Ethical challenges in clinical studies with adaptive design in oncology

Abstract

Novel immune therapies are increasingly based on the molecular differentiation of disease patterns. The related clinical studies are thus more often characterized by the so-called adaptive study designs (umbrella or basket studies including platform studies), which are continuously adjusted based on novel results. This paper analyses new study designs beyond the often-postulated need for regulation in order to identify ethical problems based on typical structural features and to—whenever possible—suggest solutions. To do so, it addresses the following topics: the relationship between social and scientific values of a study as well as aspects of the scientific validity of new forms of evidence; the inclusion of study subjects under the condition of relative uncertainty; specific challenges in the process of ethical approval, and ethical and practical challenges in the process of informing patients and receiving informed consent. Additionally, the topic of a potential risk–benefit assessment of such therapies is addressed.

Introduction

The testing and approval of new active substances for therapeutic use has evolved from a mere notification requirement for newly marketed drugs to regulatory approval after proof of efficacy, safety, and in some cases, ideally superiority of new compounds. ¹ Randomized controlled clinical trials (RCTs) with typical phases I to III are based on well-defined clinical pictures in terms of disease phenotypes. According to clear criteria, patients are enrolled in studies and randomly assigned to the respective study arms and control groups. The aim of such studies is to generate evidence on the safety and efficacy of a new therapy either in comparison to an existing (standard) therapy with or without placebo administration or a comparable therapy compared to a placebo. The latter applies only if a placebo administration in a diseased control group is ethically justifiable due to the severity of the disease with complete ignorance of the superiority of the substance to be tested (the null hypothesis).

Recent developments in biomedical research have led to a molecular differentiation of disease and patient groups as well as to therapeutic approaches that are no longer primarily individual disease patterns. ² Especially in the field of new immunotherapies, molecular and genetic properties of patients or disease processes—for example in tumors—are coming into focus as therapeutic starting points and as checkpoints for disease progress. The underlying biological structures and processes are typically found in disease phenotypes, in the case of oncology, for example, in different tumor entities. This calls into question the traditional inclusion criteria based on defined disease phenotypes. In addition, the molecular stratification of patient groups is achieved through the molecular specification of therapeutic strategies, so that large-scale studies with many participants and study arms will have to be combined to achieve the numbers of participants required for theoretically meaningful results and sufficient statistical power. The ethical challenges that arise from this are discussed below. ³ Six ethically relevant characteristics of new study designs will be considered: (1) the relationship between social and scientific values in such studies; (2) questions about the scientific validity, and thus normative reach of evidence; (3) questions of ethically balanced inclusion of study; (4) challenges for the ethical assessment of studies in the context of their approval; (5) specific challenges in patient information and informed consent; and (6) potential risk–benefit assessment.

Ethical challenges

Clarifying the balance of objectives, risks, and benefits in studies of new immunotherapies can be seen as a litmus test for a clinically successful, ethically justifiable, and socially acceptable implementation of new trial designs. How is this meant? It is in the nature of current immunotherapies to address either different molecular targets for the same disease or, conversely, the mode of action of a molecular target in different disease entities. In order to reflect this fact, new study designs have emerged in recent years that differ greatly from classical, randomized, controlled, and double-blinded trials (RCTs).

The investigation of the different, possibly also combined, molecular pathways a single (tumor) disease can be approached, is depicted in the so-called umbrella studies. The umbrella, which extends over the various approaches pursued in the study is the common disease entity. The study on how a single molecular mechanism can be used immunotherapeutically for different—albeit mostly related—disease entities, often takes place in the so-called basket studies.

The basket is created figuratively by the therapeutic strategy, in which different groups of patients with molecular subpopulations of tumors—for example, patients with colon carcinoma and patients with melanoma—are gathered together. ⁴ The advantage of this study design is that rare or very rare diseases are included under one roof, which separately are not meaningful due to their low statistical power. In basket studies, it is relatively common that different molecular groups of patients respond differently to therapeutic interventions, whereupon the study design will be adjusted to optimize the balance between risks and benefits, as well as to better achieve the defined clinical endpoints of the study. ⁴

These clinical endpoints, which ultimately form the criterion for the success of the intervention, are usually formed—even in conventional oncology trials—by surrogates, such as time to next relapse or survival time, and not by an overriding criterion of patient benefits, such as improvement in the ability to participate socially or in terms of non-standardized quality of life. While, on the one hand, adaptive designs can be considered advantageous because of their responsiveness, on the other hand, they create a situation in which the assessment of the efficiency of interventions, the relevance of negative results, and the validity of the positive results obtained by adapting the study and in the course of the study are difficult to assess from an ethical point of view. ⁵

Adaptive study designs in particular, with their prolonged adaptive duration, and with the elimination of unsuccessful approaches and the adjustment of endpoints, create a challenge that, as will be shown, can be met by new procedures of ethical evaluation and monitoring. ⁶

Social and scientific values of clinical studies

In the scientific as well as the ethical evaluation of adaptive study designs, when it comes to balancing social and scientific values, the criterion of equipoise is the point of reference. In order to better understand the ethical challenges at this point, it is worthwhile to take a look at both the concepts of equipoise and what is meant by social and scientific values.

