Barriers to recruitment when conducting a commissioned randomised controlled trial of medication versus psychological therapy for generalised anxiety disorder: some lessons learned

Randomised controlled trials (RCTs) are considered the ‘gold standard’ for evaluating the effectiveness of interventions [1]. However, conducting high-quality RCTs may prove to be challenging because of unexpected events that occur ‘en route’, leading to premature discontinuation in an estimated 25% of RCTs conducted in Europe, Canada and Australia [2‐4].

Trials may be stopped prematurely for a variety of factors, including unexpected harmful events [5] and greater-than-expected benefits of the intervention manifesting early in the trial [6]. New findings answering the primary research question or raising concerns about the safety of the intervention may also arise [7]. Although these factors play a role in the premature discontinuation of RCTs, the literature suggests that the main reason for terminating a study early is the slow and inadequate recruitment of participants [3, 8, 9]. This, along with evidence indicating that recruitment is adequate and timely in only 50% of RCTs worldwide [10], renders recruitment a key issue. In the UK, several studies have reported similar results in trials funded by two of the UK’s largest funding agencies: the UK Medical Research Council and the National Institute of Health Research (NIHR) Health Technology Assessment (HTA) Programme [9, 11, 12].

An NIHR HTA–funded trial of the anti-depressant sertraline versus cognitive behavioural therapy (CBT) for generalised anxiety (ToSCA) had to be discontinued prematurely because of a failure to recruit. Generalised anxiety disorder (GAD) is common, has distressing symptoms and often is associated with poor general functioning, substance misuse and financial difficulties [13, 14]. The 2011 UK National Institute for Health and Clinical Excellence (NICE) guidelines recommend low-intensity psychological therapies as the best initial treatment, but it is uncertain whether pharmacological or psychological therapy provides the most effective longer-term treatment for those not responding to low-intensity therapies [15]. The NIHR HTA commissioned and funded ToSCA, a UK, multi-site, phase IV RCT aiming to answer this question by assessing the clinical effectiveness at 12 months of treatment with the anti-depressant drug sertraline compared with high-intensity CBT in participants with established GAD, who had failed to respond to a low-intensity psychological intervention. The study outcomes included measures of the effects of anxiety, participant satisfaction and health economic data.

The study set-up started in August 2014, and participant recruitment was anticipated to commence in February 2015. However, participant recruitment was delayed until July 2015 because of several research governance issues, at both a national and a local level, which required consultation with both the lead Clinical Research Network and the Health Research Authority as well as local research and development (R&D) offices for the pilot sites. This was due to a query as to whether participating general practitioner (GP) practices should be registered as individual research sites, which would have been administratively burdensome given that most of the planned 360 trial participants were likely to be registered at different practices. This was addressed with a generic site-specific information form developed by working together with the lead Clinical Research Network and the Health Research Authority, and study-wide R&D approval was received 3 months after initial submission of the full set of documents, although it took a further 2 months to receive local assurances for all of the pilot sites. These delays were very time-consuming and meant that we were not in a position to start work on recruiting participants as planned and initially discussed with the pilot sites, but ultimately they were not the main reason for the failure to recruit as described in the article. The trial aimed to recruit participants via local Improving Access to Psychological Therapies (IAPT) services [16] that served as participating sites. The initial internal pilot phase of the study was planned to run for 12 months and involve four pilot IAPT sites. Unfortunately, however, at 6 months into the internal pilot in January 2016, despite varied attempts to improve participant recruitment, only five patients had been randomly assigned across all of the pilot sites against a target of 40 by this stage. After a monitoring meeting between the study team and the funders, the NIHR HTA decided to discontinue the study and asked the trial team to stop recruiting new participants and to make plans to close down the trial.

Premature discontinuation of RCTs is costly in terms of resources and a failure to answer the underlying research question and is potentially frustrating for the investigators [9, 10, 17] and for the research ethics committees and regulatory agencies that need to devote a considerable amount of time and effort to review protocols [10, 18, 19]. It also raises ethical considerations for patients who have consented to be involved. Therefore, it is very important to evaluate and learn pertinent lessons from previous trials where this has happened.

In light of our experience with the ToSCA trial, we went back to the initial trial design to assess at what points our assumptions about potential recruitment strategies were unsuccessful, and what might have been improved upon or changed in terms of the trial design, in order to give a better chance of a successful outcome. Making this information available to the scientific community may help others to learn from our experience.

The overall aim of this study was to identify reasons for the poor recruitment to the ToSCA trial which had resulted in its premature closure. Our retrospective database work indicates that only around 12% of potentially eligible patients for the trial were identified by the PWPs at the pilot sites, but the database results also indicate that only a small percentage (5%) of people assessed by the IAPT service with a GAD-7 score of at least 10 and thus possible GAD would have been eligible. In addition, of those patients who were identified by the PWPs in our study as possibly suitable for the trial, three quarters (45/60) declined to take part; the great majority did so because they did not want to be randomly assigned to receive medication.

