The TRENDS Project: development of a methodology to reliably monitor the obesity epidemic in childhood.
ABSTRACT The government has set a target to halt the rise in childhood obesity in those aged under 11 by 2010, but no system is in place to ascertain if this has been achieved. We aimed to develop a simple and reproducible methodology to monitor trends in childhood obesity.
A purposive sample of 10 primary schools and three secondary schools was selected. Children were measured with parental "opt out" consent in reception class, year 4, and year 8 (ages 5, 9, and 13 years, respectively). Measurements were compared with those obtained locally in 1996-2001. Calculations were then performed to ascertain the sample size required to confidently identify a halt in the rise in obesity using three growth measures.
A total of 999 children were measured with ascertainment of 95% in primary and 85% in secondary schools. The proportion of overweight and obese children aged 9 and 13 years had increased since 1996-2001, although only 9 year olds showed a significant rise. A general trend of an increase in obesity was observed with increasing age. Calculations showed that 1900-2400 children per age group are needed to detect a halt in the rise in obesity based on mean body mass index (BMI) standard deviation scores (SDS) by 2010 with 90% power, whereas 4200-10 500 children are needed for other measures.
We have developed a simple, cost effective methodology for accurately measuring the epidemic and recommend the use of mean BMI SDS for demonstrating if a halt has been achieved.
- SourceAvailable from: onlinelibrary.wiley.com[Show abstract] [Hide abstract]
ABSTRACT: This paper describes global (whole of Leeds) and local (super output area) analyses of the relationship between childhood obesity and many ‘obesogenic environment’ variables, such as deprivation, urbanisation, access to local amenities, and perceived local safety, as well as dietary and physical activity behaviours. The analyses identify the covariates with the strongest relationships with obesity, and highlight variation in these relationships across Leeds, thus identifying ‘at-risk’ populations. This paper seeks to demonstrate the importance of analysis at the micro-level in order to provide health planners with additional information with which to tailor interventions and health policies to prevent childhood obesity.Area 05/2008; 40(3):323 - 340. · 1.37 Impact Factor
- [Show abstract] [Hide abstract]
ABSTRACT: What is already known about this subject • Overweight and obesity prevalence estimates among children based on International Obesity Task Force definitions are substantially lower than estimates based on World Health Organization definitions. • Presence of a north–south gradient with the highest level of overweight found in southern European countries. • Intercountry comparisons of overweight and obesity in primary-school children in Europe based on measured data lack a similar data collection protocol. What this study adds • Unique dataset on overweight and obesity based on measured weights and heights in 6–9-year-old children from 12 European countries using a harmonized surveillance methodology. • Because of the use of a consistent data collection protocol, it is possible to perform valid multiple comparisons between countries. • It demonstrates wide variations in overweight and obesity prevalence estimates among primary-school children between European countries and regions. Summary Background: Nutritional surveillance in school-age children, using measured weight and height, is notPediatric Obesity 04/2013; 8(2):79-97. · 2.28 Impact Factor
- [Show abstract] [Hide abstract]
ABSTRACT: To examine the trends in BMI and adiposity among schoolchildren from Cuenca, Spain, during 2004–2010.Two methodologically comparable surveys were carried out in 2004 and 2010.Primary schools in Cuenca, a province in the middle of Spain.All schoolchildren aged 8–11 years, belonging to 4th and 5th grades from twenty schools in the Province of Cuenca, were invited to participate in both cross-sectional studies. Weight, height and body fat percentage (BF%) from bioelectrical impedance were measured with standardized procedures.The study included 550 boys and 539 girls in 2004, and 569 boys and 531 girls in 2010. In 2010, 8·1 % of children were underweight, 25·9 % overweight and 9·5 % obese. From 2004 to 2010, the prevalence of overweight rose from 21·6 % to 28·0 % (P = 0·004) and BF % increased from 22·6 % to 24·0 % (P = 0·001) among boys. No change was observed in overweight and BF % in girls, or in underweight and obesity in either sex. In boys, most of the distribution of BMI and BF % shifted to the right; in contrast, among girls no substantial change was apparent in the distribution of BMI and BF %.From 2004 to 2010 the prevalence of overweight and adiposity has continued to increase among boys. However, the obesity epidemic may have levelled off in girls. Given that the prevalence of childhood excess weight is still very high, the current evidence-based efforts to halt the obesity epidemic in Spain should be strengthened.
