, thirty-six
Dunedin, New Zealand, women in the target age range
were asked to provide detailed feedback on the ques-
tionnaire booklet regarding its layout, appearance,
instructions, clarity, ease of completion and interpretation
of questions. Of these thirty-six women, seven identified
as Ma
¯ori and five were of Pacific origin. In response to
their comments, improvements were made to the layout
of the booklet and clarity of instructions. A pilot survey of
100 women selected randomly from the electoral rolls
was also undertaken. In the pilot study, a complimentary
pen was mailed with the survey questionnaire, and up to
three reminders were provided to non-respondents. The
pilot study resulted in a 56 % response rate. Based on
the pilot study, the main survey was refined to include
financial incentives in order to improve the response rate,
and to change the final reminder from a telephone call to
a postcard in order to reach more participants. The main
survey consisted of up to four contacts with participants:
the first mail-out of the questionnaire, a thank you/
reminder postcard, and for non-respondents, a replace-
ment questionnaire and a final reminder postcard. The
mail survey procedures were modified from Dillman’s
Tailored Design Method
(15)
. Incentives included a com-
plimentary pen and an individually wrapped teabag in
the initial mail-out, and a prize draw to win one of
fourteen $NZ 100 (approximately £47) and three $NZ 200
(approximately £95) cash prizes. In addition, a random
sample of 400 women also received a $NZ 5 (approxi-
mately £2?40) unconditional incentive in the first mail-out.
Respondents were omitted from the sample if they did
not meet inclusion criteria (women aged 40–50 years) or
if they stated that they were pregnant or breast-feeding at
the time of the survey. To minimize data errors, all data
were entered twice and any inconsistencies corrected.
Ethical approval was granted by the University of Otago
Ethics Committee.
Measures
Intuitive eating was assessed using the twenty-one-item
Intuitive Eating Scale (IES)
(10)
. This scale has been
validated using data collected in four studies from 1260
women (17–61 years), with exploratory and confirmatory
factor analysis revealing a three-factor model, and with
evidence presented of internal consistency, stability over
3 weeks and construct validity. The scale measures
three factors: (i) reliance upon internal hunger and satiety
cues (e.g. ‘I stop eating when I feel full (not over-
stuffed)’); and two factors for which the majority of items
are reverse-scored, (ii) eating for physical rather than
emotional reasons (e.g. ‘I find myself eating when I
am lonely, even when I’m not physically hungry’) and
(iii) unconditional permission to eat when hungry and
what food is desired (e.g. ‘I have forbidden foods that
I don’t allow myself to eat’). Item responses were rated on
a 5-point scale ranging from 1 5‘strongly disagree’ to
55‘strongly agree’. The possible range for total IES score
was 21–105, with a higher total score corresponding to
more intuitive eating.
Frequency of binge eating was assessed by one ques-
tion, which asked how often the respondent had engaged
Intake of specific foods, frequency of
binge eating, speed of eating
Intuitive
eating
(IES)
BMI
Figure 1 Proposed model of the relationship between intuitive
eating, BMI and hypothesized mediators (intake of specific
foods, frequency of binge eating and speed of eating; IES,
Intuitive Eating Scale)
Intuitive eating and BMI in adult women 2273
in episodes of binge eating, defined as ‘eating an un-
usually large amount of food in one go and at the time
feeling that your eating was out of control (that is, you
could not prevent yourself from overeating, or that
you could not stop eating once you had started)’
(16)
.
Responses were rated on a 4-point scale including
15‘not at all’, 2 5‘less than weekly’, 3 5‘once a week’
and 4 5‘two or more times a week’. The time period
referred to was modified from that in Hay et al.’s study
(past 3 months)
(16)
to the past 12 months.
The intake of specific foods was assessed using ques-
tions from the New Zealand National Nutrition Survey
1997
(17)
. Questions asked about the usual number of
servings of fruits and vegetables consumed daily. Partici-
pants were also asked to estimate how often they usually
ate each of five food types high in fats and/or sugars
(chocolate-coated and/or cream-filled biscuits; potato
crisps, corn snacks or corn chips; cakes, scones, muffins
or sweet buns; meat pies or sausage rolls; and burgers)
using an 8 point-scale ranging from 1 5‘never’ to 8 5‘two
or more times a day’.
