Real‐world evidence to support regulatory decision‐making for medicines: Considerations for external control arms

Abstract Randomized clinical trials (RCTs) are the gold standard in producing clinical evidence of efficacy and safety of medical interventions. More recently, a new paradigm is emerging—specifically within the context of preauthorization regulatory decision‐making—for some novel uses of real‐world evidence (RWE) from a variety of real‐world data (RWD) sources to answer certain clinical questions. Traditionally reserved for rare diseases and other special circumstances, external controls (eg, historical controls) are recognized as a possible type of control arm for single‐arm trials. However, creating and analyzing an external control arm using RWD can be challenging since design and analytics may not fully control for all systematic differences (biases). Nonetheless, certain biases can be attenuated using appropriate design and analytical approaches. The main objective of this paper is to improve the scientific rigor in the generation of external control arms using RWD. Here we (a) discuss the rationale and regulatory circumstances appropriate for external control arms, (b) define different types of external control arms, and (c) describe study design elements and approaches to mitigate certain biases in external control arms. This manuscript received endorsement from the International Society for Pharmacoepidemiology (ISPE).

based on pharmacoepidemiology methods and principles, are often considered as the foundation on the use of real-world evidence (RWE) for postauthorization regulatory decision-making by regulatory bodies around the globe. By definition, RWD are data relating to patient health status and/or the delivery of health care routinely collected from a variety of sources; and RWE is the clinical evidence about the usage and potential benefits or risks of a medical product derived from analysis of RWD. 8 More recently, a new paradigm is emerging-specifically within the context of preauthorization regulatory decision-making-for some novel uses of RWE from a variety of RWD sources to answer certain clinical questions. 9 For example, enacted into law by the United States Congress in December 2016, 10  issued the framework for its RWE program, 8 laying out the details of the multifaceted approach the agency is planning to undertake, which would include demonstration projects, stakeholder engagement, and guidance documents on specific topics to assist industry to develop RWE to support FDA regulatory decisions. Similarly, within the context of regulatory decision-making by other regulatory agencies such as European Medicines Agency (EMA) and Health Canada, the concept of utilizing of RWD to support safety signal evaluation and risk management is not new, but there is an ever-increasing interest in the use of RWE to support regulatory decisions across the product life cycle including development/preauthorization stage. 11,12 There are a growing number of European Union (EU)-funded initiatives linked to RWE. 13 In addition, published in 2015, EU Medicines Agencies Network Strategy to 2020 identifies RWE as a key player in supporting regulatory decisions for safe and effective use of medicines and bringing innovation to patients with unmet medical needs. 14 In May 2019, China Center for Drug Evaluation also published draft guidance on key considerations in using RWE to support drug development. 15 In approval of new molecular entities or label expansion, substantial evidence of efficacy of medical products is required from adequate and well-controlled studies by regulatory bodies. Usually reserved for certain special circumstances, external controls (eg, historical controls) derived from RWD is also recognized by some regulatory bodies as a possible type of control arm for single arm trials to satisfy the substantial evidence standard for product approval. 8 Similar to a randomized control arm in an RCT, external control arms represent a cohort of patients established to serve as controls to an intervention arm from a clinical trial. However, unlike in an RCT, these control patients are not randomized and are selected from data sources external to the singlearm trial. In rare diseases or disease areas with high unmet need wherein randomization could be unethical or infeasible, external control arms have a growing role in supporting regulatory decisions on data generated from single-arm nonrandomized trials. 16,17 In the absence of randomization, generation of an external control arm using RWD can be challenging and is subject to certain limitations, as it is difficult to fully control for confounding. Nonetheless, many biases can be attenuated using appropriate design and analytical approaches. 18 The main objective of this paper is to improve the scientific rigor in the generation of external control arms derived from RWD. Here we (a) discuss rationale and circumstances appropriate for external control arms to support regulatory decisionmaking, (b) define pros and cons of different types of external controls arms, and (c) discuss pharmacoepidemiologic design elements and approaches to mitigate bias that can be applied in the generation and analysis of external control arms. This manuscript received endorsement from the International Society for Pharmacoepidemiology (ISPE).

