Introduction

Optimising recruitment to randomised controlled trials (RCTs) is a priority for clinicians, funding bodies, and healthcare decision-makers. Increasingly constrained resources reinforce the need for efficient generation of high-quality evidence to optimise patient care and inform service provision. Despite stringent review processes, at least half of funded RCTs fail to meet recruitment targets [1], and recruitment difficulties remain the most frequent reason for premature trial closure [2]. Not only are these difficulties associated with considerable financial waste [3], but poor recruitment can delay progress or leave unanswered clinical questions in underpowered or incomplete studies.

Attempts to unravel RCT recruitment difficulties have focused on the pivotal role of front-line clinicians as recruiters. Clinicians are on the interface of research and practice, and are often responsible for presenting the option of RCT participation to eligible patients. Clinicians’ perceptions of “equipoise” has repeatedly been cited as an assumed barrier to RCT recruitment [4,5]. Equipoise is broadly understood as a state of “substantial uncertainty” around treatment superiority [6], though there are discordant views around the degree and nature of uncertainty needed to justify randomisation [6]. Clinicians’ personal experiences and awareness of anecdotal or non-RCT evidence can disrupt their sense of equipoise. This has underpinned commentaries on why recruiting to RCTs can be so challenging—particularly in disciplines that have a propensity for swift and clear-cut decision-making (e.g., surgery) [7]. Equipoise at an individual clinician level (“individual equipoise”) can thus be hard to achieve—especially where ingrained clinical practices are under evaluation. The concept of “community equipoise” (or “clinical equipoise”) refers to the principle that there is no consensus in the clinical community about which treatment is best. In theory, this framing of equipoise enables clinicians to align their instinct to act in the individual patient’s interest with the utilitarian endeavour of benefiting future populations [8]. Although this viewpoint has been critiqued [9], clinical equipoise is widely accepted as the fundamental prerequisite that justifies an RCT. The pragmatic definition of equipoise adopted in this study framed the concept as a state where patients would be neither “advantaged nor disadvantaged” if they were to receive any of the trial treatments (as far as existing evidence would dictate) [10]. The recruiting clinician’s role was to convey this notion to eligible patients.

Previous research suggests that patients’ reasons for declining trial participation are often based on their interpretations of clinicians’ treatment preferences [11–14]; this raises questions about how successfully equipoise is conveyed to patients in practice. Interview studies have shown that clinical recruiters are not necessarily in individual equipoise [12,15–17] but assume that their personal views can be set aside during encounters with trial-eligible patients [4,18]. Although there has been extensive theoretical commentary on these issues [5,19,20], equipoise communication has not yet been investigated in practice. There is therefore little empirical evidence to inform practical guidance for clinicians, despite recognition that discussing equipoise is a core component of their roles. We set out to investigate how clinicians conveyed equipoise during recruitment appointments in ongoing RCTs, with a view to identifying practices that supported or hindered equipoise communication. A secondary objective of the study was to examine how clinicians’ practices compared with their reported intentions.

Methods

Ethical approval for this study was granted by six UK research ethics committees, with full details (e.g., dates, reference numbers) presented in S1 Text.

This was a qualitative study that adopted ethnographic approaches. Data were available from three sources: (i) audio-recorded appointments in which clinicians presented RCTs to eligible patients (“recruitment appointments”); (ii) interviews with recruiting clinicians in which perceptions of equipoise about trial treatments were explored, to enable comparison of reported intentions and actual practices; and (iii) documentary analysis of trial protocols containing the scientific evidence underlying the RCT, to aid interpretation of observed practices (e.g., assessing accuracy of information provision).

Sampling Criteria for Cross-RCT Investigation

Sampling for the present investigation involved (i) selecting the RCTs and (ii) selecting the audio-recorded appointments. A purposeful sample of four (of 16) RCTs was initially selected. Sampling was directed by the intention to include a range of clinical specialties and types of treatment comparison (e.g., trials comparing two or more existing treatments, trials comparing current practice with a new technology). Although the sampling frame included more RCTs with surgical interventions, we intentionally included two non-surgical RCTs to increase variation. Sampling was also partially informed by practical issues such as the number of audio-recorded appointments collected. Two further RCTs were later added to the sample to confirm/disconfirm emerging findings. Details of the RCTs are shown in Table 1. The balance of surgical to non-surgical RCTs in the final sample was reflective of the sampling frame.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. RCT details and associated number of appointments recorded (shown by participating clinician).

https://doi.org/10.1371/journal.pmed.1002147.t001

Audio-recorded appointments from the six RCTs were eligible for inclusion if they took place prior to any feedback to or training of the clinician (based on the recorded appointments), and if they were led by a clinician who had also been interviewed.