The TRENDS Project: development of a methodology
to reliably monitor the obesity epidemic in childhood
M C J Rudolf, R Levine, R Feltbower, A Connor, M Robinson
............................................................... ............................................................... .
See end of article for
Professor Mary Rudolf,
Belmont House, 3–5,
Belmont Grove, Leeds, LS2
9DE, UK; Mary.rudolf@
29 November 2005
Published Online First
14 December 2005
Arch Dis Child 2006;91:309–311. doi: 10.1136/adc.2005.078915
Aims: The government has set a target to halt the rise in childhood obesity in those aged under 11 by
2010, but no system is in place to ascertain if this has been achieved. We aimed to develop a simple and
reproducible methodology to monitor trends in childhood obesity.
Methods: A purposive sample of 10 primary schools and three secondary schools was selected. Children
were measured with parental ‘‘opt out’’ consent in reception class, year 4, and year 8 (ages 5, 9, and
13 years, respectively). Measurements were compared with those obtained locally in 1996–2001.
Calculations were then performed to ascertain the sample size required to confidently identify a halt in the
rise in obesity using three growth measures.
Results: A total of 999 children were measured with ascertainment of 95% in primary and 85% in
secondary schools. The proportion of overweight and obese children aged 9 and 13 years had increased
since 1996–2001, although only 9 year olds showed a significant rise. A general trend of an increase in
obesity was observed with increasing age. Calculations showed that 1900–2400 children per age group
are needed to detect a halt in the rise in obesity based on mean body mass index (BMI) standard deviation
scores (SDS) by 2010 with 90% power, whereas 4200–10 500 children are needed for other measures.
Conclusion: We have developed a simple, cost effective methodology for accurately measuring the
epidemic and recommend the use of mean BMI SDS for demonstrating if a halt has been achieved.
problem. In its recent white paper,1the government set a
target to halt the rise in obesity by 2010, but little thought has
gone into ascertaining the progression of the epidemic or
what data are needed to determine if the target has been
Relying solely on school entry measurement would be
inappropriate as the rise in obesity occurs later in childhood.2
The Department of Health is now debating whether to
monitor progress by annual ascertainment of the body mass
index (BMI) of all children,3which would require measure-
ment of height and weight. This would be an expensive
undertaking and there is an immediate need to determine
whether the aims of the government could be achieved by
developing a high quality public health monitoring system
that does not depend on the annual measurement of all
The TRENDS Project was funded in order to see if it was
possible to devise a simple, generalisable, cost effective
methodology. We aimed to do this by ascertaining current
levels of childhood obesity in marker schools, identified to
reflect local sociodemographic characteristics. We then
calculated how many children and schools would be required
to track obesity and determine if the government target was
besity in children has reached unprecedented levels,
and a plethora of initiatives (largely without any
evidence base) are being launched to tackle the
Leeds, in West Yorkshire in the United Kingdom, is ideally
suited to health services research studies as it comprises areas
with diverse socioeconomic profiles and contains a relatively
high proportion of ethnic minority groups. A purposive
sample of three secondary schools and ten primary schools
was selected for the project, with the aim of attaining a
sample that would approximate the sociodemographic make
up of the city. In order to maximise numbers, primary schools
with two class entry were selected, taking one inner-city and
one suburban school for each of the city’s five sectors. The
inner-city schools with the highest free school meal index
and largest proportion of minority ethnic pupils were chosen
in order to ensure that disadvantaged populations were well
represented. The secondary schools were selected from three
of the five sectors.
Ethics approval was obtained to measure the children with
parental ‘‘opt out’’ consent at school entry, year 4, and year 8.
