Abstract: Rare diseases present an enormous public health burden as well as substantial product development hurdles. In this talk, I will review some of FDA's programs to promote rare disease drug and biologics development. I'll provide an overview of the rare disease development pipeline and products approved by FDA's Center for Biologics Evaluation and Research, with an emphasis on cellular and gene therapies. I will discuss statistical challenges presented by applications in this space, as well as innovative statistical approaches that have facilitated approvals or late phase study designs, including novel graphical and descriptive approaches, and the role of the estimand framework in improving communication about study design. Finally, I will discuss the role Bayesian trial designs and analyses have played to date in drug and biologics regulation and opportunities for future use of these approaches.
Abstract: Achieving traditional power thresholds and conducting separate phase I through III trials are impractical when dealing with diseases that intrinsically affect a small number of individuals. For relatively stable rare diseases, a new approach known as an snSMART (small sample, sequential, multiple assignment, randomized trial) can address several issues in rare disease drug development. The snSMART allows for all participants to receive active treatment bolstering recruitment and retention and can test multiple doses and confirm the efficacy of the dose within one trial. Bayesian analyses offer an intuitive framework to share data across stages to boost the efficiency of treatment effect estimates and incorporate external control data. This talk will provide an overview of the opportunities of snSMART designs and analyses in rare disease drug development.
Abstract: Spinal muscular atrophy (SMA) and amyotrophic lateral sclerosis/motor neuron disease (ALS/MND) are rare progressive degenerative diseases that result from the loss of specific neurons that mediate motor function in the brain and spinal cord. Recently, after decades of failure, therapies have been approved that remarkably alter disease progression in both SMA and ALS/MND. Nusinersen (SPINRAZA™) and toferson (QALSODY™) have been approved because of 1) better understanding of disease biology, 2) novel therapies capable of modulating the core genetic cause of these diseases (e.g. antisense oligonucleotides) and 3) evolution in clinical trial design. Novel endpoints that incorporate change in functional status, mortality and patient drop-out in a single endpoint have been developed that accurately represent disease progression in a statistically validated way, now accepted by regulators globally. Biomarkers, such as neurofilament light chain (NF-L), which measure neuronal death, can be and have been used for ‘signal-seeking’, early dose optimization and expedited approval, all of which may serve to decrease the size, cost, time and complexity of clinical trials. Finally, the use of digital/wearable assessments-such as those derived from accelerometers placed around a subject’s limbs to assess activity-have advanced in neuromuscular disorders. Emerging ways to assess function from these sensors create the opportunity to detect changes in clinical disease at an earlier point and with fewer patients in a home environment. Further, accelerometer-based assessments can sensitively capture the regional heterogeneity of these disorders (i.e. one limb vs another) and are increasingly accepted as key (even primary) endpoints in pivotal clinical trials. However, the statistical methods to interpret these novel datasets are still emerging.
Abstract: Statisticians involved in the therapeutic development for rare diseases may encounter the setting where a placebo-controlled trial is infeasible. When this occurs, one option involves comparing outcomes in trial patients treated with the experimental therapy to historical placebo data. A key challenge in quantifying the safety and efficacy of the experimental therapy is the potential difference between the patients enrolled in the trial of the experimental therapy and the historical placebo-treated patients. Statistical methods for causal inference have been developed which may address these differences. This presentation will review a phase 3 single-arm study of ravulizumab in patients with neuromyelitis optica spectrum disorder and the causal inference statistical methods used to compare ravulizumab-treated patients with historical placebo data. Additionally, the option of using targeted maximum likelihood estimation, a doubly robust causal inference method which employs machine learning, will be discussed.
Abstract: It is important that endpoints in a clinical trial be selected with the trial’s goals in mind. In addition to complying with regulatory requirements, the chosen endpoints must have clinical significance in real world application. Multi-system diseases require treatment assessments that match the complexity of the condition; this often necessitates demonstrating a treatment effect on multiple disease components. The definition of endpoints is pivotal for multi-system diseases as multiplicity concerns may arise in diseases where treatment effects must be demonstrated across numerous disease domains. There exists limited literature comparing the statistical efficiency of different endpoint strategies across various disease areas; thus, we conduct a simulation evaluation to assess how individual domains and the relationship between multiple domains contribute to a composite treatment effect and how discordant within-patient responses impact this effect. Using the probability of study success, we evaluate which endpoint strategy provides the best statistical efficiency under different scenarios with varying responsive domains, effect sizes, and between-domain correlations. We apply our simulation framework to the complex, multisystem rheumatic disease systemic sclerosis (SSc) and evaluate several endpoint strategies including multiple testing, multi-component endpoints, composite endpoints, and win statistics.
