International Journal of Obesity (1998) 22, 520 ±528
ß 1998 Stockton Press All rights reserved 0307±0565/98 $12.00
http://www.stockton-press.co.uk/ijo
What is a healthy weight for middle aged
women?
WJ Brown, AJ Dobson and G Mishra
Research Institute for Gender and Health, The University of Newcastle, New South Wales, Australia
OBJECTIVE: To explore associations between body mass index (BMI) and selected indicators of health and well-being
and to suggest a healthy weight range (based on BMI) for middle aged Australian women.
DESIGN: Population based longitudinal study (cross-sectional baseline data).
SUBJECTS: 13 431 women aged 45±49 y who participated in the baseline survey for the Australian Longitudinal Study
on Women's Health.
RESULTS: Forty-eight percent of women had a BMI>25 kg=m2. Prevalence of medical problems (for example,
hypertension, diabetes), surgical procedures (cholescystectomy, hysterectomy) and symptoms (for example, back
pain) increased monotonically with BMI, while indicators of health care use (for example, visits to doctors) showed a
`J' shaped relationship with BMI. Scores for several sub-scales of the MOS short form health survey (SF36) (for
example, general health, role limitations due to emotional dif®culties, social function, mental health and vitality) were
optimal when BMI was around 19 ±24 kg=m2. After adjustment for area of residence, education, smoking, exercise and
menopausal status, low BMI was associated with fewer physical health problems than mid-level or higher BMI, and
the nationally recommended BMI range of 20 ±25 was associated with optimum mental health, lower prevalence of
tiredness and lowest use of health services.
CONCLUSIONS: Acknowledging the limitations of the cross-sectional nature of these data, the results ®rmly support
the bene®ts of leanness in terms of reducing the risk of cardiovascular disease, diabetes and gall bladder disease. The
®ndings are moderated, however, by the observation that both low and high BMI are associated with decreased
vitality and poorer mental health. The optimal range for BMI appears to be about 19±24 kg=m2. From a public health
perspective this study provides strong support for the recommended BMI range of 20±25 as an appropriate target for
the promotion of healthy weight in middle aged Australian women.
Keywords: body mass index; women; middle age; morbidity
Introduction
Data from Europe and the USA suggest that population levels of overweight and obesity in the developed
countries of the `western' world are growing.1,2 There
is considerable concern about the public health implications of this trend, which appears to be particularly
pronounced in women.3 In Australia, despite public
health and educational attempts to curtail the increase
in obesity, between 1980 and 1989 the average weight
of women living in Australian capital cities increased
by 3.1 kg, while in men it increased by 1.7 kg.4 In
women the most marked weight increase occurs
around the time of menopause, making middle aged
women most `at risk' of weight-related ill-health.5
The health consequences of obesity are well documented.6 ±11 Obesity is clearly associated with
increased mortality,7 and is recognised as a predisposing factor for cardiovascular disease, non-insulin
Correspondence: Dr Wendy J Brown, Research Institute for
Gender and Health, The University of Newcastle, University
Drive, Callaghan, New South Wales 2308, Australia.
Received 20 May 1997; revised 12 January 1998; accepted
20 January 1998
dependent diabetes mellitus (NIDDM), osteo-arthritis,
gall bladder disease, some sex-hormone sensitive
cancers and sleep apnoea.10 ±12 There is however an
ongoing debate about the health consequences of
being mildly to moderately overweight,6,7 and about
weight gain with increasing age.13 Current recommendations for American adults no longer condone
signi®cant increases in weight after the age of 35 y.13
The health consequences of underweight are also
relatively poorly understood, despite the fact that, on
the basis of data from the National Heart Foundation
survey5 and preliminary data from the Australian
Longitudinal Study on Women's Health (now
known as the Women's Health Australia project14),
the proportion of young women who are underweight
appears to be increasing.
Overweight and obesity are usually de®ned in terms
of body mass index (BMI, weight in kg=square of
height in metres). While there has been some controversy about the appropriateness of BMI for assessing `healthy weight,' alternative measures such as
waist : hip ratio (WHR) have not yet been clearly
linked with morbidity in women. In any event,
WHR is reported less reliably than height and
weight,15 making BMI the more appropriate indicator
for large population surveys which rely on selfreported measurements.
