To compare childhood obesity prevalence in England and the United States using different criteria.
Participants included 2- to 17-year olds in the Health Survey for England (HSE, n=33 563) and the US National Health and Nutrition Examination Survey (NHANES, n=14 540) 1999 through 2006. Mean body mass index (BMI) and prevalence of obesity were compared using the UK 1990, US 2000 Centers for Disease Control and International Obesity Task Force (IOTF) criteria.
English children at ages 2–5 years had a higher mean BMI than US children (mean difference (English minus US)=0.41 kg/m2, 95% confidence intervals (CI) 0.31–0.52). At age 8 years, mean BMI was lower in England (for ages 8–11 years, mean difference=−1.00 kg/m2, 95% CI −1.26 to −0.75; for ages 12–17 years, mean difference=−1.37 kg/m2, 95% CI −1.59 to −1.14). The IOTF criteria produced the lowest estimates of obesity prevalence. The 2000 Centre for Disease Control and Prevention (CDC) criteria produced the highest estimates in younger children and the UK 1990 criteria produced the highest in adolescents. Children aged 2–5 years in England had higher prevalence of obesity than those in the United States when using the 2000 CDC and UK 1990 criteria. US adolescents had the highest prevalence of obesity by age group using each of the three criteria.
The 2000 CDC and UK 1990 criteria give a higher prevalence of obesity in England than in the United States at ages 2–5 years; however, at age 8 years, the reverse is true. Estimates of childhood obesity prevalence rely on the criteria used, which has implications for surveillance and clinical practice.
Since 1970, the greatest increases in obesity in school children have been in North America and Western Europe (Wang and Lobstein, 2006). International comparisons are important for documenting and identifying country-level differences that might be used to suggest causal mechanisms (Cole et al., 2000). However, direct international comparisons are currently difficult because existing studies use different methods to define childhood obesity (Wang and Lobstein, 2006).
The International Obesity Taskforce (IOTF) has developed international criteria for overweight and obesity by gender from 2 to 18 years of age, derived using childhood body mass index (BMI) data from six countries. The criteria provide gender-specific values that are equivalent to a BMI of 25 and 30 kg/m2 and provide a means to compare childhood overweight and obesity prevalence internationally (Cole et al., 2000).
To our knowledge, only one study has made comparisons of childhood obesity between countries using different BMI reference criteria. In that study, childhood obesity prevalence was compared between China, Russia and the United States using IOTF, World Health Organization weight for height z-scores for <10-year olds and World Health Organization BMI percentiles for >10-year olds, as well as the US percentiles based on National Health and Nutrition Examination Study (NHANES) I (Must et al., 1991a, 1991b). Obesity prevalence and between-country comparisons varied depending on the criteria used (Wang and Wang, 2002).
In this paper, we compared mean BMI data and prevalence of overweight and obesity in children and adolescents (aged 2–17 years) from large cross-sectional surveys in England and the United States using three criteria: (1) UK 1990 reference charts for BMI from the Child Growth Foundation; (2) the US 2000 Centre for Disease Control and Prevention (CDC) BMI-for-age charts; and (3) IOTF. Our analysis had two aims: first, to compare the prevalence of overweight and obesity in England and the United States by age and gender, because the United States is leading the global epidemic of obesity and we were interested in whether childhood obesity prevalence in England (as representative of a Northern European country) had reached US levels. Second, to present these comparisons using the two country-specific methods of defining obesity.
Materials and methods
Data for these analyses came from two nationally representative cross-sectional surveys. For England, data were from repeated waves of the Health Survey for England (HSE) (http://www.dh.gov.uk/en/Publicationsandstatistics/PublishedSurvey/HealthSurveyForEngland/index.htm). Ethical approval for the HSE was obtained from all relevant Multi-Centre and Local Research Ethics Committees. Across the years included here, the number of participants aged 2–17 years varied from 2160 (in 2000) to 8114 (in 2002). Weighting was applied to take account of the underrepresentation of children in households with more than two children and the clustered, stratified multistage sample design (Sproston and Primatesta, 2003, 2004; Sproston and Mindell, 2006).
US data were from repeated waves of the NHANES. NHANES was approved by the National Centers for Health Statistics institutional review board, and all participants provided written informed consent. To account for the clustered, stratified multistage sample design, complex sampling, weighted estimates of population parameters were computed (National Center for Health Statistics, 2006).
From each study, we used data on participants aged 2–17 years for whom complete height and weight data were available from years 1999 to 2006. In HSE, 38 936 participants were eligible for inclusion in this age range and of them 33 563 (86.2%) had usable height and weight measurements. In NHANES, 15 500 participants were eligible for inclusion and of them 14 540 (93.8%) had usable height and weight measurements. Characteristics of the participants from the two surveys are shown in Table 1.