Self-reported rate of eating was assessed by a single
item, adapted from a study by Otsuka et al.: ‘How would
you describe your usual rate of eating?’
(18)
. A second item
examined the usual time spent eating the main meal of
the day.
Habitual levels of physical activity were measured using
the Rapid Assessment of Physical Activity (RAPA)
(19)
.The
RAPA provides participants with descriptions, several
examples plus diagrams of light, moderate and vigorous
activities. In the present study the diagrams were modified
to include only pictures of women and not men. Using
the diagrams and examples, participants were asked to
respond ‘yes’ or ‘no’ to items regarding their usual physical
activity: 1 5‘I rarely or never do any physical activities’,
25‘I do some light activity every week’, 3 5‘I do at least
30 min of moderate physical activity on 5 or more days
per week’, 4 5‘I do at least 20 min of vigorous physical
activity on 3 or more days per week’ and 5 5‘I lift weights
or do other muscle strengthening activities at least once a
week’
(19)
. The RAPA questionnaire has been shown in a
sample of US adults to be more highly correlated with
energy expenditure than two other self-report physical
activity questionnaires
(19)
.
Menopausal status was determined using questions from
the Australian Longitudinal Study of Women’s Health
(20)
.
Participants were asked to indicate whether or not
they had ever been diagnosed with any thyroid problems.
Self-reported current weight and height were used to
calculate BMI. Demographic questions were taken from
the 2006 New Zealand Census
(21)
. Participants’ age, highest
level of education, employment status, usual occupation
(open-ended question requesting description of their usual
job, where applicable), living circumstances/marital status,
smoking status and ethnicity were collected. In response to
the question ‘Which ethnic group(s) do you belong to?’,
women were asked to indicate from a list of nine ethnicities
all those that applied to them
(21)
. Since multiple ethnicities
could be provided (e.g. New Zealand European and Ma
¯ori),
in categorizing each participant into one ethnic group,
prioritized ethnicity was used. For example, if one of
the ethnicities selected was Ma
¯ori, this was given priority
and the participant was categorized as Ma
¯ori. Ethni-
cities were prioritized as follows: first Ma
¯ori, then Pacific
People, then Asian ethnicities, other ethnicities and finally
New Zealand European. Occupational status was coded
using the New Zealand Standard Classification of Occupa-
tions 1999
(22)
. Where occupation was not provided or was
unable to be classified, occupation was obtained from the
electoral roll. Participants were assigned a 3-digit code,
broadly falling into eight overall categories of occupations.
From these 3-digit occupation codes, socio-economic status
(SES) was estimated using the New Zealand Socioeconomic
Index 1996
(23)
where a score between 10 and 90 is genera-
ted. This was then used as a continuous SES score. If a
participant responded to a questiononspouse/partner
occupation, the higher of the two scores was used in all
analyses as an estimate of household SES.
Data analysis
Linear regression models were developed to examine the
univariate associations between demographic, health and
behavioural variables and BMI. The reference categories
for the categorical variables were: smoking status 5never,
menopause status 5premenopause, thyroid condition 5
never diagnosed, prioritized ethnicity 5New Zealand
European and physical activity 5none. All subsequent
multivariate regression models controlled for these
possible confounding variables with the addition of age
and SES. Log transformations of dependent variables
were investigated where there were issues with residual
diagnostics (normality and/or homogeneity). Variance
inflation factors were calculated to check for excessive
collinearity. Adjusted and unadjusted results are reported
as a percentage change when data were log-transformed
or as bcoefficients otherwise. Corresponding 95 % CI and
Pvalues were reported. When the two-sided Pvalue
was ,0?05, the relationship was considered statistically
significant. If a particular participant had missing items,
total IES scores and subscale scores were imputed using
the mean item score providing that 80 % or more of the
items in the scale or subscale were answered.