Abstract: Overall response was previously used as an intermediate endpoint for accelerated approval in clinical trials for multiple myeloma. However, the now high rates of response observed for combination therapies in myeloma preclude the use of response as an intermediate endpoint. Minimal residual disease (MRD) measures a deeper response and could be used as an alternative intermediate endpoint for accelerated approval. This talk will review the data from various randomized trials to support the use of MRD as an intermediate endpoint, along with the insights gained from the recent Oncologic Drugs Advisory Committee (ODAC) meeting that voted unanimously to support MRD as an intermediate endpoint.
Abstract: With advances in science and technology, new biomarker intermediate endpoints have been proposed as surrogate endpoints for long term clinical benefit endpoints. For example, minimal residual disease (MRD) has been increasingly used as a prognostic biomarker, a measure of clinical efficacy, and a guide for treatment decisions in various hematologic malignancies. Employing MRD-based endpoint in registrational clinical trials has potential to expedite drug development with cost saving, lead to smaller trials with quicker readouts, allow rapid approval from regulatory agents and benefit the patients earlier.
On March 19, 2024, the FDA granted accelerated approval for ponatinib (Iclusig; Takeda Pharmaceuticals, Inc.) for the treatment of adult patients with newly diagnosed Philadelphia chromosome-positive acute lymphoblastic leukemia (Ph+ ALL) in combination with chemotherapy, supported by data from the PhALLCON study – the first, global, Phase 3, registrational, head-to-head clinical trial. This accelerated approval was the first FDA approval based on novel primary endpoint of MRD-negative complete remission (CR), which is a composite endpoint defined in alignment with the FDA that reflects deep molecular and clinical responses and is an important prognostic indicator for long-term outcomes for patients with Ph+ ALL. To qualify this novel endpoint, several statistical analyses have been conducted to evaluate the association between MRD-negative CR and event-free survival /overall survival by utilizing both trial data and individual patient data from external trials. These analyses supported MRD-negative CR as being reasonably likely to predict long-term survival outcomes in adult patients with Ph+ ALL.
Abstract: Rare disease drug development faces unique challenges due to limited patient populations and complex disease phenotypes. Effective endpoint selection is crucial for successful clinical trials. The challenges of endpoint selection in rare diseases include balancing scientific rigor with practical feasibility. We propose leveraging advanced data analytics, including using real-world data and AI-powered tools, to assist with addressing these hurdles. By identifying meaningful endpoints and insights that accurately capture disease progression and features (e.g. genomics and phenomics findings), we can optimize and efficiently operationalize successful rare disease clinical development processes. The presentation will discuss the key considerations and case studies in defining optimal endpoints and execution strategy for rare disease studies, emphasizing the power of data-driven approaches.
ABSTRACT: Although randomized controlled trials are the gold standard for assessing the effectiveness of new treatments, they are not always feasible, particularly in rare disease contexts where recruitment is often difficult. Leveraging historical trial data can enhance the efficiency of clinical studies. In recent years, the use of propensity-score (PS) methods have gained popularity in clinical trials to incorporate external data, aiming to balance covariate distributions between trial and external sources. Among PS-based approaches, regression adjustment for the PS remains unexplored when borrowing external control data for a concurrent control arm, largely due to concerns about the correct specification of the response model. This presentation introduces the sequential borrowing-by-parts power prior, a method designed to complement the regression adjustment approach by borrowing information based on the comparability of parameters in sequential steps. To determine how much information to borrow at each step, we sequentially apply the minimal plausibility index, an empirical-Bayes-like technique. A simulation study will be presented to evaluate the proposed method’s performance across various settings, demonstrating its ability to yield minimal bias while maintaining a favorable frequentist coverage probability.