A total of 105 recruitment appointments met the study inclusion criteria and were analysed, coupled with 23 interviews with the clinicians who led these appointments. All clinicians were secondary care consultants (surgeons, oncologists, and neurologists). Interviews lasted between 21 min and 92 min. The selected data collated for this study were collected between October 2010 and December 2014. None of the 23 clinicians or 105 patients withdrew from the study (i.e., following provision of consent for the interviews or audio-recorded appointments).

The distribution of appointments per trial and clinician is shown in Table 1. In all, 83 of 105 appointments were with patients who ultimately declined trial participation (79%), 19 with patients who agreed to randomisation (18%), and three with patients where the outcome was not known (3%).

Results

Given the complex and novel approach to data collection and analysis of this study, the findings have been presented in an order that most clearly conveys the key messages to emerge from the study. A summary of clinicians’ individual perspectives on equipoise is presented first, to provide a framework for drawing comparisons between reported intentions and actual practices. The findings from the audio-recorded appointments—the focus of this study—are presented next. References are made to clinicians’ interview accounts where relevant to demonstrate how reported intentions compared with actual practices.

Interview Findings: Clinicians’ Individual Perceptions of Equipoise

Most clinicians expressed commitment to the RCT, acknowledging uncertainty amongst the clinical community about how patients should be treated (community equipoise). All clinicians acknowledged the need for evidence and indicated that they were individually “uncertain” about optimal treatment. However, interview accounts revealed a spectrum of uncertainty, ranging from complete ambivalence to strong beliefs about the superiority of a particular treatment. On the basis of the congruence of these findings with previous research [4], clinicians could be considered to fall on a continuum ranging from “clear equipoise” (equivocal views about trial treatments) to “not in equipoise” (treatment preferences and/or unbalanced views about trial treatments). All clinicians reported being uncertain or “in equipoise” on at least one occasion during the interview, but approximately half presented treatment preferences or polarised views at other stages of the interview. Clinicians were therefore divided into two groups: “more balanced” and “less balanced” (Fig 1). Those categorised as more balanced (n = 10) consistently indicated that they viewed the trial treatments in equilibrium (Box 1). Less balanced clinicians (n = 13) expressed at least one clear preference or polarised opinion about treatment superiority during their interview—either in a general sense or in relation to particular sub-groups of eligible patients (Box 1). Evidence to support categorisation of each of the 13 less balanced clinicians as such is presented in S1 Data.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Clinicians’ individual perceptions of equipoise, based on interview accounts.

https://doi.org/10.1371/journal.pmed.1002147.g001

Box 1. Interview Extracts Comparing More Balanced and Less Balanced Clinicians’ Views on RCT Treatments

More Balanced Group

R5: Um, so in other words here equipoise, as I say I think I do have equipoise here because I really do not know which is the best treatment for a patient. (R5, RCT2—more balanced group)

R1: If I thought I knew what the answer was then I probably wouldn’t be interested in being in the trial. (R1, RCT1—more balanced group)

R9: I don’t know what results this is going to be, I think there remains a degree of uncertainty. It keeps me awake at night. (R9, RCT3—more balanced group)

Less Balanced Group

R7: [My] gut reaction is probably that it [treatment x] does something and something needs to improve the outcome of these patients really, but my normal practice is not to give [treatment x] because I have to do things that have got an evidence base. (R7, RCT2—less balanced group)

R4: I think we’ll see what other trials have demonstrated. I think we’ll see that [treatment x] is better initially. (R4, RCT1—less balanced group)

R2: I’m struggling personally with what we’re going to do in the trial with the patients who for whatever reason don’t get on with [treatment y]. (R2, RCT1—less balanced group)

Regardless of their individual perceptions of equipoise, all clinicians indicated that their aim was to inform patients about the RCT without betraying their personal views. This reflected these clinicians’ appreciation for community equipoise, which in turn reinforced prior work conducted in this field [4]. Although some anticipated that they might inadvertently influence patient choice, most believed they could achieve neutral information provision:

I can be unbiased during a presentation to them. [‖] If you want me to be part of the randomised trial then I can’t, I have to be completely non-biased. (R4, RCT1—less balanced group)

I’m probably biased towards a surgical treatment slightly, but I try not to express that. (R20, RCT6—less balanced group)

The less balanced versus more balanced categorisation provided a framework for comparing clinicians’ individual perceptions of equipoise with their practices during recruitment appointments.