Regression modelling was used to assess the relation-
ship between intuitive eating (total score and subscale
scores) and BMI. It was also used to assess the relationship
between intuitive eating and the intake of specific foods
(fruits, vegetables and several high-fat/sugar foods), fre-
quency of binge eating and speed of eating. The four-step
approach of Baron and Kenny was used to explore pos-
sible mediators
(24)
. As part of this approach, the regression
of each of the above-mentioned variables v.BMI,andof
the combined effect of all these variables (intake of specific
2274 CEL Madden et al.
foods, frequency of binge eating and speed of eating) v.
BMI, was performed (Fig. 1). After establishing associ-
ations between intuitive eating and these potential medi-
ators and between intuitive eating and BMI, the final
multivariate model examined the relationship between
intuitive eating and BMI while controlling for selected
potential mediators. These potential mediators were those
variables that were statistically significantly associated with
both IES scores and BMI (frequency of binge eating and
speed of eating). The STATA statistical software package
version 10?0 (2007; StataCorp, College Station, TX, USA)
was used for all statistical analysis.
Results
Response rate and sample characteristics
Of the 2500 questionnaires mailed to potential participants,
1627 completed questionnaires were returned and forty-
seven questionnaires were returned as ‘non-deliveries’.
Twenty respondents were omitted from the sample
because they did not meet inclusion criteria. Six completed
questionnaires were excluded because there was reason to
doubt the reliability of answers (respondent indicated a
poor understanding of English, geometric patterns were
made by circling answers, respondent simultaneously
answered opposite ends of a scale or the questionnaire was
answered on behalf of someone else). The final response
rate wa s 66 % ( (1607 26)/(2500 247 220)).
Thesamplehadameanageof45?5(
SD 3?2) years, a mean
weight of 70?1(1?2) kg and mean BMI of 25?8(
SD 1?2) kg/m
2
.
Other respondent characteristics are reported in Table 1.
The sample was comparable to the general New Zealand
population in terms of SES and prioritized ethnicity, but
included a higher proportion of university educated and a
lower proportion of obese women
(21,23,25)
.
Table 2 describes the relationship between health,
demographic and behavioural variables and BMI. Res-
pondents with higher SES had a statistically significantly
lower BMI (P,0?001). Overall, there was a statistically
significant difference in BMI among the physical activity
categories (P,0?001) with a lower BMI among those
more active, and among the prioritized ethnic groups
(P,0?001). Compared with New Zealand Europeans,
Ma
¯ori and Pacific People had a higher BMI and Asian
women had a lower BMI. BMI was significantly higher for
participants who were postmenopausal compared with
those who were premenopausal (P50?016). Women
reporting a thyroid condition had a higher BMI than those
without thyroid problems (P50?005).
Regression and mediation outcomes
The IES displayed good internal consistency (Cronbach’s
a50?86). Total IES scores and all three of the IES sub-
scale scores were significantly associated with BMI in an
inverse direction after adjusting for confounding variables
(age, smoking status, menopause status, thyroid condi-
tion, prioritized ethnicity, physical activity and SES;
P,0?001). For every 10-unit increase in total IES score
(potential range: 21–105), there was a decrease in BMI
of 6?5 % (95 % CI 27?4, 25?6%; P,0?001). A 10-unit
Table 1 Demographic and behavioural characteristics of New Zealand female participants compared with national data
Characteristic (n, number of
respondents) n%* National data (%)-
BMI classification (n1581)
Underweight (,18?5 kg/m
2
)271?8
Normal range (18?5–24?9 kg/m
2
) 732 48?1
Overweight (25?0–29?9 kg/m
2
) 448 29?4
Obese ($30?0 kg/m
2
) 315 20?728?3
Prioritized ethnicity (n1594)
European and others-
-
1280 80?373?4
Ma
¯ori 181 11?412?1
Pacific People 48 3?04?6
Asian 85 5?39?9
Highest educational level attained (n1590)
Primary and some secondary school 489 30?850?2
Completed secondary school 153 9?68?9
Technical/trade school or polytechnic 438 27?623?2
University 510 32?117?7
SES (NZSEI) category (n1597)y
10–29 234 14?717?5
30–59 1065 66?760?0
60–90 298 18?722?5
SES, socio-economic status; NZSEI, New Zealand Socioeconomic Index 1996.
*Percentages may not add to 100 % due to rounding.
