Over the past two decades the prevalence of overweight and obesity has increased substantially among young children in all four UK countries.1^–5 Recent estimates suggest that at least a quarter of preschool children are overweight or obese,2^–5 similar to levels found in the USA.6 Tackling childhood obesity is a priority for the UK government4 7^–9 and there is a public service agreement target to reduce the proportion of overweight and obese children to 2000 levels by 2020.10

Although data on childhood overweight and obesity are necessary to monitor trends, differences in data collection and interpretation often prevent comparisons between areas. The national health surveys from England, Scotland and Northern Ireland all report on childhood overweight and obesity, but the dates of the surveys, age ranges and definitions of obesity are not comparable.2 3 11 The national survey in Wales collected height and weight in children for the first time in 2007.12 Recently, the Department of Health (England) developed the National Childhood Obesity Database to record the annual height and weight of all school children in Reception Year (age 4–5 years) and Year 6 (age 10–11 years). However, in 2006, only 57% of eligible Reception Year children were weighed and measured, with lower levels of overweight reported in areas with low response rates, suggesting bias.13 Regional differences in obesity have been published using the Health Survey for England, but these were based on a broad age range (2–15 years) due to small sample sizes.2 Therefore, there are no accurate assessments of regional disparities in overweight or obesity among young children in England, and differences between the four UK countries cannot be assessed because of a lack of comparable data.

Differences in the prevalence of obesity between regions may reflect discrepancies in risk factors for obesity at either the individual or area level (or both). Regional differences are also important because they may be used for resource allocation and to infer the success or otherwise of area based policies to tackle obesity. We examined country and regional differences in overweight (including obesity) in a nationally representative, contemporary cohort of children aged 3 years and determined whether the differences persisted after adjustment for individual risk factors for obesity.

METHODS

Participants

The Millennium Cohort Study (MCS) is a nationally representative prospective study of British children born in the new century. A stratified clustered sampling framework was used to over-represent children living in disadvantaged areas and from ethnic minority groups. Families eligible for Child Benefit (a universal payment to families with children) and resident in England, Wales, Scotland or Northern Ireland when their child was aged 9 months were invited to participate (response rate 72%).14 15 The original cohort comprised 18 819 children (18 553 families) born between September 2000 and January 2002. Of the 18 296 singleton infants included at the first contact, 80% (14 630) participated at the second, which occurred between September 2003 and January 2005, when the children were a mean age of 37.7 months (SD 2.5).16 There was a disproportionate loss of children from Northern Ireland, in electoral wards in England defined as “ethnic” (based on the 1991 Census, if at least 30% of residents were from an ethnic minority group) and in “disadvantaged” wards from all UK countries (the upper quartile of the Child Poverty Index).16 Main respondents (over 99% were natural mothers) were interviewed in the home at both contacts. Data from both surveys were obtained from the UK Data Archive, University of Essex. The MCS received ethical approval from the South West and London Multi-Centre Research Ethics Committees for the first and second contacts, respectively.17

Of the 14 630 singletons, 13 194 had complete and plausible child height and weight data at the second contact. Families were excluded from the analysis if the main respondent was not female (185), there were two cohort children from the same family (10), or the child had a missing height or weight (802) or a height-for-age, weight-for-age or body mass index-for-age z score ⩽−5 or ⩾5 (BMI is weight/height2) (467). Some participants had more than one exclusion criterion. Children were significantly less likely to be included in the final sample if they were from an ethnic minority group or lower income family, or if their mother was a lone parent, had a lower academic qualification or had lower socio-economic circumstances (p<0.01); however, the absolute differences were small. There were no differences in country or region of residence for those who were or were not included in the final sample (p = 0.2).

Definition of overweight

At the second contact, trained interviewers measured the children’s weights and heights without shoes or outdoor clothing. The children were weighed to the nearest 0.1 kg using Tanita HD-305 scales (Tanita UK, Yiewsley, Middlesex, UK) and height was measured to the nearest 0.1 cm using Leicester Height Measure Stadiometers (Seca, Birmingham, UK). The primary outcome measure was childhood overweight (including obesity) defined by the International Obesity TaskForce (IOTF) cut-offs for BMI, which are age and sex specific.18 There were no gender differences in early childhood overweight (p = 0.2), so results are reported together. No other measurements of body size were collected.

