METHODS

Sample

As part of Scottish National Preschool Child Health Surveillance System (NCHS-P), anthropometric data on children (height or length, weight, head circumference) are routinely collected at a number of stages during childhood, allowing longitudinal follow up of children’s growth and nutritional status. These population based data are collected from measurements of weight and height made routinely by health visitors at particular ages, and collated by the Information and Statistics Division (ISD) of the Common Services Agency, Edinburgh. All families of children of this age are expected to attend these routine checks, and >98% of families attend. The Scottish Child Health Surveillance Project Board set clinical guidelines for health professionals taking height and weight measurements at routine visits.^28,29 We examined the health records of 74 500 children who had a routine 39–42 month health review in 1998/99. The single year birth cohort in Scotland was 52 987 in 1999. The variables extracted for this cohort were age (at measurement), height, weight, deprivation category, and birth weight. Birth weight is recorded at time of birth on the birth notification certificate and entered into the information system. The deprivation category is based on postcode sector, which is recorded at the time of review. At this time, health data were accessible for approximately 80% of the preschool population of Scotland.

Definitions of undernutrition and obesity

Weight and height were used to calculate BMI (weight in kg/height in m²). In children, BMI changes with age and differs between boys and girls, so must be interpreted by comparison with appropriate reference data. We made this comparison for both undernutrition and overnutrition using the UK 1990 BMI reference data and expressed BMI as centiles and standard deviation (SD) scores.

We defined undernutrition as BMI <2nd centile (SD scores <−2.06), following clinical guidelines on UK BMI reference charts (Harlow Printing South Shields, England; Child Growth Foundation London, England). This definition represents an assessment for a group or population: sensitivity and specificity in screening for undernutrition in individual children using this definition are unclear. By definition, 2% of the reference population would be expected to have a BMI <2nd centile. Most infants and young children with failure to thrive have a low BMI which is nutritional in origin.³⁰

Obesity was defined as BMI >95th centile, and severe obesity >98th centile (Clinical BMI charts, Harlow Printing/Child Growth Foundation). These definitions have high specificity and moderately high sensitivity³¹ in identifying the fattest children within the population, and so are informative for individual children as well as populations. The definition of obesity was that which is widely used and recommended,³² and is not arbitrary. Children with BMI >95th centile are those most likely to experience co-morbidity, obesity defined in this way is likely to persist, and is associated with presence and clustering of cardiovascular risk factors.^{12,14–16,33}

Assessment of social deprivation

Deprivation was defined using the Carstairs Deprivation Category.³⁴ This is a measure of deprivation based on area of residence (calculated at Scottish postcode sector level) and derived from UK 1991 census variables for overcrowding in housing, male unemployment, low social class, and car ownership. These items were combined to create an aggregate score on a scale of 1 (most affluent) to 7 (most deprived).³⁴ This method of assessing deprivation is widely used in Scotland and strongly related to a number of health outcomes.³⁵ The children in this study were born in 1994/95, three to four years after the 1991 census used to establish exposure (deprivation) categories. The outcome data (undernutrition and obesity) were collected in 1998/99, and therefore we are assuming the social environment to which the children in this sample were exposed was as represented by 1991 census data.

BMI values for 2nd and 98th centiles in three populations

We compared the BMI values for the 2nd and 98th centiles for the sample of Scottish children aged 40 months to the UK 1990 reference (table 3). As a further comparison, the BMI values (2nd and 98th centile) of a contemporary sample of English children aged 37 months in a large prospective cohort study, the Avon Longitudinal Study of Parents and Children (ALSPAC) were calculated, using data from the Children in Focus Study. We have described this sample and the methodology elsewhere.⁹ The ALSPAC cohort reached age 37 months in 1995/96, but was slightly more affluent than the rest of the UK.³⁶

Statistical analysis

The outcome measures were overnutrition (BMI >95th and 98th centiles) and undernutrition (BMI <2nd centile). Logistic regression analysis was used to establish the unadjusted odds ratios (OR) and 95% confidence intervals (CI) for deprivation categories. These analyses were repeated adjusting for birth weight. Birth weight was categorised as follows: <2500 g, 2500–2999 g, 3000–3999 g, 4000–4499 g, and 4500 g or more. Statistical analyses were carried out using SPSS for Windows, version 10.1.

RESULTS

Sample

Children were excluded from analysis if they had missing data on gender, age, deprivation category, weight, or height (n = 9358, 12.5%) at the routine check, or if implausible BMI values were recorded (BMI SDS >+4 or <−4) (n = 658, 0.8%). The final sample was 64 484, 87% of the original sample (n = 74 500).