-Population estimates for rates of obesity in mid-age women from the New Zealand Health Survey 2006/07; prioritized ethnicity
and education level in mid-age women from the 2006 New Zealand Census; and total population NZSEI distribution from the 1991
New Zealand Census.
-
-
In our sample ‘others’ made up 13?1% (n209).
yIn NZSEI, 10 represents the lowest socio-economic group and 90 represents the highest socio-economic group. This is based on a
standard New Zealand classification of occupations, and is used as a continuous variable in all analyses.
Intuitive eating and BMI in adult women 2275
increase in total IES score in a woman with a BMI of
25?8 kg/m
2
and a height of 1?65 m (i.e. a woman typical of
our sample) would equate to a reduction of 1?63 kg/m
2
in
BMI, which is equivalent to 4?4 kg. A 5-unit increase in the
Unconditional Permission to Eat subscale (potential
range: 9–45), the Eating for Physical Reasons subscale
(potential range: 6–30) and the Reliance on Hunger and
Satiety Cues subscale (potential range: 6–30) corres-
ponded to a decrease in BMI of 3?21 % (95 % CI 24?07,
22?35 %; P,0?001), 6?62 % (95 % CI 27?49, 25?73 %;
P,0?001) and 8?25 % (95 % CI 29?49, 26?98 %;
P,0?001), respectively.
Although the cross-sectional nature of the data means
that causality cannot be ascertained, the study aimed to
test whether data were consistent with a model where
the higher BMI in less intuitive eaters could be at least
partially mediated by factors such as the intake of specific
foods, speed of eating or frequency of binge eating.
The relationship between intuitive eating and binge
eating, self-reported rate of eating, fruit and vegetable
intake, and intake of various high-fat or high-sugar foods
can be seen in Table 3.
Table 4 reports the associations between BMI and each
of the following: binge eating, self-reported rate of eating,
fruit and vegetable intake and intake of various high-fat or
high-sugar foods, while controlling for all other food and
eating-related variables. The final regression model of
intuitive eating and BMI, with frequency of binge eating
and self-reported rate of eating included as potential
mediators, is reported in Table 5. The inverse association
between total IES scores and BMI remained statistically
significant but was somewhat attenuated (25?1 % for every
10-unit increase in intuitive eating; 95 % CI 26?1, 24?2%;
P,0?001). The relationship between intuitive eating and
BMI was partially mediated by frequency of binge eating.
Discussion
The present study represents the first nationwide popu-
lation survey to explore the association between intuitive
eating and BMI. Total IES and all three subscales were
significantly associated with BMI in an inverse direction
in mid-age New Zealand women, after adjusting for
potential confounding factors. This is consistent with
suggestions that people who have a stronger awareness
of physiological signals of hunger and satiety, and eat in
response to these signals, are less likely to engage in
behaviours that may lead to weight gain (e.g. eating when
not physically hungry, binge eating) than those who
follow restrictive diet rules
(10,12)
. The body signals of
chronic dieters may have been weakened as a result of
being ignored or replaced with external diet rules;
therefore, these individuals may have lost their ability to
detect when they are hungry or full
(26)
. The inverse
association between intuitive eating and BMI is consistent
with the findings of Augustus-Horvath and Tylka
(7)
among early adult (26–39 years) and middle adult women
(40–65 years). In contrast, for 18–25-year-old women,
intuitive eating was not associated with BMI
(7)
.
If training in intuitive eating among mid-age women
could increase total IES scores by 10 units (on a scale
from 21 to 105), this would equate to a decrease of
about 4?4 kg in a woman typical of our sample. This
would require responses to less than half of the twenty-
one items to change by 1 point on the 5-point response
scale. This is a realistic and achievable improvement
(27)
.
It is noteworthy that in a 3-year prospective study of
women aged 42–52 years, a one-unit increase in reported
level of sports/exercise (on a scale of 1 to 5) was
longitudinally related to a decrease in weight of 0?32 kg
(P,0?001)
(13)
.
