Planning for the Next Pandemic: Ethics and Innovation Today for Improved Clinical Trials Tomorrow

Abstract

The coronavirus pandemic has brought public attention to the steps required to produce valid scientific clinical research in drug development. Traditional ethical principles that guide clinical research remain the guiding compass for physicians, patients, public health officials, investigators, drug developers and the public. Accelerating the process of delivering safe and effective treatments and vaccines against COVID-19 is a moral imperative. The apparent clash between the regulated system of phased randomized clinical trials and urgent public health need requires leveraging innovation with ethical scientific rigor. We reflect on the Belmont principles of autonomy, beneficence and justice as the pandemic unfolds, and illustrate the role of innovative clinical trial designs in alleviating pandemic challenges. Our discussion highlights selected types of innovative trial design and correlates them with ethical parameters and public health benefits. Details are provided for platform trials and other innovative designs such as basket and umbrella trials, designs leveraging external data sources, multi-stage seamless trials, preplanned control arm data sharing between larger trials, and higher order systems of linked trials coordinated more broadly between individual trials and phases of development, recently introduced conceptually as “PIPELINEs.”

The coronavirus pandemic highlights perceived conflict between the imperative for scientific rigor in drug development and the acute urgency of a public health emergency (Angus 2020). Different stakeholders—the public, governments, the scientific community, patients, research subjects, and drug developers—bring unique perspectives and priorities. Against the backdrop of a global health emergency, it is important to remember the ethical principles that anchor clinical research for new therapeutics and vaccines. Innovative clinical trial designs may alleviate emerging challenges, and greater cooperation and data-sharing between clinical trials can bring significant benefits. As we strive to align our efforts, it is crucial that lessons learned from current COVID-19 challenges become the foundation for effective strategies in future pandemics.

The scientific path to novel, life-saving treatment lies in a rigorous system of phased clinical trials, but this is a painstaking process that stretches over months and years. In the presence of COVID-19 contagion, acute and long-term health implications, and unprecedented mortality rate, the randomized clinical trial (RCT) gold standard may not always seem to be the fastest vehicle to speedy anti-coronavirus vaccine and therapeutic investigations compared to nonrandomized trials with smaller sample sizes. These single arm trials may attract participants more rapidly due to the absence of a control arm, given the reluctance of many to be randomly assigned to a perceived “inferior” therapy or placebo in a control arm. The magnitude of the pandemic inspires compassion for the suffering and dying, and anxious desire to spring into action, but the answer is not to skip steps to work faster, it is to leverage innovation and work smarter.

Despite the urgency of the COVID-19 crisis, existing ethical parameters and best practices for valid science still hold, and with creative problem-solving, can address immediate dilemmas. The cornerstones of research ethics—autonomy, beneficence, and justice—speak directly to public engagement with treatment and prevention of the novel coronavirus. Before a patient decides on a medical treatment or a research participant joins a study, they need accurate, current, and understandable information about the intervention. The notion of autonomy is more than just “deciding for oneself,” it is having enough information to make an informed decision. This calls for transparency and support for health and science literacy. Beneficence refers to “doing good,” either for individual improvement or meaningful contribution to scientific knowledge. Finally, justice in healthcare refers to access and allocation of resources. The just involvement of humans in research not only includes equitable distribution of risk and benefit among study populations, but also using the best scientific data and methodology available. New strategies for leveraging data sources and innovative study designs exist, as discussed in the rest of the article.

The principles of autonomy, beneficence and justice are forward-looking and implicate each other in two important ways. First, the days of paternalistic and unequal partnerships between patient/physician and subject/investigator are over. In the past four decades, numerous disease-specific grass-root movements have engaged patient communities in comprehensive advocacy training, building a wealth of multi-media educational sources for treatment management and research engagement (Sigal, Stewart, and Merino 2020). Formalized systems that support patient-partners in the design of clinical trials, such as the FDA’s Patient Representative Program or the Patient-Centered Outcomes Research Institute (PCORI) have introduced an era that expects engaged participants not passive subjects (Bromley et al. 2015). The concept of “community-engaged research” is not a fad but has become a frequent requirement for receipt of private and federal funds.

As more prominently highlighted by the pandemic, the idea that research should benefit society, introduced in the Nuremburg Code (1949) and iterated in the Belmont Report (1981), implies that scientists have an obligation to be good stewards of public resources and conduct research that will bring substantial benefit (Resnik 2016). Today’s informed, self-advocating population of patients and research participants will demand that collaborative response to a pandemic leverages infrastructure already in place and innovative methods that have been proven to efficiently advance science. Innovative clinical trial designs and the infrastructure to support them, buttressed by long-standing ethical parameters, must be built now, during the COVID-19 pandemic, for better clinical trials tomorrow.