| Rationale and regulatory circumstances
In the assessment of efficacy and safety of an experimental medical product, the presence of a comparator (control) group is critical to understanding what happens to patients with similar characteristics and who would be subject to same conditions and procedures as in the experimental treatment group, but who do not receive the experimental treatment. 19 In RCTs, randomization is conducted to allocate treatments to trial participants based on the presumption that all measured (observed risk factors) and unmeasured (unobserved risk factors) confounders would be equally distributed among the treatment arms in a study, satisfying the independence assumption of treatment assignment and ensuring that each participant has the same

KEY POINTS
• There are clinical and regulatory circumstances where randomization is impractical or infeasible or unethical to conduct.
• Usually reserved for rare diseases and other special circumstances, external controls (eg, historical controls) are recognized as a possible type of control arm for singlearm trials.
• In the absence of randomization, supporting regulatory decisions with external controls requires careful, detailed, and transparent planning and adherence to pharmacoepidemiological principles to minimize bias and confounding, and produce credible, actionable, and reproducible evidence.
• The paper discusses the rationale and regulatory circumstances appropriate for external control arms and defines different types of external control arms.
• The paper discusses specific recommendations to gauge the adequacy of the design elements of RWE in regulatory use of external control arms.
probability of receiving the experimental treatment or active control (or placebo). 19 There are some clinical circumstances where randomization is impossible to undertake-due to ethical concerns and a state of clinical  9 In another example, RWE from a transplant registry was used to provide comparison data for similar patients enrolled in a single arm trial of Zalmoxis-a cell-based treatment for a rare disorderleading to its conditional authorization from the EMA. 9 In the presence of ethical concerns or feasibility issues associated with randomization, RWD can serve as a source for external controls for efficacy and/or safety endpoints to help interpret data from a single arm trial, support expedited approval, or help label expansion of an approved therapy to additional disease states or subtypes defined by biomarkers, or by other patient and clinical characteristics. There are also other opportunities for external control arms to support regulatory decision-making-even in RCTs. For example, external controls can help augment randomized control arms in RCTs by allowing smaller numbers of patients to be assigned (randomized) to control arms. 21 These so-called "hybrid" control arms (ie, mixture of randomized controls and external controls) can also help increase the efficiency of drug-development process and may potentially allow for more resources to be used in the assessment of evidentiary gaps that may not be often otherwise addressed in traditional RCTs, including long-term outcomes, additional endpoints that are more relevant to patients and payers, or endpoints that are augmented by patient or caregiver provided information.

| Types of external control arms
External control arms are also called "synthetic" control arms as they are not part of the original concurrent patient sample that would have been randomized into the experimental or the control treatment arms as in a traditional RCT. External controls can take many forms. For example, external control arms can be established using aggregated or pooled data from placebo/control arms in completed RCTs or using RWD and pharmacoepidemiological methods. Pooled data from historical RCTs can serve as external controls depending on the availability of selected "must have" data, similarity of patients, recency and relevancy of experimental treatments that were tested, availability and similarity of relevant endpoints (eg, operational definitions and assessments), and similarity of other important study procedures that were conducted in these historical trials. It is important to note that using control data from historical RCTs still results in a nonrandomized comparison but has the advantage of standardized data collection in a trial setting and patients who enroll in clinical trials may have more similar characteristics than those who do not.
Depending on the regulatory and clinical context, RWE generated from external controls can serve as real-world benchmarks (rather than as "formal" comparators) or as real-world comparators. 9 Realworld benchmark data are useful for contextualization and to characterize the natural history of a disease, including treatment patterns and outcomes, but are more suited to support regulatory decisions when the regulatory threshold for action in the face of uncertainty is lower (eg, severe unmet need, scarcity of available patients). Realworld "formal" comparators, on the other hand, require a more stringent planning and application of RWE that closely mirrors the patient population, inclusion and exclusion criteria, design, and analytical features of the single-arm trial.
Determining whether a pharmacoepidemiological study is fit for regulatory purpose requires several considerations. 22 The choice of data source, study design, and analytics should be tailored to the intended regulatory use of an external control arm (eg, new approval of a molecular entity, label expansion, and label revision); and should be based on the clinical context of that regulatory question (eg, prevalence/incidence of the disease, clinical equipoise, expected treatment effect, standard of care options, unmet need, benefit/risk, and uncertainty threshold considerations). [23][24][25] Considerations should include the assessment of data quality and relevancy to fit intended purpose, along with other methodological (design and analytical) approaches to minimize bias and produce actionable and credible evidence for the intended regulatory purpose.