Table 2 Health, demographic and behavioural variables and their relationship with BMI among mid-age New Zealand women
Variable n% increase/decrease in BMI* 95 % CI Pvalue
Age (per year) 1512 20?03 20?35, 0?29 0?856
Prioritized ethnicity-1516 ,0?001-
-
Ma
¯ori 9?07 5?54, 12?72
Pacific People 23?17 15?74, 31?08
Asian 212?20 215?98, 28?25
Other 20?80 23?69, 2?17
SES (per unit) 1520 20?18 20?24, 20?11 ,0?001
Smoking status-1502 0?051-
-
Current smoker 0?82 21?94, 3?64
Former smoker 3?08 0?60, 5?63
Postmenopause-1509 3?02 0?55, 5?54 0?016
Physical activity-1503 ,0?001-
-
Light activity 29?43 215?20, 23?26
Moderate activity 212?78 218?33, 26?85
Vigorous activity 213?85 219?29, 28?04
Thyroid condition-1521 5?83 1?72, 10?10 0?005
SES, socio-economic status.
*Percentage increase or decrease in BMI and 95 % confidence intervals were obtained using linear regression models.
-Reference category: ethnicity 5European, smoking status 5never, menopau se status 5premenopause, physical activity 5none, thyroid
condition 5never diagnosed.
-
-
Overall effect.
2276 CEL Madden et al.
Mid-age is a time when women are at particular risk of
weight gain
(13)
. In a cohort of more than 8000 Australian
women aged 45–50 years, over a 2-year period a third of
the women gained 2?25 kg or more, nearly 18 % gained
4?5 kg or more, and there was a mean weight gain of
1kg
(28)
. Among 3064 US women aged 42–52 years, mean
weight increased by 2?1 kg over 3 years of follow-up
(13)
.
A 5-year follow-up of 17 000 Australian women, aged
35–69 years, reported that those who were initially
overweight or obese were about 20 % more likely to
experience major weight gain (5 kg or greater) than
healthy weight women
(29)
.
Given that half of our participants were overweight or
obese, a 4?4 kg decrease in body weight may be of great
practical importance. Weight losses of a similar magnitude
(4?7 kg; 6 % decrease in BMI) have been accompanied
by meaningful decreases in total cholesterol (16 %),
LDL cholesterol (12 %) and HDL cholesterol (18 %)
(30)
.
Weight gain in adult women of about 5 kg has been
associated with a 5 % greater risk of postmenopausal
breast cancer
(31)
.
Interventions that focus on deliberate energy restriction
have achieved short-term weight losses of 5–15 % of
body weight, although almost all weight is regained
within 5 years
(32,33)
. In contrast, recent interventions
among obese women suggest that training in intuitive
eating can prevent weight gain over 2 years
(9,11)
.
Since weight gain results from an energy surplus, the
inverse relationship between intuitive eating and BMI
implies a difference in either the amount and/or the type
of foods eaten. Although intuitive eating was significantly
associated with slower eating and higher vegetable con-
sumption, these effects were too small to be of practical
significance (Table 3). Interestingly, despite giving
themselves unconditional permission to eat any type of
food, intuitive eaters do not consume high-fat/sugar
foods more often than women who deliberately restrict
their food intake. Given that binge eating is a mediator of
the relationship between intuitive eating and BMI, and
that intuitive eaters are less likely to eat in the absence of
physical hunger, it is likely that intuitive eaters differ in
the amount, rather than the types, of food eaten.
Study strengths and limitations
The present study was the first using a nationally repre-
sentative sampling frame to examine intuitive eating and
its association with BMI. Particular strengths are the large
sample size, good response rate and reasonable repre-
sentativeness of the sample in terms of prioritized ethni-
city and SES, as well as the exploration of food intake
and eating behaviours in the context of the BMI–intuitive
eating relationship.