As of October 9, 2020, over 2250 COVID-19 studies have been initiated worldwide, for both therapies and vaccines, with planned enrollment of 500,000 subjects (Global Coronavirus COVID-19 Clinical Trial Tracker). Chloroquine and hydroxychloroquine are both FDA-approved to treat or prevent malaria. Hydroxychloroquine is also approved to treat autoimmune conditions such as chronic discoid lupus erythematosus, systemic lupus erythematosus in adults, and rheumatoid arthritis. The exploration of chloroquine/hydroxychloroquine, anti-malarial/anti-inflammatory agents that may reduce the cytokine storm associated with coronavirus acute respiratory distress syndrome, in a nonrandomized observational fashion illustrates the problems inherent in a hasty grab for a dubious cure.

On March 28, the FDA granted emergency use authorization to hydroxychloroquine and chloroquine to treat COVID-19 but revoked it on June 15 based on new scientific benefit/risk data suggesting that this regimen was unlikely an effective treatment. The public, healthcare providers and the scientific community were dismayed by lost time and resources. This incident raised the following issues: (i) Nonrandomized observational trials are less conclusive of safety and efficacy than RCTs, suffering from subject selection bias and unreliable effect estimates (Collins et al. 2020); (ii) Emergency use does not require data collection; (iii) Ineffective unproven therapies, if approved for use, may displace more effective therapies; (iv) Enthusiasm for chloroquine led to shortages for autoimmune disease treatment, where it is proven; and (v) Chloroquine is associated with life threatening cardiac arrhythmias, discovered to be a cause of death even in untreated COVID-19 patients. These problems reflect ethical dimensions related to compelling and valid science, resource allocation, dissemination of accurate information, cogent methodology in clinical research, avoidance of harm, and maximization of benefit.

Randomized studies can avoid some of the pitfalls of observational studies, but information must be gathered quickly and with the fewest subjects possible, especially those assigned to control therapy.

Decision analysis methods can assess the value of information acquired in earlier versus later stages of the pandemic, when early data has the potential to guide investigative pathways and impact greater numbers of future patients. The tradeoff between information quality and information timing can be explicitly modeled and optimized (Pearce et al. 2018). Technical discussion of the statistical features of adaptive designs best used to mitigate some of the risks of the pandemic related to, for example, Type I error rate control and treatment-effect heterogeneity, which might result in a large number of inconclusive or misleading trials, can be found elsewhere (Kunz et al. 2020). Stallard et al. (2020) have highlighted the benefits of adaptive design approaches available to advance clinical research within the challenging setting of a pandemic and provided a guide to the literature on statistical methods for such flexible designs with examples of COVID-19 trials.

We propose in this article that innovative trial designs, including but not limited to adaptive trials, can ethically address the tension between scientific data rigor and information timing within the pandemic context. Selected types of innovation in trial design and how such design efficiencies correlate with the ethical principles and public health benefits in the context of the pandemic are discussed for: (i) Platform trials, with multiple therapies tested, possibly at staggered times, and control data sharing (e.g., The Adaptive Platform Trials Coalition 2019; Dean et al. 2020) (ii) Designs leveraging alternative and/or external data sources to minimize control subject numbers; (iii) Multi-stage seamless trials, with adaptive selection of doses and acceleration between phases; and (iv) Higher order systems of linked trials where existing single-phase trial platforms and their components are assembled together into a larger design. These are called PIPELINES, Portfolio of Innovative Platform Engines, Longitudinal Investigations and Novel Effectiveness (Trusheim et al. 2016).

1. Platform Trials

In traditional RCTs, one primary hypothesis is tested and research participants are enrolled and randomly assigned to investigational treatment or the control/placebo group. In contrast, adaptive platform trials allow researchers to simultaneously test multiple interventions, either previously approved or experimental agents (Woodcock and LaVange 2017) against a single, possibly shared control arm, to add additional therapeutics while the platform trial is ongoing, and to “graduate” promising interventions faster or drop inefficacious treatments earlier (Antonijevic and Beckman 2018). From the subject’s perspective, beneficence is threefold: First, the efficiency of adaptive platform trials dramatically reduces the number of participants assigned to control/placebo. Second, graduating treatments and adding/dropping arms, as the accumulating evidence suggests, allows for a faster identification of effective therapeutics. Third, a platform trial with adaptive randomization permits ongoing calibration of the allocation probabilities to the different regimens on the basis of the accruing scientific data in favor of one or more treatment arms, allowing a higher chance that a patient will be enrolled on an effective therapy. We note that other adaptive designs, such as Multi-Arm, Multi-Stage (MAMS) (Sydes et al., 2009) designs that employ group sequential approach (Jennison and Turnbull 1999) to drop ineffective arms, can also increase the chances of trial participants being enrolled on effective therapy.

