Considerations for adaptive design in pediatric clinical trials: study protocol for a systematic review, mixed-methods study, and integrated knowledge translation plan

Evidence-based pharmacotherapy includes the conscientious use of the best available knowledge alongside patient values and clinical expertise to guide decision making at the bedside. According to a 2014 report by the Council of Canadian Academies, at least 50% of infants, children, and youth are prescribed at least one medicine each year [1]. Unfortunately, pediatric safety and effectiveness data are often not available for the medicines used, as clinical trials (CTs) have historically been conducted in adults. Without high-quality evidence from CTs, health care providers are forced to select and dose medicines in children based on evidence established in other populations (i.e., adults, older children, or other conditions for which the drug has been tested). Compared with adult populations, drug response in children is known to be more heterogeneous [2–5]. Children may respond differently to medicines resulting from non-linear growth in drug metabolism enzymes [2], and they often experience unique conditions (e.g. necrotizing enterocolitis, juvenile idiopathic arthritis, Kawasaki disease, Duchenne muscular dystrophy, leukemia) with unique disease etiologies compared with the adult population. The use of medicines without appropriate pediatric evidence increases variability in prescribing and can lead to over- or under-dosing; this places children at an increased risk for adverse events as well as efficacy or treatment failure. The lack of data on pediatric drug response means that approximately two-thirds of children in outpatient clinics receive an off-label (no authorized indication for this age group) medicine [6], and this increases to almost 90% in neonatal intensive care units [1, 7]. While extrapolation of effectiveness evidence obtained from adult populations may be acceptable under certain conditions, conducting more CTs with children would undoubtedly ensure that health care providers have better quality, age-appropriate evidence on the safety and effectiveness of drugs used to treat children.

Ethical and scientific reasons for the historical lack of CTs with children have been well described [1]. To overcome the challenge of having a relatively small population from which to recruit trial participants, combined with the limitations of traditional CTs, there is an urgent need for CT designs that present feasible alternatives while meeting ethical, regulatory, and methodological standards. Traditional CT design leads to protocols with rigid operating characteristics, such as total sample size, probabilities associated with fixed randomization ratios, number and composition of treatment groups, etc. In all prospective CTs, key study design elements such as the primary outcome, clinically meaningful treatment differences, and expected response variability are prespecified [8]. The success of CTs thus critically depends on the accuracy of key assumptions made prior to the recruitment of the first patient [9]. However, particularly in pediatric trials, there will often be a very limited pool of data upon which to base these essential assumptions and design decisions. Importantly, no use is made of the accumulating information through the trial, except for interim analyses to monitor safety, and effectiveness observations are analyzed only after all participants have been treated.

The principles of adaptive designs (ADs) present an alternative to traditional randomized CTs intended to increase ethical conduct and study efficiency. An adaptive design clinical trial (ADCT) is defined by the Food and Drug Administration (FDA) as “a study that includes a prospectively planned opportunity for modification of one or more specified aspects of the study design and hypotheses based on analysis of data (usually interim data) from subjects in the study” [10]. One of the defining features of an ADCT is that the possibility of adaptations is preplanned and prespecified [10], allowing for modifications to trial design characteristics in light of information accumulated during the CT. These adaptations can come in many forms, including (but not limited to) modifications to trial eligibility criteria, treatment regimens and dosing, study sample size re-estimation to enhance power based on accumulating information, the statistical analysis plan, the choice of the primary outcome, outcome adaptive randomization to assign more participants to the potentially superior treatment, or stopping of the trial if interim analysis reveals a low probability of detecting a difference between treatment and control (futility) [11, 12]. Multiple names are used for certain ADCTs, e.g., basket trials, platform trials, umbrella trials, drop the loser, play the winner, seamless phase II/III, and others [11]. ADCTs have been used at all stages of the clinical development process and can reflect either Bayesian or frequentist approaches to trial design or analysis. Adaptations are predominantly made based on data acquired from study participants, but they can also arise from external unexpected information (e.g., data from a separate pharmacokinetic study become available) [10].

