Global Food Security ] (]]]]) ]]]–]]]
Contents lists available at SciVerse ScienceDirect
Global Food Security
journal homepage: www.elsevier.com/locate/gfs
Is body mass index an appropriate proxy for body fat in children?
Colleen M. Doak a,n, Daniel J. Hoffman b, Shane A. Norris c, Maiza Campos Ponce a,
Katja Polman a,d, Paula L. Griffiths c,e
a
Department of Health Sciences, VU University Amsterdam, The Netherlands
Department of Nutritional Sciences, Rutgers, the State University of New Jersey, USA
c
MRC/Wits Developmental Pathways for Health Research Unit, Faculty of Health Sciences, University of the Witwatersrand, Johannesburg, South Africa
d
Department of Biomedical Sciences, Institute of Tropical Medicine, Antwerp, Belgium
e
Centre for Global Health and Human Development, School of Sport, Exercise and Health Sciences, Loughborough University, Leicestershire, United Kingdom
b
a r t i c l e i n f o
abstract
Article history:
Received 6 August 2012
Accepted 18 February 2013
As the global prevalence of childhood overweight and obesity spreads to low and middle income
countries, there is an increasing need for researchers to assess overweight and obesity in populations
where child undernutrition still prevails. Although BMI (body mass index) cutoffs are widely used in
research and project evaluations, they have only recently been included in WHO definitions for
overweight and obesity in children. This review describes the history of how and why BMI was
introduced as a proxy for adiposity in children, the scientific evidence and examples from epidemiological studies. Overall, BMI continues to be a valuable measure in children if the underlying
assumptions of the criteria and cut-off values are considered. However, where BMI is associated with
height, in children, we recommend using weight for height z-scores.
& 2013 Elsevier B.V. All rights reserved.
Keywords:
BMI
Overweight
Obesity
Stunting
Children
Body composition
1. Introduction
Standing height and weight are relatively easy and inexpensive measures to perform in most field settings (WHO Expert
Committee on Physical Status, 1995), and consequently, are used
as the basis for multiple indicators of nutritional status. In low
and middle income countries (LMICs), measures of height and
weight have long been used to identify acute and chronic undernutrition (WHO Working Group, 1986; WHO Expert Committee
on Physical Status, 1995; World Health Organization, 2008). As
the emerging epidemic of childhood overweight and obesity
spreads throughout developed and developing countries, the need
to measure and monitor childhood obesity has increased. Initially,
overweight and obesity (WHO Working Group, 1986; WHO
Expert Committee on Physical Status, 1995) were defined using
weight for height for age z-scores (WHZ) to identify the heaviest
children for a given height and age. These classifications continue
to be used in the literature as a more sensitive indicator of obesity
than body mass index (BMI) (Stanojevic et al., 2007). However,
WHZ has proved challenging to implement in clinical settings
since, there is no single chart that can be used to identify WHZ for
children of all ages. Furthermore, weight for height z-scores are
n
Correspondence to: VU University, Faculty of Earth and Life Sciences, Department of Health Sciences, Section of Infectious Disease, De Boelelaan 1085, 1081
HV Amsterdam, The Netherlands. Tel.: þ 31 20 598 3502; fax: þ31 20 598 6940.
E-mail addresses: [email protected], [email protected] (C.M. Doak).
not used for adults and are less familiar than BMI. A single
measure, or index, adjusted for height is preferable, and researchers and clinicians have used BMI to meet this criterion (Fig. 1).
Although BMI cutoffs are widely used in definitions of child
overweight and obesity in the literature, they have only recently
been included as such by the WHO. Table 1 shows the World
Health Organization (WHO) definitions for thinness, overweight
and obesity for children 0–5 years of age and Z5 years of age.
Namely, the z-score definitions were chosen to reflect the SD
cutoffs most closely related to the International Obesity Task
Force (IOTF) measures (Cole et al., 2007; Cole and Lobstein, 2012).
The choice to use different cut-offs for the under 5 years age
group is related to potential misclassification in children under
age 5 years as overweight or obese (de Onis and Lobstein, 2010).
This concern is particularly relevant in a context of transitioning
LMICs (Ke-You and Da-Wei, 2001). Given the potential implications for interventions, it is vital to critically evaluate the use of
BMI to identify overweight and obesity in children (2–18 years
of age).
This review considers the use of BMI from the perspective of
the LMIC context. Specifically, we will underline the methodological complexities of using BMI to explore both sides of malnutrition in a comprehensive way. Given that most research on the use
of BMI has focused on overweight/obesity, we begin with an
historical perspective, describing how and why BMI was introduced as an index of child adiposity (Section 2). In Section 3, we
review evidence for whether BMI accounts for height differences
2211-9124/$ - see front matter & 2013 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.gfs.2013.02.003
Please cite this article as: Doak, C.M., et al., Is body mass index an appropriate proxy for body fat in children? Global Food Security
(2013), http://dx.doi.org/10.1016/j.gfs.2013.02.003i
2
C.M. Doak et al. / Global Food Security ] (]]]]) ]]]–]]]
Fig. 1. shows the timeline of key events in the main events in the development and application of BMI. The application of BMI in children relates to the history of the use of
BMI in adults. Although the Quetelet index was first described in 1823, it was nearly 100 years later (1921) that an index was developed to account for the dimensions of
infants. The WHO adopted BMI cutoffs to define overweight and obesity in adults in 1995, however it was not until 2010 that de Onis and Lobstein (2010) published the
the WHO definitions for overweight and obesity using BMI for children.
