Previous data indicate a rapidly increasing prevalence of obesity and overweight among English children and an emerging socioeconomic gradient in prevalence. The main aim of this study was to update the prevalence trends among school-age children and assess the changing socioeconomic gradient.
A series of nationally representative household-based health surveys conducted between 1997 and 2007 in England.
15 271 white children (7880 boys) aged 5 to 10 years with measured height and weight.
Height and weight were directly measured by trained fieldworkers. Overweight (including obesity) and obesity prevalence were calculated using the international body mass index cut-offs. Socioeconomic position (SEP) score was a composite score based on income and social class. Multiple linear regression assessed the prevalence odds with time point (1997/8, 2000/1, 2002/3, 2004/5, 2006/7) as the main exposure. Linear interaction terms of time by SEP were also tested for.
There are signs that the overweight and obesity trend has levelled off from 2002/3 to 2006/7. The odds ratio (OR) for overweight in 2006/7 compared with 2002/3 was 0.99 (95% CI 0.88–1.11) and for obesity OR=1.06 (0.86–1.29). The socioeconomic gradient has increased in recent years, particularly in 2006/7. Compared to 1997/8, the 2006/7 age and sex-adjusted OR for overweight was 1.88 (1.52 to 2.33) in low-SEP, 1.25 (1.04 to 1.50) in middle-SEP, and 1.13 (0.86 to 1.48) in high-SEP children.
Childhood obesity and overweight prevalence among school-age children in England has stabilized in recent years, but children from lower socio-economic strata have not benefited from this trend. There is an urgent need to reduce socio-economic disparities in childhood overweight and obesity.
Over the last few decades there has been a worldwide increase in paediatric obesity affecting both developed and developing countries.1, 2 Excess adiposity at a young age is linked to immediate and long-term health risks, including increased risk of asthma3 and type 2 diabetes,4 and because of the persistence of the condition into adulthood,5, 6 increased middle-age mortality and morbidity regardless of adult weight status.7, 8, 9 Socioeconomic position (SEP) is linked to adiposity but the relationship is complex and varies with age, population, sex, ethnicity10, 11 and the type of adiposity indicator.11 A recent review of cross-sectional studies published between 1990 and 200511 found that SEP was inversely associated with children's overweight or obesity in 42% of the reviewed studies, with another 31% reporting a mixture of inverse and no associations. The choice of SEP indicator clearly influenced these relationships and parental education showed the most consistent inverse relationship with children's obesity risk. The evidence using income or social class is less conclusive and studies using composite SEP indices are few and far between.11 This contrasts with public health and epidemiological research recommendations,12, 13 which have emphasized the value of using area-level SEP indicators in health research. The importance of taking into account multiple SEP indicators has also been emphasized14, 15 because studies concentrating on a single SEP indicator fail to fully describe the relationships with excess adiposity or its proxies (for example, body mass index (BMI)). This idea is supported by the aforementioned review,11 which found that studies using multiple SEP indicators or composite indices were more likely to show inverse relationships.
We have previously documented16 a trend of expanding social disparities in children's overweight and obesity prevalence in England between 1973 and 2003, with children from less advantaged socioeconomic backgrounds (lower income and manual occupation households) showing disproportionately higher increases in prevalence than children in households with higher income or non-manual backgrounds. We do not know whether this trend, which covered a relatively short period of time (1997–2003) and was of borderline statistical importance,16 has continued in recent years. A better understanding of the relationship between SEP and children's obesity and the dynamics of this relationship over time will inform public health policies to tackle paediatric obesity and social inequalities in health.
The main objectives of this work were as follows: (a) to update the most recent overweight and obesity prevalence trends among school-age children in England using identical methodology to previously published long-term trends16 and (b) to examine the changes over time in the relationship between obesity and multiple SEP indicators.
Subjects and methods
The Health Survey for England17, 18 consists of annual nationally representative random samples of the free-living general population, drawn using multi-stage stratified probability sampling, with postcode sectors as primary sampling units and the Postcode Address File as the household sampling frame. Up to two children aged 2–15 years from each household were recruited. In this study, we used data from 1997–2007 surveys omitting 1999. Individual response rates for children ranged from 85% in 1997 to 66% in 2007. The samples had relatively few non-white children, so we included only white children in the analyses. Ethical approval had been granted by the London Medical Research Ethics Committee or the local research ethics committees before each annual data collection cycle.