In 1987, a publication by Benjamin Freedman introduced the concept of clinical equipoise. This was based primarily on divergent clinical assessments of the superiority of individual study arms, and thus fundamentally changed the previous view. ⁷ Previously the focus was on a theoretical concept of equipoise. ⁸ Theoretical equipoise means that clinicians and scientists who want to conduct a clinical trial are basically in the dark at the beginning: they should be fundamentally unclear at the outset as to which of the treatment alternatives tested in the different study arms will be superior to the others. ⁹ Only this way ethically justifiable randomization of patients into different arms of the study is possible, and randomization can take place without knowingly discriminating against individual patient groups. ¹⁰ Freedman further added the criterion of clinical equipoise to the theoretical equipoise, which ideally creates Arveil of ignorance among clinical and scientific experts. ⁷ For this to happen, it is necessary to establish honest, professional disagreement among clinical and scientific experts as to what relative medical benefits can be expected from the different treatments of the study arms. ¹¹ It is already clear here that, for ethical reasons, in order to avoid disadvantaging individual patient groups, a study will always reach an intermediate point when a treatment path proves to be superior and thus the criterion of clinical equipoise can no longer be maintained.

In principle, from an ethical point of view, it must be assumed that a clinical trial is only justified if findings can be obtained that contribute directly or indirectly to the development of new or superior therapies or to improve existing therapies with regard to their efficacy, safety, and tolerability. Secondary effects can be an improvement in effectiveness, for example by reducing intervention intervals and thus lowering patient burden, and in efficiency in terms of cost reduction. However, these two effect sizes alone are not sufficient to prove the social value of a study, because ultimately, both at the overall level and within the specific study, the overall well-being of the patients ¹¹ is always the linchpin of ethical justifiability.

Ideally, there is a balance between the social and scientific value of a study. This leads directly to considerations of how the risks and benefits of a study are to be evaluated. Study participants can derive a direct benefit, an indirect benefit related to a group of patients (group benefit), an external benefit (benefit for third parties), or a social benefit or no benefit from participation in a study. As a rule, the latter—no benefit from participation in a study—is justifiable only if there are no alternatives to the procedures tested in the study or if the studies involve healthy subjects. A negative benefit, that is, harm to patients through study-specific measures is only justifiable if the severity and expected frequency of the harm are reasonably low (e.g. risks of study-related venipuncture). The reasonableness—physical, psychological, and temporal time or by the risks of interventions and substances—is to be evaluated depending on the type and severity of the disease and the depth of intervention.

The corridor for optimizing the knowledge gain in a study is determined on the one hand by the risk-related ethical and social justifiability and on the other hand, by the achievability of the highest possible level of evidence. It is exactly this point where new adaptive study designs are striving for improvement. One reason for the criticism of classical studies is the potential that the study participants in one arm of the study may be treated sub-optimally despite the fact that clinical equipoise cannot be maintained. Adaptive study designs try to solve this problem by using preliminary study results for interim analyses, through which study protocol endpoints are reviewed and adjusted. ¹² The advantage for patients is that they are already being treated by optimized procedures at the time of inclusion in the study. However, an equity problem may arise from the fact that those study participants who were treated at the beginning of the study have a treatment disadvantage compared to patients who were included in the study at a late or very late stage, because at the beginning of the study no adaptive optimized procedures were available. The information sheet would communicate the most up-to-date gathered data to the participants, but in any case, the amount of data available varies in different stages of the study, making early participants’ information sheet lacking—by definition—the amount of information (ideally and possibly) gathered by later stages.

Evidence in adaptive study designs

A recurrent criticism of adaptive study designs is their openness to systematic bias due to continuous adjustment of interventions, intermediate goals, and endpoints. This is usually countered by specifying adaptation options as precisely as possible before the study begins. ¹³ Due to the complexity of the study design, it is not possible to detect statistical underestimation or overestimation of results. It is essential, both from a clinical-scientific and ethical point of view, to avoid the bias that can take place during observation, reproduction, and documentation of results, their analysis, and ultimately selective reporting.

