Prevalence of overweight and obese children between 1989 and 1998: population based series of cross sectional studies

Statistical analysis

StatsDirect software was used for all statistical calculations.6 We considered calculated probabilities of <0.05 to be significant and <0.001 to be highly significant. Summary statistics and confidence intervals are quoted here to two decimal places (original measurements included two significant digits) and calculated probabilities to three decimal places. Mean SD scores for height, weight, and body mass index were calculated for each year for the 3 to 4 year age group. Mean SD scores for weight were calculated for each year for the 1 to 3 month age group. We examined trends in weight, height, and body mass index with Pearson's product moment correlation for mean SD scores (weighted by the inverse of the observed variance) and year. The robustness of inferences made with these parametric and linear methods was explored by reanalysis with non-linear and non-parametric alternative methods; an alternative assumption that data exclusion constituted censorship was considered in the reanalysis.7^–9

We analysed categorical data derived from definitions of obese (body mass index SD score >1.04) and overweight (>1.64) using a χ² trend with evenly spaced scores representing the order in years from 1989 to 1998.10

Preschool children

Figure 1 shows the mean SD scores for weight, body mass index, and height for children over the 10 year period. If the distribution of weights in the study and reference populations is similar, the mean SD score from a large sample should be close to zero. We observed a significantly increasing trend in mean SD score for weight and body mass index between 1989 and 1998. Over the same period there was no increasing trend in the mean SD score for height.

Fig 1.

Mean SD scores for weight, height, and body mass index plotted against year of measurement for children aged 2.9 to 4 years. Increasing trend in scores significant for weight and body mass index but not for height (Pearson's correlation (95% CI) and P for r=0 (weighted): 0.94 (0.77 to 0.99), P<0.001; 0.93 (0.71 to 0.98), P<0.001; and 0.61 (−0.03 to 0.90), P=0.059)

• Download figure
• Open in new tab
• Download powerpoint

Table 1 shows the percentage of overweight and obese children as measured by body mass index in each year. In a population with normally distributed weights, we would expect 15% of children to be overweight (>85th centile; SD score >1.04) and 5% obese (>95th centile; SD score >1.64). This was the case for our study population in 1989, but since then the proportion of overweight and obese boys and girls increased significantly (fig 2).

Fig 2.

Annual increase in proportion of overweight and obese children; χ² for trend in overweight 71.1 (P<0.001) for boys and 33.1 (P<0.001) for girls, for trend in obesity 48.3 (P<0.001) for boys and 7.3 (P=0.007) for girls. Proportion of overweight and obese boys becomes greater than girls in early 1990s and remains so

• Download figure
• Open in new tab
• Download powerpoint

Infants

Figure 3 shows a slight increase in the mean SD scores for weight of infants during the 10 year period. However, it was close to zero (the 50th centile) and below zero for the whole period. Table 2 shows the proportion of infants with a weight above the 85th and 95th reference centiles. Throughout the study period fewer than 15% of infants weighed above the 85th centile and fewer than 5% above the 95th centile. During the same period there was an increase in the number of preschool children with a weight above the 85th and 95th reference centiles.

Fig 3.

Median SD scores for weight at 6 weeks bounded by interquartile range plotted against year of measurement for infants aged 1 to 3 months (Pearson's correlation (95% CI) 0.80 (0.35 to 0.95), P=0.005 for r=0)

• Download figure
• Open in new tab
• Download powerpoint

Alternative analyses

Reanalysis of the data with non-linear and non-parametric methods and assumptions did not change any of the inferences drawn.

Definitions of overweight and obesity

There is no consensus as to the definition of overweight and obese children. The International Obesity Task Force suggests that children over the 80th centile are overweight, as this corresponds to a body mass index of 25 at the age of 18 years in men and women, which is the adult definition of overweight.12 Our definition of overweight as being above the 85th centile has been used by others5 but is arbitrary, and data based on the 80th centile could be calculated easily. We agree with other authors that a consensus figure is required. 12 14

It is important that centiles should be based on a reference population that does not change with time.12 If, when growth curves are updated to account for changes in nutritional state, they are based on the increased weight and body mass index as found in this and other studies, 1 15 they will mask an important increase in obesity. This problem can be overcome by combining data from populations with a low prevalence of undernutrition and data that were collected before the increases in obesity now being reported. Prentice also emphasises the importance of monitoring and interpreting these changes over time.11 Data routinely collected by health authorities fulfil most of the criteria described by Bellizzi and Dietz for the development of a reference population12 and are ideal for examining change over time, as we have shown. It is also important to take into account changes in height over time. In our study, height had not increased in the preschool population. The increase in the number of children with an SD score for weight and body mass index above 1.04 must be recognised for what it is—an increase in the number of overweight and obese children and not a “normal” trend in the population. Kotani et al found that the proportion of obese children in their population had increased from 5% to more than 10% over two decades.15 In our study, a similar increase occurred in 10 years.

Effects on later health

There is evidence that obesity is likely to persist into adult life 12 15 16 and to increase the likelihood of morbidity and mortality.17 Calle et al prospectively examined the risk of death related to body mass index in over a million adults and concluded that heavier men and women in all age groups had an increased risk of death.18 Cardiovascular disease remains one of the principal causes for this excess mortality. Increased body mass index is also one of the important risk factors associated with the extent of atherosclerotic lesions in the aorta and coronary arteries in people between 2 and 39 years of age.19 Must and Strauss reviewed the risks and consequences of obesity in childhood and adolescence and concluded that an aggressive approach to prevention and treatment was required.20 Early intervention, including increased activity and reduction in high fat, high calorie foods, is important,5 and some success has been shown in such a programme.21^–23 Power et al have emphasised the importance of population based intervention to achieve this.24 In our study, the increase in the incidence of obesity occurred before the age of 4 years, and interventions should be targeted at this age group if they are to have an impact.

The National Service Framework for Coronary Heart Disease identifies the need to develop, implement, and monitor policies that reduce the prevalence of coronary risk factors in the population.25 Data that are routinely collected are important in monitoring the health of communities and should be used in the planning of community based interventions. Such data may be less accurate than those collected prospectively in carefully executed studies, but this must be balanced against the large amount of information that is readily available. In our study, valid data were collected on over 64 000 infants and children, which represented 88% of live births in the health authority and thus provided a large sample of the relevant population. The accuracy of routinely collected data must be ensured as they are a valuable source of information on population trends. Similar information obtained from other districts could be used to determine standard anthropometric measurements and trends in large numbers of infants and children. This information could be the impetus for a national programme to prevent and treat childhood obesity and its long term complications.