Reprioritizing Risk and Benefit: The Future of Study Design in Early-Phase Cancer Research

Abstract

The scientific purpose of phase I trials is to determine the maximum tolerated dose and/or optimal biological dose of experimental agents. Yet most participants in phase I oncology trials enroll hoping for direct medical benefit. The most common phase I trial designs use low starting doses and escalate cautiously in a “risk-escalation” model focused on minimizing risk for each participant. This approach ensures that a proportion of subjects will likely not receive any benefit, even if the intervention proves to be successful at appropriate doses. In this article, we propose that trial designs should employ dosing strategies that increase chances of providing benefit if the investigational agent should prove to be successful while limiting risk to reasonable levels. We then describe how adaptive trial designs can facilitate refined dose optimization based on both therapeutic benefit and toxicity, which can simultaneously decrease the risk of harm while increasing the chances of benefit.

Phase I oncology trials serve an essential role in drug development by testing novel compounds in humans. Their primary scientific purpose is to determine the maximum tolerated dose or optimal biological dose of novel agents.¹ These trials have traditionally focused on identifying dose-limiting toxicities, defined as acute adverse events that are severe in intensity and considered prohibitive for continued dosing. Phase I trials are also currently designed with the goal of minimizing the risk of toxicity for participants. A conservatively low dose of the experimental compound is given to those who are initially enrolled. This step is followed by algorithmic dose-escalation strategies for subsequent participants that aim to identify the dose level at which 33% of the sample size (usually six patients) experiences dose-limiting toxicity.² Current phase I oncology trials are designed with the assumption that higher doses result in more biological activity, making identification of the maximum tolerated dose—not objective responses or survival benefits—the primary objective of experimental inquiry.

For many drugs in development outside oncology, participants in phase I trials are paid, healthy volunteers. By contrast, phase I oncology trials have historically been viewed as too risky for healthy volunteers. Individuals eligible for oncology trials have advanced disease with limited or no evidence-based treatment options. These patients may also elect to forgo further antineoplastic treatment, instead focusing on comfort. Alternatives to trial participation include off-label use of medications effective against other cancers or “unorthodox” treatments without proven efficacy.

In this article, we will argue that current phase I oncology trial designs excessively prioritize risk minimization and, by doing so, disrespect the intentions of participants by making medical benefit extremely unlikely for those who are enrolled early in the trial and receive low doses of the drug intervention. We will further argue that, if slightly greater risks are accepted and adaptive trial designs are enacted, phase I participants can have an increased chance of therapeutic benefit at the same time that scientific integrity of the trial is preserved and the aim to limit risk of toxicities remains intact. First, we will present four prominent arguments that support the current risk-escalation model for oncology phase I trials. We will then highlight conceptual flaws in these arguments and propose counterarguments. Lastly, we will provide examples of more recent study designs that could fulfill the scientific purpose of phase I oncology trials while also respecting participants’ hopes for benefit. Although we focus on phase I oncology trials, similar principles could apply to research on other serious illnesses.

THE MERITS OF MINIMIZING RISK IN PHASE I STUDIES

Before discussing the appropriate level of risk, we will describe the chances of therapeutic benefit in phase I oncology trials. Current guidelines published by the National Cancer Institute for assessing treatment outcomes—Response Evaluation Criteria in Solid Tumors (RECIST)—define a partial response as “[a]t least a 30% decrease in the sum of diameters of target lesions, taking as reference the baseline sum diameters.” A complete response is defined as complete resolution of target lesions. Either of these outcomes would be an objective response.³ Phase I trials have historically resulted in objective response rates (ORR) of 4% to 6%, but more recent reviews have shown ORRs of 9% to 11%, with another 19% of subjects having stable disease for more than four months.⁴ A recent review of pediatric phase I trials found an ORR of 27.9% specific to hematologic malignancies.⁵ Additionally, biologically targeted agents tend to have fewer acute toxicities-adverse events compared to cytotoxic agents, with approximately 10% of subjects experiencing grade 3 or 4 toxicities, and a 0.4% to 0.5% toxic death rate.⁶ For some diseases, such as leukemia, chances of benefiting from a phase I trial have increased over time; for others, such as solid tumors, this progress has been slower. Despite these long odds, many participants in phase I oncology trials cite hope for medical benefit as their primary motivation for enrollment.⁷