Audio-Recorded Appointment Findings: Articulating Equipoise

Equipoise issues were discussed in most of the observed appointments (83 of 105, 79%). Although the term “equipoise” was generally not used, the principle was described in a variety of ways. Equipoise was most often discussed in collective (“we”) terms (79 of 83 [95%] appointments), thus indicating a tendency for clinicians to convey community (rather than individual) equipoise during patient encounters. This included references to clinical communities not knowing which treatment is best and presentations of equipoise as a global phenomenon (“nobody knows which treatment is best”). Individual equipoise (“I don’t know which is best”) was expressed in only 27 of 83 appointments (33%), though this was always in conjunction with statements of collective equipoise.

Equipoise was most commonly portrayed through statements indicating an absence of knowledge about which treatment was best (e.g., “we don’t know what’s best”) or through explanations of what the RCT was seeking to achieve (e.g., “we’re trying to find out”). Statements specifying uncertainty were also used, albeit less frequently (e.g., “we are uncertain”). Equipoise was also occasionally communicated by emphasising treatment equivalence, most often through reassurance that either/any trial treatment was equally appropriate for the patient. These statements were not mutually exclusive, with most appointments incorporating one or more of these statements (Table 2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Clinicians’ descriptions of the concept of equipoise to patients during RCT recruitment appointments.

https://doi.org/10.1371/journal.pmed.1002147.t002

Discussion

This study investigated how clinicians communicated equipoise in RCT recruitment appointments. Presenting and sustaining equipoise was found to be a challenging process that was highly entwined with clinicians’ own perceptions of equipoise. Clinicians varied in terms of their individual sense of equipoise, but were committed to the RCTs and reported that they intended to present treatments to patients in a neutral and balanced way. These expressed intentions did not always translate into practice—especially when clinicians harboured individual preferences or polarised views about treatments. Although equipoise was articulated in most appointments, observations across the RCTs revealed common practices that compromised equipoise communication. Clinicians with treatment preferences or less balanced views overrode or undermined equipoise through offering treatment recommendations, presenting imbalanced descriptions of trial treatments, or making clear their predictions about trial outcomes. These difficulties arose despite clinicians’ intentions to conceal or suspend their personal inclinations or beliefs.

A key strength of this study was its analysis of real (rather than reported) practices and events. Issues of recollection bias and selective reporting were thus minimised. Paired interview and observational data allowed novel comparison of intended and actual practices, providing potential explanations for some of the observed practices. The study drew on a large dataset derived from a range of RCTs and clinical contexts, and focused on common practices that were identified across these settings. As such, findings were not limited to a particular RCT context and are likely to be applicable more widely. A qualitative approach was crucial for this particular research question given the absence of prior empirical research and the need for an inductive approach. We made a conscious effort to look across a broad range of RCTs and focus on the practices that were common across these contexts. The reported findings are thus likely to be salient to clinicians across specialties.

The study had a number of limitations. Data were derived from RCTs that were experiencing (or were anticipated to encounter) recruitment difficulties. There was a prior intention to focus on challenging RCTs with a view to informing strategies to overcome recruitment difficulties, but our findings need to be considered further in RCTs that recruit well. Another limitation was the reliance on RCTs involving surgical interventions, which could have raised particular issues. Nonetheless, our sample also included RCTs that did not include surgical interventions, and although these were fewer in number, we did not identify any notable differences between RCT types in relation to the key findings. Integration of the recruitment sub-study (QRI) across multiple RCTs meant that some processes (e.g., study invite distribution) were organised centrally (e.g., by clinical trials units) or influenced by local issues (e.g., research nurses’ capacity and work patterns in RCT centres). We did not have a sense of how many patients and clinicians declined participation in our study (or were not approached), and as a consequence it is possible that the data captured provide insights from particular groups of patients/clinicians. We believe this is unlikely in the case of patients, as informal discussion with research nurses revealed that it was rare for patients to decline consent for audio-recording their appointments. If recordings were not captured, this tended to be due to logistical or organisation issues. However, findings may be limited by the self-selected clinicians who agreed to be interviewed and to have their appointments audio-recorded. As these individuals may have been particularly interested in improving communication or supportive of the RCT, it might be that problematic presentations of equipoise are more pronounced in appointments led by clinicians who might not have been as engaged with the research. The audio-recording process itself may have altered clinicians’ practices, though, again, any Hawthorne effect is likely to have manifested as recruiters being more mindful of concealing their views. Some of the clinicians participating in the study contributed only one appointment for observation, which limited the scope to compare clinicians’ reported and actual practices. The limitations relating to the lack of information about clinicians/patients who did not volunteer/agree to participate, and the potential influence of the Hawthorne effect, are partly countered by the fact that we identified similar challenges across clinicians/RCTs. This lends some assurance to the integrity of the key findings presented. Finally, like any interview-based study, it is possible that participants’ accounts were influenced by their knowledge of the researchers’ interests (in this case, recruitment). Clinicians were nonetheless not aware of our specific interest in investigating equipoise.