Table 1
BMI cut-offs used to define thinness, overweight and obesity in children.
Thinness
Overweight
Obesity
WHO (de Onis and Lobstein,
2010) o 5 years of age
WHO (de Onis and Lobstein,
2010) Z 5 years of age
Must et al. (1991) Z5 years
IOTF (Cole et al., 2000; Cole et al., 2007)
o 2SD BMI
42 SD BMI
43 SD BMI
o 2 SD BMI
41 SD BMI
42 SD BMI
o5th percentile BMI
485th percentile BMI
495 percentile BMI
Percentile equivalent o 18.5
Percentile equivalent of adult BMI 425
Percentile equivalent of BMI 430
in individual children or in populations. Next, Section 4 reviews
the evidence describing population differences in results when
using BMI based outcomes as compared to other measures of
adiposity. Finally, Section 5 will weigh up the policy and program
related issues of using BMI based cut-offs in children. In conclusion, we use the above results to formulate recommendations for
the use of BMI in children and to articulate future directions in
both research and policy.
2. The history of the use of BMI in children
The Quetelet body mass index (weight/stature2) was first
described by Adolphe Quetelet in 1832 as an index of body mass,
adjusted for height (Eknoyan, 2008). The Quetelet index was
designed to allow for comparisons of weight between adults of
different heights. In 1974, Ancel Keys popularized the term body
mass index (BMI) in his seminal paper in which he showed the
Quetelet index to be the best proxy for body fat percentage in
adults (Ashwell, 2011). However, it is widely recognized that
Quetelet’s BMI cannot be used in children under the age of 2 years,
given differences in body proportions between infancy and
adulthood. In 1921, to account for the problems of using the
index in infants, Rohrer (1921) introduced the ponderal index
(weight/stature3) with stature cubed as a more appropriate
adjustment for height because of the different dimensions of
infants. As with adults, the formulation of the denominator
(stature3) was chosen to adjust for stature, such that weights
from infants of different lengths could be directly compared.
Although it was widely agreed that the Quetelet body mass index
was most appropriate for adults and Rohrer’s ponderal index was
most appropriate for infants, it was not clear what formula to
apply in toddlers, young children or adolescents.
Since the mid-1980s, BMI has been used to classify overweight
and obesity globally in adults. During the 1980s, a number of
studies compared the Quetelet index (weight/stature2) to the
Rohrer’s index (weight/stature3) in children. These studies concluded weight/stature2 to be preferable in children over the age of
5. For example, Michielutte et al. (1984) found Quetelet’s BMI was
better correlated with triceps skinfolds than Rohrer’s ponderal
index in 5–12-year-old children in North Carolina. However, the
authors also noted that neither index was ideal. In fact, the
correlation between BMI and triceps skinfold was low in 5-yearold boys (r ¼0.24). Furthermore, the authors noted that neither
index was consistent by race, age or income (Michielutte et al.,
1984). For example, the strongest correlations with triceps skinfolds were in girls from the lowest income census tract whereas in
boys there was no pattern by income group. Another study by
Roche (1981) compared the BMI to the ponderal index and triceps
skinfold measures to total body fat in children and adults, from
6 to 49 years, as measured by underwater weighing. While the
authors recommended weight/stature2 as an indicator of total
body fat for girls, they noted subscapular skinfold thickness was a
better indicator for boys.
In a study of 6–74-year-old, Must et al. (1991) found BMI to be
correlated with triceps skinfolds across the age spectrum, providing justification for the use of a BMI-based reference in children
over 5 years old. Additionally, Must et al. (1991) provided
smoothed 85th and 95th percentiles of BMI for 6–74-year-old,
with results reported by race, sex and age for the NHANES I
Please cite this article as: Doak, C.M., et al., Is body mass index an appropriate proxy for body fat in children? Global Food Security
(2013), http://dx.doi.org/10.1016/j.gfs.2013.02.003i
C.M. Doak et al. / Global Food Security ] (]]]]) ]]]–]]]
population from 1971 to 1974 (Must et al., 1991). At that time,
the 85th percentile of adult BMI was being proposed for use in
identifying adults as overweight (Burton et al., 1985). In accordance with what was being proposed for adults, Must et al. (1991)
proposed using the 85th percentile for overweight and 95th
percentile for obesity for children. In addition their inclusion of
the 5th percentile provided a potential reference point for underweight. It is important to note that until recently, 2010, the WHO
did not publish BMI related reference cut-offs for children due to
concerns about the validity of BMI in children below 5 years old
(de Onis and Lobstein, 2010). Previously, the WHO recommended
using weight for the height z-score (WHZ) for assessing child
obesity (WHZ 4 þ2).
