Clinical trials (hereafter referred to as “trials”) were defined as “any type of research that studies new tests and treatments and evaluates their effects on human health outcomes… including medical interventions, drugs, cells and other biological products, surgical procedures, radiological procedures, devices, behavioural treatments and preventive care” (World Health Organization, 2019). We identified studies that reported research into the development and/or evaluation of a clinical trial:

Eligible studies needed to report the explicit use of behavioural theory, model, or framework (TMFs) in their exploration of any of these trial areas. Trials were either hypothetical or real trials, as defined below:

Types of eligible studies included:

Citations identified through the search were independently assessed via title and abstract by two authors, each then screening a select 10% sample of the others to establish consensus. Abstracts were reviewed for explicit mentions of a potential behavioural theory, model, or framework, as well as assessing initial eligibility based on the inclusion/exclusion criteria. Any disagreements regarding eligibility were discussed between reviewers to establish consensus with a third member of the team acting to resolve any disputes. Full-text papers were obtained for studies considered potentially relevant and were further assessed against the inclusion/exclusion criteria. The eligibility of a potential theory, model, or framework as behavioural (based on the definitions below) was determined at this full-text stage. Full-text papers were assessed independently by two authors with a third author helping to establish consensus if there was any disagreement.

This review aimed to synthesise summary findings across a range of different study designs to describe current research in the area. Results of the studies are presented in a narrative format, separated into sections of thematically similar findings. Quantitative data, such as year and origin of publication, are presented using frequencies.

The TMFs utilised in eligible studies were assessed for how they were applied, being categorised into one of the following three categories derived from the Knowledge to Action (KTA) framework [19]:

A total of 96 eligible publications were identified through our search (see PRISMA diagram in Fig. 1 for more details). A majority of publications (n = 48, 50%) were published between 2011–2020, and primarily originated from North America (n = 65, 68%). A further breakdown of publication year and geographic origin is available in Fig. 2.

The clinical area of the host trials associated with each publication varied. The most frequently studied clinical areas were cancers (n = 26, 27%), followed by HIV/AIDS (n = 8, 8%). A subset of publications drew from multiple host trials, either specifying the clinical areas sampled from (n = 12, 13%), whilst others did not describe the clinical areas sampled (n = 7, 7%). Most of the host trials were evaluating an Investigational Medical Product (such as a drug) (n = 24, 25%) or more than one type of intervention (n = 21, 22%). Often, the type of intervention was either not stated or there was no specific intervention under study (n = 31, 32%). A majority of studies (n = 59, 61%) were conducted within “real” trials (i.e., one that participants were actually involved in), whilst the remainder were within a hypothetical trial.

Most publications (n = 58, 60%) studied participants, potential participants, or general members of the public. The remainder studied either trial/clinical staff (n = 13, 14%) or a combination of staff and participants (n = 25, 26%). The studies were carried out in primarily high resource settings (per OECD criteria) (n = 83, 86%). A subset of the included studies (n = 42, 44%) explored participation of underserved population in trials, such as racial/ethnic minorities or women.

The vast majority (n = 93, 97%) of included studies explored some aspect of trial conduct. Included studies often explored more than one trial process at a time, reflected in the percent totals below adding to more than 100%. Within conduct, recruitment was the most widely studied process (n = 76, 79% of all studies) with retention following (n = 17, 18% of all studies). Other trial processes reported in conduct included the implementation of the intervention (n = 5, 5% of all studies), implementation of the trial generally (n = 4, 4% of all studies), or involvement of PPI members (n = 1, 1% of all studies). Other processes studied outside conduct were trial design (n = 13, 14% of all studies), analysis (n = 1, 1% of all studies) and reporting (n = 1, 1% of all studies).

Most included studies applied the TMFs prospectively (n = 85, 89%). The number of individual applications of each TMF is detailed below in Fig. 3 with 23 theories, models, or frameworks being reported only once in the included studies. Counts for historically related TMFs (e.g., Theory of Reasoned Action and Theory of Planned Behaviour) are combined. Multiple TMFs were reported in some included studies, as such the overall counts of TMFs used exceeds the number of included papers (N = 96). The most frequently cited TMF was the Theoretical Domains Framework (TDF) (n = 30, 31%), followed by the Theory of Reasoned Action/Planned Behaviour (TRA/PB) (n = 23, 24%), and Social Learning/Cognitive Theory (SL/CT) (n = 12, 13%). Use of these three most frequently cited TMFs varied overtime (see Supplementary Fig. 1). The top three TMFs applied to understand trial behaviours within underserved populations were the TRA/PB (n = 13/42), SL/CT (n = 9/42), and the TDF (n = 5/42) (Supplementary Fig. 2).

