Paper plus: Childhood obesity and fizzy drinks

Leanne Tite takes you through a cluster randomised controlled trial looking at whether consumption of fizzy drinks affects obesity in children

This month's paper is James J, Thomas P, Cavan D, Kerr D. Preventing childhood obesity by reducing consumption of carbonated drinks: cluster randomised controlled trial. BMJ 2004;328:1237. You can read the paper by going to studentbmj.com and clicking on the link.

Obesity in children is now a widespread and increasing problem with obvious health implications. It is ultimately caused by an intake of surplus energy in the diet. Although there are many culprits in children's diets that could contribute to excessive weight gain, one obvious candidate is carbonated drinks. A single carbonated drink can add a further 10% to a child's daily energy intake, so reducing solely the amount of soft drinks that a child consumes could cause a significant reduction in obesity risk. Schools make a useful environment for delivering health education initiatives to children, but previous attempts to promote healthy lifestyles in schools have not met with much success in reducing obesity rates, the authors claim. One recent but unsuccessful initiative attempted multiple interventions delivered at the same time, including dietary changes combined with extra physical activity. The current study departs from previous research by attempting to assess the effect on obesity rates of a single school based intervention, aimed at reducing pupils' consumption of carbonated drinks.

PAUL BROWN/REX

What is the study design?

The researchers used a cluster randomised controlled trial design. This means that all pupils from the schools involved were assigned to one of two study groups for the purpose of the research: the control group or the intervention group. However, only children in the intervention group actually participate in the intervention, but both groups are “tested” or measured at the beginning and the end of the study. The point of the control group is therefore to serve as a “benchmark” against which progress in the intervention group is compared. However, the vital, most important element of any randomised controlled trial is the “randomised” part. This refers to the way pupils are assigned to either the control or intervention group—that is, randomly. Human beings make extremely complex and changeable subjects to study. There really could be many hundreds of factors at any one point in time determining how many carbonated drinks a child consumes in a day and what their weight change might be over a certain period. So how can you have any confidence that it is your intervention that is responsible for any changes you observe at the end of the study?

This is where randomisation comes into its own. When you randomly assign pupils to groups, probabilistically, all the myriad other factors that affect the pupils' drinks consumption and weight change will also be randomly distributed between the control and intervention groups. This means that all the random factors in each group should even themselves out and therefore not have a greater effect on one group than the other. Only systematic group differences, that is a difference that is present for all of one group and none of the other, will show up in the data as statistically significant. As long as the study intervention is the only systematic difference between the control and intervention groups, you can say with some confidence that any difference between the groups is indeed attributable to the study intervention. This point is particularly pertinent to this study because there are two outcome measures that were used to see how much of an effect the intervention had. The first is the amount of carbonated drinks the pupils consume at the end of the study period, which the researchers attempted to directly influence through the health education lessons. The second outcome measure was change in the pupils' weight since the beginning of the study which would be only indirectly influenced by the study intervention. This makes it even more important to rule out the effect of other factors affecting the pupils' weight change during the study period.

PAUL BROWN/REX

The children in the intervention group sat through a one hour health education lesson each term for one school year during which they were taught about the negative health implications of carbonated drinks. The classes were interactive and involved the pupils in tasting games and song writing. The researchers measured the children's consumption of carbonated drinks over three days at the beginning and again at the end of the study period (after 12 months) using diaries which all children, in both the control and intervention groups, were to complete. Pupils were also measured for body mass index at the beginning of the study period and after six and 12 months.

One of the drawbacks of using diaries to measure behaviour is that the data is susceptible to respondents' inaccuracy in recording what they do. They may forget to fill out the diary or might not fill it out truthfully. The latter problem is a particular issue when respondents know how the researchers would like them to behave, in which case the diarists could be said to be responding to demand characteristics. The usual way around this problem is to blind both the participants and researchers to the respondents' study group and, where possible, to the objectives of the study. In this study, although the researchers did not tell the children which study group they had been allocated to, the children in the intervention group might nevertheless have picked up on the study objectives as a result of the extra health education lessons they received. Any of the problems associated with self report data are likely to be much greater for children, who may be particularly keen to be seen by teachers to be being good and doing the “right thing.”

Based on diary reports, the pupils in the intervention group reduced their consumption of carbonated drinks by the end of the study period by 0.6 glasses in a three day period (from an average of 1.6 glasses in three days to 1 glass) compared to an increase in the control group of 0.2 glasses. There was also a statistically significant difference between the control and intervention groups in weight change (body mass index) since the beginning of the 12 month study period. Interestingly, this difference results from the control group having 7.5% more cases of obese children at the end of the study than there had been at the beginning, rather than there being fewer obese children in the intervention group (which decreased by just 0.2%). This finding suggests that the reduction in consumption of carbonated drinks reduced the risk, by around 7.5%, of children in the intervention group becoming obese later on.

One of the biggest problems is the drop-out rate of participants throughout the study. Of all the children whose parents consented to their participation, only 36% completed the study and returned both drinks diaries. A large study drop-out rate is a problem because it can result in a biased sample. If all the children who dropped out of the study have something in common, for example, that they enjoy carbonated drinks too much to cut down and find it difficult to fill out the diaries, then a systematic bias has been introduced into the sample. Only the children who responded well to the intervention would be left, and so the intervention will inevitably look more successful than if all the children, including those who did not cut down, had also been included. But because no data were collected, it is impossible to say what the characteristics are of the children who dropped out of the study, and what effect that has had on the study outcomes. So large drop-out rates introduce uncertainty into the interpretation of the results.

Another problem is that the children's drinks consumption was measured over just six days throughout the whole of the 12 month period, at the beginning and end of the study. Just as when sampling participants, it is important when sampling behaviour in this way to get as wide or “large” a sample as possible, in order to reduce the effects of any random events that might be associated with the specific times when the pupils' carbonated drinks consumption was measured.

The researchers also note that some of the schools used the opportunity to promote drinking water during the time of the study. Because only some of the schools did this, it represents another potential source of systematic bias which might have affected some, but not other pupils' behaviour during the study period.

Given the problems outlined above it might seem that the study is riddled with major flaws. However, this is real world research and, what's more, it involves children. When doing research in real world settings, it is always impossible to completely control all the factors that might unduly affect your study. The important point is to be aware of the problems and take them into account when interpreting the findings. But real world studies also benefit from greater ecological validity, in that the researchers can be more sure that whatever intervention they test will be more likely to work well when applied to real life. Although the reductions in obesity might seem small, when you think of the minimal effort involved, and that these gains were made from just a single intervention, then the 7.5% reduction in obese children does seem more impressive. The question is though, how long after the intervention has ended would this effect last? More research is needed to assess this, and whether such interventions can be applied to other health risks in the fight against childhood obesity.