Healthy weight for middle aged women
WJ Brown et al
Where BMI is used as an indicator of `risk,' it is
useful to de®ne the BMI ranges which represent
underweight, acceptable weight, overweight or obesity, and to understand how these categories relate to
morbidity. Recent US guidelines13 suggest that a BMI
of <25 kg=m2 constitutes `healthy' weight for adults,
and in Australia the National Health and Medical
Research Council (NHMRC) has adopted the classi®cation in which acceptable (healthy) weight is represented by a BMI between 20 ±25 kg=m2.16 Classi®cation of underweight, overweight and obesity, however, is unclear. For example, data from the Nurses
Health Study indicate that women with BMI
<21 kg=m2 have the lowest risk of coronary heart
disease,17 while US dietary guidelines suggest that a
BMI <19 kg=m2 (15th percentile) may be associated
with increased risk of ill-health.13 These US guidelines also suggest that there is a risk related gradient as
BMI increases, with no clear cutpoint for distinguishing between overweight and obesity. In Canada however, BMI>27 kg=m2 is considered `obese'18 and in
Australia, BMI between 25 ±30 kg=m2 is considered
overweight, with obesity de®ned as BMI>30
kg=m2.16
In this paper we address the question `what is a
healthy weight range for middle aged women?' by
exploring relationships between BMI and indicators of
health and well-being among women aged 45 ± 49 y
who are participating in the Women's Health Australia project (WHA).19 Following initial exploration of
the associations between BMI and various medical
conditions, health care utilisation and scores on the
MOS Short Form Health Survey (SF36),20 we
assessed the prevalence of morbidity with increasing
BMI, categorised according to the Australian
NHMRC guidelines,16 with adjustment, where appropriate, for confounders such as area of residence, level
of education, smoking, exercise and menopausal
status.
Methods
Study sample
The WHA project commenced in 1996, when 41 500
women in three age cohorts, selected randomly for a
population-based cohort study, responded to a questionnaire which requested information about a wide
range of health-related issues. Details of the methods
used have been described elsewhere.21 The sampling
frame was women registered on the national Medicare
data base, which includes almost all people who are
resident in Australia, including migrants and refugees.
Participants were randomly selected, with over-sampling from rural and remote areas. There were no
exclusion criteria.
The subjects of this study were 14 205 women aged
45 ± 49 y who participated in the baseline survey for the
WHA project. The response rate was 54%. The demographic and social background characteristics of the
women are broadly representative of Australian women
in this age group, but with an over-representation of
married women and those with tertiary education.21
Questionnaire
The questionnaire requested information about demographic characteristics; medical history (for example,
Have you ever been told by a doctor that you have
hypertension, diabetes, heart disease? Have you had
any of the following procedures . . . cholescystectomy,
hysterectomy? Do you suffer from any of the following symptoms . . . back pain, constant tiredness?); use
of health services (for example, How many times in
the last twelve months did you visit your family
doctor, consult a medical specialist?); current height
and weight; health behaviours (for example, current
and past cigarette smoking, current levels of exercise,
use of hormone replacement therapy); menstrual history and the SF36 health survey. The SF36 contains
36 items which tap eight dimensions of health: role
limitations due to physical dif®culties (RP); physical
functioning (PF); bodily pain (BP); general health
perceptions (GH); role limitations due to emotional
dif®culties (RE); social functioning (SF); mental
health (MH) and vitality (VT). Summary scores for
physical and mental dimensions of health status were
also calculated from responses to the SF36 items.20
Calculation of BMI and estimation of trend curves
BMI was calculated using self-reported height and
weight, corrected following the method of Waters22
[Estimated weight ˆ 1.007 reported weight in kg;
estimated height ˆ 19.208 ‡ (0.879 height in cm)].
Subjects whose weight or height information was
missing (n ˆ 765) or who reported weight as
<30 kg, or height as <1 m or greater than 2.5 m
(n ˆ 9) were excluded from the analyses.
Crude percentages of women reporting selected
medical and surgical conditions, symptoms and indicators of health service utilisation, were calculated for
women within unit intervals of BMI: 17.5 (labelled
as 17), 17.6 ±18.5 (labelled as 18), 18.6 ±19.5 (19). . .
39.6 ± 40.5 (40), 40.6 (41). Trend curves were then
estimated to show the relationship between BMI and
the percentage of women reporting each health problem or indicator, using locally weighted regression,23
implemented via the `lowess' function in S-PLUS.24
Graphs showing the relationship between BMI and
each of the eight sub-scale and two summary scores
on the SF36 were also produced using the lowess
method. (Each line in Figure 2 therefore represents the
®tted curve for 13 431 data points).
Statistical analyses
BMI was assigned to one of ®ve categories, according
to the recommendation of the Australian NHMRC:16
521
Healthy weight for middle aged women
WJ Brown et al
522
<20; 20 ± 25; >25 ± 30; >30 ± 40; >40. Adjusted odds ratios (OR) and 95% con®dence intervals
(CI) were estimated for each of the conditions, procedures, symptoms and health care utilisation variables using multiple logistic regression, with BMI
< 20 as the reference category.