The platform approach with multiple adaptive clinical trials or subtrials, often included as intervention specific appendices to a single master protocol, has already led to a series of high-value, actionable research results: In April 2020, data from an NIH-sponsored Adaptive COVID-19 Treatment Trial (The Adaptive Platform Trials Coalition 2019), validated the clinical benefits of remdesivir, an antiviral agent, in patients hospitalized with COVID-19 (Beigel et al. 2020). A few weeks later, the similarly-designed Randomised Evaluation of COVID-19 Therapy (RECOVERY) trial in the United Kingdom demonstrated the benefits of dexamethasone, a widely used steroid (The RECOVERY Collaborative Group, 2020).

Another example is the REMAP-CAP trial (The Writing Committee for the REMAP-CAP Investigators 2020), a global platform trial adaptively testing many therapies in a factorial fashion with small number of subjects on standard of care (SOC), and adjusting randomization allocations frequently according to emerging data. If this approach is stretched to a larger collaborative scale which pools separate platform trials, then one can imagine a wider network of trials as a globe-spanning adaptive platform ecosystem more equipped to address the challenges of the next pandemic. Such an ecosystem would have to be built in advance and is discussed under the PIPELINES section below.

2. Designs Leveraging Alternative and/or External Data Sources

Data sharing is one of the biggest opportunities for improvement when one reviews the COVID-19 experience in anticipation of the next pandemic. There are already a large number of COVID-19 trials, and many are redundant or too small to be conclusive. We suggest a central funding mechanism for pilot studies that would encourage them to be reviewed and coordinated with the purpose of feeding results into larger platform trials. A major improvement could also be achieved by coordination in advance among the largest platform trials such as the World Health Organization’ Solidarity Trial (The SOLIDARITY trial; ISRCTN83971151), the National Institute for Allergy and Infectious Diseases ACTT-Trial, and the REMAP-CAP trial. Sharing of control data between such large trials, with each trial serving as an “external control” to augment the control arm of the other, could greatly reduce control arm size. Such efforts would require considerable advance planning, including the establishment of joint project governance, inter-operable computing systems, common data collection elements, and a consistent statistical approach towards adjusting for differences between trial designs. In addition, careful thought would have to be given to the rapid improvements in supportive care during a pandemic that could lead to relatively rapid improvements in control arm outcomes, potentially requiring data from early in one platform trial to be considered historical data when shared with another more recent trial, or even within the same trial. Methods for incorporation of historical control data exist (Ghadessi et al. 2020). A general methodology would have to be pre-specified and agreed between platform trial designers. Specifics of trial designs would of course have to be developed later, in response to the specifics of the new pandemic. Large platform trials with sharing of control data between them may be an efficient way to develop vaccines that typically require a very large number of subjects. It is worth considering legislation to compel sponsors to share control vaccine data in the event of a national pandemic emergency.

Sharing of control data between studies can accelerate development and reduce the number of patients on control therapy, but a roadmap of planning and methodological issues must be carefully considered (Ghadessi et al. 2020). In traditional RCTs, complete elimination of unknown confounders cannot be guaranteed, but randomization and blinding mechanisms reduce the chance of bias. A shared control arm or historical control data might lead to selection bias or a systematic difference among groups that could affect the final outcome. Such bias might originate from dissimilarity in a wide range of factors: participant demographics and baseline severity may vary; bias due to time trend shifts may occur when the time lag between external control data and the current study is large; and SOC treatment and concomitant medications may differ. These problems may be particularly challenging when faced with a pandemic where both supportive care and the SOC may be rapidly improving. Platform trials include methods to alter the SOC arm when a new therapy is confirmed safe and effective (Antonijevic and Beckman 2018). Trials with stringent inclusion and exclusion criteria, subtle selection bias, and different regional care standards might also adversely impact the study evaluation. Hence, when sharing control data or using a recent study’s control as a historical control for augmentation, the general mechanism for handling such data must be defined in advance. Important factors include: (a) inclusion/exclusion criteria, (b) type of study design, (c) well-known prognostic factors, (d) historical study quality, (e) treatment for the control group, and (f) regional effects. Ghadessi et al. (2020) also provided guidance regarding selecting historical controls for replacement or augmentation of a control arm for an existing clinical trial design, where the existing clinical trial cannot be redesigned to match closely with the selected historical control. The overall sample size of the innovative trial may be smaller than that of the traditional trial design, allowing for a faster completion of the study and accelerated path to scientific evidence (or lack thereof) in support of the development/use of new treatments.

