The Cluster Fuss Continues with Two New Studies
We keep hoping that editors and reviewers of obesity, nutrition, and physical activity studies will use a sharper eye when a cluster randomized trial comes to them. But two new publications tell us we can’t count on it yet. In both papers, the researchers claim to have proven the effectiveness of their programs. Yet neither of them has a design that’s adequate for making such claims.
The cluster fuss continues.
A Physical Activity Program for Preventing Obesity
Ivan Müller and colleagues studied the effects of four different programs for promoting physical activity in eight different South African grade schools. Each of four schools participated in a different combination of strategies. For one school, it was only a program of physical activity. Then another one received only health and nutrition education. In yet another school, children participated in both physical activity and education about health and nutrition. In the fourth school, the program was a combination of physical activity and health education. And then finally, four schools received nothing special. They were the control group.
The authors claim to have found a significant effect on BMI z-scores and skinfold thickness. But given their design, this is not possible. In every cluster with an intervention, the investigators are doing something different. This means zero degrees of freedom and zero statistical power to support any conclusions about effectiveness. None.
A “Health-Improving Active Life”
Next, Simone Teresinha Meurer and colleagues have just published a study of a program promoting physical activity and healthy diets in adults with obesity. They call it a randomized controlled community trial. But in fact it’s a cluster randomized trial with only two clusters. One group all got the program, while the other group served as the control. No individual randomization here.
So once again, having only one cluster for the test group means this study has no power to prove anything.
Why All the Fuss?
Cluster randomized studies are clearly useful. If we want to test community-based programs, more often than not, people will enroll in groups. So we often have to test them in groups. That means we need to be smart about designing these studies.
Otherwise, we’ll be feeding junk data to policy makers. Garbage in, garbage out. Without good science, good policy for obesity prevention is unlikely.
Click here for the Müller study and here for the Meurer study. For part 1 of this 2-part cluster fuss series, click here. And finally, for some excellent perspective from Brown et al on best practices for cluster randomized studies, click here.
Ladybug Cluster, photograph © Sharon Mollerus / flickr
Subscribe by email to follow the accumulating evidence and observations that shape our view of health, obesity, and policy.
February 1, 2019