As early as the 1970s emerging evidence from adult data led
researchers to advocate a single BMI cutoff for adult overweight
(BMI 4 ¼25) and obesity (BMI 4 ¼30) (Ashwell, 2011). In 1995
the WHO published these as definitions for adult overweight and
obesity (WHO Expert Committee on Physical Status, 1995). Given
the widespread use of these definitions, an alternative approach
was needed to define overweight and obesity in children that
could be applied from childhood into adolescence and relate to
adult measures, allowing clinicians to use a single measure
(although different cut-offs) to assess obesity. In 2000, the
International Obesity Task Force developed and applied a statistical smoothing method to predict BMI values appropriate for age
in children based on the adult BMI equivalents for overweight and
obesity for boys and girls aged 2–18 years to create an international reference for overweight and obesity using data from
surveys from multiple countries (Cole et al., 2000). Later, Cole
et al. (2007) also published the values equivalent to a BMI of 18.5
for thinness in children. The Must and IOTF references are both
widely used in the literature (see for example Abrantes et al.,
2003; Neovius et al., 2004; Zimmermann et al., 2004). The use of
different references and different cutoffs to define overweight and
obesity in children has made results difficult to compare. The
current challenges and solutions for improving comparability of
BMI across populations relate to the influence of height on BMI
and population differences in body composition, which will be
explored further in the following sections.
3. Does height influence BMI in children?
The principle of indices of body mass, such as BMI, is that they
allow for comparisons of body mass irrespective of height.
Although BMI corrects for height in adults, as described in
Section 2, BMI does not fully correct for height in children and
may therefore be confounded by differences in growth. Wang
et al. (2006) found BMI to be correlated to height in the US
NHANES III data for children aged 2–18 years. In particular, height
was correlated to BMI among boys aged 5–16 and girls aged 5–11
years; consistent with earlier literature (Cole, 1979, 1986). However, there are conflicting data on whether it is tall or short
children who are more likely to be classified as overweight. For
example, surveys from multiple countries show taller children are
more likely to be classified as being overweight or obese when
age-specific BMI cut-off points are used (Franklin, 1999; Wang,
2004). These associations may be biologically driven as taller
children are also more likely to have greater adiposity (Freedman
et al., 2004). It is unclear whether the growth and adiposity is
stimulated by excess energy intake alone or if the association is
driven by growth hormones and/or early maturation which could
simultaneously increase adiposity and growth.
At the same time, there is less consensus about the relationship between shortness, particularly as measured among children
who are stunted (i.e. height for age less that 2 SD of the WHO
3
reference median), and the use of BMI for overweight/obesity
classification. In cross sectional studies, some have found stunting
to be associated with overweight/obesity as measured by BMI
(Popkin et al., 1996; Fernald and Neufeld, 2007) while others have
reported no association (Mukuddem-Petersen and Kruger, 2004).
Data from longitudinal studies are also mixed. Walker et al.
(2007) found that stunting in early childhood was not related to
BMI or adiposity at age 17–18 years compared to matched
controls in a low income Jamaican cohort (Walker et al., 2007).
The study followed 9–24-month-old stunted children, who had
initially received two years of intervention (control, supplementation with milk based formula 1 kg weekly, weekly 1 h play
stimulation sessions, or both). In a South African urban cohort of
Black children (Birth to Twenty (Bt20)) stunting at age 2 was also
not related to later overweight at 9 years as assessed by BMI or
direct measures of adiposity, e.g. Dual X-ray Absorptiometry
(DXA) and skinfolds (Cameron et al., 2005). Cross-sectionally,
stunting was a significant predictor of greater BMI in 2-year-old
South Africans, but not triceps or subscapular skinfold measures
(Cameron et al., 2005). Additionally, in a longitudinal analysis of
the same children, stunting at 2 years of age did not predict BMI,
triceps skinfolds, subscapular skinfolds or body composition
(assessed by DXA) at 9 years of age.
In contrast to the findings from South Africa, stunted children
in Brazil who were also studied longitudinally gained more fat
mass than non-stunted children (Martins et al., 2004). In the same
sample, stunted children were more likely to develop central
adiposity compared to normal height children (Hoffman et al.,
2007a,b). The lack of consistency in the results will be considered
further in the following sections focusing on validity and bias in
the use of BMI in children (Section 3.1), biological and metabolic
explanations of the relationship between stunting and BMI
(Section 3.2) and population differences in height and maturation
that may explain the inconsistencies (Section 3.3).
3.1. Validity and bias
It is important to establish whether BMI is equally valid in
stunted children if it is to be used in LMIC settings. The authors of
the Bt20 study in South Africa discuss the significant cross
sectional association between stunting and increased BMI at
2 years, suggesting that this association might be explained by
the fact that height appears in the denominator of BMI and is
squared (Cameron et al., 2005). A proportionally greater reduction
in height may be exaggerated in this calculation, and result in
shorter children having larger values for BMI provided they have
the same body weight as a normal height child. In the Bt20cohort, the mean weight at 2 years approximated the CDC 2000
10th centile, whereas the mean height was less than the 5th
centile. The authors also comment that a significant correlation at
2 years was observed between height and BMI for the stunted
children (r ¼ 0.505, Po0.01) only, but not for the non-stunted
children (r ¼0.107, P40.05), suggesting that in very young
children the adjustment of BMI by height squared does not
eliminate the influence of height in a stunted population. A study
in a rural South African population of 3-year-old also observed a
high prevalence of overweight and obesity alongside a high
prevalence of stunting, suggesting that this association is not
confined to urban areas of South Africa (Mamabolo et al., 2005).
