Increasing prevalence of obesity in primary school children: cohort study - Commentary and explanation

David Ogilvie discusses what is involved in a cohort study and explains what the results of this paper mean

There is increasing public health concern about the numbers of overweight children and the impact their weight may have on their future health. In this study, an auxologist (a person who studies growth) took measurements from primary school children to see how they compared with reference standards.

What is a cohort study?

If you are a classical scholar you might remember that a cohort was a division of the Roman army - a tenth of a legion. Suppose we wanted to see how the health of soldiers changes over a period of years. We could randomly select a thousand different soldiers every year and survey their health. That would be quite useful, but it would be much more interesting to survey the same group of soldiers - a cohort - every year to see how the health of the group changes as time passes. In this study, the authors have described the health of a cohort at different ages and compared their observations with what they expected to find.

What did they expect to find?

The point to note here is that defining "normal" or reference standards for children's growth is not straightforward. It is done by plotting the distribution of height, weight, and other measurements of a large number of children. In this case, the reference standards were published in 1975 and 1990. These reference standards soon go out of date: "Don't children grow up quickly these days?"

By convention, overweight children are defined as those whose body mass index (BMI) is in the top 15% of the distribution of BMI in the population. BMI is calculated by dividing the person's weight in kilograms by the square of their height in metres. Another way of saying this is that their BMI is above the 85th centile. For each age group, the authors took the BMI which would have qualified a child in the 1990 reference population as overweight and looked to see what pro­ portion of children in this study exceeded that BMI. The same principle was applied to obese children and then to both overweight and obese children according to triceps thickness measurements.

What did they find?

If you look at the columns for 9, 10, and 11 year olds in the table you can see that quite high proportions of children had a BMI high enough to qualify them as overweight or obese - for example, 32% of 11 year old girls were defined as overweight. Remember, we are expecting to see only 15%.

Are these real differences, or are they just a chance finding? The chi­squared (X²) test is a statistical test which compares the numbers of people you have found in different categories with the numbers you expected to see in each category. The output from the test, shown in the table footnotes, is a P­value, which gives an indication of how likely it is that the difference between observed and expected results arose by chance (see box).

So the table tells us that when the children were 9, 10, and 11 they were more likely to be overweight or obese than expected, but not when they were 7 or 8. Their triceps thickness measurements, however, were not different from expected results, which is surprising. You can read more about this in the prepublication history for the paper on bmj.com.

So what?

As usual, this study seems to raise more questions than it answers. Apart from the methodological issues about reference growth standards and measuring triceps thickness, we need to put the paper in context. What will happen to these children in later life, and will the overweight children have a different long term future from the rest? These are questions which a more ambitious cohort study could answer. But we also need to know why the proportion of overweight children is increasing and what we can do about it. Different types of study are needed to answer these questions.

David Ogilvie


studentBMJ 2001;09:217-260 July ISSN 0966-6494