All ORs were adjusted for area of residence, education, smoking, exercise and menopausal status (except
that menopausal status was not included as a confounder when the adjusted OR for hysterectomy was
estimated). All confounding variables except exercise
were treated as categorical explanatory variables.
Area was categorised as `urban,' `rural' or
`remote';25 and education level as `no formal education,' `school education,' `post school certi®cate=trade
apprenticeship=diploma' or `university degree.' Four
categories of smoking were used: never smoked; now
smoke fewer than 20 cigarettes a day; now smoke 20
or more cigarettes a day; and former smoker.
Menopausal status was assigned as one of ®ve
mutually exclusive categories26: currently taking hormone replacement therapy (HRT); no HRT and had
hysterectomy=oophorectomy; no HRT and pre-menopause (still menstruating regularly); no HRT and perimenopause (some change to menstrual frequency in
the last year); no HRT and post menopause (no
menstruation for one year). Level of exercise was
determined from self-reported frequency of engaging
in `vigorous,' `less vigorous' and `work-related' exercise. Responses of never; once a week; 2 or 3 times
per week; 4,5 or 6 times per week; once every day;
and more than once every day were scored 0, 1, 2.5, 5,
7 and 10 respectively to approximate weekly frequencies of exercise, with a weighting of 5 (vigorous), 3
(less vigorous) and 1 (work related) to re¯ect exercise
intensity. The resulting score was treated as a continuous variable.
Means and CIs for the eight subscales and the two
summary scales of the SF36 were also calculated for
each BMI category, using the least square means
option of the generalised linear models procedure of
SAS.27 All means were adjusted for confounding
variables as described above. Pairwise comparisons
of the mean SF36 subscale and summary scores were
performed, for women with a BMI < 20, compared
with each of the other BMI categories. Bonferroni
corrections were used to reduce the effects of in¯ated
type 1 errors due to multiple comparisons.
Results
The means ( s.d.) for height and weight of the
13 431 subjects were 164 6.95 cm and 68.2
14.49 kg, and the mean ( s.d.) BMI was 25.7
5.28 kg=m.2
The relationships between BMI and prevalence of
speci®c medical problems (hypertension, and dia-
betes), surgical procedures (cholescystectomy and
hysterectomy), symptoms (back pain and chronic
tiredness) and health care use, are shown in Figure
1. For hypertension, diabetes, cholescystectomy, hysterectomy and symptoms, there was a monotonic
relationship with BMI. However the indicators of
health care use showed a `J' shaped curve with the
lowest utilisation occurring in women with a BMI of
24 ±25 kg=m2.
The relationships between BMI and SF36 scores are
shown in Figure 2. For three of the four physical
health subscales (PF, BP and GH) scores were highest
for BMI around 22±23, and fell markedly when BMI
was >25. For the mental health components, scores
were highest for three of the sub-scales (SF, MH, VT)
for the BMI range 21±24, with a marked decrease at
BMI values >25, particularly for VT. The graph of
the MH component summary scores showed highest
scores at BMI around 23, with lower scores for both
BMI <20 and >25 (see Figure 2).
The proportion of women in each BMI category,
and the distribution of the confounding variables
across the BMI categories are shown in Table 1.
Less than half the participants were in the `healthy'
range as de®ned by the Australian NHMRC. The
proportion of women in the highest BMI category
was greatest in women from remote areas, and the
proportion of women in the `healthy' range was
highest for tertiary educated women. There was
greater representation of smokers, and lower representation of women who had had a hysterectomy, in
the underweight category.
The adjusted ORs for four BMI categories,
relative to the reference category of BMI<20,
show that the prevalence of each of the selected
medical conditions, surgical procedures and back
pain increased progressively in each BMI category
>25, to maximum risk at BMI > 40 (Table 2). The
only exception was hysterectomy, for which the OR
was highest in the BMI >30 ± 40 category. The
most marked increases in ORs with increasing BMI,
were for hypertension, diabetes and cholescystectomy.
For chronic tiredness and the two indicators of health
care use, prevalence was lowest for women with BMI
between 20 ±25 (see Table 2).
Mean scores for each of the eight SF36 sub-scales,
and for the two summary component scores, are
shown in Table 3. Women in the BMI<20
and 20 ± 25 categories tended to score consistently higher on the physical health sub-scales and
physical component summary score. However women
in the lowest BMI category scored lower than those in
the 20 ± 25 category on the general health subscale. Similarly, for mental health, scores for women
in the lowest BMI category were lower than those in
the 20 ± 25 category on all four sub-scales and on
the mental health component summary score. Scores
were highest for women in the 20 ± 25 BMI
category (see Table 3).