Some have argued that an ORR is not an appropriate surrogate for direct medical benefit.⁸ Although not always predictive of survival benefit, ORRs are deemed by the U.S. Food and Drug Administration (FDA) to be the outcomes most attributable to the intervention under evaluation in a single-arm study. Measures such as survival, progression-free survival, and time to progression are not adequately characterized in single-arm studies and not recommended as end points for phase I trials.⁹ In accordance with the FDA Safety and Innovation Act, the FDA is permitted to use the durable objective response as one of the criterion for accelerated approval of new drugs, and there are recent examples of ORRs translating to long-term survival benefit in later trials.¹⁰ However, a recent review by Haslam et al. found that most surrogate markers in oncology studies (such as an ORR) had only low or moderate associations with survival, and they recommended caution in equating surrogate outcomes with actual survival benefit.¹¹ ORR is, then, not an ideal measure for patient benefit, but it is the most reliable scientific measure of the intervention effect in phase I trials.

Given this low chance of benefit and relatively high possibility of serious toxicities, Hey and Kimmelman highlighted three possible approaches to risk in early-phase study designs: maximax, “maximize gains by designing aggressive trials as they search for optimal components of a therapy”; maximin, “minimize harms from the worst possible outcomes by designing the least risky trials”; and some blend of these two approaches.¹² They subsequently refer to “risk escalation” as the model employing a maximin approach, and “innovative care” as the model employing a maximax approach.

Due to a long history of low response rates and entrenched ethical norms, phase I studies employ an iterative maximin strategy in which they prioritize minimization of risk at the expense of potential therapeutic benefit for many participants.¹³ Trials start with a conservatively low dose of the experimental compound and escalate slowly with subsequent cohorts. These subsequent cohorts of newly enrolled subjects will receive an identical or higher dose with criteria determined solely by the absence or occurrence of a dose-limiting toxicity within the first cohort of subjects. This risk-escalation approach results in substantial numbers of subjects receiving doses lower than what eventually becomes the recommended dose for subsequent phase II trials.¹⁴ For example, the development of imatinib for treatment of chronic myelogenous leukemia is widely considered a grand slam phase I trial because 98% of subjects who received a dose of 300 milligrams (mg) or greater achieved a complete response. However, initial dosing in this trial started at 25 mg, and 45% of subjects enrolled received a dose of less than 300 mg.¹⁵ Other studies suggest that 60% of subjects in phase I oncology trials have received doses lower than that recommended for phase II trials.¹⁶ In current phase I oncology trials, many subjects who enroll in hopes of medical benefit will receive a dose that likely has no higher chance of benefit than a placebo.

Four main arguments have been offered to support the risk-escalation strategy that prioritizes the minimization of risk in phase I trials. We will summarize each argument and then address conceptual problems. First, given the low likelihood of benefit, some argue that researchers must minimize risk of harm as much as possible; otherwise the potential harms would be disproportionate to the possible benefits. Increasing the threshold for acceptable risk could lead to increased rates of toxicity, and this increased toxicity might out-pace the increase in therapeutic benefit. Given this imbalance of risk and benefit, they argue, phase I oncology trials should continue to prioritize safety by following the risk-escalation model. In support of this argument, Koyfman et al. found in 2007 that aggressive dose-escalation study designs in phase I oncology trials were associated with increased rates of toxicity yet similar response rates.¹⁷ Notably, this study was restricted to trials of cytotoxic agents, and it is now over a decade old and does not consider current advances in targeted therapies. However, the results still merit consideration.

Phase I designs that accept more risk of toxicity to improve chances of benefit could also reinforce therapeutic misconceptions, both for patients and physicians. Green, for example, studied protocol exception requests made to institutional review boards (IRBs)—typically, requests to enroll a patient into a clinical trial for which they did not meet inclusion criteria. Green refers to requests motivated by a desire to provide therapy to an individual patient as “therapeutic misdirection,” calling this “the investigator’s counterpart to the subject’s therapeutic misconception.”¹⁸

Moreover, Miller wrote in 2000 about the “collusion of misunderstanding” in phase I oncology trials, questioning the justification of “trading on patients’ hopes” to enroll in trials that will not benefit them.¹⁹ Underlying Miller’s argument was the assumption that a small chance of benefit lacks proportionality when compared to the risk of toxicities and other inconveniences associated with trial participation. Participation in phase I oncology trials entails frequent imaging, lab draws, clinic visits, and, occasionally, additional procedures. These concerns are especially pertinent when considering that over 90% of drugs tested in phase I oncology trials will ultimately fail to come to market.²⁰ Some might argue that even if patients are willing to enroll in riskier trials, IRBs should make an independent judgment about whether the risks are reasonable, regardless of participant preferences and perceptions of acceptable risks.