In older children the evidence for a relationship between
stunting and high BMI is mixed. Among 7–9-year-old Maya
children in Mexico, Wilson et al. (2011) found that BMI was
significantly associated with a variety of adiposity measures
including percent body fat assessed by multiple measures (bioelectrical impedance analysis, z-scores of waist circumference and
skinfold measures.) Furthermore, stunting was not observed to
Please cite this article as: Doak, C.M., et al., Is body mass index an appropriate proxy for body fat in children? Global Food Security
(2013), http://dx.doi.org/10.1016/j.gfs.2013.02.003i
4
C.M. Doak et al. / Global Food Security ] (]]]]) ]]]–]]]
mediate the relationship, although it did moderate (modify) the
association between BMI and waist circumference z-scores in
children. The non-stunted children showed a stronger positive
relationship between BMI and waist circumference, suggesting
that in this sample BMI is a stronger indicator of increased waist
circumference if a child is not stunted. It was concluded that BMI
was an adequate index of adiposity in these 7–9-year-old Maya
children, although the authors cautioned for the need to measure
both stunting and obesity in countries suffering from a dual
burden of malnutrition to ensure that the child’s overall nutritional profile is understood. This paper also considered the
relationship between BMI and the adiposity measures in the
mothers of these children (75% of whom had been stunted) and
concluded that ‘‘It is not appropriate to use BMI alone to predict
percentage body fat in this sample of urban Maya women.’’
(Wilson et al., 2011: 784). In Brazilian children, Hoffman et al.
(2006) found that BMI correlates with adiposity equally well in
stunted and non-stunted children, using gold standard measures
of body composition (H218O dilution). The study compared
stunted 7–11-year-old to matched controls from shantytowns in
Sa~ o Paolo, Brazil. The authors concluded that growth retardation
did not appear to influence the relationship between body fatness
and BMI. Thus, BMI has an inconsistent and unclear relationship
to adiposity in stunted children.
3.2. Biological
Theoretically, one could adapt BMI measures to address the
mathematical issues described above that create erroneous associations between stunting and obesity (as defined by BMI).
However, given data suggesting that stunting is physiologically
associated with excess adiposity (i.e. obesity) it might not be
appropriate to mathematically adjust for height. Building on the
epidemiological evidence that linked stunting with obesity, several studies provide clinical evidence that stunting may result in
metabolic adaptations that could predispose a stunted child to
gain excess body fat. In 2000, a lower rate of basal fat oxidation
was observed in stunted children compared to normal height
children from the shantytowns of Sa~ o Paulo, Brazil (Hoffman
et al., 2000). Such metabolic adaptations following energy restriction and growth retardation are supported by a study of adults
from the Buryat, tribe of Southern Siberia who, due to nutritional
deficiencies following the break-up of the Soviet Union, are
shorter than the previous generation (Leonard et al., 2009). These
results, from two populations, representing very different age,
genetic background, culture, and socio-economic environments
show a consistent relationship between growth retardation and
altered fat metabolism.
In other analysis, Hoffman et al. (2007a,b) showed the stunted
children also deposited more fat in the central/truncal region, a
phenotype associated with chronic metabolic diseases (Peiris
et al., 1989). Therefore, independent of the argument as to
whether the use of BMI is appropriate for classifying stunted
children as overweight or obese, physiological evidence exists to
support those studies that have reported an association between
being stunted and being overweight or obese. Thus, associations
between height and BMI may reflect genuine relationships
between stunting and adiposity.
3.3. Population differences related to height and maturation
In addition to the above concerns, there are known population
differences in height, maturation and body composition that
complicate interpretation of BMI outcomes in 10–18-year-old
children. Seidell et al. (2006) analyzed the relationships between
BMI and height from seven national surveys (Brazil, China,
Indonesia, Russia, the Netherlands and the United States). In all
countries, the correlation coefficient for height plotted against
BMI was relatively high ( 0.7 or o0.8). Furthermore, height
predicted significant changes to BMI in 6 countries for boys and
3 countries for girls (Seidell et al., 2006). A comparison of the
shortest (Chinese) and tallest (Dutch) populations showed a
difference in height of 14.5 cm for males and 11.5 cm for females.
The authors applied an approach in which adult median height
was used to calculate a scaled height, using the child’s height as a
proportion of the adult median. The scaled ‘height age’ was used
to replace age in months or as the benchmark for comparing the
children’s BMI. The scaling process produced a BMI for age curve
for the Chinese survey that was parallel to, but was below the
scaled curve for Dutch children. After multiplying the Chinese age
by a factor of 0.97, the two curves lined up upon visual inspection.
Application of this procedure to data from other countries gave
similar results. These results show that existing population
differences in BMI patterns can be explained either by height,
maturation or both. Furthermore, the comparability of BMI curves
across populations can be improved by adjusting for a child’s
height relative to the median adult height in the population. A
scaling process that replaces age in months or years with a
relative ‘height age’ could address residual confounding by height
related to the mathematical and formulaic reasons as described
by Cameron et al. (2005) whilst taking into account the differences in rates of growth between children. A scaling approach,
rather than a statistical adjustment for height, would allow for
researchers to use BMI to identify associations between child
stunting and risk of adiposity (Hoffman et al., 2000; Martins et al.,
2004).
Although findings from Brazil and South Africa are contradictory, differences may be explained by the different contexts in
which the studies were carried out. The South African population
studied had very low prevalences of overweight and obesity up to
age 9 years when compared to the Brazilian study, suggesting the
South African children were living in an environment during
childhood that was not conducive to developing overweight and
obesity. Additional studies are needed to clarify the contexts in
which stunted children will develop an altered fat metabolism.