The main limitation was the cross-sectional design of
the study, which means that the observed association
does not necessarily indicate causality. Another signifi-
cant weakness was reliance on self-reported height and
Table 3 The relationship between intuitive eating and other variables (frequency of binge eating, speed of eating, intake of specific foods) among mid-age New Zealand women
Unadjusted result Adjusted* result
Variable n
% increase/decrease or unit
increase/decrease 95 % CI Pvalue n
% increase/decrease or unit
increase/decrease 95 % CI Pvalue
Binge eating-1570 21?62 21?79, 21?46 ,0?001 1505 21?65 21?82, 21?48 ,0?001
Self-reported rate of eating 1588 20?016 20?02, 20?01 ,0?001 1520 20?015 20?02, 20?01 ,0?001
Time taken to eat main meal 1579 0?01 0?008, 0?015 ,0?001 1511 0?01 0?007, 0?014 ,0?001
Fruit intake 1589 20?002 20?007, 0?003 0?419 1521 0?002 20?003, 0?007 0?369
Vegetable intake 1589 0?004 20?0002, 0?009 0?060 1521 0?008 0?003, 0?01 0?001
Chocolate-coated and/or cream-filled biscuits 1591 20?004 20?01, 0?002 0?204 1523 20?003 20?01, 0?004 0?382
Potato crisps, corn snacks or corn chips 1590 0?003 20?002, 0?009 0?246 1522 0?004 20?002, 0?01 0?179
Cakes, scones, muffins or sweet buns 1590 20?007 20?01, 20?001 0?020 1522 20?004 20?01, 0?003 0?251
Meat pies or sausage rolls 1591 0?0004 20?004, 0?005 0?881 1523 20?002 20?007, 0?002 0?322
Burgers 1591 20?0002 20?004, 0?004 0?927 1523 20?001 20?005, 0?003 0?486
*Adjusted for age, smoking status, menopause status, thyroid condition, prioritized ethnicity, physical activity and socio-economic status.
-Log-transformed data presented as % increase or decrease.
Intuitive eating and BMI in adult women 2277
weight to calculate BMI. When self-reporting, women
tend to underestimate weight and overestimate height
(34)
;
however, the mean error is small
(35)
. A study of 536
New Zealand women (35–64 years) showed that this has
little effect on analyses if self-reported BMI is used as a
continuous measure
(36)
as in the present study. A large
US study has recently concluded, from comparisons of
BMI derived from self-reported and measured values, that
the use of a continuous BMI measure derived from self-
reports can be used to estimate associations with BMI
(37)
.
Thus self-reported height and weight data are still con-
sidered valid in epidemiological studies
(38,39)
. Use of
actual measures, while preferable, would have greatly
limited the sample size and representation, and therefore
the generalizability of the study findings.
The study would have benefited from a more compre-
hensive assessment of eating habits, including estimation
of energy and nutrient intakes, and a more detailed
measure of binge eating than the one-item measure
used. However, this would have greatly increased
respondent burden. Previous research on intuitive eating
and the non-dieting approach has focused on women.
The relationship between intuitive eating and BMI in
adult men is unknown.
Conclusions
Women who follow external diet rules or eat in response
to emotional and situational influences may be more likely
to have a higher BMI than those who eat in accordance with
their body signals. The association between BMI and
intuitive eating appears to be partially mediated by fre-
quency of binge eating. If these observations are confirmed
in longitudinal studies and larger intervention trials, they
may highlight a promising approach to weight management
and weight gain prevention. Women who binge-eat may
particularly benefit from training in intuitive eating.
Acknowledgements
This work was supported by funding from the Department
of Human Nutrition, University of Otago. There are no
conflicts of interest to declare. C.E.L.M., C.C.H. and A.G.
had a major role in the writing and interpretation of data;
S.L. and C.E.L.M. designed, pre-tested and piloted the
questionnaire, and conducted the mail survey; C.E.L.M. and
A.G. conducted the analyses; C.C.H. and A.G. had a major
role in study design. C.C.H. is the Principal Investigator,
and conceived of and supervised the study. All authors had
full access to all data and can take responsibility for the
integrity of the data and the accuracy of the data analysis.
All authors have seen and approved the contents of the
submitted manuscript. The authors thank Fiona Hyland
for data entry.
References
1. Jeffery RW, Drewnowski A, Epstein LH et al. (2000) Long-
term maintenance of weight loss: current status. Health
Psychol 19, Suppl. 1, S5–S16.