Figure 3 How often each TMF was used in our sample

In addition to what TMFs were reported and how often, we also assessed how each TMF was applied according to three key KTA phases [19]. Of the 96 included studies, the majority (n = 40, 42%) applied the TMFs to “Identify Problem”. These studies were exploratory in nature, utilising the TMFs more generally to probe into aspects of one or more trial processes. The ‘Identity Problem’ studies were often explorations of larger trial processes like recruitment rather than specifying specific behaviours within the overall process e.g. identification of patients or seeking/providing consent. The other two KTA phases, “Assess barriers/facilitators to knowledge use” and “Select, tailor, implement interventions”, were represented equally in the included studies (28 studies each, 29%). The former focused on more specific targets for change, identifying behavioural determinants for discrete trial processes. Although the “Assess barriers/facilitators to knowledge use” studies stopped short of developing interventions, in some cases they proposed potential strategies to enhance facilitators and reduce barriers to their target behaviours. The third KTA phase, “Select, tailor, implement interventions”, was most often represented by studies applying the selected TMFs to plan and execute interventions targeting trial processes. This could include adapting existing interventions or evaluating whether interventions successfully influenced the behavioural determinants targeted. In some studies, TMFs were applied across all three KTA phases (e.g. TDF used across all three phases), whilst others were applied to one or two of the phases. The mapping of TMFs across the KTA phases is presented in Fig. 4 below.

Figure 4 What KTA phase each TMF was used in

Trial conduct was the primary process category reported within our included studies. A vast majority focused on recruitment and retention to trials, with recruitment research far outpacing retention. This abundance of recruitment-focused research is consistent with findings from previous reviews, both within and outside behavioural science applications [8, 9, 16, 20]. However, efforts to ensure research on retention is also offered appropriate attention has gained traction in recent years [8, 21–23]. Outside of conduct, trial design, the planning of the trial, was the next most frequently studied area, but still small in scale in comparison to the study of trial conduct processes, recruitment and retention.

In terms of clinical area, those studying trials within cancers were the majority. These findings may not be surprising given cancer is a particularly difficult clinical area to conduct trials within but also an area with several committed funding streams. One analysis of nearly eight-thousand cancer trials found that 39% of these failed to complete due to poor recruitment [24].

Trials suffer from a lack of diversity in their sample populations, an issue that carries ethical and scientific consequences to the validity of trial results [25]. Recent efforts to correct this imbalance are important steps forward but may lack demonstrable impact if they fail to meaningfully improve both the number and experiences of underserved participants within trials [25]. Our results reflect a significant portion of behaviourally-informed methods research focused on this area. Behavioural TMFs applied in the studies typically accommodate multiple socio-ecological levels in their constructs, as such they may be particularly well-suited to explore the complex psychological and socio-economic factors influencing poor participation from underserved groups [13, 14, 26]. Many of the investigations identified in our sample went past exploratory work into the development and testing of interventions aimed at improving underserved groups’ participation.

A number of included studies applied TMFs to develop behaviour change interventions, representing the ‘Select, tailor, implement interventions’ phase of the KTA cycle. Despite some interventions demonstrating effectiveness in improving trial processes like recruitment, they are rarely replicated – a generic issue in the improving recruitment and retention space [8, 9]. It is difficult to ascertain why so many potentially promising interventions appear to fade into obscurity past these initial publications. It may be that any improvements were seen as minimal, particularly when considering the fiscal and opportunity costs associated with further refining and/or implementing these interventions [27]. Another scenario may be that these interventions are implemented, but only within the limits of the trials units that participated in their development. As with clinical improvements, methodological improvements are slow to reach practice and require that their developers promote their implementation to wider user bases, along with the “buy-in” from these user bases that motivates them to adopt innovations [6, 7]. It is essential that these interventions are replicated and evaluated for acceptability and feasibility with trialists. The grounding of interventions in behavioural TMFs can facilitate this work, anchoring their mechanisms and outcomes in shared theoretical language [12, 28].

Finally, most of the included studies were hosted within “real” trials, gathering insights from participants and/or staff from active or completed trials rather than asking about a proposed or otherwise hypothetical trial. This is notable because of the literature available that suggests hypothetical decisions/behaviours can be different from actual behaviour [29, 30]. To frame that within trials, there are likely different theoretical considerations to explain/predict the behaviour of individuals that are faced with the actual consequences of enrolling/conducting a trial versus when they are contemplating it as a possibility. Care should then be exercised when designing/interpreting investigations of hypothetical versus real behaviours within trials, particularly whether mechanisms will generalise to real-world trial processes [30, 31].