Healthy weight for middle aged women
WJ Brown et al
523
Figure 1 Relationship between body mass index (BMI) (in intervals of 1 kg/m2) and crude percentage of women reporting medical
problems, surgical procedures, symptoms and health care utilisation.
Healthy weight for middle aged women
WJ Brown et al
524
Figure 2 Relationship between body mass index (BMI) (in intervals of 1 kg/m2) and means for SF36 sub-scale and summary scores for
women aged 45±49. (RP ˆ role limitations due to physical dif®culties; PF ˆ physical functioning; BP ˆ bodily pain; GH ˆ general health
perceptions; RE ˆ role limitations due to emotional dif®culties; SF ˆ social functioning; MH ˆ mental health; VT ˆ vitality).
Table 1 Demographic and health related characteristics of women in the ®ve body mass index (BMI) categories
BMI Range
na
%
Area (n ˆ 13358)
urban (%)
rural (%)
remote (%)
Education (n ˆ 13269)
no formal (%)
school education (%)
trade/diploma/cert (%)
university degree (%)
Smoking (n ˆ 12922)
never (%)
<20 per day (%)
>20 per day (%)
ex-smoker (%)
Menopause (n ˆ 11791)
HRT (%)
hysterectomy (%)
pre-menopause (%)
peri-menopause (%)
post-menopause (%)
Exercise score (n ˆ 13243)
mean s.d.
a
w2 or F value
P
<20
>20 ^<25
>25 ^<30
>30 ^<40
>40
938
7.0
5971
44.5
4027
30.0
2177
16.2
297
2.2
8.1
6.1
8.2
47.0
43.6
39.1
28.3
31.0
31.1
14.5
17.0
18.7
2.0
2.3
3.0
w2 (8) ˆ 61.0
<0.001
6.2
6.6
7.1
9.2
36.4
44.2
47.9
51.7
31.6
30.7
29.1
18.7
22.6
16.3
13.9
11.1
3.2
2.3
2.1
1.1
w2 (12) ˆ 181.7
<0.001
6.8
8.7
10.4
5.8
44.7
49.8
41.8
43.8
29.8
27.5
30.1
31.1
16.5
12.7
15.9
16.9
2.3
1.3
1.7
2.4
w2 (12) ˆ 62.1
<0.001
7.4
4.9
7.6
6.7
7.6
42.3
37.0
47.4
45.9
45.3
30.7
32.8
29.4
29.3
29.4
17.6
22.7
13.8
15.4
16.2
2.0
2.5
1.9
2.5
1.6
w2 (16) ˆ 125.3
<0.001
16.4 15.1
15.9 13.4
n varies due to missing data for some characteristics.
14.7 12.7
13.1 12.8
11.3 10.7
F(4,13239) ˆ 28.2
<0.0001
Healthy weight for middle aged women
WJ Brown et al
Table 2 Percentage of women and odds ratios (OR) and 95% con®dence intervals (CI) for selected medical and surgical conditions,
symptoms and health care utilisation by body mass index (BMI) category, after adjustment for area of residence, education, smoking,
exercise and menopausal status.a
Medical conditions
Diabetes
Heart disease
Hypertension
Surgical procedures
Cholescystectomy
Hysterectomya
Symptoms
Often having back pain
Often feeling tired
Health care utilisation
Visited family doctor >5 times a year
Consulted medical specialist 3 times a year
a
n
BMI category
Prevalence
OR
95% CI
12
80
122
127
47
18
108
96
70
12
103
810
916
806
164
<20
20 ± 25
>25 ± 30
>30 ± 40
>40
<20
20 ± 25
>25 ± 30
>30 ± 40
>40
<20
>20 ± 25
>25 ± 30
>30 ± 40
>40
1.6
1.4
3.2
5.9
19.3
2.0
1.7
2.4
3.2
4.3
10.6
13.3
22.8
37.5
61.3
1.0
1.1
2.4
5.0
16.0
1.0
0.9
1.1
1.6
2.0
1.0
1.2
2.3
4.3
9.5
±
0.6±2.0
1.3±4.6
2.7±9.3
8.1±31.7
±
0.5±1.5
0.7±2.0
0.9±2.9
0.9±4.5
±
1.0±1.6
1.8±2.9
3.4±5.5
6.8±13.4
36
319
352
338
70
160
1151
982
678
69
<20
20 ± 25
>25 ± 30
>30 ± 40
>40
<20
20 ± 25
>25 ± 30
>30 ± 40
>40
3.7
5.5
9.3
16.3
26.0
16.9
18.3
24.3
31.0
21.6
1.0
1.4
2.4
4.3
7.1
1.0
1.1
1.5
2.0
1.4
±
0.9±2.0
1.6±3.4
2.9±6.2
4.5±11.5
±
0.9±1.4
1.2±1.8
1.6±2.5
1.0±1.9
164
1012
891
573
84
187
941
767
511
85
<20
20 ± 25
>25 ± 30
>30 ± 40
>40
<20
20 ± 25
>25 ± 30
>30 ± 40
>40
16.4
16.5
22.4
26.3
31.5
20.7
15.9
19.9
24.9
29.0
1.0
1.0
1.4
1.6
1.6
1.0
0.8
0.9
1.1
1.7
±
0.8±1.3
1.1±1.7
1.3±2.0
1.2±2.3
±
0.7±0.9
0.7±1.1
1.0±1.5
1.2±2.3
251
1360
1137
804
143
135
716
562
395
62
<20
20 ± 25
>25 ± 30
>30 ± 40
>40
<20
20 ± 25
>25 ± 30
>30 ± 40
>40
30.0
24.6
31.4
38.4
50.7
8.5
5.8
7.7
9.9
15.0
1.0
0.9
1.1
1.7
2.8
1.0
0.8
1.0
1.3
1.6
±
0.7±1.0
0.9±1.3
1.3±2.0
2.1±3.9
±
0.7±1.0
0.8±1.2
1.1±1.7
1.1±2.3
ORs not adjusted for menopausal status.