A possible first source of error for the integrity and validity of adaptive studies is that complete blinding, that is, not knowing which patient will receive which treatment is difficult to maintain. This applies in particular if the study is adjusted based on interim results. It is ultimately always possible for experts to make informed guesses as to which study arm a new participant is currently assigned to, especially since some of the studies involve very small numbers of patients. In addition, since there are often no effective treatment alternatives, especially in the area of rare tumor diseases, an additional bias may arise from the fact that, in the clinical setting, those patients are preferably considered for study participation who can be assumed to benefit from participation in the study. Finally, all of this additionally ensures clinical trials with adaptive designs’—especially when it comes to (immune) therapies for rare tumor diseases—limited replicability and reproducibility, which may compromise generalizability. ¹⁴ Ultimately, the degree of systematic bias is inversely proportional to the degree of scientific validity of the evidence generated. This has, as is easily overlooked, not only an effect on the quality of the study results. Evidence also has (a) a justifying function in the weighing of benefits and risks and (b) a justifying function in the introduction, application, use, and reimbursement of innovative procedures in standard care.

The ethically balanced inclusion of study participants

An ethically significant advantage of adaptive study designs is the minimization of potential harm to study participants through a research strategy that consistently exposes fewer study participants to the stresses and risks of study participation. Through continuous optimization of the studies, more participants are treated under more favorable conditions, that is, with lower treatment risks and increased prospects of success. Against the background of the bias in the inclusion of patients who have a good chance of benefiting from participation in the study, it is central to create mechanisms to ensure a fair and ethically acceptable recruitment strategy. This in turn is a particular challenge for clinical researchers who are involved in the standard care of patients.

As a trial progresses with increasingly optimized treatment strategies, it may become increasingly, clinically and ethically, problematic to restrict the use of a superior therapy to pre-selected patient groups, especially when patients are enrolled late in a trial. If the trial shows—in different stages—better results from particular drugs, then we can conclude the drugs’ superiority. The use of the superior therapy to preselected patient groups “may” occur not because of an oversight but rather as part of the trial itself—when, for example, we consider the specific group of patients needed for trials in cases of rare diseases such as stage III non-small-cell lung cancer patients. In such cases, the participants need to be a pre-selected group, otherwise, the efficiency of the drug cannot be determined with valid levels of certainty. Particularly when patients who are enrolled late in a trial may benefit from results obtained with patients enrolled earlier. It may be argued that the progress that would otherwise be made step by step in successive studies is summarized here in a single study. However, it can be an actual challenge when participants in one and same study provide their consent to study participation under different conditions, especially with regard to risks and benefits.

This equity gap within a study currently poses a considerable ethical challenge, particularly in the context of clinical research in tumor therapies. On the one hand, patients—especially those who have not been treated or are refractory to therapy—are a particularly vulnerable group with a per se high willingness to give consent. Often a study is seen as the saving straw and is not infrequently presented as such. On the other hand, especially in the field of immunotherapy, adaptive designs can also be used to find the least effective dosages and combinations of immunotherapeutic agents and to eliminate them from the active study arms. Not all patients may be able or willing to accept this as the primary (extraneous) benefit of their participation in the study at a very early stage of the study. ¹⁴ Thus, the initial promise of adaptive study designs that participants would have to bear fewer burdens and risks under the sign of less uncertainty, is put into perspective. The complex design of studies in the field of immunotherapy increasingly requires that a larger number of patients remain in basket and/or umbrella studies for a longer period of time in order to obtain reliable sample sizes for interim analyses. ¹⁵

Challenges to the ethical assessment of studies

In the previous sections, we have already discussed core elements of adaptive trial designs in the field of immunotherapy—with a particular focus on tumor diseases—which are relevant for the ethical evaluation of these studies in the context of approval and the granting of ethical votes. The potential future changes in the context of a single clinical trial alone are currently difficult to map using established procedures of the responsible ethics committees. Thus, it can lead to the opening of new study arms, closure of existing study arms, change of inclusion and exclusion criteria for study participants, adjustment of the study population with regard to molecular subtypes of their disease, doses of test substances, endpoints and combination of endpoints, and changes in discontinuation criteria and randomization (including changes in study arms), ⁴ which may not be fully covered by a single vote of an ethics committee or by a series of amendments.