Much of the literature around clinical equipoise has been theoretical, although there have been some empirical studies that have focused on clinicians’ reported difficulties with negotiating equipoise. These studies indicate that clinicians often assume that personal struggles with equipoise can be set aside during patient encounters and have no bearing on recruitment [4,18]. However, studies reporting patient perspectives indicate that their decisions around trial participation can be guided by their interpretations of what their clinician prefers or recommends [11–14]. There is also evidence to suggest that patients construct these interpretations based on perceived cues, rather than direct statements [14]. The present study bridges these bodies of evidence by providing insight into what actually happened during particular RCT recruitment appointments. This has enabled a detailed examination of how equipoise is put into practice, and identification of commonly recurring ways in which equipoise communication can be compromised.

Some of our findings are supported by previous research involving observations of recruitment appointments. These studies were largely nested in single RCTs and reported on trial-specific challenges clinicians encountered. For example, the potential for loaded terminology to disrupt equipoise has been reported in RCTs of laryngeal and prostate cancer [30,31]. The scope to look across RCTs helped us to distinguish particular aspects of discussions that are prone to loaded terminology or misrepresentation, such as portrayal of current practice and descriptions of the RCT question [30]. These areas could be amenable to generic training or guidance that encourages clinicians to carefully articulate the information specified in the trial protocol.

Though there are few multi-RCT studies of communication practices, one study by Jenkins and colleagues included a content analysis of how frequently particular trial concepts were mentioned in recruitment appointments [32]. Although “equipoise” was not specifically considered, the authors reported that “uncertainty” was mentioned in almost all of the 82 appointments observed. By contrast, we found equipoise discussion was omitted in 20% of the observed appointments. This discrepancy may be explained by differences in sample sizes and clinical contexts. Jenkins and colleagues observed fewer clinicians’ practices (n = 5) and focused solely on oncology trials. Oncologists are more accustomed to discussing issues of equipoise and uncertainty, compared to clinicians working in other specialties, such as surgery [7]. The greater diversity of clinical specialties and greater number of clinicians observed (n = 23) in our study may explain the comparatively higher proportion of recruitment appointments that omitted discussion of equipoise/uncertainty. Other studies have also shown that equipoise is often omitted from RCT recruitment discussions, though these studies also specifically measured the frequency of statements of “uncertainty” [33,34]. Reported presentations of equipoise in this study were based on clinicians’ own terms, rather than preconceived assumptions. This confirmed that statements of equipoise often extend beyond mere mention of uncertainty, as suggested in prior descriptions of how clinicians discuss equipoise in interviews [35].

Introducing the concept of equipoise to patients can be a challenge in its own right, as indicated by our findings that some clinicians were inclined to omit the concept from recruitment appointments. The substantial literature on clinicians’ management of medical uncertainty may be particularly helpful for illuminating explanations for the omission of equipoise. Uncertainty is a complex, multifaceted phenomenon that is pervasive throughout the medical literature. The definition of equipoise used in this study most closely aligns with the notion of scientific ambiguity—uncertainty in clinical recommendations on the basis of insufficient scientific evidence [36]. There has been remarkably little research that investigates how scientific ambiguity is operationalised in routine practice, although the small body of evidence that does exist has highlighted two broad strategies adopted by clinicians: creating an illusion of certainty [37–39] and transferring uncertainty to others (such as the patient) [37,38,40,41]. Our observations of clinician-initiated recommendations and tendencies to gear appointments towards patients choosing a treatment may therefore reflect strategies used to address uncertainty in routine practice, and underlie why recruitment needs to be recognised as a different activity [42]. There are opportunities for future research to compare recruitment appointments with routine consultations to explore issues arising from these different contexts.