| Pharmacoepidemiological design considerations
In most pharmacoepidemiological research using existing RWD, treatment assignment cannot be randomized; thus, the receipt of treatment may be dependent on multiple factors-including patient sociodemographic and clinical characteristics, insurance status, prescriber preference, and geographic or institution related variations in the practice of medicine. In the absence of randomization, the design of an external control arm with RWE should be constructed in view of the intended regulatory use, regulatory requirements, clinical context, timeline for evidence generation, and availability of appropriate and sufficiently high-quality data sources.

| Contemporaneous, historical, or hybrid cohorts
As summarized in Table 1  gender and, race/ethnicity) than the single-arm trial populations. 29 As usually the research question is to compare the outcomes between experimental arm and control arm, the external control arm may not need to represent the general patient population but T A B L E 1 Cohort and data collection options for an external control arm generated using real-world data (RWD)  While the use of advanced analytical and statistical approaches in generating RWE using external control arms is not within the main focus of this paper, it is important to note that further advanced analytical options may also be considered to mitigate confounding. 25,30,31 Examples include propensity score matching or inverse probability of treatment weights. 32,33 It is, however, likely that the number of covariates available for analytical adjustment may be limited by the number of covariates concurrently available both in the clinical (trial) data set and in RWD. Depending on the clinical context and endpoints to be assessed, the intensity and frequency of follow-up visits in real-world settings should also be evaluated carefully to ensure that data required for endpoints (eg, imaging data, laboratory test results) would be available and assessed in a similar fashion as in the single-arm trial to mitigate performance bias.

| Sources of RWD and target patient population
Endpoints used in clinical trials may not always be available or assessed in similar fashion in real-world setting. For example, in oncology, clinical parameters or imaging data needed to assess objective response rate or progression free survival may not be always available in RWD sources.
Therefore, to mitigate detection bias, developing and validating endpoints that can be used in real-world studies are critical. In addition, certain biases related to differential censoring or loss-to-follow-up (eg,

| CONCLUSIONS
The potential benefits of establishing external controls using RWE to support regulatory decision-making include providing evidence in circumstances when conducting traditional RCTs is unethical, impractical, or infeasible; supporting evidence development of marketed medical products for label expansions; increasing efficiencies of evidence development for regulatory purposes and expediting access of medical products to patients; and enabling supplemental evidence development that is more relevant to patients, providers, payers, and policy makers. The growing emphasis on the role of RWE in decision-making by regulatory agencies has fueled a new optimism to achieve these goals and provided an impetus for a new regulatory framework.
The acceptance of external control arms by regulatory agencies to support specific regulatory decisions may differ across therapeutic areas and clinical contexts. Several disease areas can uniquely benefit from the use of RWD in establishing external controls, given the randomization-based challenges associated with investigating rare diseases or diseases with high unmet need, particularly in view of the ever-increasing number of patient subpopulations defined by specific genetic mutations or biomarkers. 44,45 In the absence of randomization, supporting regulatory decisions with external controls requires careful, detailed, and transparent planning and adherence to pharmacoepidemiological principles to minimize bias and confounding, and produce credible, actionable, and reproducible evidence. In light of the changing regulatory landscape, continued efforts by stakeholders are needed to harmonize principles for regulatory use of RWD for external control arms. We start by offering specific recommendations to gauge the adequacy of the design and analysis of RWE in regulatory use of external control arms.

ETHICS STATEMENT
The authors state that no ethical approval was needed.

ACKNOWLEDGMENTS AND DISCLAIMERS
This manuscript received endorsement from the International Society for Pharmacoepidemiology (ISPE). The authors acknowledge Thomas Jemielita for assistance with review of the manuscript. No specific funding for this work was provided, and no specific product is involved. This article reflects the views and opinions of the authors and should not be construed to represent the views and opinions of the U.S. government or the U.S. Food and Drug Administration.