Some have also argued that dose-escalation schemes that start with low doses provide important information about the lower end of the therapeutic curve. As described by Hey and Kimmelman, the main purpose of early-phase trials is to determine the “necessary and sufficient components of an effective and approximately optimal intervention,”²¹ which Kimmelman previously described as an “intervention ensemble.”²² They further provide three arguments for the importance of a complete and thorough initial discovery process that identifies the appropriate ensemble for subsequent testing. First, determining the optimal treatment ensemble (which includes information about the lower end of the dosing curve) is critical to ensure that participants in later-phase trials or clinical care are not injured because of toxicities that could have been identified in this initial discovery step. Second, identifying the ideal intervention ensemble is essential to “establish necessary conditions for clinical equipoise”²³ that justify subsequent randomization in later-phase trials. Third, they argue that having incomplete information on the ideal ensemble could predispose later studies to failure, whereas the same agent administered at the appropriate dosing, timing, and route of delivery (in other words, ensemble) might have truly provided benefit.

In describing the maximax, or “innovative care,” strategy, Hey and Kimmelman highlight that by targeting the ideal dose initially, researchers run the risk of losing vital information about the efficacy of lower doses: “[E]ven in a best-case scenario where the first test produces a significant response, the innovative care strategy returns little information about the contours of the ensemble landscape.”²⁴ Without an understanding of the lower bounds of therapeutic efficacy, they argue, valuable information is lost that could better inform future studies, decrease exposure to unnecessarily high doses when lower doses would suffice, and save costs long term by identifying lower levels of intervention that are equally effective.

In a third argument supporting the current risk-escalation strategy, some argue that prioritizing safety is more likely to sustain future drug-development efforts and protect the greater research community. Unexpected and major harms occasionally result from experimental interventions, despite strong preclinical evidence of safety. For example, deaths during trials and secondary cancers temporarily derailed the field of gene therapy.²⁵ Additionally, the use of CD28 superagonists led to severe illness and multiorgan failure in the first six subjects to receive this therapy.²⁶ When such catastrophic events take place, they obviously harm the individual subject. These events can also taint the public’s view of the entire research enterprise, potentially dissuading individuals from enrolling in future trials, and the public from supporting early-phase trials.²⁷

Fourth, some argue that, despite the participants’ primary motivations to seek therapeutic benefit, the individual preferences and values of potential participants should not influence study design. Adherents to this view argue that study design is a scientific endeavor and the preferences and values of potential subjects should not dictate it. Rather, trial design should strive to maximize the likelihood of reaching scientific aims in ethically permissible ways.²⁸ In considering the previously discussed catastrophic toxicities, one could argue that such events demonstrate why it is a mistake to consider clinical trials to be private transactions between researchers and patients rather than societal transactions with potential effects that extend far beyond the clinician-researcher–patient-participant relationships. Although it might be rational for a participant to want to enroll in a risky trial, it might be irrational for stakeholders in the broader research community to allow such trials to proceed.

REEVALUATING THE RISK-BENEFIT TRADE-OFF

Although the four arguments in favor of a risk-escalation approach that prioritizes the minimization of risk address important aspects of the risk-benefit trade-off in phase I oncology trials, we believe they contain conceptual problems, to which we will respond in order. The first argument proposes a lack of proportionality: using higher initial doses would make the risk-benefit ratio too unfavorable for reasonable people to accept; therefore, study designs should aim to minimize risks even if this decreases (or precludes) the possibility of benefit. However, we argue that proportionality is contextual. To address this argument, we must first discuss what defines “benefit.” The most obvious benefit to participants would be an improvement in survival. It has been clearly shown that chances of survival benefit from phase I oncology studies are low but certainly not nonexistent, and occasional trials result in grand slams. However, phase I trials are not designed to demonstrate survival benefit, and the most reliable outcome measurements are either ORR or stable disease. It is true that an ORR or stable disease might not prolong the participant’s life in all cases, but both are reasonable markers of success in absence of another option.