The above studies clearly demonstrate the potential for height to
influence the results, particularly in young children ( oage 2
years). This methodological issue can be clarified by researchers
by testing whether weight/stature2 is associated with height for
the study population used. However, it is important to note a
scaling approach would be particularly challenging in multiethnic populations such as exist in many countries. Until the
methodological uncertainties are resolved, and the potential of
height-age scaling to control for height and maturation further
explored, we recommend using WHZ instead of BMI wherever
BMI is associated with height in children.
4. Population differences in interpreting BMI
Researchers have assumed that, based on the correlation
coefficient, the association between BMI and adiposity is sufficiently high for use in epidemiological studies. Namely, the
assumption is that BMI would appropriately distinguish differences in adiposity within a given population. However, epidemiological associations are sensitive to the choice of the cutoff
used to define overweight and obesity. As described above, BMI
may be inappropriate in a short or stunted population. For
instance, in 2-year-old South Africans, stunting was a significant
predictor of greater BMI, but did not predict triceps or subscapular skinfold measures (Cameron et al., 2005). Furthermore, in
Asian populations established BMI cut-offs do not identify those
Please cite this article as: Doak, C.M., et al., Is body mass index an appropriate proxy for body fat in children? Global Food Security
(2013), http://dx.doi.org/10.1016/j.gfs.2013.02.003i
C.M. Doak et al. / Global Food Security ] (]]]]) ]]]–]]]
most at risk. Asians, for example, have a greater risk for chronic
disease outcomes at a lower BMI, leading researchers to call for a
different set of BMI cut-offs in Asian populations (Low et al.,
2009). Indian infants in particular have been found to have a
greater fat mass at any given weight as compared to British
infants (Yajnik et al., 2003), suggesting that in Indian populations
a potentially unhealthy body composition is already evident
at birth.
Could differences in age or body composition of various study
samples distort the findings of epidemiological studies that rely
on BMI to define overweight and obesity? Analysis of this
question was explored in a review of interventions testing
prevention programs targeting overweight and obesity in school
aged children. For example, Doak et al. (2006) found interventions
reporting both BMI and skinfold measures were more likely to be
deemed effective when assessed by skinfold than by BMI based
outcomes. Further analysis showed the conclusions of the review
itself were influenced by the inclusion of studies reporting
skinfolds and adiposity, which were more likely to be deemed
effective, as compared to studies reporting on BMI outcomes
(Doak et al., 2009). Additionally, conclusions about effectiveness
differed depending on how BMI was used to evaluate the intervention effect. For example, an intervention could effectively
reduce the prevalence of overweight and obesity but not significantly shift the mean BMI z-scores (Doak et al., 2009). These
findings show that the result of a study may be influenced not
only by the use of BMI but also by how BMI is used to evaluate it.
It is important to note that in spite of the limitations, studies
show a strong correlation between BMI and body composition. In
South Africa, Cameron et al., (2009) studied 8–11-year-old prepubertal Black children and showed BMI was more strongly
associated with the DXA assessments of fat mass compared to
waist circumference. Wang et al. (2007) also showed correlations
between BMI and DXA measures of adiposity in Chinese children.
The study included 2493 children aged 6–18 years from a
population-based cohort of twin pairs. In this relatively lean,
rural Chinese population, BMI and waist circumference were
highly correlated with each other and were good surrogates of
total body fat, trunk fat, and percent body fat in prepubertal
children of both genders as well as pubertal girls. However, in
pubertal boys, both BMI and waist circumference overestimated
total and trunk fat, especially percentage of body fat. The above
results show, on the one hand, strong correlation coefficients
between BMI and adiposity measures, and on the other hand,
spurious outcomes when overweight or obesity are defined using
BMI. These differences may be explained by age, height and body
composition of the study populations. Epidemiological studies are
needed to compare dichotomous classifications of overweight and
obesity based on BMI versus other measures of adiposity.
5. The use of BMI: program and policy implications
There are a number of policy and programmatic implications
of using BMI instead of WHZ. Concerns about the validity of BMI
in younger children led the WHO to apply different BMI cut-offs
in children above and below 5 years old (de Onis and Lobstein,
2010). This decision has implications for population prevalence of
malnutrition. In a group of children with BMIs similar to the
reference population, the prevalence of overweight and obesity
would be 2.3% ( 4 þ2 SD) in 4-year-old but 16% ( 4 þ1 SD) in
5-year-old. However in both age groups the prevalence of thinness would remain the same, i.e. 2.3% ( o 2 SD). Thus, the BMI
cut-off in children above 5 years old could distort public health
priorities towards concerns of overweight and obesity and creates
5
an artificially higher prevalence of overweight and obesity in
older versus younger children.
A key challenge to using BMI is in the need for consistency and
convergence for measures of ‘‘thinness’’ as well as overweight and
obesity. The WHO has only recently defined thinness in children
using BMI cutoffs (WHO, 2010). Thus, BMI is widely studied and
applied in research reporting overweight or obesity in children
(Abrantes et al., 2003; Neovius et al., 2004; Zimmermann et al.,
2004). In contrast, few researchers have studied BMI to assess
thinness in children. Given the lack of any clear rationale for using
BMI, most researchers and practitioners continue to use weight
for height z-score measures to study thinness in children (Briend
et al., 2011; Radhakrishna et al., 2010). Similarly, food supplementation programs evaluate intervention effectiveness using
WHZ (Inayati et al., 2012) whereas overweight and obesity
prevention programs use BMI z-scores or percentiles (Story
et al., 2012). The current trends preclude a holistic perspective
in analysis. Namely, the use of BMI has introduced methodological complexities for assessing child nutrition comprehensively,
including thinness, overweight and obesity.
