Let’s learn about randomized control trials in research

When research scientists create a study to improve the mental health of adolescents, how can they know if it really works? Simply seeing that study participants improved afterward isn’t enough, because that could have happened for other reasons — such as the simple passage of time, changes in personal life, or even the fact that they were receiving more attention than usual from being part of the study. The most reliable way to cut through these uncertainties, and science’s answer to this problem, is the Randomized Controlled Trial, or RCT.

The idea is quite simple. Participants are divided into two groups by lottery. Group 1 receives the intervention—the program or treatment being tested. Group 2 is the control. A control may receive nothing or a different version of the intervention depending on the study’s intent. At the end, researchers compare the two groups to see what changed. The control group helps us understand whether the change was in fact caused by the intervention. The lottery is what makes the method so reliable. Because the groups are formed randomly, they tend to be similar to each other. So, if one group improved more than the other, the likelihood that it was due to the intervention is much greater.

In an ideal RCT, all environmental and sociodemographic factors are distributed identically between groups — such as income, age, prior treatments, etc. This way, any observed difference in outcomes between treatment and control would imply the existence of a causal relationship between the intervention and the result within the study. For this reason, RCTs are the primary study design for identifying causal relationships between interventions and outcomes. They are currently considered the best tool for understanding the impact of a given intervention on health outcomes in medicine, and increasingly in the social sciences.

Why does this matter for mental health?

Imagine we are creating a virtual program to be applied with adolescents. Or that we are carrying out an intervention in schools with teachers. How do we know if the program is actually improving mental health?
In mental health, it’s very easy to think something worked when it actually didn’t. The simple fact that a person receives attention can already make them feel better, regardless of the program itself. Without a comparison group, we can’t separate the real effect of the program from this “attention effect.”

When we talk about an “attention effect,” we’re pointing to something quite straightforward but powerful: the simple act of being noticed, listened to, and cared about can, all by itself, make someone feel better. This isn’t about the specific techniques or content of a mental health program; it’s about the human experience that comes with participating in any kind of study or intervention.
Imagine an adolescent who is struggling. They’re invited into a program. Someone calls them by name, asks how they’re doing, listens without judgment, and maybe checks in on them week after week. Even if the program’s “active ingredient” is ineffective, that young person might still report feeling less alone, more hopeful, or less distressed — simply because someone showed up for them in a consistent, caring way.

There is also another risk that RCTs help identify: the iatrogenic effect. This means that, in some cases, an intervention may worsen a participant’s condition rather than improve it such as unintentionally increasing a study participant’s anxiety or distress or other mental health symptoms rather than reducing it. Without a control group for comparison, this type of harm can go completely unnoticed. That is why the RCT is considered so important in health research: it gives the researchers greater confidence in saying that an intervention truly makes a difference; and that the difference is a positive one.

Psychosocial Support Program in Schools

A concrete example is the Psychosocial Support Program in Schools (PSP), which is being implemented in the Brazilian state of Pernambuco and will soon arrive in Rio Grande do Sul, in public state school networks.

In Rio Grande do Sul, Juntô, the Brazil initiative of the Stavros Niarchos Foundation (SNF) Global Center for Child and Adolescent Mental Health at the Child Mind Institute is seeking to understand the best delivery format. For this reason, some schools will be selected by lottery to receive the program in different formats, online or in-person. Before and after, assessment tools will be applied to measure changes in teachers’ knowledge about mental health, attitudes, and referral networks for mental health issues. In this way, we will be able to understand the program’s effectiveness in improving mental health indicators, as well as the best format for its implementation.

Limitations of RCTs

Despite being considered the best type of evidence, RCTs also have important limitations. First, they are expensive and difficult to implement, which often restricts their application. There are also ethical concerns, since it is not always possible or acceptable to randomly determine who does or does not receive a given intervention.

Another issue is the difficulty of generalizing results. The fact that something works in an experiment does not mean it will work the same way in other contexts, with other populations. This happens because real-world conditions differ from the controlled conditions of a study. Differences in context and in how factors interact can significantly alter the final outcome. There is also a dependence on large samples for results to be more reliable. When a study is small, results can be unstable, and small changes in the data can completely alter the conclusions.

RCT Basics

Want to know more about RCTs? Here are some core methodological components in our mini-dictionary!

Randomization: The probabilistic assignment of participants to the different arms of a study. This procedure aims to ensure that, in expectation, the groups are equivalent with respect to all relevant variables, both observable and unobservable. Consequently, any systematic differences in outcomes can be causally attributed to the tested intervention, rather than to confounding factors.

Causal inference: A variable can be considered a cause of an outcome if, after controlling for relevant confounding factors, the probability of the outcome occurring is greater in the presence of the variable than in its absence. The central challenge in empirical causal inference is precisely the identification and control of these confounding factors. In observational studies, this requires prior substantive knowledge and adequate statistical modeling. In RCTs, randomization acts as a mechanism that distributes these factors in a balanced way between groups, eliminating the need to explicitly observe them.

Blinding: The concealment of treatment allocation from participants, healthcare professionals, and outcome assessors, in order to reduce biases related to expectations, differential behavior, or subjective measurement of results.

Allocation concealment: The process of preventing the randomization sequence from being known prior to participant allocation, so as to avoid potential selection bias.

Statistical power calculation: Before implementation, the sample size required to detect an effect of clinically relevant magnitude with a given probability is defined.

About the authors

Núbia Stephane dos Reis Ribeiro and Ana Beatriz Araújo Santa Cruz Goyanna are members of Juntô Jovem, the youth council in Brazil of the Stavros Niarchos Foundation (SNF) Global Center for Child and Adolescent Mental Health at the Child Mind Institute.

Read more articles from the Our Voices blog

Black consciousness and the silent weight on mental health

My Message to Young People: Take Up Space

For schools to become spaces of healing and belonging

“Nothing about us without us”: youth build the future of mental health in Brazil