Although these chances of medical benefit are low, rarely do risk-benefit calculations solely entail likelihood of objective toxicities versus objective responses. The decision-making process is also driven by emotion, hope, and the values of decision-makers. Patients with end-stage cancer often mention their desire to “try everything” for the benefit of themselves or others. For example, parents with end-stage cancer may want to try everything in order to allay guilt about leaving their children behind, especially if they have young children.²⁹ Enrollment might also provide individuals with a sense of purpose or hope for patients that is performative as they navigate pain and anxiety at the end of life. Thus, participants may perceive benefits of enrollment beyond objective responses. This argument of disproportionality, then, ignores the interests of the individuals considering enrollment. The vast majority of participants enroll in phase I trials with the hope or expectation of clinical benefit, and many physicians offer enrollment as a treatment option.³⁰

This argument also ignores the alternatives to enrollment in phase I oncology trials. Potential subjects have advanced cancer with few or no evidence-based treatment options. When provided the option of a small chance that something great will happen versus a guarantee that something bad will happen, many will take the small chance despite the risk entailed. In fact, many patients at the end of life still opt for treatment rather than palliation alone, and these patients also tend to accept greater risks for a small possibility of benefit. In one study, “more than 90% of patients said they would still participate in the study even if the experimental drug had severe adverse effects, including a 10% chance of dying”—a level of risk almost never observed in oncology phase I trials.³¹ Whether such exposure to increased risk is appropriate depends on the context of the decision and the amount of risk. When faced with certainty of death versus a small chance of benefit, some people will, understandably, choose the small chance even with extremely long odds. In fact, enrolling such a patient in a trial with a starting dose so low that therapeutic benefit is biologically implausible (as in the risk-escalation model) is likely disproportionate from the perspective of the patient. They are assuming risks with virtually no chance of medical benefit, and they might even feel deceived based on their likely motivations for enrollment. Additionally, this argument assumes that better dose and schedule calibration employed by more sophisticated study designs will not lead to improved benefits, ignoring recent successes that led to the FDA’s “breakthrough therapy” classification, which expedites the clinical trials’ testing and the FDA review and approval process for experimental drugs that meet certain criteria.³²

This first argument supporting the current risk-escalation approach also addresses a lack of understanding or appreciation of likely outcomes. The bioethics literature is replete with examples of research participants and physicians demonstrating such therapeutic misperceptions.³³ However, these misperceptions are related to the informed consent process and individual characteristics of those involved, not trial design per se. If dosing strategies shifted toward innovative study designs that entailed more risk for participants, more work would be necessary to ensure that additional risks are conveyed and appreciated prior to enrollment. This should not, however, preclude modifications to study designs in phase I cancer trials.

Lastly, this initial argument questions whether the perceptions and preferences of potential research participants should influence IRB decision-making. In part, we agree with this argument. Individuals’ preferences for an extremely risky intervention should not unduly influence the IRB’s approval of a trial. However, we argue that individuals’ preferences provide important context for these ethical deliberations. To explore this argument, consider a hypothetical scenario in which an IRB is considering a proposed phase I oncology study that currently requires twice weekly clinic visits for enrolled patients and daily laboratory tests for a period. Now suppose that a large body of research demonstrates that frequent clinic visits and laboratory tests are perceived by cancer patients at the end of life to be extremely burdensome, and many cancer patients regret enrollment because of this burden. This data informs the ethical deliberation about balancing patient burdens with the scientific purpose in this study. We argue that considering participants’ views about perceived benefits and their openness to risk can similarly provide useful context to IRBs. This contextual information will not justify an unjustifiable study, but it can provide guidance to IRBs as they deliberate on questions of potential harm and benefit.

The second argument for a risk-escalation approach that prioritizes the minimization of risk focuses on threats to the greater research enterprise (and societal benefits of this research) if phase I trials collect incomplete data to inform a treatment ensemble for use in later-phase studies. To be clear, our proposed changes to early-phase studies cannot be justified if they lead to negative consequences for future research subjects, patients, or the drug development process itself. Some might argue that prioritizing therapeutic benefit could lead to a next generation of cancer drugs with surprise toxicities after FDA approval, resulting from insufficient safety assessment in the rush to establish efficacy. However, phase I trials are effective at identifying acute toxicities; later-phase studies and postmarketing surveillance are essential for identifying toxicities that are rare or delayed. A point in favor of our proposal to modify trial designs is that providing higher doses to more participants would lead to more data on these acute toxicities, not less.