Measurement of height and weight
In the HSE, height and weight were measured in participants' homes by trained researchers. Height was measured using a portable stadiometer. Informants were asked to remove their shoes. One measurement was taken, with the participant stretching to the maximum height and the head aligned in the Frankfort horizontal plane. Height was recorded to the nearest millimetre. Weight was measured using Soehnle, Seca and Tanita electronic scales with a digital display. Participants were asked to remove their shoes and any bulky clothing. Weight was recorded to the nearest 100 g.
In NHANES, height and weight were measured in the body measurement room of the NHANES Mobile Examination Centre by trained health technicians. Height was measured using a fixed stadiometer. One measurement was taken, with participants instructed to breathe in and stand as tall as possible and the head aligned in the Frankfort horizontal plane. Height was recorded to the nearest millimetre. Weight was measured using a Toledo electronic scale with a digital display. Participants were asked to wear only underwear, disposable paper gowns and foam slippers. Weight was measured in pounds and automatically converted to kilograms, recorded to the nearest 100 g.
The UK 1990 age-related BMI reference curves cover the age range from birth to 23 years, and the curves are presented as nine centiles in the format of Cole (0.4th, 3rd, 10th, 25th, 50th, 75th, 90th, 97th and 99.6th centiles) (Cole et al., 1995). The reference sample of children was based on data from 11 surveys on gender, age, height and weight collected between 1978 and 1990 for 15 636 boys and 14 899 girls, aged from 33 weeks to 23 years. Exact age was calculated from the dates of birth and the dates of the measurements for all participants. Summary centile curves were fitted to the data using Cole's LMS (lambda mu sigma) method and penalised likelihood (Cole et al., 1995).
The US 2000 CDC BMI-for-age charts for ages 2–20 years were developed with data from five national health examination surveys between 1963 and 1994 and from limited supplementary data. Smoothed percentile curves were developed in two stages: first, selected empirical percentiles were smoothed with a variety of parametric and non-parametric procedures; in the second stage, parameters were created to obtain final curves, additional percentiles and z-scores (Kuczmarski et al., 2002).
The IOTF criteria for obesity were based on international survey data from six large nationally representative cross-sectional growth studies in Brazil, Great Britain, Hong Kong, The Netherlands, Singapore,and the United States from birth to 25 years of age (Cole et al., 2000). For each of the surveys, centile curves were drawn that at age 18 passed through the widely used criteria of 25 and 30 kg/m2 for adult overweight and obesity. The resulting curves were averaged to provide age- and gender-specific criteria from 2 to 18 years.
We estimated the prevalence of overweight and obesity in children and adolescents according to the three sets of criteria from the growth curves described above. Using the UK 1990 and 2000 CDC BMI-for-age reference charts for BMI, we classified as overweight/obese those who had an age–gender-specific BMI above the 85th percentile, and classified as obese those who had a BMI above the 95th percentile. Using the IOTF classification, we used participants' defined age–gender-specific overweight and obesity criteria. This meant that for each of the three methods, each participant was graded as being of recommended weight, overweight/obese or obese. Where growth charts referred to months rather than to years of age, we used the nearest mid-year figures. Using the appropriate weighting for each study, as described above, we estimated the proportion of participants in each country who were above the overweight and obesity criteria from the three growth charts. In addition, we compared the trends in mean BMI and prevalence of obesity using the IOTF criteria for 2-year periods across the 8 years for all children combined. All analyses were conducted using Stata SE 10.1 (College Station, TX, USA).
Mean BMI values for children and adolescents in England and the United States, by age in years and with 95% confidence intervals (CI), are shown in Figure 1 and Table 1. As expected, the distribution of BMI is positively skewed for all age groups and for both countries. At younger ages, English children had higher mean BMI than their peers in the United States but this pattern was reversed in older children. Specifically, higher mean BMI was found in English compared with US children at ages 2–5 years (mean difference England minus United States=0.41 kg/m2, 95% CI 0.31 to 0.52) but lower mean BMI in English compared with US children/adolescents from ages 8 to 17 years: at ages 8–11 years the mean difference was −1.00 kg/m2 (95% CI −1.26 to −0.75) and at ages 12–17 years the mean difference was −1.37 kg/m2 (95% CI −1.59 to −1.14). At ages 5–7 years the mean BMI was similar in the two countries (mean difference=0.09 kg/m2, 95% CI −0.07 to 0.25).
There were gender differences in the ages at which there were the greatest differences in BMI between English and US children. In girls, the biggest differences in BMI were at age 4 years, when English girls had a mean BMI 0.42 kg/m2 (95% CI 0.25–0.60) higher than US girls, and at age 13 years, when English girls had a mean BMI −1.62 kg/m2 (95% CI −1.88 to −1.36) lower than US girls. In boys, the biggest differences were at age 2 years, when English boys had a mean BMI 0.64 kg/m2 (95% CI 0.43–0.84) higher than US boys, and at age 16, when English boys had a mean BMI −2.40 kg/m2 (95% CI −2.69 to −2.12) lower than US boys.
