Keywords

1 Introduction

Antimicrobial resistance (AMR) has been described as one of the major threats to public health in the twenty-first century and sparked concerns about the possibility of a post-antibiotic era, in which many infections are no longer treatable (WHO 2014; General Assembly of the United Nations 2016; Davies et al. 2013). Such a development would not only have detrimental effects on patient care, but is also likely to significantly affect public health, agriculture, as well as economic and national security (PCAST 2014; O’Neill 2014; Adeyi et al. 2017; OECD 2018).

Over the past few years, the challenge of AMR has increasingly been recognised by policy makers. The World Health Organization (WHO) developed a global action plan on AMR (WHO 2011, 2015), the United Nations General Assembly passed a declaration on the issue (General Assembly of the United Nations 2016) and the G20—a forum of the world’s major economies—has placed AMR on its agenda (Bundesministerium für Gesundheit 2011). In addition, considerable investments have been made to strengthen and incentivise the development of new antibiotics (Boucher et al. 2013; Drive-AB-12 2018; Rex 2014; Outterson et al. 2016a, b).

Since one of the major causes of AMR is the overprescription and overconsumption of antibiotics, comprehensive response plans to preserve antibiotic effectiveness do not focus solely on the development of new drugs. They also include a wide range of measures, including surveillance, infection control and, crucially, the promotion of the “rational” use of antibiotics (WHO 2014). Rational use programmes primarily focus on avoiding unnecessary prescriptions of antibiotic treatments, for example in the context of viral respiratory infections. However, they also include practices that involve delaying or withholding access to antibiotics that are known to be beneficial. These practices can place some patients at risk of harm and thereby lead to conflicts between two important ethical goals: promoting patients’ clinical interests and preserving antibiotic effectiveness for future use.

The ethical conflict between these goals has so far not been adequately addressed. In this paper, we describe programmes for rational antibiotic use and show when they require placing some patients at risk of harm for the benefit of managing the existing pool of effective antibiotics. We argue that ethical guidance on acceptable risks in clinical research can help to develop acceptable programmes for rational antibiotic use. Based on a recognized ethical framework for risk and risk-benefit evaluations in clinical research, we propose a novel framework that allows clinicians and policy-makers to evaluate when rational antibiotic use is (or is not) ethically justified.

2 Rational Use of Antibiotics

The promotion of “rational” antibiotic use – sometimes referred to as “prudent” or “appropriate” use – is a key component of many response plans to AMR (Bundesministerium für Gesundheit 2011; UK Department of Health 2013; Public Health Agency of Canada 2014; The White House 2014; Australian Department for Health 2015). The WHO provided the most widely cited definition of the rational use of medicines in general, which includes the rational use of antibiotics. According to WHO, this requires that “patients receive medications appropriate to their clinical need, in doses that meet their own individual requirements, for an adequate period of time, and at the lowest cost to them and their community” (WHO 1985). Elsewhere, the WHO defined appropriate antibiotic use as the “cost-effective use of antimicrobials which maximizes clinical therapeutic effect while minimizing drug-related toxicity and the development of antimicrobial resistance” (WHO 2011). Most concise is the definition provided by the Alliance for the Prudent Use of Antibiotics (APUA), which defines prudent use as “the right drug for the right condition for the right amount of time” (Wilson and Tan 2010).

Notwithstanding some differences in meaning, these definitions all delineate types of antibiotic use that are clearly “irrational”—that is, obvious instances of overuse or misuse where antibiotics are prescribed or taken even though they are not the appropriate treatment. For example, antibiotics are an appropriate treatment for “strep throat”, which is caused by a bacterium susceptible to certain antibiotics. However, they are not an appropriate treatment for most sore throats, which are caused by viruses against which antibiotics are not effective.

There is a wide array of interventions designed to reduce antibiotic prescribing, and evidence suggests that restrictive measures, which limit the availability of antibiotics, have a higher chance of success than information campaigns or educational interventions that teach prescribing clinicians about AMR (Charani et al. 2011; Davey et al. 2013). In many instances, such restrictions appear to be a promising and effective way of reducing antibiotic use. For instance, recent research has shown that antibiotic prescribing in primary care can be restricted substantially, without a negative effect on clinical outcomes, through a practice called delayed prescribing (Schuetz et al. 2009; Spurling et al. 2013; Little et al. 2014). Here patients for whom antibiotics are unlikely to be an appropriate treatment—say most patients with sore throats—receive a prescription that they can use at the pharmacy if their symptoms do not improve on their own within a few days. Receiving such a prescription gives patients reassurance, even though only few end up filling it. A Cochrane review concluded that delayed prescribing offers no additional risks to the patient, while offering similar patient satisfaction when compared to immediate prescribing (Spurling et al. 2013).

