Treatment preference and recruitment to pediatric RCTs: A systematic review

Background Recruitment to pediatric randomised controlled trials (RCTs) can be a challenge, with ethical issues surrounding assent and consent. Pediatric RCTs frequently recruit from a smaller pool of patients making adequate recruitment difficult. One factor which influences recruitment and retention in pediatric trials is patient and parent preferences for treatment. Purpose To systematically review pediatric RCTs reporting treatment preference. Methods Database searches included: MEDLINE, CINAHL, EMBASE, and COCHRANE. Qualitative or quantitative papers were eligible if they reported: pediatric population, (0–17 years) recruited to an RCT and reported treatment preference for all or some of the participants/parents in any clinical area. Data extraction included: Number of eligible participants consenting to randomisation arms, number of eligible patients not randomised because of treatment preference, and any further information reported on preferences (e.g., if parent preference was different from child). Results Fifty-two studies were included. The number of eligible families declining participation in an RCT because of preference for treatment varied widely (between 2 and 70%) in feasibility, conventional and preference trial designs. Some families consented to trial involvement despite having preferences for a specific treatment. Data relating to ‘participant flow and recruitment’ was not always reported consistently, therefore numbers who were lost to follow-up or withdrew due to preference could not be extracted. Conclusions Families often have treatment preferences which may affect trial recruitment. Whilst children appear to hold treatment preferences, this is rarely reported. Further investigation is needed to understand the reasons for preference and the impact preference has on RCT recruitment, retention and outcome.

If patients have a preference for treatment offered in an RCT they may decline randomisation to access treatment outside the trial. The external validity of an RCT may be compromised if patients with treatment preferences decline to participate, and bias is possible if uneven numbers of participants drop-out or cross-over between treatment arms [19,20]. Preferences can also affect adherence to treatment arms in RCTs where blinding to trial interventions is not possible [21,22]. Trials recruiting adult patients have reported treatment preference as a barrier to recruitment [23][24][25] evidence in relation to the ways in which preferences for trial interventions might affect recruitment and retention in pediatric trial settings [18].
Systematic reviews investigating the effects of treatment preference in RCTs have largely focused on trials recruiting adult patients [26,27]. A systematic review published in 2005 investigated the effects of participants' and professionals' preferences on recruitment, retention, and treatment outcomes. This review extracted data from 34 RCTs, but only four of the included trials had recruited pediatric-patients. Preferences were found not to significantly affect trial validity, but preferences did influence patients' willingness to participate [26]. The second systematic review published in 2008 focused on musculoskeletal trials, extracting data from 18 RCTs none of which recruited pediatric patients [27]. This review investigated the effect of preference on attrition and outcomes but did not investigate the effect of treatment preference on recruitment. It found that patient preferences for treatment were associated with treatment effects.
We cannot assume treatment preferences will have the same impact on recruitment to pediatric trials as has been shown in adult trials. Pediatric trials involve the combined preferences of parent(s), patient and recruiting clinicians, in addition to a more complex consent process [28,29]. There will also be variation in the extent to which young people participate in decision-making and the recruitment process, depending on the nature and severity of their illness [30][31][32][33][34][35]. The purpose of this systematic review was to identify pediatric RCTs where treatment preferences are reported, and describe the impact of preference on recruitment and retention.

Methods
A review protocol was developed and registered with PROSPERO: https://www.crd.york.ac.uk/PROSPERO/display_record.asp?ID= CRD42015015942. The review protocol also included methodology relating to the syntheses of qualitative data extracted from papers identified via this systematic literature search, which will be submitted for publication separately [36].

Study eligibility and inclusion criteria
Scoping exercises were used to define and refine relevant search terms using the PICOC model: Population, Intervention, Comparison, Outcomes and Context [37]. Qualitative sub-studies embedded in RCTs or quantitative primary and secondary outcome papers were eligible for inclusion if they reported RCTs recruiting new-borns, children and young people aged 0-17 years to an RCT, in any clinical area. Eligible papers were also required to report treatment preferences for all or some of the participants/parents. Database searches were limited to 1950-2014 inclusive.

Search strategy
A search strategy was developed with guidance from University of Bristol data specialists (NIHR/CLAHRC West and Cochrane Collaboration group), the search strategy can be found in Supplemental Information, Appendix A. Database searches of MEDLINE, CINAHL,  Authors were contacted to establish whether full RCT results had been published, two provided copies of their papers [38,39] and three confirmed that they had not [40][41][42][43].