All measurements were carried out by one school nurse (AC)
who received auxological training. The children wore light
clothing only and were measured in privacy. The following
were recorded: height to 0.1 cm with a freestanding
stadiometer (Raven, Dunmow, UK); weight to 0.1 kg; body
composition by body impedance (Tanita, West Drayton, UK);
and waist circumference 4 cm above the umbilicus.
Ethnicity was determined from school records and self
report. The Townsend score was used to assign a measure of
socioeconomic status (SES) to each child based on postcode
of residence. Figures were compared with those for Leeds in
the 2001 census.
BMI was calculated, and data were converted to standard
deviation scores (SDS).4 5Percentages of obese and over-
weight children were calculated using current clinical
criteria6based on the UK 1990 growth references (.91st
centile for overweight; .98th centile for obese). The
percentage of obese children was also calculated using the
International Obesity Task Force (IOTF) criteria.7Data were
analysed by age, SES, and ethnicity. The latter was
categorised as Caucasian, Asian, or other, as numbers were
too small to allow for meaningful analysis of other ethnic
The main purpose of the study was to establish a
methodology for monitoring by sample rather than by whole
Abbreviations: BMI, body mass index; IOTF, International Obesity Task
Force; SDS, standard deviation scores; SES, socioeconomic status
population. The most important requirement was to deter-
mine the size of the sample required to demonstrate a halt in
the rise in obesity. No guidance has been delivered to
determine either what constitutes a rise or what indicates it
has been curbed. We used the data that we had obtained to
perform sample size calculations in order to determine the
number of individuals and schools that would be required to
confirm that no change in obesity had occurred between 2004
and 2010. Calculations were analogous to those used for a
non-inferiority (equivalence) clinical trial, where measures of
obesity are ‘‘equivalent’’ over time if they are within some
pre-defined threshold or ‘‘margin of equivalence’’. We chose
the more stringent criteria of 90% power to minimise the risk
of a type II error, alongside a 5% level of significance.
We therefore considered three conservative scenarios,
bearing in mind that we had previously shown that mean
BMI SDS increased by 0.00–0.17/year (depending on age and
sex) between 1996 and 19984:
N a difference in mean BMI SDS of no more than 0.15
between 2004 and 2010,
N a difference in the proportion of obese children (BMI
.98th centile) of no more than 2% between 2004 and
N a difference in the proportion of obese children (using
IOTF cut offs) of no more than 2% between 2004 and 2010.
We also compared our growth data with data from a
previous local cohort of children in order to see if there had
been an increase in obesity over time. These measurements
had been carried out during the APPLES project, a
randomised controlled trial of a health promotion pro-
gramme to prevent obesity in primary schools.8 9Children
in the intervention and control arms in this study were
followed as a cohort between the years 1996 and 2001.2 5
A total of 999 children were measured with ascertainment of
95% in primary and 85% in secondary schools. Only eight
parents opted to exclude their children from the study (two
in reception class and six in secondary school). Three primary
school children could not be measured due to plaster casts or
wheelchairs, six were abroad, and 30 children were absent.
The secondary schools had problems in organisation on the
days visits were arranged and 62 pupils were not measured.
Analysis confirmed the sample was representative of the
city’s sociodemographic and ethnic profile.
The results are shown in table 1. Mean BMI SDS
(p=0.001) and the percentage of obese (p=0.017) and
overweight children (p=0.003) all increased significantly
with age. Figures had also increased compared with our
equivalent data from 1996–2001, with a statistically sig-
nificant rise observed in the percentage of overweight
children (BMI .91st centile) in year 4. There was also a
small increase in the SD of mean BMI SDS between 1996–
2001 and 2004 for both year 4 and year 8 children, raising the
possibility that a concurrent increase in underweight children
may be occurring. Subgroup analysis showed no differences
by ethnicity or SES.
Sample size calculations to determine the number of
individuals and schools required to confirm no increase in
obesity since 2004 for each scenario are shown in table 2. The
results show that between 1900 and 2400 children per age
group/year would be needed to detect a halt in the rise in
obesity by examining the change in mean BMI SDS by 2010
with 90% power, whereas 4200–10 500 children are needed if
other measures are to be considered.