The concept of equipoise is poorly demarcated in the literature, and, as such, the findings of this study need to be interpreted in light of the assumptions we have made and the stances we have taken. Equipoise can be considered in two contexts: first, as a phenomenon that justifies an RCT proposal and, second, as a rationale for randomisation presented during an RCT recruitment appointment. This study assumed that public funding of the RCTs indicated there was sufficient justification for the RCTs, although questions have also been raised around whether equipoise should take account of considerations such as social values and cost of treatment [43]. Definitions of equipoise in the context of the RCT recruitment appointment can also vary. Rather than an inductive grounded theory investigation [29], we specified a priori a broad and pragmatic definition of equipoise as a state where a patient would be neither advantaged nor disadvantaged if they were to receive any of the RCT treatments under investigation, based on existing evidence. We assumed that a patient who is clinically eligible for an approved RCT should be equipped with evidence-based information that ensures they are aware that the interventions under consideration would be safe, equivalent, and appropriate for them. This does not, however, account for external or patient-related non-clinical factors that can alter this balance. Future research needs to examine how patients’ values, beliefs, and preferences are managed in the RCT recruitment appointment, and the implications this has for recruitment outcomes. Focusing on recruiters’ explanations and information provision in isolation risks creating an oversimplified picture of what is likely to be a complex set of interactions. Our findings may also have oversimplified some issues through our pragmatic labelling of clinicians as more or less balanced. This labelling acted as a proxy for individual equipoise but does not take account of the continuum of views we identified. Our findings should thus be viewed as an initial step, in need of future refinement.

The findings of this study have practical implications. Identification of commonly recurring communication challenges provides opportunities for targeted strategies to improve practice. The practices that obfuscated equipoise communication varied in nature and would require different solutions. Treatment recommendations were the most extreme practices that contradicted equipoise. Recommendations were often rationalised in terms of patients’ characteristics, presented by clinicians who had alluded to being uncomfortable with particular aspects of trial eligibility criteria. Previous research has also shown that clinicians’ perceptions of equipoise can waver at the limits of eligibility criteria [4]. Like any aspect of clinical practice, patients’ suitability for trial interventions needs to take into account individual patient circumstances. However, difficulties arise when patients are steered against RCT participation on the basis of clinical or biological characteristics that are not stated as formal exclusion criteria. Filtering patients on these bases can have implications for the generalisability of RCT findings and interfere with eligible patients’ autonomy to make decisions about trial participation. One solution to untangling these difficulties might be to promote open discussion about discomfort or concerns about trial eligibility criteria during trial initiation meetings and site visits [4,42]. The QRI uses evidence from audio-recorded appointments and interviews to air these issues during investigator feedback and training sessions [22]. Recording detailed reasons for patients not entering a trial (e.g., in screening logs) will also lend some transparency to trial conduct. These reported details will need to go beyond ambiguous statements such as “clinician discretion”. Another alternative is for nurses or other professionals to undertake recruitment. While nurses do increasingly take on this responsibility, clinicians need to be accustomed to presenting equipoise and uncertainty comfortably: patients will still look to clinicians for advice, irrespective of whether clinicians are actively recruiting into an RCT. Furthermore, there is encouraging emerging evidence that it is possible for clinicians to learn recruitment skills [22,44,45].

This study found that clinicians often shared their opinions and predictions about RCT outcomes with patients. This raises questions around what should be expected of clinicians who hold the dual roles of physician and RCT recruiter. Avoiding personal views may be difficult in situations where patients request an opinion or recommendation. Trials that involve treatments that differ to usual care pose particular challenges, as clinicians need to have a degree of faith in “the new” to justify disrupting usual practice. Conversely, clinicians may also need to justify use of older, less innovative treatments in trials that evaluate novel interventions that have already diffused into routine practice without a supporting evidence base. Consistently communicating the concept of community equipoise throughout the appointment is one potential solution to responding to recommendation requests. The principles underpinning Chard and Lilford’s distinction between absolute equipoise and effective equipoise may also be useful here, in which equipoise around a single outcome (absolute equipoise) is distinguished from equipoise created by balancing multiple considerations (e.g., financial costs, side effects, effectiveness, quality of life) (effective equipoise) [5]. This alternative approach to viewing equipoise may provide a helpful framework in practice for balancing the advantages and disadvantages of trial treatments in efforts to sustain equipoise.

