Adaptive Designs for Clinical Trials in Nephrology

Introduction

CKD affects over 850 million people worldwide.¹ The condition incorporates multiple individual etiologies, complex disease trajectories, and several rare complications, such as calciphylaxis.² Barriers to conducting successful clinical trials in kidney diseases may variously include low-incidence conditions, slow disease progression, paucity of surrogate end points, poor tolerance of placebo, lack of evidence for existing practices, and uncertainty about potential treatment effect sizes. Adaptive trial methodologies and the use of perpetual trial platform infrastructure could address some of these and increase the likelihood of clinically meaningful outcomes in kidney disease research.

Adaptive Clinical Trials

Adaptation in clinical trials³ facilitates accommodation of the unknown in trial design with methodological rigor, often resulting in improved efficiency. Trial design involves decisions on the basis of assumptions about a range of factors, including those related to participant biology, risk of disease progression, the most promising intervention, and anticipated intervention effects. Fixed elements in trial design are established before trial commencement, whereas adaptive elements are modified on the basis of a prespecified approach to new information. High-quality adaptive trials are designed in light of multiple trial simulations that are used to define the trial operating characteristics that optimize rigor and trial efficiency.

Multiple elements of a trial can be adapted (Figure 1B).⁴ Adaptive elements are often incorporated into trials embedded in perpetual platform infrastructure to enhance trial efficiency.⁵ Adaptive trials feature preplanned modifications to various elements of the trial on the basis of interim data analysis. Platform trials establish trial infrastructure to be used for more than one trial question, and adaptive trials are often hosted in platform structures.

Figure 1.

Adaptive elements in the BEAT-Calci platform trial design. (A) BEAT-Calci trial design. (B) Adaptive trial elements and implementation in the BEAT-Calci trial. BEAT-Calci, Better Evidence and Translation for Calciphylaxis; HD, hemodialysis; MCO, medium cutoff; MgCit, magnesium citrate; STS, sodium thiosulphate; Vit K1, vitamin K1.

Key features for trial integrity and interpretability include transparent, prespecified descriptions of which trial elements will be modified and how, along with evaluations of the effect these adaptations may have on study validity. High-quality adaptive trials typically use considerable pretrial evaluation to determine the trial operating characteristics through simulations of trial design decisions and operations, such as participant selection and stratification, recruitment rate, organization of study arms, plausible treatment effects, treatment interactions, and effect of interim analyses and decision rules on statistical power and the ability of the trial to determine true effects.

Case Example: Better Evidence and Translation for Calciphylaxis Trial

Calciphylaxis (calcific uremic arteriolopathy) is a devastating rare disease affecting predominantly people suffering from kidney failure, with a reported incidence of 0.05%–4%.⁶ The disease is characterized by progressive, painful, cutaneous lesions, with reported 12-month mortality rates of 45%–80%.^7,8 For many years, only a small number of randomized controlled trials (RCTs) were initiated, largely testing varied surrogate end points over short durations. Recruitment challenges led to termination of several trials, although two trials were recently completed and presented in an abstract format: a phase 2 RCT of vitamin K1 (phytonadione) treatment (NCT02278692) involving 26 participants on hemodialysis and a phase 3 trial of SNF472 (hexasodium fytate) involving 71 participants (NCT04195906). With complex and poorly understood pathophysiology and no established disease-altering treatments, management approaches are arbitrary and inconsistent.⁶

The Better Evidence and Translation for Calciphylaxis (BEAT-Calci) multidomain, randomized, international, adaptive, platform trial (NCT05018221) illustrates how adaptive techniques can address some of the challenges of evidence generation in kidney disease conditions (Figure 1). The BEAT-Calci trial is designed to evaluate the effectiveness of a range of interventions for the treatment of calciphylaxis in patients with kidney failure receiving hemodialysis. Two domains are currently open, evaluating existing pharmacotherapies and dialysis membranes. The primary end point is an eight-point ordinal categorical scale that encompasses clinically meaningful outcomes of calciphylaxis, including wound healing and the avoidance of new lesions, amputation, or death at 12 weeks, with secondary evaluation at 26 weeks. Other outcomes include quality of life, pain score, infection, all-cause hospitalization, cost-effectiveness, and long-term outcomes.