Discussion
What is a healthy weight range for middle aged
women? The results presented here suggest that the
answer is BMI between about 19 and 24. However, it
depends to some extent on which conditions are being
considered. High BMI was associated with increased
reporting of hypertension, diabetes, cholescystectomy
and back pain; while BMI between 19 and 24 was
associated with optimal mental health and vitality and
lowest levels of tiredness and use of health services.
An important contribution of this study is the
analysis of BMI by single unit intervals. With a
sample size of 13 431 we were able initially to
avoid arbitrary categories of BMI and to produce
graphs of prevalence of self-reported morbidity and
SF36 scores. These graphs clearly illustrate a graded
increase in prevalence across the entire BMI range for
some variables (for example, hypertension, cholescystectomy); while for others there was increased prevalence (or decrease in SF36 score) above a BMI of
around 24 kg=m2 (for example, tiredness, medical
consultations, general health, mental health).
Subsequent analyses were conducted to adjust for the
effects of area of residence, education, smoking, exercise and menopausal status, which are potential confounders between some of the conditions and BMI.
However, the more re®ned analyses did not alter the
main ®ndings. We did not adjust for alcohol intake
because only 1.1% of the women reported that they
consumed more than two alcoholic drinks per day.
525
Healthy weight for middle aged women
WJ Brown et al
526
Table 3 Means and 95% con®dence intervals (CI) for the eight sub-scales and the two summary scores of the SF36 by body mass
index (BMI) category, after adjustment for area of residence, education, smoking, exercise and menopausal status.
SF36 score
RP
PF
BP
GH
RE
SF
MH
V
Physical component summary
Mental component summary
n
BMI category
Mean
95% CI
P-value*
912
5833
3945
2121
289
909
5851
3951
2131
292
919
5915
3991
2155
294
897
5752
4004
2098
291
908
5814
3940
2114
290
937
5970
4026
2176
297
931
5918
3996
2155
295
931
5920
4004
2156
295
861
5574
3778
2034
280
861
5574
3778
2034
280
<20
20 ± 25
>25 ± 30
>30 ± 40
>40
<20
20 ± 25
>25 ± 30
>30 ± 40
>40
<20
20 ± 25
>25 ± 30
>30 ± 40
>40
<20
20 ± 25
>25 ± 30
>30 ± 40
>40
<20
20 ± 25
>25 ± 30
>30 ± 40
>40
<20
20 ± 25
>25 ± 30
>30 ± 40
>40
<20
20 ± 25
>25 ± 30
>30 ± 40
>40
<20
20 ± 25
>20 ± 30
>30 ± 40
>40
<20
20 ± 25
>25 ± 30
>30 ± 40
>40
<20
20 ± 25
>25 ± 30
>30 ± 40
>40
81.8
82.5
80.1
74.3
73.1
87.7
88.3
85.3
79.9
72.9
73.3
73.2
70.3
65.7
61.8
73.3
75.6
72.5
67.1
58.5
77.7
80.6
77.9
76.6
72.4
81.4
83.3
82.5
80.1
77.3
72.3
74.5
73.4
72.5
70.4
59.7
61.6
58.6
55.1
50.7
50.7
50.6
49.4
46.9
44.6
46.7
48.3
47.5
46.7
44.9
79.3 ± 84.5
81.2 ± 83.8
78.6 ± 81.6
72.4 ± 76.1
68.5 ± 77.6
86.3 ± 88.9
87.6 ± 88.9
84.6 ± 86.1
78.9 ± 80.8
70.6 ± 75.2
71.5 ± 75.0
72.3 ± 74.0
69.3 ± 71.3
64.5 ± 66.9
58.8 ± 64.8
71.8 ± 74.7
74.9 ± 76.4
71.7 ± 73.4
66.0 ± 68.1
56.0 ± 61.0
75.0 ± 80.4
79.2 ± 82.0
76.4 ± 79.4
74.7 ± 78.5
67.8 ± 77.0
79.7 ± 83.1
82.4 ± 84.1
81.6 ± 83.5
78.9 ± 81.3
74.3 ± 80.3
71.0 ± 73.6
73.8 ± 75.1
72.6 ± 74.1
71.5 ± 73.4
68.2 ± 72.7
58.2 ± 61.2
60.8 ± 62.4
57.8 ± 59.5
54.1 ± 56.2
48.1 ± 53.3
50.0 ± 51.4
50.2 ± 50.9
49.0 ± 49.8
46.4 ± 47.4
43.4 ± 45.8
45.8 ± 47.6
47.8 ± 48.8
47.0 ± 48.0
46.1 ± 47.3
43.4 ± 46.5
±
0.65
0.20
0.0001
0.0008
±
0.35
0.001
0.0001
0.0001
±
0.92
0.0014
0.0001
0.0001
±
0.002
0.36
0.0001
0.0001
±
0.035
0.89
0.47
0.047
±
0.035
0.21
0.19
0.018
±
0.001
