A note regarding the special issue on innovative design and analysis of complex clinical trials and opportunities for future research

The special issue on innovative design and analysis of complex clinical trials (Volume 30, Issue 6) is a compilation of invited papers from speakers at the April 2019 Duke-Industry Statistics Symposium on “Innovative Design and Analysis of Complex Clinical Trials for Drug and Device Developments.” This issue covers a wide variety of topics and is a recommended read for statisticians who are practitioners in the field of drug development. Some common themes in this issue include dose finding trials utilizing both efficacy and toxicity (adaptive) platform trials, and practical considerations for designing Bayesian sequential or adaptive trials. For statisticians working on immunotherapies or cell therapies, four papers on dose finding trials address some challenging issues germane to these areas. In one paper, the trial design accounts for late-onset toxicities in Maximum Tolerated Dose (MTD) determination (Andrillon et al. 2020). In recognition of the limited understanding of dose-response relationship, several papers proposed dose finding trials that consider both efficacy and safety outcomes for dose selection (Lin and Ji 2020; Yin and Yuan 2020; Li et al. 2020). Furthermore, emerging topics on platform trials are talked about in two papers. The paper by Bai et al. (2020) explores multiplicity issues for sharing a control arm while in Ivanova et al. (2020) discusses key elements of utilizing cross-over design under a master protocol. Last but not the least, adaptive designs were discussed in various settings, including paediatric studies (Psioda and Xue 2020), enrichment trials (Joshi et al. 2020; Wang et al. 2020) and sequential designs (Ivanova and Qaqish 2020; Wei et al. 2020), to name a few. The number of statistical issues that emerge in clinical trial designs and operations are innumerable. Yet, the wide variety of topics covered by this issue is not expected to span across all innovative designs of interest and its aspects.

One prominent topic not illustrated in this special issue is on incorporating real-world evidence (RWE) in the design and analysis of clinical trials. Trialists are interested to expand the RWE utilization throughout the clinical development program including in support of new drug approval. The explicit mention of the use of RWE in drug development in the 21^st Century Cures Act (Public Law 114–255, 2016) has increased interest in this area. However, it is to be noted that Sec. 3022 of the 21^st Century Cures Act (Public Law 114–255, 2016) only mentions the use of RWE in support of approval for new indication for already approved drugs and to help support or satisfy post-approval study requirements. Scientific publications, regulatory discussions (e.g. Food and Drug Administration (FDA) 2017, FDA 2019a and FDA 2019b) and cross-functional discussions are becoming increasingly abundant in this fast-developing area. It is imperative to reflect on the basic principles that affect the successes or failures of utilizing RWE and statisticians’ role in facilitating sound study design within this changing regulatory environment. Prominent among the list of issues is the ability to make causal estimates of effect and how data and associated biases make that challenging. Of note, the Journal of Biopharmaceutical Statistics (JBS) is still accepting manuscript submissions for a special issue on RWE and the submission instructions can be found on ASA connect website or ASA Biopharmaceutical section discussion forum.

This special issue does not cover discussions relevant to estimands within the context of innovative designs. Whether or not a design is complex and innovative, one important element is the trial objective and how this is translated into a question of interest. The ICH E9 (R1) addendum on estimands (European Medicines Agency (EMA) 2020) came into effect in the EU on July 30, 2020 and has been formally implemented by Health Canada since July 21, 2020. This addendum promotes clarity in defining the treatment effect of interest posed by the clinical question of interest (estimand) and in describing the method of estimation, so that the study objective and statistical analysis can be aligned. Envisioning intercurrent events and addressing these events and their potential impact on the estimand is a key aspect of the discussion in ICH E9 (R1). In the context of complex innovative trials, such as trial designs that require “borrowing” information from previously completed trial (see, for example, Psioda and Xue 2020), clearly identifying the clinical questions, the associated estimand and intercurrent events is challenging and requires more thought and research. The adoption of different analysis strategies mostly depends upon the therapeutic area, scientific question of interest and the complexity of the trial design. In the adoption of a Bayesian analytic method, for example, what is the implied estimand when one borrows information? Statisticians, as the subject matter experts in the study team, should proactively discuss potential intercurrent events and the associated relevant analysis to be employed to precisely answer the clinical question of interest. Furthermore, associated sensitivity analysis that explore the robustness of the main analysis should be considered. With the increasing attention on estimands from global regulatory agencies and the growing experience from practitioners, JBS is currently seeking manuscript submissions for a special issue on estimands, in the hope of facilitating the sharing of experience and complementing what is currently available from the literature. Details on this special issue can be found on ASA connect website or ASA Biopharmaceutical section discussion forum.

Advancing methodologies in risk-benefit assessment also merits further investigation. More and more, regulatory agencies are putting emphasis on structured risk-benefit assessments in reviewing submission dossiers (U.S. Department of Health and Human Services, Food and Drug Administration, Center for Drug Evaluation and Research (CDER), Center for Biologics Evaluation and Research (CBER) 2017) and in evaluating periodic safety updates (U.S. Department of Health and Human Services, Food and Drug Administration, Center for Drug Evaluation and Research (CDER), Center for Biologics Evaluation and Research (CBER) 2016). Benefit-risk (BR) assessment can be made at the patient level or by using summary statistics of benefit and risk. When the trial design is complex and innovative, assessment of benefit and risk within the trial may also be complex. Trials featuring borrowing of control information from previous trials may require assessing BR or elements of the BR across the different trials and a clear understanding of the safety and efficacy estimands that are components of the BR assessment. Mt-Isa et al. (2013) have reviewed 47 different benefit-risk approaches extracted from identified reviews and papers. As the number of therapies that are available to patients rises, the number of healthcare systems that require structured health technology assessments is also expected to grow. Hence, innovative benefit-risk methodologies and discussions that build on the work described in Mt-Isa et al. (2013) will likely be of interest to a broader group of statisticians working in this area.

Another area of importance to regulatory agencies is safety signal surveillance and detection. This could be initiated during the drug development phase where safety information is limited to clinical trials data and pre-clinical data. Safety signal surveillance and detection also occurs in the post-approval phase in post-approval commitment trials such as safety registries or from data that arises from patients’ prescription use where, as the number of drug users expands, the chance of observing rare drug-related events increases. Innovative designs such as hybridization of a randomized trial with a real-world registry in addition to innovative statistical methods that can be used to detect and visualize possible safety signals, especially rare events, in safety databases will provide solutions to this challenging area. Careful consideration of the clinical question of interest, the estimands, the adverse event data collection process and the intercurrent events and confounding factors in the design and analysis of safety registries are important. Statistical discussions on this matter can provide some needed recommendations. Furthermore, wider use of meta-analytic methods that are robust when the number of events is rare will also contribute to addressing an increasingly common issue in safety signal analysis (Hong et al. 2020).

Again, we congratulate the guest editors, Professors Anastasia Ivanova and Yuan Ji, and the organizers of the conference, Drs. Xiaofei Wang and Anna Maria Masci, on a special issue that promotes innovations in the design and analysis of complex trials. Drug development is a multiplex enterprise. Addressing some elements of trial design without addressing other elements may pose challenges to the adoption of complex innovative designs during clinical development discussions. It is hoped that the additional questions posed here will promote further research in these areas and elicit solutions that could help these designs be more broadly applicable. The journal’s upcoming special issues on estimands and on real world evidence are looking to provide more insight on these important areas of interest.