Adaptive designs typically have lower expected sample sizes or lower realized sample sizes compared with fixed designs, for the same statistical power (e.g., by reducing sample sizes up to 60% compared to “fixed” sample sizes) [13]. However, their maximal sample size is typically larger or can yield improved information on how an intervention should be applied (e.g., to which subgroup, or at which dose) [9, 14]. ADCTs may also identify ineffective interventions earlier than traditional CTs, thus reducing overall subject burden (i.e., the number of children required to be involved in a study) and trial costs [8]. The judicious and widespread use of approved medicines for children requires evaluation beyond what CTs require for marketing authorization. Regulators in Europe have suggested that, particularly where extrapolation cannot be used with certainty, there may be a role for ADCTs early in the drug development pathway, but that discussions should be held with regulators early in the protocol development process [15]. A report published in 2006 indicated that regulatory experience with ADCTs was limited [16]. Furthermore, the opinions of health technology assessment bodies that determine which medications will be covered (or on formulary) for particular conditions based on evidence from ADCTs have not been evaluated.

The principles of AD were introduced in the 1950s [17] and have been applied previously in some pediatric CTs with some success. In fact, an analysis of 24 pediatric CTs conducted between 1963 and 2005 with sequential designs showed a significant reduction in sample size relative to what would have been calculated with a traditional fixed-sample design [13]. More recently, there has been discussion among European Union (EU) regulatory bodies with respect to the validity of extrapolation from adults to children [18]. Within this discussion, it has been recommended that during drug development pediatric investigation plans be adaptive and integrated into the development plans for adults [15].

The PREMILOC trial in particular provides a strong example of the potential benefits of AD [19]. The results of this study identified a significant increase in survival without bronchopulmonary dysplasia in preterm babies given low-dose hydrocortisone administered within the first 10 postnatal days compared with placebo. The researchers performed sequential interim analyses estimating parameters of efficacy and safety to avoid recruiting additional participants once either efficacy or futility had been established [19]. This adaptive approach allowed researchers to stop the trial as soon as a clear signal of efficacy had been demonstrated, eliminating the need to randomize potentially hundreds of babies to receive placebo had the trial continued until a traditionally determined sample size was obtained.

While ADCTs appear to be advantageous to child health researchers, the scientific (including statistical), regulatory, ethical, and logistical considerations for ADCTs in children have yet to be established. Currently, it is not known when, how, and what combinations of adaptations work best to answer which research questions. Previous work by Hatfield et al., who evaluated the characteristics of AD approaches used in registered CTs between 2000 and 2014, reported variability in the type and combinations of adaptations applied in CTs involving all populations [20]. However, this study did not extract data on age groups from included trials. CT registries, ethics boards, and regulators report seeing an increase in the use of AD [20], but barriers to implementing ADCTs exist. Kairalla et al. formally reviewed barriers to ADCTs in 2012 [8]. Barriers to the uptake of ADCT not specific to child health include a lack of knowledge and expertise among researchers, few case studies, and a lack of infrastructure support for the planning of ADCTs [8]. Moreover, statistical analysis methods for AD are unconventional and not widely understood, and although potentially desirable compared with traditional designs, AD itself is not free of ethical issues. These barriers create uncertainty and hesitation among researchers and funding agencies as to how study methods and findings will be perceived by the community. Concerns regarding the introduction of bias and the control of type I and type II error rates have also contributed to hesitation among research funders and peer reviewers [8]. Furthermore, these frequentist concerns may also outweigh a more evidence-driven, Bayesian calculus of sample size, resulting in unnecessarily large samples and increased recruitment efforts. As education and expertise are reported barriers, developing and disseminating evidence tailored to the needs of various stakeholders on ADCT considerations is essential to familiarize researchers, ethics boards, regulators, journal editors, and families on the appropriate use of AD in CTs globally.