Country and region of residence

The child’s country and region of residence at the second contact was categorised as England, Wales, Scotland or Northern Ireland and, within England, as one of the following nine regions: London, East Midlands, East, North East, North West, South East, South West, West Midlands, or Yorkshire and Humberside.

Individual risk factors for overweight

Individual risk factors for overweight were based on maternal self-report. At the first contact, the child’s ethnicity was reported by the mother (out of 16 possible choices) and classified according to guidelines from the Office for National Statistics,19 maternal socio-economic circumstances were categorised according to the National Statistics Socio-economic Classification,20 maternal education was defined as the highest academic qualification attained, and lone motherhood status was defined as being a lone mother when the child was aged 9 months. Family income reported at the second contact was used unless missing, when values from the first contact were substituted (1589). At the first contact, mothers reported their prepregnancy weight and current height, and those with BMI ⩾25 were classified as overweight (including obesity). Mothers also reported their age at their first live birth, whether they smoked any cigarettes during the pregnancy with the cohort child, their child’s birthweight, duration of breastfeeding, and age at which the infant was introduced to solid foods. At the second contact, mothers reported the number of hours the child watched television or videos daily.

Statistical analysis

All analyses were conducted using STATA statistical software, version 9.2 SE (StataCorp, College Station, TX, USA), with survey commands to account for the clustered sampling framework and obtain robust standard errors. Weighted percentages were derived and logistic regression analyses were conducted using sample and non-response weights to allow for the clustered sampling design and non-response between contacts. Wald tests were used to obtain p values. Four logistic regression analyses were conducted to calculate odds ratios for childhood overweight at age 3 years by country and, separately, English region of residence. The first was a univariable regression analysis. The second analysis only adjusted for the child’s ethnicity. The third analysis also adjusted for the following socio-demographic characteristics: maternal socio-economic circumstances, family income, maternal highest academic qualification, lone motherhood status, and age at first live birth. The fourth analysis adjusted for the socio-demographic characteristics already listed as well as the following other risk factors for overweight: maternal prepregnancy overweight, smoking during pregnancy, birthweight, breastfeeding duration, introduction of solid foods, and television viewing daily.

RESULTS

Socio-demographic characteristics and other risk factors for overweight in children and their families from each UK country are shown in table 1 and from each English region in table 2. Overall, 18.0% (2411) of children were overweight and 5.0% (691) were obese at age 3 years; however, this varied by country and region of residence (table 3). At the country level, the order of prevalence of early childhood overweight (including obesity), ranked lowest to highest, was England, Scotland, Wales and Northern Ireland. Within England, the East had the lowest prevalence of overweight and the North East the highest. In univariable analyses, children from Northern Ireland and Wales were more likely to be overweight than children from England. Within England, children from the East and South East were less likely to be overweight than children from London. There were no differences in overweight between Scotland and England, or the other English regions and London, before or after adjustment.

Among children from Northern Ireland and Wales, the odds ratios attenuated slightly but remained significant after sequential adjustment for the child’s ethnicity, individual socio-demographic characteristics and other risk factors for overweight. Within England, the odds of children from the East or South East being less likely to be overweight compared to children from London were maintained after adjustment for the child’s ethnicity and the additional socio-demographic characteristics. However, after adjustment for all individual risk factors, there was no longer any significant difference in overweight between children from the South East and children from London.

DISCUSSION

Three-year-old children in Northern Ireland and Wales were more likely to be overweight than children in England and these differences were maintained after adjustment for individual risk factors for overweight. Within England, children from the East were less likely to be overweight than children from London both before and after adjustment for individual socio-demographic characteristics. Children from the South East were also less likely to be overweight than children from London after adjustment for the child’s ethnicity and other socio-demographic characteristics.