0.114
0.79
0.15
±
0.017
0.17
0.0001
0.0001
±
0.71
0.0003
0.0003
0.0001
±
0.0006
0.103
0.96
0.06
* P-values for pairwise comparisons with BMI < 20, after Bonferroni correction for multiple comparisons (that is multiplying the
pairwise P value by 4).
RP ˆ role limitations due to physical dif®culties; PF ˆ physical functioning; BP ˆ bodily pain; GH ˆ general health perceptions; RE ˆ role
limitations due to emotional dif®culties; SF ˆ social functioning; MH ˆ mental health; V ˆ vitality.
In addition to replicating previous ®ndings of direct
relationships between BMI and hypertension, coronary heart disease, diabetes and gall bladder disease,6,10,17 we found similar relationships with
symptoms of tiredness and back pain, which are
among the most common symptoms reported as
health concerns by women.28 We also found higher
rates of hysterectomy in women with BMI > 25,
which we are unable to explain, because we did not
ask the reason for the hysterectomy. The lower hysterectomy rate in the group with BMI > 40 could be
due to surgical risk in very obese women, however
this trend was not statistically signi®cant.
Associations between overweight and obesity and
psychological well-being have also been well documented10,29,30 and are con®rmed by this study. However, while there were very few women in this study
with BMI < 17 kg=m2, there was an association
between moderately low BMI (17±18) with poorer
mental health and indicators of tiredness and reduced
vitality. As reduced mental well-being and increased
tiredness in thin women may be a re¯ection of underlying illness, we repeated the analyses, excluding
those women who reported unwanted weight loss of
5 kg or more in the last six months (5% of the sample).
This did not alter the main ®ndings. Thus while
Healthy weight for middle aged women
WJ Brown et al
acknowledging that some forms of mental illness,
such as depression, can be a precursor of weight
loss,31 it may also be true that thin women experience
more tiredness as a result of low energy intake. The
cross-sectional nature of the data, which could be
viewed as a major limitation of the present study,
mean that questions of cause and effect cannot be
answered at this stage. However, as the study moves
into its longitudinal phase, the role of BMI in predicting changes in mental and physical health will be
explored.
A second limitation of the study is that all the
information, including the height and weight measurements, as well as symptoms, medical conditions and
health service use, were self reported. Women are
known to underestimate their weight and over-estimate their height, so that estimated BMI values tend
to be lower than actual ones.32 To correct for this bias
we calculated BMI using adjustments to heights and
weights based on comparisons of self reported and
actual measurements of 4727 women in the risk factor
prevalence study conducted by the Australian
National Heart Foundation in 1989.22
On the issue of self report, a previous Australian
study has shown that there may also be under-reporting of health service use,33 but this appears to occur
uniformly across the range from low to high frequency of use. Provided the bias is unrelated to
BMI, it would be unlikely to signi®cantly affect the
relationship between BMI and health service use. In
the WHA study it will be possible, in due course, to
assess the magnitude of any such bias, by linking the
self reported survey data to Medicare claims data for
the study participants.