3. Multi-Stage Seamless Designs and Sample Size Adjustments

Responding to urgent need, high-profile vaccine developers such as Moderna and Pfizer with BioNTech jumped rapidly into large scale Phase III trials after selecting one dose of the vaccine candidate based on limited short-term immune response and safety data from small Phase I trials. Skipping phase II dose ranging studies without robust evaluation of the optimal dose and/or schedule runs the risk of failing Phase III due to suboptimal dose selection. Alternatively, Seamless phase II/III designs can mitigate such risk without substantially increasing the development timeline by seamlessly combining Phase II (proof-of-concept and/or dose ranging) with Phase III (confirmatory) objectives into one study (Chaturvedi et al. 2014; Chen, Gesser, and Luxembourg 2015; Reeves et al. 2020). The seamless merging of distinct development phases and their objectives into one study for the same investigational agent is distinct from platform trials, which cycle multiple agents in and out of the given trial, which may or may not also be seamless. Seamless trials may be used for both novel agents and for repurposing of previously approved agents. Major advantages include removing the time gap between development phases and reducing total sample sizes by leveraging the totality of data across the study parts while maintaining adequate scientific rigor (Bretz et al. 2006).

Another adaptive strategy is sample size re-estimation during the course of the trial, that is, starting with a smaller trial and increasing the sample size only if needed. Many treatments do not show efficacy due to insufficient enrollment of research participants. This can happen in the absence of a robust body of evidence to allow reliable assumptions on the magnitude of treatment effect and/or its variability (Mayer et al. 2019), a likely scenario in a pandemic. Even with good planning, unexpected study results may occur. There may be an unexpectedly lower efficacy compared to previous trials, or the assumptions made for the sample size calculation prove to be inaccurate. An alternative approach to address uncertainty in expected treatment effect is Group Sequential Design (Jennison and Turnbull 1999). This design allows early stopping for efficacy or futility at a pre-specified interim analysis in case that observed efficacy is different than anticipated at the beginning of the trial. Uncertainty in treatment effect is a high risk in a fast-changing pandemic with a novel infectious agent, where even the microbiologic behavior and clinical manifestations of the disease are not yet well understood. Streamlining sample sizes also addresses the problem of oversized trials which are expensive, raise the cost of resulting treatments, require longer timelines to obtain conclusive results, and divert resources that could have been applied to other research.

Patients are desperate for better medicines; key advantages of multi-stage seamless adaptive designs are the ability to accelerate the access to a successful treatment and vaccine, while maintaining scientific rigor, due to combining development stages into a single clinical study (Stallard et al., 2020) and using resource allocation more efficiently because of the flexible sample sizes.

4. Organizing Clinical Trials Into Higher Order Linked Development Systems: PIPELINEs

In previous sections we described the importance of high order multi intervention specific appendices combined into a single master protocol, such as platform trials, for addressing some challenges of a pandemic. Due to the scope of the problem, differences in health systems and populations, and other factors, future pandemics may require an even higher level of organization than platform trials; a model described as a PIPELINE (Trusheim et al. 2016). PIPELINEs assemble existing single-phase trial platforms and their components into a larger system of linked trials. Linkages between trials may occur between phases or across related therapies, often with predefined decision rules. An example mentioned above would be linking numerous pilot studies into later stage confirmatory trials, so that selected therapies graduate from pilot studies into confirmatory trials. A PIPELINE may combine basket trials, umbrella trials, and platform trials with seamless graduation across the clinical evidence spectrum. Basket trials allow a single therapy to be studied in multiple indications at once, generally with a common predictive biomarker or pathophysiologic mechanism, whereas umbrella trials allow multiple drugs to be studied in parallel in one indication (Woodcock and LaVange 2017; Antonijevic and Beckman 2018). PIPELINEs may extend to post-approval data and real world data. Higher order trials, whether platform trials or PIPELINES, require a much more complex organizational structure. In this section we will describe key aspects of such structures. These organizational principles also apply to sharing of control data between platform trials, as discussed earlier. We note that the concept of PIPELINEs is very broad, and they can be assembled in many ways according to the characteristics of the problem. If a PIPELINE were to be deployed for a pandemic, much of it would need to be set up in advance, that is, a generalized PIPELINE for handling pandemics would have to be considered. Architects of the PIPELINE would have to decide what sort of trials would be part of the overall system, and how they would be linked in a PIPELINE fit for purpose.