Table 4 Combined effect of the intake of specific foods, speed of eating and frequency of binge eating on BMI among
mid-age New Zealand women (n 1434)
Food intake and eating
behaviours
% increase/decrease
in BMI* 95 % CI Pvalue
Binge eating 7?46 6?08, 8?85 ,0?001
Self-reported rate of eating 1?87 0?50, 3?26 0?007
Time taken to eat main meal 20?15 21?56, 1?29 0?840
Fruit intake 20?51 21?48, 0?47 0?307
Vegetable intake 1?00 20?03, 2?04 0?057
Chocolate-coated and/or cream-filled biscuits 0?36 20?34, 1?06 0?313
Potato crisps, corn snacks or corn chips 20?74 21?57, 0?09 0?080
Cakes, scones, muffins or sweet buns 21?44 22?22, 20?66 ,0?001
Meat pies or sausage rolls 2?44 1?31, 3?58 ,0?001
Burgers 1?74 0?48, 3?02 0?007
*Adjusted for age, smoking status, menopause status, thyroid condition, prioritized ethnicity, physical activity and socio-economic status.
Table 5 Relationship between intuitive eating (IES score) and BMI controlling for potential mediators (intake of specific
foods and eating behaviours) among mid-age New Zealand women (n 1441)
Food intake and eating
behaviours
% increase/decrease
in BMI* 95 % CI Pvalue
IES score 20?51 20?61, 20?42 ,0?001
Binge eating 4?26 2?83, 5?70 ,0?001
Self-reported rate of eating 0?820?42, 2?04 0?199
IES, Intuitive Eating Scale.
*Adjusted for age, smoking status, menopause status, thyroid condition, prioritized ethnicity, physical activity and socio-economic status.
2278 CEL Madden et al.
2. Wardle J & Beales S (1988) Control and loss of control over
eating: an experimental investigation. J Abnorm Psychol
91, 35–40.
3. Birch LL, Fisher JO & Davison KK (2003) Learning to
overeat: maternal use of restrictive feeding practices
promotes girls’ eating in the absence of hunger. Am J Clin
Nutr 78, 215–220.
4. Ogden J (1995) Cognitive and motivational consequences
of dieting. Eur Eat Disord Rev 3, 228–241.
5. French SA & Jeffery RW (1994) Consequences of dieting to
lose weight: effects on physical and mental health. Health
Psychol 13, 195–212.
6. Korkeila M, Rissanen A, Kaprio J et al. (1999) Weight-loss
attempts and risk of major weight gain: a prospective study
in Finnish adults. Am J Clin Nutr 70, 965–975.
7. Augustus-Horvath CL & Tylka TL (2011) The acceptance
model of intuitive eating: a comparison of women in
emerging adulthood, early adulthood, and middle adult-
hood. J Counsel Psychol 58, 110–125.
8. Foreyt JP & Goodrick GK (1993) Weight management
without dieting. Nutr Today 28, 4–9.
9. Bacon L, Stern JS, Van Loan MD et al. (2005) Size
acceptance and intuitive eating improve health for obese,
female chronic dieters. J Am Diet Assoc 105, 929–936.
10. Tylka TL (2006) Development and psychometric evaluation
of a measure of intuitive eating. J Couns Psychol 53,
226–240.
11. Hawley G, Horwath CC, Gray A et al. (2008) Sustainability
of health and lifestyle improvements following a non-
dieting randomised trial in overweight women. Prev Med
47, 593–599.
12. Hawks S, Madanat H, Hawks J et al. (2005) The relationship
between intuitive eating and health indicators among
college women. Am J Health Educ 36, 331–336.
13. Sternfeld B, Wang H, Quesenberry CP et al. (2004) Physical
activity and changes in weight and waist circumference in
midlife women: findings from the Study of Women’s Health
Across the Nation. Am J of Epidemiol 169, 912–922.
14. Elections New Zealand (2010) Enrolment statistics. http://
www.elections.org.nz/ (accessed February 2010).
15. Dillman DA (2007) Mail and Internet Surveys. The Tailored
Design Method. Hoboken, NJ: John Wiley and Sons.
16. Hay PJ, Mond J, Buttner P et al. (2008) Eating disorder
behaviors are increasing: findings from two sequential
community surveys in South Australia. PLoS One 3, e1541;
available at http://www.plosone.org/article/info%3Adoi%
2F10.1371%2Fjournal.pone.0001541
17. Russell DG, Parnell WR, Wilson NC et al. (1999) NZ Food:
NZ People. Key Results of the 1997 National Nutrition
Survey. Wellington: Ministry of Health.