The study population was drawn from a whole
population data base, without exclusion of the very
sick or in®rm. While the response rate for the study
was 54%, the participants were similar to women of
similar age in the general population, with respect to
smoking34 and exercise levels.35 There was, however,
slight over-representation of married women (80.7%
compared with 77.1% in the population) and of
women with higher levels of education (10.1% with
a university degree compared with 5.0% in the population).21 Previous studies have shown that women
who are not employed or who have lower levels of
education, are more likely to be overweight or obese,3
so it is likely that our sample includes fewer women
with high BMI than the general population. Moreover,
women with higher levels of education are likely to
have better preventive health care, so the prevalence
of health problems may be under-estimated. The
effect of these biases may be that we have slightly
over-estimated the optimal range for BMI.
As there is a dearth of information about the health
of women living in non-metropolitan areas of Australia, and because we needed good estimates of
health status and health service utilisation, women
living in rural and remote areas were over-sampled. In
light of the tendency for women from these areas to be
more overweight, we adjusted for area of residence in
the analyses.
Generally the results presented here, and recently
elsewhere,6;7 support the bene®ts of leanness, in terms
of reducing the risk of cardiovascular disease, diabetes
and gall bladder disease. In this context, the 1995 US
move toward de®ning `healthy' BMI as <25 kg=m2 is
supported.13 However, support for the bene®ts of
being lean should be tempered by the ®nding that
BMI below about 19 kg=m2 is associated with
decreased vitality, increased tiredness, poorer mental
health and increased use of health services. Establishing the direction of causality will need further investigation, which will be possible from this study in the
future.
As the mean BMI of women in this study
(25.7 kg=m2) was above the upper limit of both current
US and Australian recommendations for `healthy'
weight, it is timely that the NHMRC has recently
proposed the development of national strategies for
the prevention of further increase in overweight and
obesity in Australia.36 The data presented here con®rm
that for middle aged women, BMI in the range
19 ±24=25 kg=m2 is associated with optimal health.
Acknowledgements
The research on which this paper is based was conducted as part of the Australian Longitudinal Study on
Women's Health (The Universities of Newcastle and
Queensland). The successful completion of the ®rst
stage of this project would not have been possible
without the work of the entire research team. Our
special thanks go to Jean Ball for her expert assistance
with data management, and to our research assistants,
Lyn Adamson, Phoebe Bissett and Joy Ellem. We are
grateful to the Department of Health and Family
Services (Australian Commonwealth Government)
for funding.
References
1 Manson JE, Colditz GA, Stampfer MJ. Parity, ponderosity,
and the paradox of a weight-preoccupied society. JAMA 1994;
271: 1788 ± 1790.
2 Laurier D, Guiget M, Chau NP, Wells JA, Valleron AJ.
Prevalence of obesity: A comparative survey in France, the
United Kingdom and the United States. Int J Obes 1992; 16:
565 ± 572.
3 Carpenter L, Bartley M. Fat, female, and poor (commentary).
Lancet 1994; 344: 1715 ± 1716.
4 Bennett S. Cardiovascular risk factors in Australia: Trends in
socioeconomic inequalities. J Epidemiol Commun Health
1995; 49: 363 ± 372.
5 Risk Factor Prevalence Study Management Committee.
Survey No 3, 1989. National Heart Foundation of Australia
and Australian Institute of Health: Canberra, 1990.
6 The Australasian Society for the Study of Obesity. Healthy
weight Australia±a national obesity strategy. Published by the
authors: Sydney, NSW, 1995.
7 Manson JE, Willett WC, Stampfer MJ, et al. Body weight
and mortality among women. N Eng J Med. 1995; 333,
677 ± 685.
527
Healthy weight for middle aged women
WJ Brown et al
528
8 Colditz GA, Willett WC, Rotnitzky A, Manson, JE, Weight
gain as a risk factor for clinical diabetes mellitus in women.
Ann Intern Med. 1995; 122, 481 ± 486.
9 Lapidus L, Bengtsson C, Larsson B, Pennert K,
Rybo E, Sjostrom L. Distribution of adipose tissue and
risk of cardiovascular disease and death: A 12 year follow
up of participants in the population study of women in
Gothenburg, Sweden. Br Med J Clin Res Ed. 1984; 289,
1257 ± 1261.
10 National Institutes of Health Consensus Development Panel.
Health Implications of Obesity. Ann Intern Med. 1985; 103,
1073 ± 1077.
11 Pi-Sunyer FX. Medical hazards of obesity. Ann Intern Med.
1993; 119, 655 ± 660.
12 Bray GA. Effects of obesity on health and happiness. In: KD
Brownell, JP Foreyt (eds) Handbook of eating disorders:
Physiology, psychology and treatment of obesity, anorexia
and bulimia, Basic Books: New York, 1980.