The Health Survey for England measuring methodologies and instrumentation were identical across time and has been described in detail previously.16, 18 In brief, height was measured using Harpenden stadiometers (Chasmors, London, UK) and weight was measured using electronic digital scales (Tanita, Akita, Japan). For comparability with previous prevalence estimates,16 weight was adjusted for children's clothing using the same method as before.16, 19 Exact age was used to calculate prevalence rates, but was grouped into 5–7 and 8–10 years for the presentation of results.
Household income was converted to equivalized annual household income adjusted for the number of persons in the household by using McClements scoring system.17, 18 To examine whether SEP has a cumulative effect on obesity prevalence over time, we developed an aggregate SEP score based on the family's position on income (quintiles) and social class scales (I, II, III-manual, III-nonmanual, IV and V). The Registrar-General's social class is based on occupation of the head of the household.20 For each indicator, children were assigned between zero (lowest income quintile, social class V) and four (highest income quintile) or five (social class I) points. The resulting score ranged from zero (lowest SEP) to nine (highest SEP). We grouped SEP score to low (0–3), medium (4–6) and high (7–9) to give similar size groups. Similar composite scores of SEP score have previously been used by us21 and others.22
Data handing and statistical analyses
Overweight (including obesity) and obesity rates were computed using age–sex-specific international BMI cut-offs23 corresponding to adult BMI cut-offs of 25 and 30 kg m−2 at 18 years of age. We computed the year-on-year prevalence of overweight and obesity between 1997 and 2007. The Health Survey for England years were also merged in pairs (1997/8; 2000/01; 2002/3; 2004/5; and 2006/7) to increase the statistical power. The baseline year was 1997 as household income data were not available before then. We compared rates for surveys from 2004 to 2007 with the published16 2002/3 rates by using the likelihood ratio tests. The prevalence trends of 1997–2007 were tested using multiple logistic regression with obesity or overweight as the binary outcome, time point as exposure (reference: 2002/3), and social class, income, sex and age group as covariables.
We tested for, but failed to find, interactions of time point by sex and by age group (5–7 and 8–10 years) and second-order interactions between time point, sex and age group. Hence, the remaining multivariable analyses were age group and sex-adjusted, but not stratified by age and sex. The change over time in the relationship between prevalence and SEP was tested by the following: (a) fitting linear interactions between socioeconomic indicators (income and SEP score) and time point, (b) comparing the prevalence between SEP groups at each time point and (c) calculating the adjusted odds for overweight and obesity by SEP group at each time point using multiple logistic regression.
Sample sizes and missing values
There were 19 203 children between the age of 5 and 10 years (9718 boys) in the included Health Survey for England years. After the exclusions of 2582 non-White children and another 1350 with invalid BMI (missing height or weight information, or both) the core dataset comprised 15 271 children (7880 boys). Owing to 1622 missing values on income or parental social class, SEP analyses included 13 649 cases (6902 boys). Table 1 summarizes the characteristics of the sample by 2-year time point.
Overall prevalence trend
Figure 1 shows the trends in overweight and obesity prevalence from 1997 to 2007, with both levelling off after 2002. Table 2 presents the overweight and obesity prevalence between 2002/3 and 2006/7 by age and sex. With the exception of obesity in boys aged 5–7, neither overweight nor obesity prevalence changed noticeably during the period.
Table 3 confirms this with the multivariable-adjusted odds for being overweight (0.99, 95% confidence interval 0.88–1.11) or obese (1.06, 0.86–1.29) in 2006/7 were not significantly different from 2002/03.
Socioeconomic differences in the trends
Overweight and obesity prevalence by income and SEP score at each time point are shown in Figures 2 and 3 respectively, where disparities tended to increase over time. In Figure 2, the gradient in prevalence between high and low income groups was significant for overweight in boys (P=0.04) and girls (P=0.003) in 2006/7, and for obesity in girls in 2002/3 (P=0.001), 2004/5 (P=0.005) and 2006/7 (P=0.04). The SEP score gradient in Figure 3 also increased over time, reaching significance in 2006/7 for boys' overweight (P<0.001) and obesity (P=0.002), when obesity prevalence in the low SEP score group was twice than that in the higher groups. Similar gradients in girls' obesity prevalence were seen in 2002/3 (P<0.001), 2004/5 (P=0.03) and 2006/7 (P=0.03).
There was a strong interaction between the SEP score and time point for both overweight (P=0.002) and obesity (P=0.02). Figure 4 shows the sex and age-adjusted odds of overweight and obesity by SEP score group and time point relative to 1997/8, and only the low SEP score group showed rising trends over time.
Using nationally representative data, we have examined the 10-year time trends of overweight and obesity prevalence in children by SEP. Our findings suggest that despite stabilization of the overall prevalence between 2004 and 2007, social disparities continue to grow at the expense of children in lower SEP groups.