These findings are in line with the results of a Cochrane review suggesting that poor recruitment to RCTs can be significantly affected by factors influencing the recruiting clinicians [23]. A major problem within the trial appears to have been a low rate of identification of potential participants by the PWPs. We investigated this during the pilot phase and identified several contributing factors, including the fact that the PWPs were clinically stretched and most did not see the research project as a priority for their time. Fletcher et al. [17] suggested that funding-protected research time and having extra staff support [23] could improve recruitment. We managed to secure some additional funding for the lead PWPs (one lead PWP per pilot site) but unfortunately this did not have a significant positive impact on recruitment. Another factor contributing to the low identification rate of potential participants by the PWPs may have been that they are trained to deliver a low-intensity CBT intervention which may have given them a bias, whether conscious or not, towards psychological therapy. There is a potential tension between clinical and research roles when asking clinicians to recruit to research studies [17]. As the recruiting clinicians often also have a gatekeeper role, they may approach only patients whom they personally deem suitable to take part and not all potentially eligible patients for a trial. Additionally, there is evidence for an association between clinicians’ orientation towards research and successful recruitment of patients for clinical trials [23]. Training programmes for clinicians recruiting for RCTs consisting specifically of workshops over one or two consecutive days covering trial-specific and generic research issues may also contribute to enhanced recruitment [24].

Although we thought we had discussed the trial in some detail with the IAPT services involved, it might have been helpful to have discussed the implications of study eligibility criteria more thoroughly in advance with key stakeholders, such as senior clinicians/managers, in order to anticipate potential differences in perspective between clinicians and researchers about the appropriateness for referral to the trial. An alternative might have been to involve members of the research team more directly in participant recruitment (e.g., by placing them within the PWP clinics) but this would have had significant resource implications and was not encouraged by the IAPT services involved.

As only 5% of the 7602 patients passing through the IAPT pilot sites were identified in the database search as potentially eligible for recruitment to the trial, this would have made the whole process very tight. Furthermore, a significant number of the 5% might have not fulfilled all of the study eligibility criteria in terms of not having major depression or having another more significant anxiety disorder, further reducing the population of GAD patients who met eligibility criteria for the study. For example, if one extrapolates from our pilot study, it is possible that three quarters of the 366 deemed potentially eligible may have not wanted to take part, reducing the sample to merely 91 people.

The likelihood that many of the 5% would have another disorder may correlate with the concerns held by the IAPT supervisors about what is the most significant disorder affecting each individual patient. Most GAD is known to be comorbid with other axis 1 disorders [15]. The study criteria would have excluded at baseline assessment all those with comorbid depression and patients considering a comorbid anxiety disorder more significant than their GAD, but it would have been of benefit to get these more accurate figures, not only for recruitment to this trial but also in terms of assessing the relevance of the trial eligibility criteria for the overall service provision for GAD. Shortage of eligible patients (i.e., identification of fewer eligible patients than originally anticipated) has previously been reported as a major obstacle to recruitment in relation to the design and conduct of RCTs [25‐27].

In addition, the PWPs stated that their patients were often reluctant to consider a design in which they might be randomly assigned to medication rather than a psychological intervention. Patients’ treatment preferences have been widely reported as key barriers to recruitment [28‐31], and a meta-analysis of patient views about receiving psychological or pharmacological treatment for a range of psychiatric disorders indicated a significant preference for psychological treatment [32]. This patient view might have been amplified by the fact that they were already receiving treatment within a psychological therapy service.

Prior experience of treatment has been listed by Stiggelbout and de Haes [33] as a factor shaping patients’ treatment preferences. It has also been demonstrated that the way in which the information about the interventions is presented or ‘framed’ may play a role in shaping patients’ treatment preferences [34, 35]. We attempted to address this factor by suggesting appropriate approaches for the presentation of the trial and the planned interventions as well as suitable responses which the PWPs might give to patient concerns about receiving treatment with anti-depressants, but we had little success. The fact that the medication (i.e., the anti-depressant drug sertraline) was available outside of the trial and that patients were able to get a prescription for it from their GP if they wished might have also played a role in patients’ reluctance to take part in the trial. It would have been very helpful to conduct semi-structured qualitative interviews with both patients and PWPs to further explore these issues around poor recruitment, but unfortunately there was no funding available to do this.

An alternative research design may be appropriate to investigate the main clinical question which remains to be answered, namely whether high-intensity CBT or medication has a better longer-term impact for people with clinically significant GAD. This might be carried out by recruiting potential participants from primary care rather than subsequent to referral for a low-intensity treatment in an IAPT setting, as primary care is where initial discussions about whether patients would prefer a pharmacological or psychological approach usually take place and they may be more prepared to accept randomisation to either at this stage. This might mean circumventing the low-intensity treatment indicated at step 2 in the NICE guidelines, and our database work indicated that quite a large number of people with a GAD-7 score of at least 10 improved with three or more sessions of low-intensity therapy (Fig. 3), but an alternative design might be for those in the psychological therapy arm to be initially offered low-intensity therapy and to proceed to high-intensity therapy only if they did not improve. In addition, it would probably be clinically appropriate to ask people being stepped straight up to a step 3 psychological intervention whether they would be interested in taking part in the trial, which was not the case with the design used in the ToSCA trial. Each research design has its pros and cons, but a lesson from this trial might be not to propose or commission a randomised trial with too many restrictions which may affect recruitment as this can affect both participant recruitment and the generalisability of any results.