Calls to apply different BMI cut-offs for Asian populations in
adults also have important and broad implications for the use of
BMI in Asian children. For example, Virani (2011) has recently
provided cut-offs for Asian-Indian children, creating centile
equivalents for BMI cutoffs using the IOTF methodology. However, the Virani (2011) cutoffs are not appropriate for other
populations and may also not be appropriate for lower income
Indian children. Furthermore, a separate set of BMI based references for specific populations runs contrary to the WHO arguments supporting the use of international standards and
references. Additionally, the use of separate references for Indian
children raises the question of appropriate BMI cutoffs for thinness. If international BMI cutoffs of 25 for overweight and 30 for
obesity are too high for Indian or Asian adults, the BMI cutoff of
18.5 as an indicator of thinness may also be too high. On the other
hand, if short or stunted children have an inappropriately high
BMI for mathematical reasons, stunted children may have an
artificially high BMI. Thus, stunted children who are also truly
thin may be missed by lowering the reference cut-off for thinness.
6. Conclusions
Overall, we find that BMI continues to be a valuable measure
of adiposity in children if used with caution. First, it is important
that researchers are aware of the potential for height to influence
the results, particularly in young children ( oage 2 years). As
stated previously, this methodological issue can be clarified by
testing the equation for BMI in the study population to confirm
that weight/stature2 is not associated with height for the study
population used. Where BMI is associated with height in children,
we recommend using WHZ instead of BMI. In stunted children
waist circumference and skinfold measures can help to provide a
complete picture of body shape and body composition at the
population level. Third, the comparability of BMI curves across
populations can be improved by adjusting for a child’s height
relative to the median adult height in the population. Studies are
needed to clarify whether this process can be used to improve the
predictive value of BMI as a measure of adiposity. Whether or not
BMI is an appropriate measure of thinness is under-studied.
Furthermore, when reporting results, it is important to keep in
mind the underlying assumptions of the cut-offs used when
reporting prevalence results for overweight and above and for
thinness. Specifically, it is important to note that the BMI cut-offs
may overstate overweight and obesity relative to thinness,
particularly in children over 5 years of age. In conclusion, if the
Please cite this article as: Doak, C.M., et al., Is body mass index an appropriate proxy for body fat in children? Global Food Security
(2013), http://dx.doi.org/10.1016/j.gfs.2013.02.003i
6
C.M. Doak et al. / Global Food Security ] (]]]]) ]]]–]]]
limitations of BMI are recognized, it can be effectively used as an
indicator to document emerging overweight and obesity in
children. It is critical that the same methods of classifying BMI
into either underweight or overweight risk categories are applied
consistently over time so as to interpret change within the
population. It is also important to better inform policy makers
to appreciate the potential of current WHO BMI cut-offs to
overstate obesity and overweight relative to underweight, particularly in children above the age of 5. Nevertheless, when used
together with other indicators, BMI can help to identify the
appropriate programmes and policy implications for children in
resource poor settings. It is important to note that the need for
reporting additional indicators may be a barrier to the use of BMI
in some instances. However, where possible, the inclusion of
additional indicators in studies from different populations can
help to improve the comparability of results based on BMI alone.
References
Abrantes, M.M., Lamounier, J.A., Colosimo, E.A., 2003. Comparison of body mass
index values proposed by Cole et al. (2000) and Must et al. (1991) for
identifying obese children with weight-for-height index recommended by
the World Health Organization. Public Health Nutrition 6 (3), 307–311.
Ashwell, M., 2011. Charts based on body mass index and waist-to-height ratio to
assess the health risks of obesity: a review. The Open Obesity Journal 3, 78–84.
Briend, A., Van den Broeck, J., Fadnes, L.T., 2011. Target weight gain for moderately
wasted children during supplementation interventions—a population-based
approach. Public Health Nutrition 11, 1–7.
Burton, B.T., Foster, W.R., Hirsch, J., Van Itallie, T.B., 1985. Health implications of
obesity: an NIH consensus development conference. International Journal of
Obesity 9 (3), 155–170. (ABSTRACT).
Cameron, N., Wright, M.M., Griffiths, P.L., Norris, S.A., Pettifor, J.M., 2005. Stunting
at 2 years in relation to body composition at 9 years in African urban children.
Obesity Research 13 (1), 131–136.
Cameron, N., Jones, L.L., Griffiths, P.L., Norris, S.A., Pettifor, J.M., 2009. How well do
waist circumference and body mass index reflect body composition in prepubertal children? European Journal of Clinical Nutrition 63 (9), 1065–1070.
Cole, T.J., 1979. A method for assessing age-standardized weight-for-height in
children seen cross-sectionally. Annals of Human Biology 6, 249–268.
Cole, T.J., 1986. Weight/heightp compared to weight/height2 for assessing adiposity in childhood: influence of age and bone age on p during puberty. Annals
of Human Biology 13, 433–451.
Cole, T.J., Bellizzi, M.C., Flegal, K.M., Dietz, W.H., 2000. Establishing a standard
definition for child overweight and obesity worldwide: international survey.
British Medical Journal 320, 1240–1243.
