Clinical trials: research questions and study design | ПРЕЦИЗИОННАЯ ОНКОЛОГИЯ

Clinical trials: research questions and study design

BRAF targets in melanoma. Biological mechanisms, resistance, and drug discovery. Cancer drug discovery and development. Volume 82. Ed. Ryan J. Sullivan. Springer (2015)

Clinical research specifically applies to research involving human participants. The goal of clinical research is often thought of as that of advancing the goals of medicine, either by development of new therapies or by deepening understanding of disease mechanisms [22]. An individual physician undoubtedly has a fiduciary duty to her patient, which entails upholding widely accepted principles such as beneficence, non-malfeasance, and autonomy [23]. Though the ultimate goal of clinical research is to move the field forward, these endeavors sometimes put the physicianresearcher at odds with this highly individualized therapeutic contract.

In this section we will review the ethical aspects of clinical research in therapeutic interventions, those that pertain to the conception, design, and implementation of clinical trials in oncology, with a focus on the issues relevant to randomized, controlled study design and special issues for early phase trials in oncology.

Advantages of randomized controlled trials (RCTs)

RCTs represent the cleanest study design, capable of minimizing the sources of systematic error due to confounding or bias which would otherwise lead to either a) erroneous rejection of the null hypothesis when in fact it is true (the false-positive case) or b) failure to reject the null hypothesis despite sufficient evidence to do so (the false-negative case). RCTs, as compared to other epidemiologic study designs (i.e. case-control studies, cohort studies, and non-randomized single-arm trials) inherently contain several features that help to minimize these potential errors.

First, the principle of random assignment helps to minimize or eliminate differences in the study arms that would otherwise be present due to selection bias or certain disease or clinical characteristics that would predispose a participant to be more or less likely to receive the treatment or control. That is, sicker patients may be more likely to be enrolled on a treatment arm if not randomly assigned [24]. Second, the use of concurrent control groups enables timely comparison of treatment and non-treatment groups as opposed to the confounding that arises when historical controls are used [25, 26]. Third, the use of power and sample size calculations to minimize type I and type II errors are valuable features of statistical analyses of RCTs [27]. Some RCTs also use blinding (of participants, investigators, or both) to further eliminate unconscious sources of bias [24].

Although it is hard to argue that statistical rigor and the goal of reducing systematic error inherently pose ethical issues, the practice of randomization, the use of control groups (particularly those employing placebo controls), and even blinding have been debated in the ethics community for many years. In this next section we will explore the ethical considerations surrounding the concept and design of randomized controlled trials, the role of equipoise in assessing appropriateness of an investigation, and the use of control groups in clinical trials.

Ethical considerations in the design of RCTs

Value (value to participants versus to society)

One of the primary requirements of ethical clinical research is arguably that it involves a test or treatment of inherent value by either (a) improving the health of trial participants and future patients or (b) contributing to the body of knowledge encompassing pathophysiology, diagnosis, or treatment of a particular disease [21, 28]. Emanuel et al. argue that research that is likely to produce nongeneralizable results, results that cannot be translated into clinical practice, or knowledge that cannot be disseminated to the scientific community would not meet the minimum ethical standard of value [21]. As we move forward into an increasingly strained research funding environment, identifying questions that provide high scientific or social value will become ever more important.

Scientific validity

Research lacking scientific rigor has a questionable ethical basis. Scientific validity is comprised of: a sound hypothesis, established methods, rational statistical analyses, adequately powered studies, and a feasible study plan [19, 21] The importance of scientific validity arises from the principles of judicious use of limited resources and minimizing the risk of exploitation of research subjects [29]. In the case of molecularly targeted therapies, one can argue that a study examining a novel agent targeting a known pathway or oncogene, reasonable efficacy in preclinical or animal studies, and one that has a measurable biomarker would constitute sound biological rationale. These requirements would ensure that participants are exposed to a favorable risk-benefit ratio, and that the results of such research will be useful to future patients.

Fair participant selection

There are numerous historical examples of exploitation of vulnerable populations. Out of these controversies, fair selection of participants has arisen as a key component of clinical research ethics. The major elements of fair selection are (a) a sound scientific basis upon which to define a study population, (b) appropriate exclusion of participants who would incur an unreasonably high risk of harm from study participation, (c) enrichment of a study population with individuals most affected by a particular condition, and (d) appropriate safeguards to ensure the protection of vulnerable populations [29].

