What are some of the limitations of RCTs in mental health research?

Many scholars, mainly from the medical field, have previously recognized the limitations of RCTs (e.g. Hannan, Reference Hannan2008; Booth & Tannock, Reference Booth and Tannock2014). Some limitations of a purely positivistic academic perspective are that it:

1. (1) dismisses techniques and practices with greater external validity, such as observational studies and ethnographic work;

2. (2) reduces changes (and impact) to those measured by a set of questionnaires, thus failing to acknowledge that any given behavior is the result of a complex set of interactions with others around as well as with very specific environmental circumstances;

3. (3) assumes that experimental knowledge gathered from a very specific context is applicable and generalizable to different settings; and thus

4. (4) usually fails to acknowledge the value of context and culture.

Evidence from clinical reflections, observations, ethnographic work, case studies and qualitative data are rarely taken into account when assessing the impact of an intervention, even though they have greater external validity than RCTs and are useful for acknowledging complexity of outcomes. The need to simplify outcomes in RCTs not only affects evaluation research but also influences intervention design. Interventions are becoming simpler, briefer and mainly focalized on one or two risk factors (i.e. those that can be measured and quantified). In addition, regardless of the low external validity of RCTs, interventions tend to be disseminated widely across communities, countries and cultures. Even though wide dissemination might be considered cost-effective, local culture and context-specific factors tend to be ignored when making these decisions. More importantly, once RCTs have been conducted and interventions proven impactful, its content and shape ‘freezes’ in time: any major change to the intervention might revoke its evidence-based status, and new funding and effort is required to establish updated experimental impact. Freezing interventions in space and time is imposed through strict certifications, adherence protocols and fidelity checks. Therefore, granting evidence-based status based on results from RCTs assumes that different communities and contexts are identical (and what works somewhere would work elsewhere) and fails to acknowledge that societies evolve rapidly in time.

Are RCTs an efficient tool in a rapidly changing world?

Our work has mainly centered on the prevention of drug and alcohol use during childhood and adolescence in Latin America, and thus, the following examples will be specific to this issue. One example of the rapid evolution of society is the changes in marijuana use legislation around the world (e.g. legalization of marijuana in some states of the USA and in Uruguay, for example). These rapid societal changes are modifying perceptions of risk and attitudes about this drug (Schuermeyer et al. Reference Schuermeyer, Salomonsen-Sautel, Price, Balan, Thurstone, Min and Sakai2014). Current policies are also reducing the credibility of preventive interventions (Haggerty et al. Reference Haggerty, Santaella and Caulkins2015) and thus organically changing content of many interventions and the perceived need for their implementation.

One well-known intervention is the Life Skills Training Program (LST), which has been commonly implemented in both high- and low-income countries (Botvin & Griffin, Reference Botvin and Griffin2004, Reference Botvin, Griffin, Murta, Leandro-França, Brito dos antos and Polejack2015). This intervention is designed to improve social skills such as communication and emotional regulation in adolescents in order to prevent mental health difficulties later in life (e.g. substance use). These skills seem universal and applicable to most countries and cultures. However, there is no empirical evidence suggesting that the skills are indeed universal, and given that RCTs of the intervention have been mainly conducted in the USA, it is impossible to establish generalizability of findings. Most importantly, LST was developed 30 years ago, when adolescents had a very different pace of living, different concerns and environmental influences. There are currently no studies evaluating if the intervention is still relevant to the needs of adolescents today. In addition, recent RCTs of LST report small or no effects at all (Vicary et al. Reference Vicary, Henry, Bechtel, Swisher, Smith, Wylie and Hopkins2004, Reference Vicary, Smith, Swisher, Hopkins, Elek, Bechtel and Henry2006; Gorman, Reference Gorman2011; Luna-Adame et al. Reference Luna-Adame, Carrasco-Giménez and Rueda-García2013). A potential explanation is that the evidence of efficacy from previous RCTs might have been fixated to a very specific context and point in time. If developers incorporate changes in the original intervention to ‘update’ its content and techniques, then a new wave of expensive RCTs will be needed to grant evidence-based status. By the time this experimental information is available, society would have probably evolved and new changes to the content of the intervention will be needed (as well as new evidence of efficacy). In Fig. 1, we provided a diagrammatic conceptualization of this vicious cycle of gathering experimental evidence that is not necessarily timed around societal changes.

Fig. 1. Diagrammatic conceptualization of the evidence-based cycle. Note 1. The gray sphere delineates what can be measured and tested empirically. Note 2. The process of establishing evidence is rigid and iterative. Dynamic societal changes (e.g. needs, structures and attitudes) are difficult to measure from a fully positivistic stance. These changes and the methods that could be used to understand them (e.g. experiential, qualitative and ethnographic data) are left out in the cycle of establishing evidence for interventions. Once societal changes are perceived by developers and policy makers, interventions are adapted and the process for establishing evidence starts once again.

Our work with the Strengthening Families Program 10–14 (SFP 10–14; Molgaard et al. Reference Molgaard, Kumpfer and Fleming1997) in Panama provides an example of how qualitative information can overcome some of these limitations. The SFP 10–14 is an evidence-based intervention (often listed in compilations as one with ‘promising evidence’) with several trials documenting its effectiveness, most of which have been conducted in the USA. In between 2010 and 2012, the United Nations Office on Drugs and Crime invested in training facilitators to deliver the intervention in Panama in order to prevent adolescent substance use. Instead of conducting an RCT, we opted for evaluating the appropriateness of its content in this context, its potential for implementation and parents’ experiences after taking part in the intervention using qualitative methods (Mejia et al. Reference Mejia, Ulph and Calam2014, Reference Mejia, Ulph and Calam2015). Although our methodology might not be sufficient for establishing efficacy/effectiveness based on current guidelines, it allowed an in-depth, participant-driven, comprehensive understanding of the impact of the intervention in these communities. We were also able to gather practical information that could inform larger-scale implementation. This would not have been possible by solely conducting an RCT.