Measures of obesity in Leeds school children measured in 2004 and in those measured in 1996–2001
Reception classYear 4Year 8
APPLESTRENDS (2004) APPLES (1996–98)TRENDS (2004)APPLES (2001)TRENDS (2004)
Number of children
Age range (years)
% Overweight children
with BMI .91st centile
% Obese children with
BMI .98th centile
% Obese using IOTF
Mean BMI SDS
SD of mean BMI SDS
SDS (95% CI)
Mean % body fat
by BIA (95%CI)
14.7 (11.3 to 18.7)
21.8 (18.0 to 26.1)*
24.0 (18.6 to 30.1)14.8 (12.0 to 17.9)*
20.3 (16.0 to 25.2)
–7.1 (4.7 to 10.2) 6.0 (4.2 to 8.2)10.3 (7.6 to 13.6)9.8 (6.8 to 13.7) 12.9 (8.8 to 18.0)
– 4.7 (2.8 to 7.4)3.3 (2.0 to 5.0)6.7 (4.5 to 9.6)5.1 (2.9 to 8.1)8.0 (4.8 to 12.3)
– 0.16 (0.05 to 0.28)0.21 (0.12 to 0.30)0.38 (0.26 to 0.50)0.35 (0.22 to 0.48)0.53 (0.37 to 0.69)
0.35 (0.27 to 0.43)*
0.6 (0.5 to 0.71)*
0.63 (0.49 to 0.76)
0.55 (0.35 to 0.75)
–––18.9 (18.0 to 19.7)23.0 (22.0 to 24.1)*
20.7 (19.5 to 21.9)*
?Intraclass correlation coefficients for BMI SDS calculated across schools.
*Significant differences (p,0.05) between 1996–2001 and 2004.
BIA, bioelectrical impedance analysis.
required to detect a halt* in the rise in obesity with 90%
power (a 5%) using three measures of obesity
Numbers of children in each school year
(5 year olds)
(9 year olds)
(13 year olds)
*Data shown in table 1 from 2004 were used as the baseline figures.
310 Rudolf, Levine, Feltbower, et al
Since this study was devised, a heated debate has developed
regarding the value or harm of routinely identifying children
as obese. The government has now suggested a policy of
measuring all school children annually with feedback to
parents in the hope that this will encourage families to
action. This proposal has been met with opposition by
paediatricians, the National Screening Committee, and
school nurses. The objections relate to the cost of universal
screening, the lack of any evidence that informing parents is
effective as an intervention, and the potential harm that may
accrue to obese children, for example in terms of their self
esteem or bullying.
Our results suggest an alternative to universal monitoring
by demonstrating a cost effective methodology for accurately
measuring the epidemic of obesity, using a minimal number
of schools. It also indicates that the use of routinely collected
data is quite inadequate for monitoring. Apart from their
inaccuracy,10 11measurements are only taken at school entry,
when levels are low compared with later on in childhood.
Reliance on this data could be falsely reassuring.
We found that the conduct of the study was straightfor-
ward in primary schools, with an excellent ascertainment of
95%, providing confidence in the quality of our results. Data
collection was more problematic in secondary schools as they
tended to be more disorganised in their cooperation and the
lower ascertainment rate reduces confidence in the results.
However, despite the difficulties, we would argue that it is
important to capture this age group as it is at this age that
obesity gets a grip and tracks most strongly into adult life. We
need to refine the process rather than take the option of
dropping secondary schools from a monitoring programme.
Our sample size calculations were based on a number of
hypothetical scenarios such that a halt in the increase in
obesity, however defined, would be considered to have
occurred if the margin of equivalence was no more than 0.15
SDS for mean BMI or a 2% increase in the proportion of obese
children. These are clearly arbitrary figures; the sample size
estimates depend not only on the length oftimeover which the
levels of obesity are being monitored but are also highly
sensitive to changes in the tolerance limits. However, our data
from 2004 can be used as a baseline for future studies, and
ultimately provide an indication of the numbers of children
who need to be sampled and measured in order to determine
whether the rise in childhood obesity has begun to level off.
