Example: Why having a control group is essential for establishing a causal relationship
This randomised controlled trial measured the impact of an intervention that was designed to reduce stunting.
Between 2010 and 2014, it was observed that:
- the control group (without the intervention) saw their stunting levels increase from 68.2% to 74.8%;
- the stunting levels of the intervention group remained almost stable.
Results from secondary analysis of Tubaramure, a food-assisted integrated health and nutrition programme in Burundi
(Source: Leroy J.L., Olney D., Ruel M., 2016. Tubaramure, a Food-Assisted Integrated Health and Nutrition Program in Burundi, Increases Maternal and Child Hemoglobin Concentrations and Reduces Anemia: A Theory-Based Cluster-Randomized Controlled Intervention Trial. The Journal of Nutrition, 146(8), 1601-1608, https://doi.org/10.3945/jn.115.227462.)
- It is only by comparing the delta of pre- and post interventions for the control group and the treatment group that a robust conclusion can be reached: the intervention reduced stunting levels by 6.4 percentage points (74.8-68.2) – (64.3-64.1).
- Without a control group, the measurements “before and after intervention” for the treatment group could be compared, leading to the conclusion that the intervention had no impact (increase of 0.2 pp).
- Without baseline measurements, the populations “with and without” interventions could be compared, leading to the conclusion that the intervention had a very high impact (reduction of 10.5 pp).
In the last two cases, an incorrect conclusion has been reached.