The rich individual- and family-level information collected in this large contemporary, nationally representative cohort of UK children provided an opportunity to examine regional differences in overweight at age 3 years and assess whether they were dependent on individual risk factors for overweight. Previous analyses of regional differences in childhood overweight have not taken into account the individual characteristics of the children and their families within the regions, partly due to a lack of data.

The MCS was based on a stratified clustered sampling framework suggesting that children from some regions were more likely to be included in the sample than others. Within each country, wards were divided into three strata: “ethnic”, “disadvantaged” (both defined previously) and “advantaged” (all remaining wards). The sample was then selected separately for each stratum.15 It is possible that children from a region may not be representative of the entire region because only certain wards were selected; however, the analyses were conducted using both survey commands and sample weights to take into account the sampling design. If children moved regions between the first and second contacts, influences on their body mass may derive from their original region of residence rather than the region they resided in at age 3 years. However, only 3.6% (476) of the children and their families moved regions between contacts. There was also attrition between contacts and approximately 9% of children had incomplete or implausible height, weight or BMI data at the second contact. Although exclusion of these children may have biased the estimated prevalence of overweight or obesity, the proportions of children who were overweight or obese in the MCS are similar to those reported in surveys of young children from all four UK countries.1^–5 Furthermore, sample and non-response weights were used to correct for the non-response and make the results generalisable to the UK population.

Despite the high prevalence of early childhood overweight and obesity across all four UK countries, it is difficult to directly compare levels using the national health surveys. The MCS is therefore a valuable resource for comparing levels of overweight and obesity between countries. The MCS data at age 9 months and 3 years will be enhanced in planned contacts when the children are aged 5 and 7 years; this will complement data collected at the regional level, as in the National Childhood Obesity Database (NCOD), and through national health surveys. There are also limited data on national levels of obesity in preschool age children across resource-rich countries,21 which prevents between-country comparison. National surveys defining childhood obesity by the IOTF cut-offs will also enhance international comparison.18

The NCOD and Health Survey for England (HSE) examined regional differences in childhood overweight. Although the NCOD results should be interpreted with caution because of possible bias associated with low response rates, the Reception Year (age 4–5 years) children from Yorkshire and Humberside had the lowest prevalence of overweight or obesity and children from the North East had the highest.13 In 2002 the HSE found that boys aged 2–15 years from London had the highest prevalence of overweight or obesity, while those from Yorkshire and Humberside had the lowest. In contrast, girls aged 2–15 years from the South East and South West had the lowest prevalence of overweight or obesity, while those from the North East had the highest.2 In the MCS, children from the North East had the highest levels of overweight or obesity, which is consistent with previous results; however, children from the East and South East had the lowest levels. There was no evidence of differential uptake in the collection of the MCS children’s height and weight measurements at the second contact by region in England (data not shown), so regional differences are unlikely to be due to selective non-measurement of overweight or obese children. Discrepancies between studies could be attributed to the ages of the children or sampling.

Differences in early childhood obesity between areas found in the MCS, and reported by others, may be due to differences in characteristics of the individuals living there, features of the country or region itself, or the implementation of policies there. Our findings suggest that country and region of residence may influence early childhood overweight over and above individual characteristics. These influences may derive from features of the environment in these locations or the implementation of policies which affect children’s weight gain. Although there is potential for residual confounding due to unmeasured individual characteristics, we adjusted for a wide range of risk factors for overweight that have been identified in the literature.22 23 Regions which have low levels of early childhood overweight or obesity may have environmental, cultural or policy-related (eg, infrastructure) characteristics that provide or support opportunities for physical activity and/or access to a healthy diet. Further work is needed to measure environmental and community-level factors in the East and South East and determine how these areas may promote the adoption of a more active and healthy lifestyle. Our results also suggest that area-level policies should continue to support environmental changes that remove barriers to physical activity or healthy eating for young children.

In conclusion, differences in early childhood overweight between UK countries and English regions appear to be largely independent of individual risk factors, suggesting that environmental differences may be important in mediating overweight in early childhood. This provides support for policies that target the “obesogenic” environment and promote opportunities for families with young children to be physically active and access healthy foods.