The top three TMFs reported in our sample were the Social Cognitive Theory (formerly Social Learning Theory), the Theoretical Domains Framework, and the Theory of Planned Behaviour (formerly the Theory of Reasoned Action). The representation of these three TMFs is similar to our past review on application of behavioural science to recruitment and retention and appears consistent with the TMFs most often used in research on healthcare behaviours [14, 16, 26]. Although these TMFs are well-represented in the literature, it is still important to interrogate if these TMFs are accurately and cohesively representing the behavioural influences and mechanisms of clinical trial processes. How to go about choosing a TMF is a challenging endeavour, and one that is often predicated on the experience of the researchers, what they are familiar with, and what has been used previously [12, 13, 26]. This can lead to perpetualisation of TMFs without proper assessment of their purpose of fit or whether their use leads to improved effectiveness in interventions [12, 13]. Even if TMFs are only applied in exploratory roles, some are lacking in their abilities to frame factors outside of individual actors [26]. In complex systems, like trials, it is imperative to incorporate all interacting components within a system that can exert an influence on an individuals’ behaviour [12, 13]. Ideally, all aspects of a specified behaviour should be explicable through a chosen TMF and all constructs of a TMF found to contribute to these explanations [12]. When work moves into developing interventions, the proposed mechanisms of change for the intervention should be explicitly linked to the explanatory constructs of the TMF [13, 26, 31]. Further, any evaluation of an intervention’s effectiveness should select outcome measures that assess whether the appropriate theoretical constructs were influenced, via the proposed mechanism of action, leading to the desired change in behaviour [11, 26, 33–35]. Many of the included studies within our review do not clearly report information about theory selection, application, or evaluation, if at all. This often led to confusion in when a TMF was applied, how much of it was applied, and what trial-related behaviours researchers were trying to target [12, 33]. Without proper specification of the target behaviour, it is unclear to those outside the research team what explicit behaviours within larger conceptual processes, such as “recruitment”, “participation”, or “enrolment”, are being targeted in the investigation. Such ambiguity deters attempts to collate findings across papers, except at the highest level of processes [28, 36]. With so many individual, distinct behaviours existing within processes like recruitment, collapsing them under one umbrella term fails to illustrate the heterogeneity of behaviours being studied. Lack of specification also precludes attempts to replicate findings/interventions or to adapt interventions to other populations and contexts [11–13, 33, 34]. This lack of reporting clarity is not unique to trials methodology but is particularly burdensome to trialists that do not have the time or possibly experience to tease out essential details on intervention operationalisation, and how they might be able to refine implementation to suit their individual trial centre circumstances [7, 12, 33, 37].

This review has several strengths. Our search strategy was designed to be as inclusive as possible, of both behavioural TMFs and trial processes. The volume of results returned from the database search makes it unlikely that much published literature was missed. Incorporation of a forward-backward citation analysis further reduces this possibility. One potential limitation to our search strategy is the search terms utilised to capture processes outside of conduct. The review utilised search terms generated for previous reviews of recruitment and retention research but did not benefit from similar resources outside those areas. Instead, we added search terms based on key papers within trial processes outside of recruitment and retention to ameliorate this potential gap.

The review benefitted from prior work to identify behavioural TMFs and we referenced their results extensively when including TMFs [14]. We also utilised Davis et al’s definition of theory and applied it to potential TMFs outside their list, further ensuring all relevant TMFs were captured [14]. This process, however, relied on checking the provided references for the TMF cited in the included study in order to verify if the criteria for behaviour and theory were met. There were some instances in which the cited TMF references were insufficient to make an informed decision on the TMFs inclusion and so were excluded.

This work has highlighted the wide-ranging application of behavioural science to processes within clinical trials. However, it also reinforces a focus of trials methods research being on recruitment and retention. Future activity should consider other trial related behaviours, such as dissemination of trial results to participants, when employing a behavioural lens. Efforts to standardise how behaviourally-focused research is conducted and reported are ongoing and need to be implemented within the trials methodology space [12, 36, 38]. This will be essential in synthesising evidence on factors underlying trial participation, and related processes, and how best to improve it. Further clarity in TMF application will allow for hypothesis testing of theoretical constructs, either providing evidence of fit or demonstrating lack of utility, both providing a more succinct repertoire of potential TMFs to choose from in future studies [26, 32]. Ongoing research aims to provide agreement on which TMFs can be considered further for their utility in understanding trial processes. Extending the methodological application of behavioural TMFs across trials (and processes) in a range of settings, recruiting diverse populations and testing a variety of interventions will further develop the knowledge base on how behavioural science can be leveraged.