13 Department of Agriculture, US Department of Health &
Human Services. Report of the Dietary Guidelines Committee
on the Dietary Guidelines for Americans. 1995 http:==www.
nal.usda.gov=fnic=Dietary.dgreport.html.
14 Women's Health Australia. Data book for the baseline survey
of the Australian Longitudinal Study on Women's Health
(main cohort). December 1996. Research Institute of Gender
& Health, University of Newcastle. 1996.
15 Rimm EB, Stampfer MJ, Colditz GA, Chute CG, Litin LB,
Willett WC. Validity of self-reported waist and hip circumferences in men and women. Epidemiology 1990; 1, 466 ± 473.
16 National Health & Medical Research Council. Report of the
one hundredth session. Australian Government Publishing
Service: Canberra, 1985.
17 Manson JE, Colditz GA, Stampfer MJ, et al. A prospective
study of obesity and risk of coronary heart disease in women.
N Eng J Med. 1990; 322, 882 ± 889.
18 Health Promotion Directorate, Health Services and Promotion
Branch. Report of an expert group. Canadian Guidelines for
Healthy Weights. 1988.
19 Brown W, Bryson L, Byles J, Dobson A, Manderson L,
Scho®eld M, Williams G. Women's Health Australia: Establishment of the Australian Longitudinal Study on Women's
Health. J Women's Health. 1995; 5, 467 ± 472.
20 Ware JE, Kosinski M, Keller SD. SF 36 Physical and
mental health summary scales: A user's manual. Boston
MA: The Health Institute, New England Medical Centre,
1994.
21 Brown WJ, Bryson LB, Byles JE, Dobson AJ, Lee C, Mishra
G, Scho®eld M. Women's Health Australia: Recruitment for a
national longitudinal cohort study. Women & Health, 1998;
28(8) in press. Under editorial review, available from the
authors.
22 Waters A. Assessment of self-reported height and weight and
their use in the determination of body mass index. Australian
Institute of Health & Welfare: Canberra, 1993.
23 Cleveland WS. Robust locally weighted regression and
smoothing scatterplots. JASA. 1979; 74, 829 ± 836.
24 Mathsoft. S-PLUS Reference Manual Volume 1. Seattle,
Washington, (October 1995), 1995.
25 Department of Primary Industries and Energy, Department of
Human Services and Health. Rural, remote and metropolitan
areas of classi®cation: 1991 Census edition. Australian Government Publishing Service: Canberra, 1994.
26 Dennerstein L, Smith AM, Morse C, Burger H, Green A,
Hopper J, Ryan M. Menopausal symptoms in Australian
women. Med J Aus. 1993; 159, 232 ± 236.
27 SAS Institute Inc. SAS=STAT Users guide, version 6. (4th edn)
Vol 2. SAS Institute Inc: Cary, NC 1989.
28 Brown WJ, Doran FM. Women's health: Consumer views for
planning local health promotion and health care priorities. Aust
J Public Health. 1996; 20, 149 ± 154.
29 Stunkard AJ, Wadden TA. Psychological aspects of severe
obesity. Am J Clin Nutr. 1992; 55, Suppl 524 ± 532.
30 Sullivan M, Karlsson J, Sjostrom L., et al. Swedish
Obese Subjects (SOS)±an intervention study of obesity.
Baseline evaluation of health and psychosocial functioning
in the ®rst 1743 subjects examined. Int J Obes. 1993; 17,
503 ± 512.
31 McCarthy M. The thin ideal, depression and eating disorders
in women. Behar Res Ther 1990; 28, 205 ± 215.
32 Stewart AW, Jackson RT, Fort MA, Beaglehole R. Underestimation of relative weight by use of self-reported height and
weight. Am J Epidem. 1987; 125, 122 ± 126.
33 McCallum J, Lonergan J, Raymond C. The NCEPH record
linkage pilot study: A preliminary examination of individual
Health Insurance Commission records with linked data sets.
National Centre for Epidemiology and Population Health:
Canberra, 1993.
34 Australian Bureau of Statistics. National Health Survey 1995.
(First Results). ABS Catalogue No 4392. Australian Government Publishing Service: Canberra, 1996.
35 Bauman A, Bellew B, Booth M, Hahn A, Stoker L, Thomas
M. Towards best practice for the promotion of physical
activity in the Areas of New South Wales. NSW Health Dept
(Sydney) Centre for Disease Prevention and Health Promotion: Sydney, 1996.
36 National Health & Medical Research Council Expert Panel.
Acting on Australia's weight: a strategic plan for the prevention of overweight and obesity. Australian Government Publishing Service: Canberra, 1997.