18. Otsuka R, Tamakoshi K, Yatsuya H et al. (2006) Eating fast
leads to obesity: findings based on self-administered
questionnaires among middle-aged Japanese men and
women. J Epidemiol 16, 117–124.
19. Topolski TD, LoGerfo J, Patrick DL et al. (2006) The Rapid
Assessment of Physical Activity (RAPA) among older adults.
Prev Chronic Dis 3, 1–8; available at http://www.cdc.gov/
pcd/issues/2006/oct/06_0001.htm
20. Australian Longitudinal Study on Women’s Health (2007)
Fifth Survey for Mid-Age Women. Newcastle, NSW: The
Research Institute for Gender and Health, University of
Newcastle.
21. Statistics New Zealand (2006) Census 2006. Wellington:
Statistics New Zealand.
22. Statistics New Zealand (2001) New Zealand Standard
Classification of Occupations, 1999 (NZSCO99). Wellington:
Statistics New Zealand.
23. Davis P, McLeod K, Ransom M et al. (1997) The New
Zealand Socioeconomic Index of Occupational Status
(NZSEI). Wellington: Statistics New Zealand.
24. Baron RM & Kenny DA (1986) The moderator–mediator
variable distinction in social psychological research:
conceptual, strategic, and statistical considerations. J Pers
Soc Psychol 51, 1173–1182.
25. Ministry of Health (2008) A Portrait of Health: Key Results
of the 2006/07 New Zealand Health Survey. Wellington:
Ministry of Health.
26. Birch LL & Fisher JO (2000) Mothers’ child-feeding
practices influence daughters’ eating and weight. Am J
Clin Nutr 71, 1054–1061.
27. Cole RE & Horacek T (2007) Effectiveness of the ‘My Body
Knows When’ intuitive eating non-dieting weight manage-
ment pilot program. J Am Diet Assoc 107, A90.
28. Williams LT, Young AF & Brown WJ (2006) Weight gained
in two years by a population of mid-aged women: how
much is too much? Int J Obes (Lond) 30, 1229–1233.
29. Ball K, Crawford D, Ireland P et al. (2003) Patterns and
demographic predictors of 5-year weight change in a multi-
ethnic cohort of men and women in Australia. Public
Health Nutr 6, 269–280.
30. Seim HC & Holtmeier KB (1992) Effects of a six-week,
low-fat diet on serum cholesterol, body weight, and body
measurements. Fam Pract Res J 12, 411–419.
31. World Cancer Research Fund/American Institute for Cancer
Research (2007) Food, Nutrition, Physical Activity, and the
Prevention of Cancer: a Global Perspective. Washington,
DC: AICR.
32. Institute of Medicine (1995) Weighing the Options: Criteria
for Evaluating Weight-Management. Washington, DC:
National Academy Press.
33. Wadden TA (1993) Treatment of obesity by moderate and
severe caloric restriction: results of clinical research trials.
Ann Intern Med 119, 688–693.
34. Waters A (1993) Assessment of Self-Reported Height and
Weight and their Use in the Determination of Body Mass
Index. Canberra: Australian Institute of Health and Welfare.
35. Engstrom JL, Paterson SA, Doherty A et al. (2003) Accuracy
of self-reported height and weight in women: an integra-
tive review of the literature. J Midwifery Womens Health
48, 338–345.
36. Stewart AW, Jackson RT, Ford MA et al. (1987) Under-
estimation of relative weight by use of self-reported height
and weight. Am J Epidemiol 124, 122–126.
37. Stommel M & Schoenborn CA (2009) Accuracy and
usefulness of BMI measures based on self-reported weight
and height: findings from the NHANES and NHIS
2001–2006. BMC Public Health 9, 421–443.
38. Spencer EA, Appleby PN, Davey GK et al. (2002) Validity of
self-reported height and weight in 4808 EPIC-Oxford
participants. Public Health Nutr 5, 561–565.
39. Burton N, Brown W & Dobson A (2010) Accuracy of body
mass index estimated from self-reported height and weight
in mid-aged Australian women. Aust N Z J Public Health
34, 620–623.
Intuitive eating and BMI in adult women 2279
