Bayesian Inference for a Principal Stratum Estimand to Assess the Treatment Effect in a Subgroup Characterized by Post-Randomization EventsBaldur Magnusson & Frank Bretz (Novartis)January 31The treatment effect in a specific subgroup is often of interest in randomized clinical trials. When the subgroup is characterized by the absence of certain post-randomization events, a naive analysis on the subset of patients without these events may be misleading. The principal stratification framework allows one to define an appropriate causal estimand in such settings. Statistical inference for the principal stratum estimand hinges on scientifically justified assumptions, which can be included with Bayesian methods through prior distributions. Our motivating example is a large randomized placebo-controlled trial of siponimod in patients with secondary progressive multiple sclerosis. The primary objective of this trial was to demonstrate the efficacy of siponimod relative to placebo in delaying disability progression for the whole study population. However, the treatment effect in the subgroup of patients who would not relapse during the trial is relevant from both a scientific and regulatory perspective. Assessing this subgroup treatment effect is challenging as there is strong evidence that siponimod reduces relapses. Aligned with the draft regulatory guidance ICH E9(R1), we describe in detail the scientific question of interest, the principal stratum estimand, the corresponding analysis method for binary endpoints and sensitivity analyses.
Wearable and Implantable Technology (WIT) with Biopharmaceutical ApplicationsCiprian Crainiceanu (Johns Hopkins University, Department of Biostatistics)February 20Wearable and Implantable Technology (WIT) is rapidly changing the data analytic landscape due to their reduced bias and measurement error as well as to the sheer size and complexity of the recorded signals. In this talk I will review some of the most used and useful sensors in the ever-expanding WIT analytic environment and their potential impact on Biopharmaceutical research. I will describe the use of accelerometers, heart and glucose monitors, as well as their combination with ecological momentary assessment (EMA) for improved patient reported outcomes. Several case studies highlighting the application of WIT in clinical trials will be provided. I will introduce an array of scientific problems that can be answered using WIT and describe methods designed to analyze the WIT data from the micro- (sub-second-level) to the macro-scale (minute-, hour- or day-level) data. Based on a better understanding of the WIT data, I will show how the design of experiments can be improved for specific Biopharmaceutical interventions.
Incorporating Innovative Techniques Toward Extrapolation and Efficient Pediatric Drug DevelopmentMargaret Gamalo (Eli Lilly & Company)March 28The conduct of pediatric clinical trials is legally required, monitored, and encouraged in major geographic areas such as the United States and Europe. However, because pediatric patients are considered vulnerable populations, they should only be enrolled as research subjects in a clinical trial if enrolling adult subjects will not be able to the answer the scientific question related to health and welfare of children. Thus, there is an ethical obligation to build the foundation for the use of pediatric extrapolation and related innovative analytical strategies with appropriately designed pediatric and adult clinical trials to reduce the amount of, or general need for, additional information needed from children to reach conclusions. This webinar will focus on extrapolation and innovative applications of clinical trial designs, analytic strategies to more efficiently leverage prior information, and modelling approaches that impact the data required to determine efficacy of an investigational drug in pediatrics. The planning of pediatric trials and regulatory interactions related to required pediatric studies and the expectations for innovative analytics are also discussed.
Basket Trials in OncologyMithat Gonen (Memorial Sloan Kettering Cancer Center, Biostatistics Service)April 26Cancer drug development has undergone fundamental changes over the past decade. Key features of this change have been emergence of targeted treatments and checkpoint inhibitors, utilization of larger expansion cohorts in Phase I trials and an increasing interest in developing combination regimens. Statistical design of early-phase trials have tracked these changes as well. In this webinar we will discuss single-arm Phase II trial designs for precision medicine, focusing on basket trials with a focus on novel methods and a balanced view of their advantages and disadvantages.
Platform Trials - Software and Operational SolutionsKyle Wathen (Janssen)May 16In recent years, platform clinical trials have gained substantial support. They provide an efficient way of testing multiple compounds in a single and consistent framework. In the traditional setting, each new compound requires independent proof-of -concept, phase 2A and 2B trials to be designed and conducted. Patients are often randomized between the new treatment and control, with the same control utilized in many studies. In contrast, platform clinical trials utilize a master protocol with each new compound added through an intervention specific appendix. This innovative approach leverages the common control, patient outcomes and recruitment centers to improve efficiency in terms of operations, analysis and decision making. In this presentation, I will begin with the key aspects of platform trial including terminology, protocol structure overview and discuss some of the key advantages. Next, I will discuss some of the intricate details and complications involved in simulating a platform trial with multiple compounds. In the last part of the presentation I will introduce an R package that is in development for simulating platform trials.
Generating and Harnessing RWE, HIT, and AI in the Era of Big DataKelly Zou, Jim Li & Nikuj Sethi (Pfizer, Upjohn Division)June 11Real world evidence (RWE) is defined by the 21st Century Cures Act as “data regarding the usage, or the potential benefits or risks, of a drug derived from sources other than randomized clinical trials.” RWE is derived from analysis of real world data (RWD), which can be claims and transactions for healthcare resource utilization, electronic health records, surveys, linked datasets and digital data collected outside a traditional randomized control trial. To harness real world evidence (RWE) across multiple health information technology (HIT) data sources and big data HIT repositories, artificial intelligence and digital innovations can be applied for meaningful insights. Examples on noncommunicable diseases (NCD) are used as illustration. To enhance the effectiveness and efficiency of health care delivery for NCDs, it is important to understand the risk factors for disease progression, treatment patterns, and healthcare utilization. Fruitful collaborative research opportunities exist across different healthcare stakeholders, including academia, industry and government, to leverage RWD for gaining valuable insights. Bioethical considerations are also presented.