Prevalence of overweight and obesity
Tables 2 and 3 show the prevalence of overweight/obesity and obesity for England and the United States in age- and gender-specific groups according to each of the three criteria. Using the 2000 CDC criteria, the prevalence of obesity in 2- to 5-year olds (girls and boys combined) reflected the pattern for mean BMI: prevalence of obesity in English children was 14.2% (95% CI 13.1–15.3) and in US children in this age group it was 11.9 (95% CI 10.2–13.6; mean difference=2.3%, 95% CI −0.3 to 4.3). The mean difference was similar for the UK 1990 criteria (mean difference=2.0%, 95% CI 0.0–3.9). There was little between-country difference in the prevalence of obesity in this age group in either gender using the IOTF criteria (Table 3).
Estimates of obesity for older children were consistent with patterns of mean BMI. Although the actual estimates of obesity prevalence varied considerably by the criteria used in both countries, the mean difference in prevalence was similar for each criteria. For example, at ages 12–17 years, the prevalence of obesity was markedly lower in English than in US adolescents using the 2000 CDC criteria for obesity (difference=−7.4%, 95% CI −5.8 to −9.0), using the IOTF criteria (difference=−6.6%, 95% CI −4.7 to −8.5) and the UK 1990 criteria (difference=−8.3%, 95% CI −10.0 to −6.5). US adolescents (aged 12–17 years) had the highest prevalence of obesity by age group and by country using each of the three criteria, with the highest estimate being 26.4% using the UK 1990 criteria.
Within country and gender groups, patterns by age differed according to the criteria used. For example, applied to English boys each of the three criteria suggested a different pattern by age: 2000 CDC criteria indicated a higher prevalence of obesity in 2- to 5-year olds (15.8%, 95% CI 14.2–17.3) than in 12- to 17-year olds (9.0%, 95% CI 8.1–9.9) but the UK 1990 criteria indicated the opposite: a lower prevalence of obesity in 2- to 5-year olds (14.7%, 95% CI 13.2–16.2) than in 12- to 17-year olds (17.7%, 95% CI 16.5–19.0). For the same groups, IOTF criteria showed almost no difference between 2- to 5-year olds and 12- to 17-year olds, with estimated prevalences of obesity of 6.0% (95% CI 5.0–7.1) and 6.1% (95% CI 5.3–6.9), respectively. The same pattern of differences, according to the different criteria, was evident for obesity in English girls. In US boys and girls, all three criteria identified increasing prevalence of obesity with age but the relative proportions were different: in US girls, for example, the 2000 CDC criteria identified 11.7% (95% CI 9.5–13.8) of 2- to 5-year olds and 16.0% (95% CI 14.1–17.9) of 12- to 17-year olds as obese, whereas the UK 1990 criteria identified a slightly lower percentage of 2- to 5-year olds (10.9%, 95% CI 8.9–12.9) but a higher percentage of 12- to 17-year olds (24.3%, 95% CI 22.0–26.6) as obese.
Figures 2 and 3 show how the criteria for obesity vary by age and gender. These illustrate why we see the differences in patterns by age, gender and country as described above. For example, in boys the IOTF criteria are higher than either of the two country-specific criteria at all ages, resulting in a lower prevalence of obesity. In contrast, for girls the IOTF criteria are similar to that of the UK 1990 growth curve 95th percentile from ages 2 to 7 years, higher than both the country-specific criteria until age 15, and is similar to the 2000 CDC curve 95th percentile (but not the UK 1990 curve) at ages 16 and 17.
We analysed the mean BMI and obesity prevalence (using only the IOTF criteria) for age groups 2–5, 6–11 and 12–19 years, by 2-year periods to check whether there were between-country differences in trends across time in BMI and obesity prevalence (Supplementary Table 4). There is no evidence of a difference in trends between the countries over time. However, the numbers within each time period strata are small. There is also no evidence of a difference between consecutive 2-year periods for the prevalence of obesity or mean BMI.
Mean BMI differed between the countries by age, with English children having a higher mean BMI than their US peers at ages 2–5 years but lower mean BMI at 8 years and above. For public health, the prevalence of obesity is particularly important, but because population distributions of BMI are positively skewed, the patterns of mean BMI do not necessarily reflect patterns of obesity. We found that US adolescents (ages 12 through 17) had the highest prevalence of obesity of all age groups and in either country by all three criteria (UK 1990, 2000 CDC and IOTF).