Undernutrition and obesity

The prevalence of undernutrition (girls 3.3%, boys 3.2%), obesity (girls 8.0%, boys 9.0%), and severe obesity (girls 4.1%, boys 4.4%) was significantly higher than expected from the UK 1990 reference values of 2%, 5%, and 2% respectively (χ² goodness of fit, p < 0.001 in all cases). Table 1 shows prevalence by deprivation category. Undernutrition and obesity were significantly associated with deprivation category, with a significantly higher prevalence of both undernutrition and severe obesity (BMI >98th centile) in the more deprived children. Within each deprivation category the prevalence of children with BMI >98th centile exceeded those <2nd centile, with the exception of the deprivation category 7 (most deprived) where prevalences were similar (4.8% v 4.7%; table 1). The cumulative prevalence of undernutrition and obesity in the most deprived groups 6 and 7 was 8.4% and 9.5% respectively compared to the least deprived group 1 (6.9%).

Table 2 shows the unadjusted and adjusted (for birth weight) odds ratios and 95% confidence intervals for undernutrition and obesity by deprivation category. Scottish children aged 3–4 years in the most deprived category (7) had a 30% higher risk of undernutrition compared to those in the least deprived category (1) after adjusting for birth weight (AOR 1.29, 95% CI 1.03 to 1.62). The same group (category 7) had a 43% higher risk of obesity compared to the least deprived after adjusting for birth weight (OR 1.43, 95% CI 1.1.6 to 1.77). Undernutrition was most evident in the deprived categories (6 and 7), while the risk of obesity was higher in the middle categories (4 and 5) in addition to the most deprived (7).

BMI: 2nd and 98th centiles in three populations

Table 3 shows the absolute BMI values (kg/m² ) for the 2nd and 98th centiles for the sample of Scottish children, the UK 1990 reference data, and the ALSPAC study. The BMI value for the 2nd centile in the Scottish children was approximately 0.5 kg/m² lower than the equivalent for the UK 1990 reference, and 1.0 kg/m² lower than that of the ALSPAC sample of English children. The BMI value for the 98th centile in the Scottish children was 1.0 kg/m² higher than the equivalent for the UK 1990 reference and similar to the ALSPAC sample of English children. These data show a kurtotic distribution of BMI in the Scottish population, consistent with a higher prevalence of Scottish children in both extremes of the BMI distribution.

DISCUSSION

Main findings and implications

The present study has shown, for the first time, the coexistence of undernutrition and obesity associated with social deprivation in contemporary children from the developed world. Recent evidence has shown the phenomenon of coexistence of undernutrition and obesity in developing countries and within families.^2–4 In the present study, Scottish children aged 3–4 years in the most deprived families had a 30% higher risk of obesity and 50% higher risk of undernutrition when compared to children in the least deprived group, and these remained significant after adjusting for birth weight. The data presented in this study show a higher prevalence of undernutrition at age 3–4 years in the most deprived groups. For obesity at age 3–4 years there was a higher prevalence in both middle deprivation categories (4 and 5) and in the most deprived groups (6 and 7). Within each deprivation category the prevalence of children >98th centile exceeded those <2nd centile with the exception of the most deprived group. Thus obesity was more common in middle and low socioecomomic groups, while undernutrition was more confined to the most deprived groups. The cumulative prevalence of undernutrition and obesity in the most deprived group was higher in the least deprived (9.5% v 6.9%).

Undernutrition

With the strength of association between socioeconomic status and birth weight,²⁶ it was important to analyse the data with an adjustment for birth weight. A low birth weight and lack of catch up growth may have explained some of the risk of undernutrition at age 3–4. The odds ratio for undernutrition in deprivation category 7 (OR 1.29, 95% CI 1.03 to 1.62) shows that the association between social deprivation and undernutrition at age 3–4 remained significant after adjusting for birth weight. This relation may be a reflection of growth patterns of children living in poorer circumstances,^24,25 with inadequate nutrition and a higher risk of household food insecurity. The phenomenon of household food insufficiency and the potential adverse effects on nutritional status and child development have recently been described in school age children from the USA.³⁷ With the high nutritional requirements for growth, younger children are more vulnerable to undernutrition and its consequences.^22,24 Undernutrition persisting through childhood may have significant effects on cognitive development, school achievement, and later health.^22,37 However, the specific long term health and developmental risks to children in more affluent developed countries are less well defined and warrant further study.^23,30,38

Obesity

Obesity, even in young children, confers increased risk of short and longer term morbidity,^11,33 and is particularly strongly associated with presence and clustering of cardiovascular risk factors and type 2 diabetes.^12–15 Moreover, the immediate psychosocial consequences for obese children may increase the risk.^11,17,19 An association between Townsend score (an index of social deprivation) and obesity in children of school age has previously been described for English children.⁶ The present study showed that risk of obesity is much higher than expected and associated with social deprivation at an early age. The indices of nutritional status used in the present study reflect energy balance (energy intake − sum of energy outputs), and are largely determined by energy intake and levels of physical activity. The epidemic of childhood obesity which occurred in the UK during the 1990s resulted largely from increasing physical inactivity coupled with consumption of energy dense diets. However, the specific factors which explain the associations between obesity, undernutrition, and social deprivation in childhood need further study. There are marked differences in nutritional quality (energy and micronutrient density) of the diet in Scottish children from different deprivation categories,³⁹ and this may be partly responsible for the greater risk of malnutrition in young children in the most deprived families.