Following principles of integrated knowledge translation (IKT) [21], we have partnered researchers with knowledge user organizations (e.g., various Canadian CT networks, including those with as well as without a specific pediatric focus, and with CT networks in the USA and Europe) for this study. Principles of IKT mean that researchers and knowledge users/partners are involved in all stages of the research to ensure scientific, practical, and policy relevance. In this study, these knowledge users, researchers, and partner organizations have been involved in defining the research plan and will aid in interpreting the findings and disseminating the deliverables. The objective of this study is to identify, evaluate, synthesize, and disseminate regulatory, ethical, logistical, and statistical facilitators and challenges of AD application in pediatric CTs. These data will inform the development of educational materials regarding if and how AD can be used to increase efficiency in answering clinical research questions. Moreover, in true IKT fashion, some of our collaborating partners are study authors.

This study will address previously identified barriers to the adoption of AD in child health CTs (incorrect use, lack of knowledge, few cases, lack of support, ethical challenges, regulatory evaluation, methodological and statistical issues) [8] through a systematic review and mixed-methods study. To increase appropriate use of AD in CTs, this study will include a comprehensive IKT component.

Within the realm of pediatric CTs, it is critical to facilitate efficient multi-center CTs in order for studies to be adequately powered to evaluate drug safety and effectiveness. The findings of this project will determine if and how AD can answer research questions more efficiently in pediatric populations to facilitate multi-centre trials and minimize the burden that such research places on patients and their families. The support of national training infrastructure and methodological expertise will be necessary to encourage researchers to apply AD. The commonly multi-centre nature of CTs in child health warrants agreement between regulators and research ethics boards of the suitability and challenges of innovative CT design. We recognize and acknowledge that agreement does not automatically imply that the correct conclusions have been reached, and we will use this possibility as an opportunity to create materials and interventions for education. Our project team will engage with regulators in Canada, the USA, and Europe to evaluate their perceptions and training needs for reviewing CTs using AD. If ADCTs can indeed be shown to make better use of scarce funding and CT resources, they will most certainly allow more research questions to be answered with more high-quality evidence. More relevant and reliable evidence will mean that health care providers will have access to the appropriate, population-specific evidence they require to ensure the safe and effective use of medicines in children.

Based on previous studies [8, 35] and discussions within the research community, we know there is a knowledge and training gap on the use of AD in CTs. Appropriate use and successful uptake of AD will require changes to perceptions of the clinical research community and updated policies — changes that must be based on evidence obtained systematically from a variety of stakeholders. Paramount is the need to identify when and how AD can increase CT efficiency. By creating new and by synthesizing existing knowledge, this study will lead to increased confidence and familiarity with AD in CTs for the research community and the public. This may, in turn, increase the likelihood that ADCTs will receive funding, since reviewers can then be directed to evidence on the strengths and weaknesses of ADCT. We aim to change the thought process and support how CTs are designed; this will have an impact extending far beyond child health research.

Potential limitations of the systematic review include the impact of publication bias. As positive and novel findings are more likely to be published, this may skew the analysis of included studies. While large-scale efforts to reduce publication bias (e.g., negative results journals, open reviewing) are beyond the scope of this project, we will endeavor to objectively critique the literature and report these limitations in resulting deliverables. Furthermore, planned studies that may have encountered the most significant barriers to implementing an AD design for a CT may not have begun at all (e.g., planned trials unable to overcome ethical or regulatory hurdles for the proposed adaptations); these would not be reflected in the published literature. Secondly, there are various views on the ethical considerations for CTs involving children [36, 37]. It is feasible that disagreements in ethical and scientific principles for involving children in research could be encountered during our focus groups. We will present all arguments provided in our deliverables to provide a clear picture of all opinions presented. Through involvement of a diverse group of stakeholders, including the public, we expect and encourage alternative views on the application of innovative trial methods to maximize the breadth and scope of our education materials. Our research will be useful only to the extent that the community can access it. Ensuring that our deliverables reach those planning CTs will also be a challenge. To this end, we have engaged a diverse group of national and global partners to help facilitate dissemination, and we plan to meet bimonthly with our partners to ensure that our deliverables meet their needs and the needs of their networks. Finally, it is inevitable that there will be some identified barriers (e.g., funding models that do not account for flexible sample sizes) for which we may be unable to propose solutions. These will be noted and will set the agenda for future projects.