ABSTRACT: Conducting randomized controlled trials for medications targeting rare diseases presents significant challenges, due to the scarcity of participants and ethical considerations. Under such circumstances, leveraging real-world data (RWD) to generate supporting evidence may be accepted by the regulatory agency. Constructing external control arm (ECA) from RWD for a single arm trial has been conducted occasionally. A complication in this design is that patients from RWD may be eligible at multiple time points. Most studies approach this by selecting one time point as the index date for ECA patients. Here, we propose a novel design for externally controlled trials that permits the inclusion of ECA patients at various entry points. Accompanying this design, we make recommendations of statistical methods to account for measured confounders, limited sample size, within-subject correlation, and potential overdispersion inherent in count data. Furthermore, we present an idea for the blinding process for this type of study. We have conducted a series of simulations to assess the performance of the design and statistical methods in terms of bias, type I error, and efficiency, as compared to the approach of selecting only one entry per ECA patient. The study and parameter setup were based on a hypothetical case inspired by a rare disease study. The results indicate that allowing multiple entries for ECA patients can lead to enhanced performance in many aspects. It provides a controlled type I error, robustness against certain model misspecifications, and a moderate power improvement compared with selecting a single entry per ECA patient.
ABSTRACT: In drug development for rare diseases, randomizing patients to placebo may be challenging. It could raise ethical concerns or make patients reluctant to participate. Therefore, the use of external control data becomes critical due to the limited patient population and lack of randomized control arm. However, effective utilization of this data requires addressing several limitations. In this discussion, we will explore common frameworks for performing external control analysis, highlighting key considerations and potential risks. Additionally, we will discuss some innovative ways to handle missing data and how they can reduce bias in our studies.
Abstract: Since it was proposed decades ago, seamless two-stage designs with treatment selection have been recognized to save sample size and time to enhance the development of effective treatments. These designs have been successfully implemented to accelerate the drug development process. When the primary endpoint takes a long time to mature, patients may have incomplete follow-up for the primary endpoint at the interim analysis for treatment/dose selection. For ethical and enrollment considerations, it is desirable to switch patients with partial follow-up to the selected dose after the interim. However, it has been perceived in the existing literature that treatment switching for seamless two-stage designs would undermine the validity of such methods and inflate the type I error rate. In this paper, we investigate such designs under treatment switching and show that the p-value combination methods can provide strong control of the type I error rates under realistic conditions. Simulation studies were performed to assess the validity and other operating characteristics of seamless two-stage designs subject to treatment switching under realistic conditions.
ABSTRACT: Giant Axonal Neuropathy (GAN) is an ultra-rare chronic neurodegenerative disease, with less than 100 individuals diagnosed in US. The disease typically presents around in the first several years of life with decline in motor function, and it is characterized by rapid progression of motor disability. An open label, escalating single administration study to assess the safety and efficacy of gene transfer using the vector scAAV9/JeT-GAN delivered through modified intrathecal lumbar delivery to the brain and spinal cord of patients with GAN was sponsored and conducted by NIH. Pfizer, as a collaborator, established the statistical framework for analysis of the trial efficacy data. We selected Bayesian approach as the primary for analysis of the data. Hierarchical models for repeated measures were used to estimate posterior distributions for the change in disease progression slope with respect to the efficacy endpoints at each dose level, as compared with the mean pretreatment slope across the dose groups. This study case provides a useful example of handling the analysis in ultra-rare disease setting, where information is sparse.
Abstract: Dose escalation strategies are primarily focused on identifying either the maximum tolerable dose (MTD) or the optimal biological dose (OBD) to recommend for further development in Phase II. The objectives of both strategies are primarily to limit the risk of trial participants experiencing an unnecessarily toxic dose. However, since most therapies can be re-administered to patients, limited attention has been paid to the issue of underdosing. The emergence of gene therapies, which seek to modify a patient’s DNA composition via virus (or equivalent) delivery systems, introduces significant risk to underdosed patients, as the immunogenicity-inducing nature of these therapies might bar them from receiving future therapeutics using the same technology. In this presentation, we seek to demonstrate how study designs can be used to address this issue, as well as compare between existing strategies.