We would recommend that change in mean BMI SDS is
used as the principal measure to determine if the epidemic
has been halted, as it utilises data from a cross-section of the
whole population rather than taking a minority who reach a
somewhat arbitrary cut off. It not only requires a substan-
tially smaller sample size compared to other measures but
also is probably more effective at monitoring the effects of
preventive strategies that are likely to have more of an impact
on the mildly overweight than those at the extreme. The
standard deviation of this measure provides some informa-
tion regarding the distribution within the population, and
could alert to the possibility of increasing numbers of
Discerning the percentage of obese children provides
additional information, although these figures need not be
used to ascertain if the government target has been met as
this requires much larger numbers. The IOTF criterion for
obesity is the more stringent measure and is useful for
international comparison, although clinicians tend to prefer
the clinical cut off of the 98th centile. Waist circumference
and body composition measurements are also of interest but
require more study before they can be considered useful;
waists are difficult to measure accurately and showed no
increase over time despite the increase in BMI SDS. Body
composition was problematic as it cannot be measured in
young children, although paediatric standards have recently
been devised for a portable device.12
We conclude that the TRENDS experience has enabled us
to initiate a public health monitoring system using a
representative sample of less than 10% of schools. It is
important to emphasise that a primary school is too small a
unit to assess change in obesity levels even if universal
monitoring was introduced, but accurate information using
our methodology will be available for each Primary Care
Trust in Leeds as well as by SES and ethnicity. Our approach
can be easily implemented, is cost effective, and could be
extended across the country to determine if government
targets are met over the next 5 years.
M C J Rudolf, M Robinson, University of Leeds and East Leeds PCT,
R Levine, The TRENDS Project, East Leeds PCT, Leeds, UK
R Feltbower, Centre for Epidemiology and Biostatistics, University of
Leeds, Leeds, UK
A Connor, School Nurse Service, East Leeds PCT, Leeds, UK
Competing interests: none declared
1 Department of Health. Choosing Health: making healthier choices easier.
London: The Stationery Office, 2004.
2 Rudolf MCJ, Sahota P, Barth JH, et al. Increasing prevalence of obesity in
primary school children. BMJ 2001;322:1094–5.
3 Armstrong D. Obesity - a growing issue, CMO Update March 2005
4 Rudolf MCJ, Cole TJ, Krom AJ, et al. Growth of primary school children: a
validation of the 1990 standards and their use in growth monitoring. Arch Dis
5 Rudolf MCJ, Greenwood DC, Cole TJ, et al. Rising obesity and expanding
waistlines in school children: a cohort study. Arch Dis Child 2003;89:235–7.
6 Rudolf MCJ. The obese child. Arch Dis Child E&P, 2004;89:ep57–62.
7 Cole TJ, Bellizzi MC, Flegal KM, et al. Establishing a standard definition for
child overweight and obesity worldwide: international survey. BMJ
8 Sahota P, Rudolf MCJ, Dixey R, et al. APPLES: process and impact evaluation
of a primary school based obesity prevention programme in the UK. BMJ
9 Sahota P, Rudolf MCJ, Dixey R, et al. APPLES: a primary school based
randomised controlled trial to reduce obesity risk factors. BMJ
10 Gully KS, Aitken S. Can the local prevalence of childhood obesity be
determined from routinely available data? Faculty for Public Health, University
of Wales Annual Scientific Meeting, Edinburgh, June, 2004.
11 Jones SE, James-Ellison M, Young S, et al. Monitoring trends in obesity in
South Wales using routine data. Arch Dis Child 2005;90:464–7.
12 Tanita. Body fat charts. West Drayton, UK: Tanita UK, 2005.
What is already known on this topic
N Obesity in childhood has reached unprecedented levels
in recent years
N The government has set a target to halt the rise in
childhood obesity by 2010, but no system is in place to
ascertain if this has been achieved
What this study adds
N We have developed a simple, cost effective methodol-
ogy for accurately measuring the obesity epidemic
N We recommend the use of mean BMI SDS for
demonstrating if a halt in the rise in obesity is achieved
The TRENDS Project311