Screening and data extraction
Each title and abstract was screened independently for inclusion by two researchers using the systematic review platform Covidence [44]. Discrepancies were documented, discussed and resolved in regular meetings by reviewers and a senior member of the study management team (EC) to ensure eligibility criteria were understood and screening queries resolved consistently. At the full text review stage papers were read in chronological order by two researchers (LB and AB, HK, RL or RP). Author(s) extracted relevant numeric data and/or descriptive reports of treatment preference into an Excel template (see Supplemental Information, Appendix B).

Impact of treatment preference -RCTs with non-randomised preference arms
Eight papers reported RCTs which used non-randomised 'preference arms' in addition to randomised treatment arms from the outset [51,53,61,63,72,73,84,102]. All of these trials reported the number of eligible families declining randomisation arms because of a preference for treatment, this ranged from 11 to 55%. One of these trials was extended by two years to increase recruitment to randomised trial arms [53]. Two additional trials introduced preference arms because families declined participation because of preferences for treatment [48,76].

Clinician preferences for trial treatments
Most studies did not comment on why families held a treatment preference, but six papers reported different forms of clinician preference for a particular treatment which may have influenced patient preference [28,41,63,77,84,85]. Two trials stated that staff experienced discomfort with children's medication/intervention being selected by a process of randomisation [28,84], one highlighted that 'consent was more likely when the recruiting physician was a member of the research team' [85] and in another, a parent whose child was randomised to a splint treatment arm was told the day after randomisation by a clinician outside the RCT that 'all buckle fractures need to be casted' [77]. Finally, one trial reported that parents who refused randomisation did so because of; 'a desire to have decisional control, and they trusted their physician's choice of treatment more than a computer's choice' [109]. These findings suggest that recruiters and treating clinicians may be an important influence on parent and patient treatment preferences when families consider RCT participation.

Discussion
To our knowledge, this is the first systematic review that has specifically investigated whether treatment preference influences recruitment into pediatric trials. The review has shown that families often have preferences for treatment at recruitment, and some families consent to trial involvement despite having preferences for a specific treatment. The number of eligible families declining participation in an RCT because of preference for treatment varied widely: From 2 to 70% in feasibility RCTs, from 2 to 50% in conventional main RCTs, and from 11 to 55% in trials with preference arms. Declining accrual rates and a loss of clinical equipoise led to the closure of two trials [48,74], and two required extensions because of slow recruitment [53,87].
Several trials included in this systematic review introduced preference arms to improve recruitment. Patient preference trials (PPTs) and comprehensive cohort designs [110,111], (in which participants with a preference are offered their treatment of choice, and those 3-4yrs Reported attitudes of parents whose child took part in a speech and language therapy RCT comparing immediate treatment and watchful waiting. Harth 1990 [83] Secondary Primary paper Van Asperen 1992 [101] 6mths-3yrs Double-blind, placebo-controlled trial of ketotifen, a new and unlicensed (for Australia) oral asthma drug. Hissink Muller 2011 [40] Secondary (poster presentation) Primary  without a preference have their treatment allocated randomly) offer the opportunity to investigate the effects of preference on recruitment, validity and treatment outcome [26,27,112]. Although this is one way of dealing with patients' preferences for treatment, this design has a number of disadvantages. PPTs often require larger numbers of patients. In extending trial duration to meet recruitment targets for the randomised arms, they may reduce external validity and generalisability of results. Also, such designs do not necessarily improve informed consent [53,110,111,[113][114][115].
A key strength of this review is that a large number of papers were screened for inclusion by two reviewers at all stages in the review process. This review was enriched by the inclusion of a wide range of papers, including data from papers reporting primary trial outcomes, and papers reporting qualitative findings on patient or parent experiences of trial involvement, and reasons for decline, consenting and recruitment. Limitations include the fact that seven papers reported findings from multiple trials in one paper [28][29][30]39,48,52,70], and many of the papers reporting qualitative findings did not include full CONSORT flow diagrams, therefore data on those who were lost to follow-up or withdrew due to preference could not be extracted. The effect that treatment preference has on retention in pediatric trials requires further investigation. If trial acronyms or references were provided in secondary papers, we carried out a search for each related primary RCT outcome paper, but only 18/28 additional papers were located. Data relating to 'participant flow and recruitment' was not always reported consistently in primary RCT outcome papers. One paper reported that 76 participants were allocated to treatment arms, but only 68 then entered the RCT, presumably eight withdrew postrandomisation but reasons for this were not provided [46]. A lack of standardised detail in the reporting of recruitment and retention methodology in RCTs has also been highlighted previously in a systematic review of behavioural interventions recruiting dyads (adult patients and their support person) [116].
Parental reasons for strongly held treatment preferences include concerns about side effects and attitudes towards new 'experimental' or 'placebo' interventions [55,117,118]. Although altruism is often cited as a reason for RCT participation, there is also poor parental understanding of the process of randomisation and perceived personal benefit for their child [14,119]. In pediatric trials, parents and children are often both involved in receiving information about the trial and making a decision about whether to take part, with support from a recruiting clinician [120,121]. Our findings showed that parents' preferences are reported more frequently than children's preferences. Only nine papers reported child preference, even though the majority of included trials were conducted with children and young people who were old enough to assent to RCT involvement and express their views on treatment.
Children's preferences for treatment differed from parental views on three occasions [28,42,53]. Older children and teenagers have reported different views from their parents on the acceptability of treatment and participation in asthma research protocols [122]. This is not consistent with guidance suggesting young people's voices need to be more widely heard [35,123], or approaches to communication which aim to support personal autonomy instead of isolated 'independence' of choice in decision-making [124,125].
Although this systematic review was not seeking to report clinician preference for treatment in pediatric RCTs, a small number of studies did report that members of the recruiting/treating teams held preferences. The impact of clinician preference has been described as affecting pediatric trials [26,126]. In one trial 63% of parents said the doctors recruiting them had influenced their decision to participate [55]. Clinician preference has also been shown to influence recruitment in adult trials [4,[127][128][129]. More research should be carried out to investigate the influence of recruiting professionals' preferences for treatment on the decision-making process of families.