4.1. Governance Issues: Balancing Central Coordination With Individual Flexibility

Fighting pandemics involves several arrays of stakeholders. At the geographical level, international, national and regional needs differ, as do public health systems and regulatory requirements. Governance structures must determine designs, select and enforce standards, define decision criteria, and construct networks. They must also implement infrastructure, from information systems to legal support. Key players must be represented at each level in this governance model, for example, patients, payors and sponsors (Antonijevic and Wang, 2018). Decision criteria may require different layers with proper communication. One layer will address needed improvements at the global project level. The other layer would represent individual and local stakeholder needs, allowing for local flexibility and creativity. Persinger (2014) described this problem and proposes specific strategies.

4.2. Importance of Decision Analysis to Guide Adaptations in Higher Order Systems

Due to the complexity of the pandemic and disparate needs of the stakeholders, good principles for decision making are paramount. Discussions with stakeholder representatives must take place from the beginning. Principles of sound decision-making in higher order systems have been described (Antonijevic et al. 2018) beginning with defining context and values. Once values are identified and the problem framed, development options need to be identified. Complex clinical trial simulations are important in this context and would support objectivity of decision-making rules in the presence of multiple stakeholders.

Another aspect of robust and adequate decision criteria is optimality. Platform trials and PIPELINES invoke decision making at the portfolio level, which is starkly different from the individual trial level (Beckman et al. 2011; Chen and Beckman 2009; and Patel et al. 2013). The optimal portfolio decision or design is often not optimal for individual comparisons. Sponsors want success of their products in comparison to the competition, individual countries will have different needs focused on public health. Clearly, two layered decision criteria, consistent with two-layered governance is necessary for higher order systems. Beckman et al. (2018) discussed decision making from the perspective of multiple stakeholders, while Antonijevic and Wang (2018) proposed specific methodology for satisfying multiple stakeholders’ requirements in contexts of individual trials vs platforms and portfolios.

4.3. Mechanism to Replace Control Group When SOC Care Evolves

Platform trials and PIPELINES involve a perpetual process, where some individual products or indications “graduate” and proceed into the later stages of development, and some therapeutics eventually get approved. Approved products may become the new SOC, requiring a change in the control arm. These changes should be anticipated, pre-planned, and addressed by the design. As discussed in the scenarios above, sophisticated layering of scientific process that relies on the evolution and evaluation of the most current data, fulfills the promise of “just” participation in scientific research based on the best methodology available.

5. Conclusions

Although COVID-19 introduces challenges for clinical research, medical experts, public health officials, and the general public, the ethical foundation of the relationship between these parties remains the same. The patient’s well-being is the primary goal of medical treatment, whereas the imperative of clinical research is to gather valid scientific data. These might appear as competing aims, but in truth, the first relies on the second. Best practices for ethical research can be clearly correlated to innovative and more efficient clinical trial designs that share all stakeholders’ ultimate goals in a pandemic: quickly find effective and safe therapeutics to treat the infection, minimize the casualties and offer protection to the world population against future pandemics via vaccines and successful treatments. The Belmont Report demands researchers ensure justice through careful selection of study aims and study design that will generate robust findings without unduly endangering research participants. That is exactly what a creative innovative trial design sets out to do.

As the next pandemic unfolds, innovative trial designs can promote effective and safe treatments while maintaining scientific rigor and honoring ethical principles for human research subjects. Prior to the next pandemic, we should lay the groundwork for more extensive control data sharing between adaptive platform trials, potentially even including higher order structures such as PIPELINEs. In the race to effective treatments for the next pandemic, multiple stakeholders may together benefit from the design of a common “ecosystem for major trials” instead of overlapping and competing clinical studies. If separate designs are desired by distinct organizations, a mechanism for sharing control arm data among the major large confirmatory trials could be designed upfront to bring speed and enhance the efficiency of the RCTs. Further, these larger trials could be also systematically linked to early pilots. The discussions on the design of the ecosystem(s) would include all stakeholders, such as patient communities, health authorities, payors, and the pharmaceutical and biotechnology industries. The result would offer a more streamlined approach without redundant underpowered trials, with a link between the learning and the confirmatory phase, and with more coordination and data sharing between major confirmatory trials. One could imagine the WHO, or a presidential commission on pandemic preparedness calling for engagement into such a large scale and complex system of clinical trials. Some details would have to be customized for each particular pandemic, but a general structure could be in place in advance. Leveraging innovative designs to reduce redundancy among large confirmatory trials and among control groups, to reduce the number of inconclusive studies and to get the most information from limited global resources, honors the spirit and principles for the ethical conduct of clinical research while advancing rigorous and rapid progress in discovering new treatments and vaccines against the next pandemic.