Skip to main content
World Psychiatry logoLink to World Psychiatry
. 2025 Sep 15;24(3):436–437. doi: 10.1002/wps.21359

Has the time come to stop using control groups in trials of psychosocial interventions?

Pim Cuijpers 1,2
PMCID: PMC12434354  PMID: 40948066

Control conditions are an essential element in evidence‐based mental health. In randomized trials, these conditions are needed to make sure that the improvements found in people who receive an intervention are not caused by other factors, such as spontaneous recovery. But control conditions in mental health research trials have many problems.

Drug trials are relatively simple, because they almost always can make use of pill placebo control conditions. This allows to examine the exact contribution to the effects of the specific substance in the pill, compared to the same pill without the active substance, in double‐blind designs. However, the outcomes of these trials can also be biased. For example, adverse effects of medications in psychiatry are common, and often reveal who receives the active drug in a randomized trial, thus breaking the masking. The outcomes may also be affected, for example, by sponsorship bias and the use of “placebo run‐in” methods 1 . Furthermore, pill placebo cannot be used in trials examining psychosocial interventions.

One of the most used control conditions in these latter trials is waitlist, in which participants receive the intervention after the experimental group has finished 2 , 3 . One advantage is that in the end all participants receive the intervention. A disadvantage is, however, that waitlist controlled trials result in larger effects of an intervention than, for example, usual care control groups, and may overestimate the true effect of the intervention 4 , 5 . It is not clear why the effects are larger – maybe because of increased expectations, but that is not certain 2 .

Usual care is another control condition that is often chosen 6 . Trials with such control groups have the advantage that they can show what an intervention can add to the care that is already available. The main disadvantage, however, is that usual care can vary considerably. It can be delivered in primary care, in specialized care, in general medical settings, or in other settings. The care delivered within each of these settings also varies considerably across countries and regions. For example, usual care delivered in a primary care setting can be a relatively extensive treatment in high‐income countries, but in low‐ and middle‐income countries it typically means no treatment at all. To solve this, trials sometimes use “enhanced usual care”, in which the care is delivered as it should be in ideal circumstances. However, this does not solve the issue of the heterogeneity of usual care across settings and countries, making the results of trials difficult to generalize to other settings and communities than where the trial was conducted.

Another type of control condition is “no treatment”, in which participants are not offered care at all. However, every person has the right to access treatment, whether or not he/she participates in a trial. This means that everyone always has access to “usual care”. “No treatment” may, therefore, be better considered as a specific type of usual care. It suffers from the same heterogeneity as other usual care conditions, because it varies considerably across countries and communities.

“Psychological placebos” have in principle the possibility to mask participants for the condition they have been assigned to 2 , 3 . However, when a psychological placebo is delivered in such a way that it is a credible treatment, it can be quite effective in itself. For example, cognitive behavioural therapy for depression does not significantly differ from non‐directive counseling, which is often used as a psychological placebo condition 7 . If, on the other hand, psychological placebo is not delivered as a credible treatment, then patients know that they are in the control group, the masking is broken, and expectancies will be reduced. In this case, patients in the control group may be disappointed and demoralized, and the effects that are found in the trial can be more an artefact than a true outcome of the intervention. “Psychological placebos” should, therefore, preferably be avoided in trials, or at least be used very cautiously.

So, there is not really a satisfactory control condition that can be used when examining the effects of psychosocial interventions. But, over the past decades, an increasing number of randomized controlled trials have examined the effects of many psychosocial interventions. For example, in the field of psychotherapies for depression, more than 400 randomized trials have been conducted with all kinds of control conditions. Overall, these therapies have comparable effects, can be administered in all kinds of formats and in many different target groups 5 . At some point new controlled trials add very little new knowledge to what is already available. In psychotherapies for depression that seems to be the case already, but also in other research areas there are several dozens and sometimes hundreds of trials available on psychosocial interventions, such as in the treatment of anxiety disorders, substance use disorders and psychotic disorders.

When we know that a treatment works in comparison to a control group, and that has been confirmed in multiple randomized trials, we could step away from controlled trials altogether and focus on other clinical questions that are also relevant. New treatments can be compared to established treatments in randomized trials without a control group to examine if they have comparable effects. Fractional factorial designs do not need a control condition either and allow to examine effective components of psychosocial interventions 8 . Stepped wedged designs can be used to examine the effects but also how to implement interventions in routine care. And we can focus on the collection of large, high‐quality datasets in routine care, with as many relevant predictors as possible, in order to identify who benefits from which treatment and further develop personalized interventions. Using techniques such as propensity score matching 9 , we can estimate the effects of an intervention by accounting for all covariates that may predict its outcome.

In summary, the most frequently used control conditions in trials of psychosocial interventions overestimate the effects (waitlist), are extremely heterogeneous by design and cannot be generalized to other settings or countries (usual care and no treatment), or cannot be delivered as they should (psychological placebos). In randomized controlled trials there is no good alternative for such control groups, and researchers should carefully consider the advantages and disadvantages of each type of control before deciding which one to use. At some point, however, we have enough controlled trials, and new ones do not add very much to what is already known. For several research areas, this moment is there or very close. This allows to step away from conventional randomized trials with control groups and to examine other questions that are relevant to improve outcomes for patients.

REFERENCES


Articles from World Psychiatry are provided here courtesy of The World Psychiatric Association

RESOURCES