Conclusions
This systematic review shows that treatment preference can be a barrier to recruitment to pediatric RCTs. In some cases this can result in the need to change the design of the trial (introduction of preference arms), extend recruitment or result in trial closure. Further investigation is needed to understand the impact treatment preference has on retention, and on the outcomes under investigation in pediatric trials. Exploration of the reasons for parent and child preferences would also be beneficial to ensure that families are fully informed when making decisions about RCT participation.

Conflicts of interest
The authors have no conflicts of interest relevant to this article to Primary 9-12mths Compared ventilation tubes (VT) and watchful waiting (WW) in the management of patients with otitis media with effusion. The generaliszability of randomised patients with eligible non-randomised patients was studied via preference arms. Weinstein 2013 [76] Primary Preference arms added after 3yrs of recruitment.

10-15yrs
The effectiveness of bracing, compared with observation in preventing progression of the curve to 50°or more in idiopathic scoliosis patients, with equivalent non-randomised preference arms. Van Wijk 2014 [66] Secondary Primary paper Van Wijk 2014 [102] 4.5-6.5mths Primary: Effectiveness of helmet therapy for positional skull deformation compared with the natural course of the condition Secondary: Assess parents' decision for helmet therapy in infants with skull deformation.
a Primary papers were defined as those reporting primary RCT outcome(s). Secondary papers were those reporting embedded/related studies (e.g. qualitative) describing patient/parent experience of trial involvement, reasons for decline, consenting and recruitment.   Barratt 2013 258 (33%) Not reported. 9 (26%) of non-responders reported concern with being in either the intervention or control group, but only 37/305 non-responders replied to question.

No
Bauchner 1996 648 (total eligible not reported) Not reported.

No
Forsander 1995 38 (93%) Not reported Immediately after randomisation 3 families in the control arm reported that they would have preferred the family therapeutic care arm.

n/a children under 3yrs
Hissink Muller 2011 Not reported Not reported. 41% participating parents reported a preference for therapy with methotrexate and etanercept and 6% had hoped against assignment to this group. Primary aversion was highest (25%) in the prednisone group [40]. Declined trial n = 38 (29%) [96].  disclose.

Financial disclosure
The authors have no financial relationships relevant to this article to disclose.

Funding source
This work was undertaken with the support of the Medical Research Council (MRC) ConDuCT-II Hub (Collaboration and innovation for Difficult and Complex randomised controlled Trials In Invasive procedures -MR/K025643/1).

Contributors' Statement Page
All authors approved the final manuscript as submitted and agree to be accountable for all aspects of the work.