Randomization and the concept of equipoise

The notion of uncertainty regarding the superiority of a novel treatment as compared to the standard of care (which may be either no treatment or a previously established and accepted therapy) is considered essential to the justification for an RCT. When inferiority in any treatment arm is identified, whether at the outset or in the midst of the trial, the investigation should be halted and access to the superior therapy for participants provided [30].

The definition of equipoise becomes more complicated when we consider whether equipoise exists on a personal or community level, also known as “theoretical” or “individual equipoise” and “clinical equipoise,” respectively. The former, first defined by Charles Fried, can be thought of in terms of the individual doctorpatient relationship, and implies that a physician may (only) recommend a clinical trial to a patient if the physician does not deem one treatment superior [31]. Arthur Schafer, however, suggests that even if a physician does not remain in a true state of equipoise throughout a study, a full explanation of risks and benefits to the participants allows them to make an informed choice, thereby preserving and ensuring equipoise [32]. This rationale neglects to account for the inherent trust that patients place in their physicians, and the potential for subtle coercion or even suggestion of a particular course to a patient who may be in a vulnerable state owing to advanced illness. Although medicine has been moving towards a more patient-centered and shared approach to decision-making [33], shifting the burden of determining equipoise to patients appears morally questionable.

These individualized and imperfect definitions of equipoise exclude the broader and perhaps more utilitarian implications of “clinical equipoise,” which Freedman argues exists when “there is no consensus within the expert clinical community about the comparative merits of the alternatives to be tested,” thereby removing the personal biases and preferences of individual physicians from the decision to move forward with investigation [30]. Freedman asserts that when true clinical equipoise exists among a medical community regarding a treatment or intervention, only then is a randomized controlled trial ethical.

In reality, it is likely that equipoise needs to exist on both a personal and community level. At the patient level, individual or personal equipoise would be requisite in order for a physician to recommend a trial to a particular patient, but clinical equipoise within the medical community may be necessary to justify initiation of a trial in the first place [34]. There are certainly situations where one of these may be met in absence of the other, or neither of the conditions is fulfilled.

Despite the popularity of clinical equipoise among the scientific community as the justification for conducting RCTs, there are several problems with relying solely on this concept. Miller and Joffe recently proposed that equipoise fails to offer resolution to five key problems in determining the appropriateness of RCTs [35]:

  1. The ill-defined nature of equipoise itself. Even with an expanded view of equipoise (that of “clinical equipoise” as proposed by Freedman), it remains a difficult entity to clearly define. For example, what level of agreement must there be in a medical community to infer true uncertainty? What is the quality of evidence on which “expert opinions” are based must be present in order to assume that the consensus is a valid one?
  2. Emphasis on expert opinion. Seeking the expertise of experienced members of a medical community does not necessarily guarantee a balanced, unbiased review of the therapies under investigation, nor does it ensure that the opinions themselves are based on prior evidence from other RCTs.
  3. Pitfalls of defining therapeutic efficacy based on surrogate outcomes. This is particularly relevant to the conception of oncology clinical trials, where equipoise is often considered lost due to encouraging results of phase I and II trials, which typically use response rates as a surrogate endpoint. Miller and Joffe argue that we must be careful not to conflate tumor response with improvements in overall survival or even quality of life [35, 36, 37, 38]. An example illustrating this is that of bevacizumab, a monoclonal antibody against vascular endothelial growth factor (VEGF), in metastatic breast cancer. While it initially garnered FDA approval on the basis of improvements in tumor response and progressionfree survival, a meta-analysis revealed that there was no overall survival benefit, prompting the FDA to remove breast cancer as an indication for this agent [39].
  4. High costs of novel therapies. Given the rising costs of cancer care, at least partially attributable to novel agents [40], there is an argument to be made for using an RCT even when true equipoise does not exist to be certain that a treatment provides sufficient benefit to justify its often high cost. This calls into question the physician’s ethical duty to the patient versus the medical community’s obligation to society at large to help curb healthcare costs.
  5. Early termination of RCTs based on interim analyses. The concept of equipoise mandates that a trial be brought to a close if the balance tips, even minutely, in favor of one treatment over another [30, 41]. This has the potential of overstating the benefits of a particular therapy, and underreporting of adverse events [35].