Through partial approvals up to defined interim analyses, adaptive designs would, in the view of the ethics committees, rather have the character of a sequence of partial studies. A succession of amendments would also be problematic, as in particular the requirements for patient recruitment and informed consent would be affected by changes in the risk–benefit profile and would have to be adapted to an extent that would go beyond a simple amendment. What at first appears to be a regulatory hurdle, surprisingly unfolds when one takes a look at the underlying text of—for example—the law of Clinical Trials Regulation (Regulation (EU) No. 536/2014) in this regard under §(48). ¹

Apart from the reporting of suspected unexpected serious adverse reactions, there may be other events that are relevant in terms of benefit–risk balance and which should be reported in a timely manner to the Member States concerned. It is important for subject safety that, in addition to serious adverse events and reactions, all unexpected events that might materially influence the benefit–risk assessment of the medicinal product or that would lead to changes in the administration of a medicinal product or in overall conduct of a clinical trial are notified to the Member States concerned. Examples of such unexpected events include an increase in the rate of occurrence of expected serious adverse reactions which may be clinically important, a significant hazard to the patient population, such as lack of efficacy of a medicinal product, or a major safety finding from a newly completed animal study (such as carcinogenicity).

The crucial point here is that the predictability of risks and benefits, which is not given ex ante by complex adaptive designs, would have to be repeated in the course of the study. From an ethical point of view, this means a combination of staggered approvals of studies based on interim individual arms with ongoing ethical monitoring of the study for the assessment of the benefits and risks of planned adaptations. The latter requires at least a prospective adjustment of the resources of the organized ethics committees. This applies both in terms of the number of processes to be assessed as well as in relation to the dynamics of clinical research. Therefore, an adaptation of the instrument of Data and Safety Monitoring Boards (DSMBs) to a “Data, Safety and Ethics Monitoring Board” could be a first step in the direction of scientific-ethical monitoring that is subject to reporting requirements. At present, it is being pointed out at conferences and congresses—for example, in the working group of medical ethics committees—that adaptive study designs could also be understood in the sense of clinical development programs.

Complex adaptive studies in the field of (oncological) immunotherapy, which also increasingly include drugs for novel therapies (advanced therapy medical products), such as cell, gene cell, and gene therapeutics, therefore have the advantage that adaptations in the design would not have to be represented in a sequence of amendments, but rather as an integral part of the development program. Thus, the regulatory timeline remains doable as long as no cell therapeutics or genetically modified organisms are used, which is increasingly likely to be the case. Moreover, the core of the actual ethical challenge lies elsewhere. ¹⁶

A positive ethical vote can only go so far as the weighing of risks and benefits for study participants, which must also be made against the backdrop of the social and scientific value of a study, and be based on a well-defined and concrete, that is, fully known procedure. Prospective ethical evaluations under the condition of possibility are normatively not sufficiently robust since they are regularly under the sign of uncertain prognosis or uncertainty per se.

Specific challenge for informed consent

It may at first seem surprising that an element of clinical studies, which is the point of entry into the research activity, is dealt with at the end of this article. The rationale is that the patient information and thus the basis for the patient's informed consent is, in principle, a condensate of the study protocol with all its intended social and scientific values, the procedures for establishing validity and evidence for the ethically balanced inclusion of study participants in the context of randomization ¹⁷ and finally the professional and ethical harmlessness. This means that all the aspects discussed above are prerequisites for dealing with the question of how innovative adaptive study designs on the inclusion of study participants through informed consent are relevant. The topic is to be treated pointedly, but by no means subordinated. The interests of the patients, the risks and harms they can expect in the context of a study, the potential direct benefit, the group benefit, or just the external benefits, as well as the requirement for complete and unreserved information, are central to the ethical and full disclosure of study designs. ¹⁸

Patients who agree to participate in the study are not necessarily able to foresee how the study arm in which they will later find themselves will develop. Thus, a specific consent to a well-defined intervention is not possible at all, but rather—comparable with the procedure for the transfer of data and biomaterial for biobanks—requires a very broadly defined informed consent, which may also extend to possible adaptations in the study protocol, and may be necessary at the time of the study—that are not yet known at the time of consent. Therefore, it must be ensured that potential study participants fully understand the differences between the very broad consent required of them under relative uncertainty and the specific consent required in the context of clinical trials with static inclusion and exclusion criteria and predefined interventions.

Study participants may be subjected to different treatments, modifications, or substitutions of test substances. This must also be presented fully and unreservedly in informed consent forms. Thus, potential study participants can weigh up their consent properly. Finally, it must be clearly communicated that those participants who are enrolled in the study at a later stage may expect a higher benefit.