Toxicity, however, is only one component of this treatment ensemble. The second argument also proposes that providing fewer participants with potentially subtherapeutic doses will result in insufficient data on ideal dosing, schedule, pharmacokinetics, and pharmacodynamics of the investigational agent. For example, consider the best-case scenario for the innovative care strategy as described by Hey and Kimmelman: researchers choose a starting dose that they believe to be close to the appropriate treatment dose based on preclinical studies, and the first group of participants demonstrate a dramatic response to the investigational agent. The authors note that “[e]fficacy trials can then immediately proceed using that ensemble.”³⁴ However, this need not be the case. It would be ethical and feasible to provide a subsequent cohort with a lower dose or different timing of administration if the researchers believed this information was essential for subsequent studies, especially if it is possible that the lower dose will provide similar efficacy. Alternatively, some of this data could be collected in other phases of study design, when participants have less severe disease and might provide a more representative understanding of pharmacokinetics and pharmacodynamics. The presentation of risk escalation as the only study design capable of generating this data is inaccurate.

The third argument for the current risk-escalation approach rests on the need to protect the greater scientific enterprise from societal reputation damage should a catastrophe occur due to a higher starting dose. The disastrous examples most commonly referenced focus on gene-transfer studies and CD28 super-agonists; however, it is probable that neither scenario was related to dosing but, rather, a response to the treatment modality itself. This argument rests squarely on the power of anecdote, and even the argument’s proponents concede that such catastrophes in phase I oncology trials are exceedingly rare. Yet these same proponents will simultaneously dismiss examples of phase I success stories as mere anecdote that should not influence this ethical discussion.

In addition, this third argument raises the important question of whether trial enrollment should be considered a private transaction between clinician-researcher and patient-participant, as opposed to a social necessity that requires consideration of impacts on the broader research enterprise. This argument accepts that it might be rational for a participant to want to enroll in a trial with higher risks, but it might not be appropriate for the research community to allow the participant to enroll. This argument is compelling for trial designs that clearly pose significant and striking risks at the outset. However, the changes to trial design we are recommending do not fit this mold. We argue that phase I trial designs should aim to limit the number of participants receiving subtherapeutic doses (should the intervention prove effective) by starting with slightly higher doses and modifying doses over time using data collected from the trial itself. We are not arguing for drastic changes to the safety profile of phase I trials. Instead, we argue against the current approach of minimizing risk without any consideration of potential benefit.

Finally, the fourth argument in favor of an approach that prioritizes the minimization of risk questions whether preferences or values of individuals should be considered in trial design. This argument furthers the question of whether the decision to enroll in a trial should be considered a private transaction or an event with broad implications for the greater research enterprise, and hence implications for the rest of society. While informed consent is critical in ensuring that participants have been respected in their decision-making process and supported to make ethically appropriate decisions, it is not a cure-all. As we mentioned, even if it is rational for participants to enroll in risky studies, it might not be ethical to allow them to enroll if the outcomes would have significant negative ramifications beyond their individual experiences. However, the answer to this question should not be “either/or,” but rather “both/and.” History is replete with examples of research that was performed for societal benefit without consideration of the individual participant’s interests. This history of atrocities and scandals is largely responsible for launching the modern bioethics movement. We argue that ethical research should seek to balance the public interests of society and the private interests of the individual participant. If either component is removed from this equation, then research becomes ethically fraught.

Our argument focuses on the risk of modifying dosing strategies in phase I oncology trials, not the risk of the intervention itself. Preposterous interventions based on faulty science should not be permitted. Interventions with an unreasonably high level of predictable risks should not be permitted. However, reasonable interventions with higher risk of toxicity based on dosing strategy should be permitted so long as participants understand and appreciate their chances of harm and benefit and this change does not hinder the scientific purpose of the study. Based on this reasoning, we argue that the current risk-escalation (maximin) model of early-phase study designs should shift toward an innovative care design (maximax). Such an approach better balances The Belmont Report’s endorsement of participant self-determination and the pursuit of benefits as well as the protection of participants and societal interests.

NEED FOR INNOVATION IN PHASE I ONCOLOGY STUDY DESIGNS

Currently, two main categories of dose escalation methods predominate phase I trials: rule-based designs (3 + 3 design, and accelerated titration models such as “rolling six”) and model-based designs (such as continual reassessment models).³⁵ Rule-based designs utilize low initial doses to minimize risk of harm for research subjects.³⁶ Model-based designs, conversely, accommodate sequential learning from accumulating data. These model-based designs can be formulated to optimize harm versus benefit trade-offs and thereby improve the expectation of benefit for trial participants if the therapy proves to be effective. Additionally, model-based designs are more amenable to incorporation of biomarker-guided strategies that address the intrinsic heterogeneity of tumors in a population. If subpopulation heterogeneity is monitored in phase 1 trials, efforts can be made to improve outcomes for participants with favorable subtypes of cancer by pursuing more aggressive treatment strategies for these individual participants.

