A controlled trial combined with a prospective cohort study

Sometimes the nature of a research question and its setting prohibit particular study designs for logistical or ethical reasons. Steven Reid explains a recent study that followed up a randomised controlled trial

This month's paper is Walker S P, Chang S M, Powell C A, Simonoff E, Grantham-McGregor S M. Effects of psychosocial stimulation and dietary supplementation in early childhood on psychosocial functioning in late adolescence: follow-up of randomised controlled trial. BMJ 2006 Jul 28. doi:10.1136/bmj.38897.555208.2F.
You can read it by clicking here.

Why do the study?

In developing countries, linear growth retardation is a common problem affecting as many as 40% of children under 5 years of age. It is defined as height for age less than two standard deviations below reference values, and the most common cause is malnutrition. Follow-up studies have found that these children are more likely to have learning difficulties and behavioural problems. What is not clear, however, is whether these problems are a direct consequence of growth failure or simply due to poverty and deprivation.

Walker and colleagues wanted to find out whether they could improve the development of these growth impaired children by supplementing their diet and providing them with additional social stimulation. In 1991, they did a controlled trial in Jamaican children with growth impairment aged 9 to 24 months. They found that after two years both an improved diet and weekly play sessions with a health worker had beneficial effects on the children's cognitive development.¹

In this paper the researchers describe a follow-up study they did when the children reached 18 years to determine whether the early intervention had long term effects on emotional and social development. In effect, they were combining a controlled trial with a prospective cohort study.

How was the study done?

As the researchers were following up participants from an earlier study, it is important to look back at how the original trial was done. It was an experimental trial in which the researchers assigned different interventions to children with growth impairment in Kingston, Jamaica.¹ In all, 129 children were recruited from a house to house survey in poor neighbourhoods.

Before starting an experimental study it is important to calculate an appropriate sample size so that you can be confident that you have sufficient statistical power to detect meaningful differences between the groups. Although the researchers were able to enrol the number of children required for the outcomes in the original study, it is not clear that they would have had a large enough sample for the outcome measures they were interested in 16 years later.

Eligible children whose parents had given consent were assigned to one of four groups, receiving dietary supplements, social stimulation, both interventions (as the researchers wanted to know if any effects were additive), or neither (the control group). The sample was stratified by sex and age, characteristics that by themselves might affect the outcome. Stratification ensures that these characteristics are distributed equally between groups.

The researchers state that the children were assigned to the intervention groups randomly by labelling every first child a control, every second child was allocated to dietary supplementation, and so on. The aim of randomisation is to rule out the likelihood that there are differences between the groups that may affect the outcome and produce misleading or biased results. The problem with allocating by rote is that the researcher can foresee the intervention group that each child will be assigned to, which may introduce a bias. For example, a researcher might postpone the enrolment of a severely malnourished child if the next intervention on the list was social stimulation rather than food supplements.

What happened to the participants in the study?

A health worker visited all of the participants weekly for two years. Details of the interventions are given in the table. No placebo was given to the children not receiving dietary supplementation as it would have been unethical to provide a malnourished child with a true placebo such as a low calorie supplement. After the initial trial the researchers attempted to trace the children at the ages of 7 and 11 years to assess their development and then again for the present study in late adolescence (17 to 18 years).

What outcomes were measured?

The researchers were interested in whether this group of adolescents had emotional or behavioural problems or learning difficulties, and what effect supplements and stimulation may have had. To measure these outcomes, they used questionnaires administered by an interviewer and answered both by the adolescents and their parents. These questionnaires measured a number of variables including anxiety, depression, self esteem, attention deficits, and antisocial behaviour. Importantly, they had been used extensively in other studies and were considered both valid and reliable.

The interviewers were supposed to be blind to which intervention each participant received but the adolescents and their parents were obviously not. Blinding is important because knowledge of whether supplements or stimulation were given may have influenced the way in which outcomes were assessed. Unfortunately, participants in research will often reveal to the researcher what intervention they were given even when asked not to.

What were the results?

Crucial to the success of all trials is to minimise the loss of participants for adequate follow-up. This is because people who drop out of studies often differ from those remaining. Keeping hold of the entire sample is particularly difficult when the follow-up is over a number of years. In this study, the researchers traced 103 (80%) of the original children after 16 years, which is an impressive achievement. Three quarters of those lost to follow-up had migrated, and we know that people move abroad for a variety of reasons some of which may have been associated with the outcome measures of the study.

One way of dealing with drop outs is to analyse the results of the study according to the intention to treat principle, which means including all participants in the analysis regardless of whether or not they were followed up. This can be difficult, or indeed impossible, if you have no outcome measures at all for those who have dropped out. Simple outcomes, such as mortality, are much more readily obtainable than the results of complex symptom inventories used in this study. So in this case the investigators have looked at the measures taken at enrolment to see if there were any differences between the children that were not followed up and the others. They did find some small differences and adjusted for them in their analysis.

They also noted that the drop out rates were the same in each intervention group.

For their primary findings, the researchers combined all of the individual outcomes to come up with an overall measure of psychosocial functioning. They found that although food supplements had no benefit, social stimulation had a positive effect that was statistically significant. On looking at the outcomes individually the researchers found that adolescents who had received social stimulation reported less anxiety and depression, better self esteem, and their parents reported fewer problems with attention. Dietary supplementation had no effect and there were also no differences in other outcomes they looked at, such as contact with the police or judiciary, sexual behaviour, or disciplinary problems at school.

What does this study mean?

Walker and colleagues found that social stimulation of Jamaican children with growth impairment led to benefits in their emotional wellbeing and fewer problems with attention in adolescence, but a weekly food supplement had no effect. These results fit with their previous studies, which showed that both social stimulation and food supplements were beneficial in early life but by the time the children reached 11 years, the effects of the food supplement were not sustained.^{1 2}

An experimental study such as this one is difficult to do for logistical reasons and because of the deprived setting. The researchers did particularly well in maintaining the study over 16 years, with only a modest loss to follow- up. A randomised controlled trial would be the ideal design to answer the questions posed by the researchers. Although the randomisation process used here was flawed, they did compare the characteristics of the intervention groups at both enrolment and follow-up to ensure they were comparable. Another limitation is the possibility that the sample may have been too small, and important effects may have been missed, a type II error.

Concluding remarks

Despite some limitations, this follow-up study of a controlled trial provides us with some important information. A programme of regular, stimulatory play produced long term psychological benefits for children with growth impairment. The quality of a study is determined not only by its validity but also its generalisability. Generalisability refers to how much the study findings would apply to a larger population outside of this study setting. In this case we can be fairly confident that the children in this study would be representative of children with growth impairment elsewhere in the developing world. Also, the researchers managed to recruit every child considered eligible for the study. The next question is whether an intervention such as this could be translated into a real world setting and, importantly, would it be affordable?

Competing interests: None declared

Steven Reid, consultant psychiatrist, Department of Liaison Psychiatry, St Mary's Hospital, London W2 1PD
Email: steve.reid@nhs.net


studentBMJ 2006;14:309-352 September ISSN 0966-6494

1. Grantham-McGregor SM, Powell CA, Walker SP, Himes JH. Nutritional supplementation, psychosocial stimulation, and mental development of stunted children: the Jamaican study. Lancet 1991;338:1-5.
2. Chang SM, Walker SP, Grantham-McGregor S, Powell CA. Early childhood stunting and later behaviour and school achievement. J Child Psychol Psychiatry 2002;43:775-83.