Cole, T.J., Flegal, K.M., Nicholls, D., Jackson, A.A., 2007. Body mass index cut offs to
define thinness in children and adolescents: international survey. British
Medical Journal 335 (7612), 194.
Cole, T.J., Lobstein, T., 2012. Extended international (IOTF) body mass index cutoffs for thinness, overweight and obesity. Pediatric Obesity 7 (4), 284–294.
de Onis, M., Lobstein, T., 2010. Defining obesity risk status in the general childhood
population: Which cut-offs should we use? International Journal of Pediatric
Obesity 5 (6), 458–460.
Doak, C.M., Visscher, T.L., Renders, C.M., Seidell, J.C., 2006. The prevention of
overweight and obesity in children and adolescents: a review of interventions
and programmes. Obesity Reviews 7 (1), 111–136.
Doak, C., Heitmann, B.L., Summerbell, C., Lissner, L., 2009. Prevention of childhood
obesity—what type of evidence should we consider relevant? Obesity Reviews
10 (3), 350–356.
Eknoyan, G., 2008. Adolphe Quetelet (1796–1874)—the average man and indices
of obesity. Nephrology Dialysis Transplantation 23 (1), 47–51.
Fernald, L.C., Neufeld, L.M., 2007. Overweight with concurrent stunting in very
young children from rural Mexico: prevalence and associated factors. European Journal of Clinical Nutrition 61 (5), 623–632.
Freedman, D.S., Thornton, J.C., Mei, Z., Wang, J., Dietz, W.H., Pierson Jr, R.N.,
Horlick, M., 2004. Height and adiposity among children. Obesity Reviews 2,
846–853.
Franklin, M.F., 1999. Comparison of weight and height relations in boys from
4 countries. American Journal of Clinical Nutrition 70, 157S–162SS.
Hoffman, D.J., Martins, P.A., Roberts, S.B., Sawaya, A.L., 2007a. Body fat distribution
in stunted compared with normal-height children from the shantytowns of
Sa~ o Paulo, Brazil. Nutrition 23 (9), 640–646.
Hoffman, D.J., Sawaya, A.L., Martins, P.A., McCrory, M.A., Roberts, S.B., 2006.
Comparison of techniques to evaluate adiposity in stunted and nonstunted
children. Pediatrics 117 (4), e725–e732.
Hoffman, D.J., Sawaya, A.L., Verreschi, I., Tucker, K.L., Roberts, S.B., 2000. Why are
nutritionally stunted children at increased risk of obesity? Studies of metabolic rate and fat oxidation in shantytown children from Sa~ o Paulo, Brazil.
American Journal of Clinical Nutrition 72 (3), 702–707.
Hoffman, D.J., Martins, P.A., Roberts, S.B., Sawaya, A.L., 2007b. Body fat distribution
in stunted compared with normal-height children from the shantytowns of
Sa~ o Paulo, Brazil. Nutrition 23 (9), 640–646.
Inayati, D.A., Scherbaum, V., Purwestri, R.C., Wirawan, N.N., Suryantan, J., Hartono,
S., Bloem, M.A., Pangaribuan, R.V., Biesalski, H.K., Hoffmann, V., Bellows, A.C.,
2012. Combined intensive nutrition education and micronutrient powder
supplementation improved nutritional status of mildly wasted children on
Nias Island, Indonesia. Asia Pacific Journal of Clinical Nutrition 21 (3),
361–373.
Ke-You, G., Da-Wei, F., 2001. The magnitude and trends of under- and overnutrition in Asian countries. Biomedical and Environmental Sciences 14 (1–2),
53–60.
Leonard, W.R., Sorensen, M.V., Mosher, M.J., Spitsyn, V., 2009. Comuzzie AG
Reduced fat oxidation and obesity risks among the Buryat of Southern Siberia.
American Journal of Human Biology 21 (5), 664–670.
Low, S., Chin, M.C., Ma, S., Heng, D., Deurenberg-Yap, M., 2009. Rationale for
redefining obesity in Asians. Annals, Academy of Medicine, Singapore 1,
66–69.
Mamabolo, R.L., Alberts, M., Steyn, N.P., Delemarre-van de Waal, H.A., Levitt, N.S.,
2005. Prevalence and determinants of stunting and overweight in 3-year-old
black South African children residing in the central region of Limpopo
Province, South Africa. Public Health Nutrition 8, 501–508.
Martins, P.A., Hoffman, D.J., Fernandes, M.T., Nascimento, C.R., Roberts, S.B., Sesso,
R., Sawaya, A.L., 2004. Stunted children gain less lean body mass and more fat
mass than their non-stunted counterparts: a prospective study. British Journal
of Nutrition 92 (5), 819–825.
Michielutte, R., Disker, R.A., Corbett, W.T., Schey, H.M., Ureda, J.R., 1984. The
relationship between weight-height indices and the triceps skinfold measure
among children age 5–12. American Journal of Public Health 74, 604–606.
Mukuddem-Petersen, J., Kruger, H.S., 2004. Association between stunting and
overweight among 10–15-y-old children in the North West Province of South
Africa: the THUSA BANA study. International Journal of Obesity and Related
Metabolic Disorders 28 (7), 842–851.
Must, A., Dallal, G.E., Dietz, W.H., 1991. Reference data for obesity. 85th and 95th
percentiles of body mass index (wt/ht2) and triceps skinfold thickness.
American Journal of Public Health 53, 839–846.