An example is the model-based, adaptive trial developed by Thall and Cook that explored the use of thrombolytic agents in patients with acute ischemic stroke. This Bayesian dose-finding trial was devised to identify a dose with “good” reperfusion rate (benefit) without excessive hemorrhage (harm).³⁷ The design was guided by a utility function that expressed each dose level as a function of response and toxicity rates. Dose selection was outcome adaptive, such that the next dose level was chosen to maximize the beneficial outcome. In other words, accumulated data on risks and benefits were used to identify the “best” dose for subsequent participants. Other model-based adaptive designs have been proposed for dose-selection trials intending to optimize toxicity and efficacy.³⁸ Additionally, theoretical considerations have established performance benchmarks for accuracy in dose selection.³⁹ More recently, methods of design have been developed for identifying dose levels that are both safe and effective in phase I and II trials, and these designs have been implemented in trials studying a combination of targeted therapies and immunotherapies.⁴⁰ In rule-based study designs, there is a binary choice between prioritizing risk reduction or maximizing potential benefit. In adaptive study designs, both risk and benefit can be incorporated into a function that identifies subsequent doses.⁴¹ Despite these advances in trial-design methodology, most current phase I studies employ rule-based designs. Rule-based designs are simple to implement, but they do not fully support the participant’s goal of enhancing potential therapeutic benefit, nor do they make the best use of resources or maximize the speed of drug development.

Our argument for reprioritizing therapeutic benefit in phase I trial design is not an argument for unfettered access to unproven treatments, nor is it an argument for reckless disregard of individual safety in favor of potential therapeutic benefit. Rigorous evaluation of novel compounds by the FDA and strict criteria for enrollment in clinical trials remain of paramount importance. Additionally, we are not arguing that every patient with advanced disease should be eligible for enrollment in phase I trials. Rather, phase I trials should be presented as one option to eligible patients during a robust discussion of risks, benefits, and the patient’s goals and values.

We recognize that altering phase I study designs will be accompanied by practical barriers. Model-based designs require statistical expertise and specialized infrastructure at the enrolling institutions, for example. Additionally, modifying dosing strategies in phase I trials as we have proposed could expose more participants to risk of toxicity, and implementation of these proposed changes will require agreement and collaboration of all involved stakeholders, including physician-investigators, sponsoring agencies, pharmaceutical companies, IRBs, and research participants. Furthermore, the ethics of adaptive trial designs have been called into question, especially in later-phase trials.⁴² Nonetheless, the status quo for phase I oncology trials is sorely in need of modernization.

Many patients with advanced cancer who enroll in phase I trials are motivated by the hope for therapeutic benefit. Recent data show that ORRs in phase I oncology trials continue to slowly improve, especially for specific types of cancer. With the further integration of next-generation sequencing and targeted therapies, these numbers are likely to continue improving. However, the vast majority of phase I oncology trials currently employ rule-based designs such that a large proportion of subjects receive a dose that would be ineffective even if the intervention were subsequently proven to be successful. Additionally, great inefficiencies in the drug development process persist.⁴³ As we have highlighted, the arguments in favor of the status quo are flawed and insufficient.

There are signs that patients and others are growing restless with the wait for novel, effective therapeutics.⁴⁴ In the absence of forethought from the medical research community, society will act. This could result in diminishment of FDA oversight authority and in approval of incompletely tested therapies. Similarly, this could result in further advancement of “right-to-try” legislation that could harm the drug development process and delay or impair the ability to ensure safety and efficacy of novel drugs.⁴⁵

Investigators, IRBs, and pharmaceutical companies should employ dosing strategies that deemphasize risk minimization; the emphasis should instead be on taking steps that promote potential therapeutic benefit while limiting risks (with the caveat that trials of agents that are both “first in human” and “first in class” may require a more conservative approach initially). Such reprioritization will better respect the motivations of participants who enroll in hopes of medical benefit without interfering with the scientific purpose of these studies. Much of the ethics literature on phase I oncology trials has focused on ensuring that subjects understand that they are unlikely to benefit. Perhaps it is time to shift the focus toward actually improving chances of benefit. ♦