Abstract: Gene therapies aim to address the root causes of diseases, particularly those stemming from rare genetic defects that can be life-threatening or severely debilitating. While there has been notable progress in the development of gene therapies in recent years, understanding their long-term effectiveness remains challenging due to a lack of data on long-term outcomes, especially during the early stages of their introduction to the market. To address the critical question of estimating long-term efficacy without waiting for the completion of lengthy clinical trials, we propose a novel Bayesian framework that selects relevant data from external sources to improve long-term efficacy estimates. In this presentation we will provide theoretical insights of the method, along with simulations and an application to clinical trials of a gene therapy.
Abstract: Chimeric antigen receptor (CAR) T-cell therapies has demonstrated efficacy and manageable safety in the treatment of hematologic malignancies. Real-world external control arm (ECA) studies have been used in supporting CAR T-cell therapy single-arm clinical trials for regulatory submission in pre-marketing setting. ECA representative of target indication and standard of care was created and compared with CAR T-cell treatment arm to generate evidence of comparative effectiveness of the treatment of interest versus conventional therapies in the real-world. The current approaches of using ECA with methods to optimize the utility of RWE and mitigate challenges and bias of using RWD will be discussed.
Abstract: To date, neurologists principally have two options to identify a safe and efficacious ALS treatment: (i) Randomized experiments designed to establish a causal relationship between a drug and a ALS-related outcome, which are often conducted in specific subpopulations, using non-human subjects, and/or in artificial laboratory environments. (ii) Uncontrolled, non-randomized observational studies (e.g., on humans) designed to estimate a conditional association between a drug and an ALS-related outcome in natural environments. Because it is often unethical and/or infeasible to conduct randomized experiments, many fields (e.g., environmental epidemiology, social science) have traditionally relied on the observational study approach. The aggressive monotonically debilitating nature of ALS, with the rich database of patients, both provide specific opportunities for innovation. The goal here is to embed an observational study into a hypothetical inexpensive RCT, sometimes referred to as studies of "digital twins". We will describe the operational strategy to assess causality from a non-randomized data set (Bind and Rubin, 2019, 2020, 2021), which involves (i) a conceptual stage, (ii) a design stage, (iii) a statistical analysis stage, (iv) a sensitivity analysis stage, and (v) a summary stage.
Abstract: Generalized pairwise comparisons (GPC) have been proposed to simultaneously analyze several outcomes of any type (discrete, continuous, possibly censored).1 GPC are especially useful when the outcomes can be prioritized from clinically most important to least important, and/or when clinical thresholds are deemed relevant for some of these outcomes (for instance, survival gains should exceed x months to be considered clinically worthwhile). In randomized clinical trials comparing some experimental treatment to control, GPC consist of comparing all possible pairs of patients, one taken from the experimental group and one taken from the control group. Each pair is classified as favorable, unfavorable or neutral for the highest priority outcome. Neutral pairs are then classified using the next outcome of lower priority, and the process is repeated until all outcomes have been analyzed. The Net Treatment Benefit (NTB) is defined as the difference between the proportions of favorable pairs and unfavorable pairs. GPC is particularly attractive for rare diseases because utilizing multiple outcomes can significantly increase statistical power compared to analyzing a single “primary” outcome.² Furthermore, the flexible prioritization of outcomes and the setting of clinical relevance thresholds make GPC well-suited for patient-centric analyses.²
Abstract: A platform trial is an innovative clinical trial design that uses a master protocol (i.e., one overarching protocol) to evaluate multiple treatments in an ongoing manner and can accelerate the evaluation of new treatments. However, the flexibility that marks the potential of platform trials also creates inferential challenges. Two key challenges are the precise definition of treatment effects and the robust and efficient inference on these effects. To address these challenges, we first define a clinically meaningful estimand that characterizes the treatment effect as a function of the expected outcomes under two given treatments among concurrently eligible patients. Then, we develop weighting and post-stratification methods for estimation of treatment effects with minimal assumptions. To fully leverage the efficiency potential of data from concurrently eligible patients, we also consider a model-assisted approach for baseline covariate adjustment to gain efficiency while maintaining robustness against model misspecification. We derive and compare asymptotic distributions of proposed estimators in theory and propose robust variance estimators. The proposed estimators are empirically evaluated in a simulation study and illustrated using the SIMPLIFY trial.