The results are in agreement with studies from other counties showing that, at least in the developed world, the paediatric obesity epidemic has been slowing in recent years. Recent trends in overweight or obesity prevalence have stabilized or reversed in France (3–1424 and 7–925 years), Switzerland (6–13 years)26 and Sweden (girls 10–11 years).27 In the United States of America the prevalence of high BMI for age among children or adolescents has not changed significantly between 2003–2004 and 2005–2006 or between 1999 and 2006.28 There are several possible explanations for this finding. At the individual and family level it may relate to the extensive media attention that issues around obesity have received in recent years29 and a subsequent rise in body weight awareness and health consciousness. At the policy level it could be because of anti-obesity policies and policy targets, such as the English government's 2004 Public Service Agreement, thereby setting a target to halt the year-on-year increase of obesity in children under 11 years of age30 and the French National Nutrition and Health Programme.31 However, no formal evaluations of such interventions exist and we can only speculate as to the likely reasons.
Evidence on time trends of socioeconomic disparities in prevalence is less clear. A major US study10 concluded that income-based socioeconomic disparities in childhood overweight weakened between 1971–1975 and 1999–2002. We cannot directly compare these results with ours because of the differing time periods, but Wang and Zhang's conclusion10 is not supported by their own data for boys of 2–9 years, wherein a strong income gradient was present only for the most recent time point 1999–2002 (low SEP: 17.4%, middle SEP: 15.0%, high SEP: 9.7%, P<0.05). In contrast, a French study24 concluded that social disparities in overweight for children of 3–14 years did not increase during 1999–2007, a time period similar to ours. Our own socioeconomic gradient became stronger in 2004/5 and 2006/7 for boys' overweight and obesity as well as girls' overweight, and this was more pronounced for SEP score (Figures 3 and 4) than for income (Figure 2), supporting previous suggestions that social disparities in health12, 13, 14 or specifically paediatric obesity15 are best described using multi-domain or composite SEP indicators.
Although the previous Health Survey for England data suggest that lower SEP children are less likely than middle and high SEP children to have a healthy diet18, 32 or to exercise or take part in sports,18, 32 our data do not discriminate between possible explanations for the growing social disparities over time that we document. If we assume that health policy interventions and family/individual health awareness are responsible for the stabilization of the prevalence trend in less deprived children, the expansion of socioeconomic disparities may partly be due to difficulties to reach and communicate health messages to families from lower socioeconomic groups. Previous research shows that higher socioeconomic groups tend to follow recommendations for health behaviours33 and respond more actively to health-related media messages34 than lower socioeconomic groups. It is possible that the recent exponential increase in obesity and obesity-related (diet and physical activity) media messages29 has been received more positively by non-manual than by manual families. As lower socioeconomic groups tend to be wary of measures and messages aimed at changing their lifestyle because they see them as ‘nanny-statism’ that erodes their autonomy,35 it is possible that policies targeting children's eating and physical activity habits have been perceived less favourably by lower-income and social-class groups.
The main strengths of our study are the objective weight and height measuring methodology that remained the same across survey years, the consistency across time sampling methodology, the nationally representative and relatively large samples of children and the availability of more than one SEP indicator. On the other hand, we acknowledge that our study also had several limitations. The US study by Wang and Zhang10 showed that there are important interactions between ethnicity and SEP, but our analyses did not include children from minority ethnic groups because of their small numbers. The decline in response rates in recent years may have introduced respondent bias as individuals (and presumably families) from lower SEP groups are more likely to be both non-responders in survey research36, 37 and obese or overweight.38 Further response bias may have been introduced by the considerably higher percentages of BMI missing data in the 2000s (8–14%) compared with 1997/98 (2%). Refusal to co-operate with weight measurements, which was one of the main reasons for missing BMI data, increased from 1.0% in 1997/8 to 3.0% in 2002/3 and 3.7% in 2006/7. If families, who rejected participation in the survey and children, refused to do the weight measurement then they were more likely to be obese, the obesity rates we reported may be an underestimation of the actual prevalence in most recent years.
The previously documented rapid upward trends in obesity and overweight among primary school age children have levelled off between 2002 and 2007. Over the same time period socioeconomic disparities have widened. Our results support the need for urgent action to tackle socioeconomic inequalities in children's health.
The Health Survey for England is funded by the English Department of Health and the Health and Social Care Information Centre. This work was not supported by any specific source of funding. We thank the families and children who participated in the Health Survey for England.