The common occurrence of particular equipoise-related difficulties across trials suggests that these challenges are amenable to generic training that could be incorporated into undergraduate and postgraduate medical curricula—particularly given that clinicians are increasingly expected to discuss issues of uncertainty and evidence. Recommendations for improving communication may also be adopted if clinicians are able to consider and apply these in the context of an RCT they are working within. A current programme of applied research has been developed to address recruitment difficulties by delivering tailored interventions as RCTs proceed [22]. Interventions applied within RCTs to date have included training around some of the issues reported in the present study [45]. Our findings also raise the possibility of targeting training to recruiters based on their individual perceptions of equipoise expressed at the trial’s outset (e.g., through interviews). Further research will need to investigate whether training can lead to sustained change in the practices of less balanced clinicians.

This study did not intend to make causal inferences between recruiters’ practices and trial participation outcomes, though there is potential for future research to examine possible associations. There is also potential to consider how categorisation of recruiters as more or less balanced is associated with recruitment outcomes, although this would run counter to our interpretation of individual equipoise falling on a continuum. Such efforts will need to consider the complex nature of patient decision-making, which can be influenced by an array of factors within and outside the recruitment appointment. Future research may also consider patients’ interpretations of appointment events—particularly their perspectives on equipoise and trial participation immediately following appointments, and levels of informed consent.

Supporting Information

Acknowledgments

Collaborating members of the RCT study groups are shown below.

ACST-2 study group: Alison Halliday (chief investigator), Jean-Pierre Becquemin, Anna Belli, Marc Bosiers, Piergiorgio Cao, Christina Davies, Jenny L. Donovan, Alastair Gray, Michael Gough, Elizabeth Hayter, Peter Leopold, Sumaira McDonald, Jonathan Michaels, Borislava Mihaylova, Richard Peto, Steven Robertson, Peter Rothwell, Rachael Scott, Dafydd Thomas, Frank Vermassen.

By-Band-Sleeve study group: Chris A. Rogers, Robert Andrews, Jane M. Blazeby (chief investigator), James Byrne, Jenny L. Donovan, Jamie Kelly, Graziella Mazza, David Mahon, Hamish Noble, Barnaby C. Reeves, Janice L. Thompson, Sarah Wordsworth, Richard Welbourn.

Chemorad study group: Jane M. Blazeby (chief investigator), Paul Barham, Sara T. Brookes, Tom Crosby, Jenny L. Donovan, Stephen J. Falk, S. Michael Griffin, William Hollingworth, Andrew D. Hollowood, Richard Krysztopik, Wyn Lewis, Jo Nicklin, Christopher Streets, Sean Strong, Dan Titcomb, Geraint Williams.

CSAW study group: David Beard (co-chief investigator), Andrew Carr (co-chief investigator), Jonathan Cook, Cushla Cooper, Benjamin Dean, Jenny L. Donovan, Alastair Gray, Stephen Gwilym, Andrew Judge, Naomi Merritt, Jane Moser, Jonathan Rees, Ines Rombach, Julian Savulescu, Irene Tracey, Karolina Wartolowska.

OPTIMA prelim study group: Rob Stein (chief investigator), John Bartlett, David Cameron, Amy Campbell, Peter Canney, Jenny L. Donovan, Janet Dunn, Helena Earl, Mary Falzon, Adele Francis, Peter Hall, Victoria Harmer, Helen Higgins, Louise Hiller, Luke Hughes-Davies, Claire Hulme, Iain Macpherson, Andreas Makris, Andrea Marshall, Christopher McCabe, Adrienne Morgan, Sarah Pinder, Christopher Poole, Elena Provenzano, Daniel Rea, Nigel Stallard.

POUT study group: Alison Birtle (chief investigator), Rik Bryan, James Catto, John Chester, Jenny L. Donovan, Ann French, Emma Hall, Chris Harris, Mark Johnson, Rob Jones, Francis Keeley, Tony Kirkbank, Roger Kockelbergh, Rebecca Lewis, Michelle Newton, Thomas Powles, Rachel Waters, Andrew Winterbottom.

Considerable thanks are given to the patient and clinician participants who agreed to be interviewed and audio-recorded in each RCT. We would also like to thank the research nurse teams who facilitated the recording and interview processes. The following individuals are acknowledged for their support of this work: Kerry Avery, Rebecca Barnes, Carmel Conefrey, Samantha Husbands, Nicola Mills, Katrina Turner, Ceri Rowlands, and Paul Whybrow.

The views expressed are those of the authors and not necessarily those of the Medical Research Council, Arthritis Research UK, Cancer Research UK, the National Health Service, the National Institute for Health Research, or the Department of Health.