Response-Adaptive Randomization

Accumulating data are used to change the intervention assignment ratio for new participants as the trial progresses.⁹ The current Pharmacotherapies Domain of the BEAT-Calci platform is currently evaluating three active agents against a shared placebo arm, reducing the overall sample size needed compared with three separate trials, each with their own placebo comparator. At predetermined intervals, interim analyses are conducted by an unblinded statistician to determine whether any agents have met a prespecified standard of proof for benefit or futility. After each interim analysis, the randomization ratio is altered from the initial 1:1:1:1 ratio to favor promising treatment arms in accordance with accumulated response. The main study team, sites, and participants are blinded to the altered randomization ratio. Pretrial simulation models inform the trial operating characteristics to ensure that the repeated evaluations do not increase the rate of chance findings. Agents meeting the standard of proof are removed from further testing, and beneficial agents become the standard of care in the trial. Meanwhile, the trial continues to enable further testing of other agents.

Sequential Multiple Assignments

Sequential multiple assignments or rescue randomization is used in the Current Pharmacotherapies domain, in a systematized reflection of common clinical practice. Participants who are not responding at 4 weeks are randomized to an additional active blinded treatment that they are not currently receiving. All participants whose conditions are not improving at 12 weeks receive all active treatments until week 26. The underpinning conceptual framework is that the 4-week placebo period will produce separation for effective treatments while maintaining reflections of clinical practice.

Addition of New Domains

The platform framework of the trial allows for trials in multiple domains of care to be assessed.¹⁰ Currently, the platform includes two randomized domains—the Pharmacotherapy and Dialysis Membrane domains—with participants able to participate in one or both. The platform can additionally accommodate new domains to address novel treatment options or research questions within the overarching trial infrastructure. The effect of incorporating new interventions on the existing trials is simulated before their commencement to determine the effect of commencing new trials on the efficiency of existing trials.

Statistical Considerations

The Pharmacotherapies domain was the first domain within the platform. Thousands of simulations encompassing multiple scenarios were run to define the trial operating characteristics. The design has a simulated type 1 error that is controlled at or below 2.5%. Further simulations were conducted to model the addition of the Dialysis Membrane domain to the existing domain to confirm it would not compromise the efficiency of the existing domain. Enrollment continues within the domains until a predefined success or futility threshold is met.

Participant Preferences for Active Treatments

For participants who value randomization to active treatments, the BEAT-Calci trial offers (1) response-adaptive randomization, increasing the probability of being randomized to a promising active intervention; (2) randomization to three active arms or a single control arm; and (3) certainty of receiving active treatment if not improving at week 4, all within the Pharmacotherapeutics domain. In addition, the multidomain structure means individuals can participate in two trials concurrently.

Despite the attractive features of adaptive designs, there are several challenges for consideration in adaptive trial design and conduct. Intensive statistical input is required for simulations and for the ongoing conduct of the trial. There are operational complexities with continuous upskilling of the operations team to ensure robust data monitoring, efficient adaptive response processes, and communication to participants and investigators. Platform trial infrastructure requires maintenance funding, which can be at odds with current research funding models. Funding bodies could consider flexible and sustainable operational support for adaptive platform trials in situations where there is a prima facie efficiency case.

Evidence generation in rare kidney diseases, such as calciphylaxis, can move beyond observational studies or small RCTs through global collaboration and innovative, adaptive trial methodology. Adaptive designs, along with thoughtful consideration of clinical care practices and patient priorities, may lead to better acceptance, delivery, and interpretation of trial outcomes.

Disclosures

Disclosure forms, as provided by each author, are available with the online version of the article at http://links.lww.com/JSN/E827.

Funding

Meg Jardine is supported by an Australian Government NHMRC Investigator Grant.

Author Contributions

Conceptualization: Meg J. Jardine.

Funding acquisition: Meg J. Jardine, Rathika Krishnasamy.

Investigation: Meg J. Jardine, Rathika Krishnasamy.

Methodology: Meg J. Jardine, Rathika Krishnasamy.

Writing – original draft: Meg J. Jardine, Rathika Krishnasamy.

Writing – review & editing: Meg J. Jardine, Rathika Krishnasamy.