Assessing When Electronic Health Records or Claims Databases Are Fit for a Specific Research Question or Regulatory PurposeCynthia Girman (CERobs Consulting, LLC)September 27Background: There is escalating interest in the use of real-world data (RWD), particularly insurance claims and electronic medical records, not only for safety but also for effectiveness evaluation. The Food & Drug Administration released a framework for public comment in December, 2018 which focused on randomized trials in clinical practice, potential observational designs, and data quality aspects for the use of real-world evidence (RWE) in regulatory and clinical contexts.Objective: To review the FDA framework on RWD and provide a structured study-specific context by which to assess the feasibility of using RWD for a specific research question or regulatory purpose (new indication or expanded labeling). Description: There is no ‘one size fits all’ approach to qualifying a data source for broad research or regulatory purposes because the feasibility of using RWD depends on the decision for which the results will be used, the anticipated effect size and the quality of the data critically needed to address the specific research question. Accuracy and reliability of data (including extent of missing data) to define specific elements of the research question, such as the population, intervention, comparator and outcome, fundamentally drive whether RWD in electronic health record or claims databases can be useful for a specific decision. Ascertainment and adequate capture of these elements along with the sample size and anticipated treatment effect size leads to a practical approach for assessing the feasibility of RWD in the context of the specific research question. Discussion will consider perspectives on the use of RWD for internal pharmaceutical company decision-making as well as the use of RWD for regulatory labeling or new indications.
Approaches for Estimating Treatment Effect in Principal Strata Where an Intercurrent Event Confounds the Measure of Primary Interest, with ExamplesBohdana Ratitch (Eli Lilly), Michael O'Kelly (IQVIA) & Ilya Lipkovich (Eli Lilly)October 16In a randomized experiment, post-randomisation events can obstruct the estimate of the effect of a treatment upon an outcome. The difficulties posed by a post-randomization event depend upon what is desired to be estimated (i.e. upon the estimand). For example, if it is desired to estimate efficacy in the absence of a rescue medication, then the event of rescue renders the outcome unascertainable post-rescue. Another experiment may aim to estimate quality of life – in that case the event of death can require special thinking as to how to take death into account. In these and other cases, the obstructing event delineates patient subsets, selected post-randomization, with distinct patterns of outcome of interest. To deal with the resulting selection bias when the event is affected by treatment, one route that acknowledges the role of the confounding event is to estimate the treatment effect within subject subsets referred to as principal strata (Frangakis and Rubin, 2002) formed in a way that removes confounding after conditioning on stratum membership. For example, one can estimate “the effect of treatment on quality of life at time K in a principal stratum of subjects who would survive to time K, no matter to which treatment they had been randomized”, with the corresponding estimand being clearly interpretable with a well-defined outcome of interest for all subjects within the stratum. The biggest challenge is how to identify the effects within the strata given that stratum membership is only partially observed: while we can observe whether the event happened to a subject on their assigned treatment, we cannot observe whether they would have had the event if randomized to the other treatment, which is required to define principal strata membership. This webinar covers a range of approaches for identifying and estimating treatment effects within principal strata that emerged in the last 10 years focusing on key assumptions behind each method that make identification possible. An approach based on multiple imputation that is gaining popularity is illustrated with a real-world example.
Leveraging Longitudinal Data in Drug Development via Mixed-Effects ModelingJosé Pinheiro (Janssen)November 22Longitudinal data are routinely collected in drug development, but often not fully utilized in primary and secondary analyses (which tend to rely on single time endpoints, such as change from baseline, response at Week X, etc.). While the focus on a single time point may be driven by regulatory requirements in a “confirmatory” setting, (parametric) longitudinal modeling utilizing the totality of observed data offers an opportunity to achieve substantial data analysis efficiencies at the “learn” stage of drug development. Some of the challenges in modeling longitudinal data are the correlation and non-constant variance often observed in this type of measurements. Mixed-effects models provide a powerful tool for analyzing longitudinal data, flexibly representing both their inherent correlation and non-constant variance patterns. This presentation will provide an introductory overview of linear and non-linear mixed effects models, illustrated with examples of longitudinal data analysis in drug development and other application areas (including R code to fit and evaluate the models).
Model-Based Meta-Analysis (MBMA): Harnessing Public External Data to Improve Internal Decision-MakingMatthew Zierhut (Janssen)December 12As landscapes for the treatment of many diseases have grown increasingly competitive and costs to conduct clinical trials have soared, pharmaceutical companies are striving to utilize all available information to make critical development decisions earlier and with more accuracy. Specifically, when developing a new compound, it is critical to understand the landscape of available (or potentially available) treatments, as well as the probability that a specific compound could compete successfully in this market. Model-based meta-analysis (MBMA) is an analytical tool that allows one to assimilate publicly available clinical trial results into a predictive model, which can be used to simulate various scenarios and to project the likelihood of achieving differentiable efficacy and/or safety. This webinar is intended as an introduction to MBMA and, using case studies, aims to demonstrate the value of MBMA through its application across various therapeutic areas.