Potential risk–benefit assessment

As was discussed earlier, due to the complexity of the study design, it is not possible to detect statistical underestimation or overestimation of results. However, it is possible to do a potential risk–benefit analysis for adaptive study designs by implementing the literature supporting the proposed innovative study. This method, originally introduced by CIRAO^®2 Quebec as a socioeconomic impact evaluation, maintains that by building on the actual data from similar cases, researchers can come up with a hypothetical case study, which can, in principle, determine the potential risks and benefits of the drug/treatment to be studied. CIRANO's method provides the potential direct (individual and health care) and indirect (social) benefits and risks. For example, direct benefits to the individuals could be savings in expenses, which would have otherwise been needed for medications or caretakers, or hospital stays, whereas the healthcare systems can potentially benefit from savings earned from patients staying shorter at hospitals. According to CIRANO, calculating such cost savings can be done by means of data gathered from hospitals, relevant literature, or government databases in the form of information on, for example, costs per patient per hospital admission, or the number of admissions.

The indirect benefits that range from reducing the general morbidity of disease to improvement of quality of life, similarly, can be calculated, for which the CIRANO team proposes the “willingness to pay-approach.” ³ As is mentioned, this method is suggested as a potential risk–benefit analysis and is mainly based on the existing data. In other words, using relevant literature on the analog drugs, substances, or procedures, the CIRANO team proposes a hypothetical risk–benefit analysis. This is not to say whether the participants are willing to pay for the drug on trial, but to come to a conclusion based on the willingness to pay approach for similar substances. Particularly in solidaritarian healthcare systems, the main direct effect for the individual could be taken into account in the form of quality of adjusted life years or in an assessment of the patients’ ability to lead their life which can be explained based on qualitative interviews and standardized tools.

Concluding remarks

Currently, ethical analyses of biomedical innovation often point to the need for regulation. In this way, evaluative approaches often become normative too quickly. With regard to innovative procedures in the field of clinical research and development, normative analysis must always weigh the conflicting interests and evidence. Despite the epistemological challenges that concepts of evidence pose, ethical approaches that are primarily oriented toward evidence-based solutions to systematic problems are more realizable. This is even more true since the evidence base is generally not as volatile as the situation of interest. The aim of this article was therefore to provide insights into the ethically relevant structural features of adaptively designed, complex studies in order to point out:

(A) conditions of the realization of the social and scientific values of clinical research in the light of new study designs,

(B) the ethical particularities of the validity of findings from adaptive study designs,

(C) the ethically relevant effects of the structural features of these study designs on the role of the study participants,

(D) the challenges for an ethical (and legal) evaluation of adaptive study designs, and

(E) the novelty of the conditions for informed consent of potential study participants.

Where the ethical tradeoffs are critical to whether and how clinical research and development can proceed, pragmatism is called for in addition to analysis. ¹⁹ The theory-saturated admiration of the problem with a simultaneous absence of proposed solutions can be a challenge in light of the opportunities, and new approaches for complex, adaptive studies in the field of immunotherapy. Therefore a conclusion for the practice at this point follows:

1. Opportunities and risks of new—especially adaptive—study designs shed new light on fundamental ethical questions, but do not raise fundamentally new ethical questions.

2. Besides a difficult balance between social and scientific values of adaptive study designs, these also stand out due to specific structural features that systematic bias have. Only an offensive and open approach and the implementation of tools for detection of bias, especially in the selection of patients and the study adaptation will help to prevent bias or at least make it transparent. The fact that the new study designs, due to the dynamic adjustments of intermediate and endpoints, are difficult to replicate, and thus the generalizability of the results suffers, may possibly be resolved in the study design by focusing on only essential adjustment, omitting possible but dispensable changes, and by sharpening and standardization of statistical tools.

3. A major problem is the fact that adjustments are not fully predictable. This leads, in addition to a reassessment of opportunities and risks on an ongoing basis, to specific challenges in recruiting and educating potential study participants.

4. The characteristics of adaptive study designs currently represent a challenge in the context of issuing ethics votes. Between a restrictive position, there seem to be two possibilities: the admissibility of studies as fragmentation of adaptive designs into single studies or the submission of a series of complex and comprehensive amendments, and the technical-pragmatic solution of treating such studies as clinical trials as part of a development program. New tools need to be developed for complex adaptive studies with a potentially long duration. In addition to interim partial approvals, the expansion of Drug and Safety Monitoring Boards (DSMBs) to become “Drug, Safety, and Ethics Monitoring Boards” should be considered.

5. The particular vulnerability of certain patient groups—such as patients with fully treated or refractory tumor diseases—requires a special effort to ensure that informed consent is given under a realistic assessment of the risks of participation in the study. In particular, in complex study design, new forms of informed consent, for example, in the sense of oriented to intermediate goals, will be required in the foreseeable future.