A key finding of our analyses is that there were marked differences in obesity prevalence in each country and by age and gender when applying each of the three different criteria. However, general differences in prevalence between England and the United States were consistent irrespective of which criteria were used, except at younger ages, at which for all three criteria we found English children aged 2–5 years had a higher prevalence than US children, which was least marked for the IOTF criteria.
These findings have important implications for public health surveillance, clinical practice and research into childhood obesity. For example, the higher prevalence of obesity in young children in England compared with the United States might suggest that efforts to prevent childhood obesity in England should focus on surveillance and intervention research in preschool children. Furthermore, using IOTF criteria, the overall prevalence of childhood obesity in both England and the United States in all age groups was considerably lower than using either of the country-specific criteria, which is in keeping with studies that have found that the IOTF criteria have poor sensitivity for obesity (Reilly et al., 2000, 2010; Zimmermann et al., 2004). Using the IOTF criteria, one would be likely to estimate the magnitude of the childhood obesity epidemic as far smaller than when using country-specific analyses.
Strengths and weaknesses of the study
A key strength of this analysis is the use of data collected as part of nationally representative surveys with standardised methods. We have used three established methods of classifying children and adolescents as overweight/obese and obese. International comparisons of the prevalence of obesity have relied on published studies from single countries using the classification system and years of coverage that were used in the initial publication (Wang and Lobstein, 2006). Our findings suggest that relying on country-specific classifications is likely to present problems in terms of making meaningful comparisons between as well as within countries.
A limitation of this study is that the prevalence estimates by age and gender are pooled over an 8-year period, during which time the prevalence of obesity increased and more recently has appeared to plateau (although longer-term trends are required before it can be concluded that the epidemic has reached its peak in these countries) (Ogden et al., 2008; The Information Centre, 2008). The relatively small numbers of obese children/adolescents in each age group by year make it impossible to draw conclusions about annual differences between countries; however, there appeared to be no difference between the two countries in trends over time when we examined the data in 2-year periods. The mean and median age of the 2- to 5-year-old English children was slightly older than that of US children; this is unlikely to explain the differences seen in mean BMI because older children in the 2- to 5-year age group have decreasing BMI. We have not attempted to distinguish how the prevalence of obesity differs between the two countries in relation to their ethnic and socioeconomic profiles, because the two surveys assess these factors differently.
Two studies suggest that all three criteria for defining childhood obesity used here have similar and high specificity but that the IOTF criteria may have lower sensitivity when compared with a gold standard of high body fat (Reilly et al., 2000, 2010; Zimmermann et al., 2004). It could be argued that the best method for defining childhood obesity should be that which best predicts adverse obesity-related short- and long-term outcomes. We therefore recommend that the IOTF and one or more country-specific criteria be used and compared in research studies until it is clear which provides the best method for identifying children at greatest adverse consequences from obesity.
US adolescents aged 12–17 years had the highest prevalence of obesity by age group compared with England using each of the three criteria assessed. In contrast, English children aged 2–5 years had higher mean BMI and higher prevalence of obesity. It is unclear what factors drive these between-country differences, as there is currently a lack of high-quality studies exploring country-level differences in risk factors for obesity and examining how these explain the observed differences. Differences in the age of adiposity rebound and onset of puberty could account for some of the differences in mean BMI and obesity; both have been documented to occur earlier in US compared with European populations (Wang, 2004; Euling et al., 2008; Wyshak and Frisch, 1982). However, it has not been possible to examine this further using these data.
Our results demonstrate very marked differences in the prevalence of childhood obesity by age, gender and country when different, established methods for defining childhood obesity are applied to the data. For public health surveillance, clinical practice, research and to permit meaningful international comparisons in prevalence of childhood obesity, it would be ideal to identify a gold-standard method of defining childhood obesity and to apply this method universally. Until there is clear evidence to support adopting one method to apply to all data, both IOTF and one or more country-specific criteria should be reported in future country-level surveillance systems and research. This would make international comparisons feasible and contribute to our understanding of the best method for defining childhood obesity, while preventing selective use of different criteria for political or publication reasons.
No funding was received for the design and conduct of the study; for collection, management, analysis and interpretation of the data; and for preparation, review or approval of the manuscript. No specific funding was received for the work. IAL is funded by NIHR, the UK National Institute for Health Research, as part of PenCLAHRC, the Peninsula Collaboration for Leadership in Applied Health Research and Care. RRK is supported by the South West Public Health Training Scheme. DAL is funded by the Higher Education Funding Council for England. DAL has grants from the UK Medical Research Council, the Wellcome Trust and the British Heart Foundation, although this work was not directly funded by any of these. RJ is funded from a Career Development Fellowship supported by the National Institute for Health Research. The views expressed in this publication are those of the authors and not necessarily those of the NHS, the National Institute for Health Research or the Department of Health.
About this article
Supplementary Information accompanies the paper on European Journal of Clinical Nutrition website (http://www.nature.com/ejcn)