Control groups

Assignment of the appropriate concurrent control group is a central issue to the design of any clinical trial, especially that of an RCT. The basic purpose of any control group is to give the experimenter a basis for comparison and a way to measure efficacy of the treatment under investigation. Miller makes the distinction between absolute and relative efficacy, where the former is a pure measure of the effect of the intervention in a given clinical scenario, whereas the latter measures if and to what degree the treatment being studied is better (or equal/worse) to a standard therapy [42]. Similarly, the concepts of superiority and equivalence (non-inferiority) trials are pertinent to the discussion of control groups. Superiority trials are designed to demonstrate that the treatment under study results in a better clinical outcome than the control, whereas non-inferiority trials are designed to show that the new treatment is not worse than the standard.

Control groups are classified by the FDA into five major categories: placebo concurrent controls, active treatment concurrent controls, no treatment concurrent controls, dose-comparison concurrent controls, and external controls [43]. In this section we will discuss placebo and active treatment controls as the two most commonly encountered types of controls in oncology RCTs.

  1. Placebo concurrent controls. A placebo is an intervention (usually pharmacologic) which appears indistinguishable from the intervention under investigation, but is inactive and not expected to impact the natural course of the disease. In a RCT setting, it is the closest approximation of the “counterfactual,” or what would have happened to the intervention group in the absence of the intervention [24, 44]. Placebo controls used in a double-blinded, randomized setting are also the best measure of absolute efficacy of a particular intervention, and are used to demonstrate superiority [42]. In cases where no established therapy exists to treat the disease under investigation, it would be ethically sound to use placebo controls.

However, the controversy arises when there are known therapies for a disease state. Proponents of placebo controls assert that even if a previously established treatment is available, there may be a methodological rationale for the use of placebo controls; cases where an established therapy may behave different in a subset of patients, have previously demonstrated historically marginal benefits, or have a significant side effect profile [42]. Additionally, disease for which the natural history is indolent may be more amenable to a placebo-control design.

Miller proposes three areas that should be addressed in order to accurately assess the burden of a placebo control: severity of illness, likelihood of harm, and duration of harm. It is widely accepted that placebo-controlled trials are ethically suspect for severe, life-threatening conditions that will likely result in death or serious disability, or in cases where withholding any form of treatment has a significant chance of long-lasting harm to participants [25, 26]. Critics of placebo controls also argue that intolerable suffering, even if self-limited and non-life threatening ought not to be allowed to be studied in this manner. Emanuel and Miller refer to placebo-controlled trials of ondansetron in the 1990s, when other anti-emetics (such as metoclopramide) had already been proven superior to placebo for chemotherapy-induced nausea [45, 46]. Trials of this nature, where participants are denied symptomatic relief with an established and effective nature, are not ethical. Oversight by external and institutional review boards is essential to ensure the appropriateness of RCTs that use placebo controls, and rigorous review of informed consent procedures are particularly important in this scenario. Provisions should be in place to ensure that specific criteria are established to quickly remove participants from trials if they suffer significant harm or suffering, and placebo administration should be kept as limited as possible [45].

  1. Active treatment concurrent controls: These studies are designed to compare a new treatment to an established, typically standard treatment. Active treatment controls are a test of relative efficacy, and can be used in either superiority of noninferiority trials. A methodological weakness of active or positive controls is that although the new treatment may or may not be better than the standard of care, the difference between the two is likely of lesser magnitude than the difference between the new treatment and placebo. Power and effect size calculations usually dictate, then, that a larger sample size is needed to detect an effect size in the case of an active controls trial [42]. Proponents of active control trials assert that placebo-control design is inferior because the theoretical counterfactual in the latter case is not one of clinical interest. This is particularly true in oncology clinical trials for stage IV disease, where participants receiving standard therapy as an active control are more representative of a non-protocol patient, who would likely not be receiving “no treatment.”

[contact-form-7 id=»5168″ title=»Контактная форма 1″]


WordPress: 86.53MB | MySQL:1040 | 2,802sec