Neovius, M.G., Linné, Y.M., Barkeling, B.S., Rossner, S.O., 2004. Sensitivity and
specificity of classification systems for fatness in adolescents. American
Journal of Public Health 80 (3), 597–603.
Popkin, B.M., Richards, M.K., Montiero, C.A., 1996. Stunting is associated with
overweight in children of four nations that are undergoing the nutrition
transition. Journal of Nutrition 126 (12), 3009–3016.
Peiris, A.N., Sothmann, M.S., Hoffmann, R.G., Hennes, M.I., Wilson, C.R., Gustafson,
A.B., Kissebah, A.H., 1989. Adiposity, fat distribution, and cardiovascular risk.
Annals of Internal Medicine 110, 867–872.
Radhakrishna, K.V., Kulkarni, B., Balakrishna, N., Rajkumar, H., Omkar, C., Shatrugna, V., 2010. Composition of weight gain during nutrition rehabilitation of
severely under nourished children in a hospital based study from India. Asia
Pacific Journal of Clinical Nutrition 19 (1), 8–13.
Roche, A.F., 1981. Grading body fatness from limited anthropometric data.
American Journal of Clinical Nutrition 34 (12), 283l–288l.
Rohrer, F., 1921. Der Index der Körperfülle als Maß des Ernährungszustandes [The
index of corpulence as measure of nutritional state]. Munchener Medizinische
Wochenschrift 68, 580–582. (in German, Abstract).
Seidell, J.C., Doak, C.M., de Munter, J.S.L., Kuijper, L.D.J., Zonneveld, C., 2006. Crosssectional growth references and implications for the development of an
international growth standard for school-aged children and adolescents. Food
and Nutrition Bulletin 27 (4), S189–S198. (supplement).
Story M., Hannan P.J., Fulkerson J.A., Holy Rock B., Smyth M., Arcan C., Himes J.H.,
2012. Bright Start: Description and main outcomes from a group-randomized
obesity prevention trial in American Indian children. Obesity (Silver Spring),
20 (11), 2241–2249.
Stanojevic, S., Kain, J., Uauy, R., 2007. The association between changes in height
and obesity in Chilean preschool children: 1996–2004. Obesity (Silver Spring)
15 (4), 1012–1022.
Virani, N., 2011. Reference curves and cut-off values for anthropometric indices of
adiposity of affluent Asian Indian children aged 3–18 years. Annals of Human
Biology 38 (2), 165–174.
Walker, S.P., Chang, S.M., Powell, C.A., 2007. The association between early
childhood stunting and weight status in late adolescence. International Journal
of Obesity (London) 31 (2), 347–352.
Wang, H., Story, R.E., Venners, S.A., Wang, B., Yang, J., Li, Z., Wang, L., Liu, X., Tang,
G., Xing, H., Xu, X., Wang, X., 2007. Patterns and interrelationships of body-fat
measures among rural Chinese children aged 6–18 years. Pediatrics 120 (1),
e94–e101.
Wang, Y., Moreno, L.A., Caballero, B., Cole, T.J., 2006. Limitations of the current
World Health Organization growth references for children and adolescents.
Food and Nutrition Bulletin 27 (4), S175–S188, supplement.
Wang, Y., 2004. Epidemiology of childhood obesity—methodological aspects and
guidelines: What is new? International Journal of Obesity and Related
Metabolic Disorder 28 (3), S21–S28.
W.H.O. Working Group, 1986. Use and interpretation of anthropometric indicators
of nutritional status. Bulletin of the World Health Organization 64, 929–941.
WHO Expert Committee on Physical Status, 1995. The use and interpretation of
anthropometry: Report of a WHO Expert Committee. World Health Organization, WHO Technical Report Series 854, Geneva.
Please cite this article as: Doak, C.M., et al., Is body mass index an appropriate proxy for body fat in children? Global Food Security
(2013), http://dx.doi.org/10.1016/j.gfs.2013.02.003i
C.M. Doak et al. / Global Food Security ] (]]]]) ]]]–]]]
World Health Organization, 2008. Training course on child growth assessment. C:
Interpreting growth curves. World Health Organization, Geneva /http://www.
who.int/childgrowth/training/enS, p. 14, (accessed 05.08.12.).
Wilson, H., Dickinson, F., Griffiths, P.L., Azcorra, H., Bogin, B.A., Varela Silva, I.,
2011. How useful is bmi in predicting adiposity indicators in a sample of Maya
children and women with high levels of stunting? American Journal of Human
Biology 23 (6), 780–789.
Yajnik, C.S., Fall, C.H., Coyaji, K.J., Hirve, S.S., Rao, S., Barker, D.J., Joglekar, C.,
Kellingray, S., 2003. Neonatal anthropometry: the thin-fat Indian baby. The
7
Pune Maternal Nutrition Study. International Journal of Obesity and Related
Metabolic Disorders 27 (2), 173–180.
Zimmermann, M.B., Gübeli, C., Püntener, C., Molinari, L., 2004. Detection of
overweight and obesity in a national sample of 6–12-y-old Swiss children:
accuracy and validity of reference values for body mass index from the US
Centers for Disease Control and Prevention and the International Obesity Task
Force. American Journal of Clinical Nutrition 79 (5), 838–843.
Please cite this article as: Doak, C.M., et al., Is body mass index an appropriate proxy for body fat in children? Global Food Security
(2013), http://dx.doi.org/10.1016/j.gfs.2013.02.003i