BMI distribution and public health implications

The observation that both undernutrition and obesity are associated with social deprivation may have implications for preventive strategies. The traditional paradigm has been that shifts in the entire distribution of BMI (down) are desirable, and that population based approaches (in obesity prevention, for example) are necessary, rather than approaches which target high risk groups.^40,41 However, if the pattern of malnutrition in deprived children resembles a U shaped curve with both undernutrition and obesity prevalent, success in shifting the population distribution (of BMI, for example) downwards could potentially have adverse implications for those children at the lower end of the distribution. Our analysis suggests differences in BMI distribution between the three samples we compared (Scotland; UK reference; ALSPAC) with the Scottish sample having a lower BMI value at the 2nd centile and higher BMI value at the 98th centile. In contrast the BMI values for the 2nd and 98th centile in the ALSPAC sample of English children were higher than the UK 1990 reference data. This suggests differences in these population groups, which may be greater than might have been expected based on methodological differences between datasets.

Strengths and weaknesses of the present study

The present study was epidemiological, cross sectional, and limited to identifying associations, and so cannot definitively confirm causality between social deprivation and malnutrition. The strength of using population data lies in the sample size, power, and wide coverage/representation of the population. The value of using routinely collected data is increasingly being recognised.^6,10 However, there are a number of limitations with using routinely collected data not under a strict research protocol. There are issues of missing data and data quality which are usually greater than studies with strict protocol design. From the sample of 74 500 children in the present study, those missing data for one or more of the variables deprivation category, height, weight, or gender equated to 9358 (12.5%), and there were very few with invalid BMI data (658, 0.8%). There was a small bias of more missing data from the most deprived groups (8.2% in category 7 v 6.2% in category 1), but this is likely to result in the differences between deprivation groups being underestimated rather than overestimated.

Our definitions of obesity were well founded, in that above approximately the 91st centile for BMI on the UK 1990 reference, any definition of obesity based on BMI centile cut off has moderately high sensitivity (low to moderate false negative rate), but high specificity (low false positive rate).³¹ In addition, obesity defined as BMI >95th centile has a strong tendency to persist, and is associated with a number of measures of clinical/biological significance (such as psychological health; social, educational, and economic outcomes; and presence, clustering, and maintenance of cardiovascular risk factors).¹¹ At present there is no gold standard for the definition of undernutrition in child population groups, and the sensitivity and specificity of any available definition is unknown. The definition used here (BMI <2nd centile) is suitable for comparison between groups or populations, but it cannot definitively identify individual children as undernourished. The WHO expert committee recommends using the observed standard deviation of height/weight or BMI SDS distribution to assess the quality and spread of height and weight survey data.⁴² With accurate age and anthropometric measurements in a population, the SDS of the observed BMI SDS distribution should be relatively constant and close to 1.0 for the reference distribution (ranging within 0.2 units). In this study the standard deviation of the BMI SDS distribution equated to 1.14; close to 1, but indicating a flatter, bell shaped distribution. An important limitation of the study was not being able to adjust for other potential risk factors for under and overnutrition in children, notably maternal BMI, smoking, and catch up growth.

There is currently no gold standard for assessing social deprivation, although a combination of area based and individual measures (for example, maternal education, income) provides a more accurate assesment.^43,44 The present study was limited to the use of the area based Carstairs Deprivation Category, a widely used measure strongly linked to morbidity and mortality in Scotland.³⁵ There was a time difference between the assessment of deprivation category (from 1991 census), the birth of the children (1994/95), and outcome measures (taken 1998/99), but such a difference is small and inevitable in longitudinal studies of this kind.

Conclusions

We have shown that undernutrition and obesity coexist in preschool children in Scotland and are strongly associated with social deprivation. Further research is needed to identify other risk factors involved in the development and persistence of malnutrition in the children, and to clarify the long term effects. Effective public health strategies for improving child health will encompass environmental, behavioural, and biological determinants of childhood undernutrition and obesity,^1,40 some of which may be interrelated, and focus on improving nutritional status of the child population.