So understood, rational use of antibiotics is ethically justified because it promotes patients’ clinical interests while also preserving antibiotic effectiveness for future use. Not only are present patients spared the side effects of unnecessary treatment, avoiding such treatment also curbs the spread of antibiotic resistance and thereby helps to ensure that existing antibiotics remain effective to treat future patients. Indeed, because rational use so understood promotes the interests of present as well as future patients, clinicians and policy-makers are ethically required to ensure that patients use “the right drug for the right condition for the right amount of time” (Wilson and Tan 2010).

3 Ethically Challenging Instances of Rational Use

In addition to avoiding unnecessary antibiotic use, rational use programmes embrace practices that restrict access to antibiotics even when treatment would reduce a given patient’s health risk (Millar 2012). One example for this are programmes that ask clinicians to factor resistance thresholds in the community when they suggest antibiotic treatment to their patients. For example, clinical practice guidelines for treating community-acquired pneumonia recommend the use of macrolides—a particular class of antibiotics—unless 25% or more of the relevant pathogens are resistant to macrolides in the given local community (Mandell et al. 2007). Only then should clinicians switch to more effective second-line drugs (Mandell et al. 2007).Footnote 1 When considering the entire patient population, this is a sensible recommendation, since it is to be expected that macrolides—the first line of treatment—remains effective in most patients if resistance is below 25%. However, a delayed switch to second-line drugs negatively affects clinical outcomes in some patients, notably those with severe pneumonia. Some authors estimate that a similar guideline has carried a mortality risk of more than 1% for all patients in the past (Daneman et al. 2008). Rational use practices of this type effectively delay or withhold a proven beneficial antibiotic in order to curb the spread of antimicrobial resistance. This places some present patients at risk for the sake of future patients.

In these and comparable cases, the rational use of antibiotics confronts clinicians with a trade-off between promoting their present patients’ best clinical interests and helping to maintain effective antibiotic treatments for future patients (Kollef and Micek 2014). This trade-off seems particularly stark for several reasons (Littmann et al. 2015; Littmann and Buyx 2018). First, it is difficult to measure the contribution of any individual course of antibiotics to the emergence of AMR in a population, so that estimates about the benefits of rational use programmes are relatively uncertain. It seems likely, however, that delaying or withholding any given antibiotic treatment makes no more than a very small contribution to curbing AMR and, in some cases, no contribution at all. For example, a patient might develop resistant bacteria but die before transmitting these to other people or the environment. As a result, rational use programmes that involve delaying or withholding antibiotic treatment are neither necessary nor sufficient for addressing AMR. Moreover, the benefits of preserving antibiotics may accrue to unknown people in the potentially distant future. By contrast, the benefits of antibiotic treatment for the present patient are much easier to measure and can therefore be predicted with greater certainty. Antibiotic treatment also has immediate benefits that accrue to known or “identified” persons. These differences readily explain why a typical clinician who is tasked with delaying or withholding antibiotics from her present patients with the goal of tackling AMR would feel an acute ethical conflict (Cohen et al. 2015).

Bioethicists have started to discuss this ethical conflict (Selgelid 2007; Battin et al. 2009; Millar 2012; Oczkowski 2017; Littmann and Buyx 2014; Littmann et al. 2015). However, at both the clinical and policy level, there is no ethical guidance on when rational use programmes that involve delaying or withholding antibiotics and hence compromise the best clinical interests of present patients are acceptable. Moreover, many policy documents lack explicit discussion of any ethical challenges, and those that include ethical considerations are narrowly disease- and pathogen-specific (Bundesministerium für Gesundheit 2011; WHO 2011; Davies et al. 2013; Andreasen 2014). What is therefore needed is a more specific analysis of whether and when it is acceptable to restrict antibiotics use with the goal of curbing AMR, even when doing so poses risks to present patients and thereby compromises their best clinical interests.

4 The Analogy to Clinical Research

To address this question, it is helpful to consider more generally whether and when it is justified for clinicians not to promote the best clinical interests of their present patient. Standard professional guidance takes a relatively restrictive stance on this issue. For example, the World Medical Association states that clinicians owe their patients “complete loyalty”, and that they may place the interests of others above those of the patient only in “exceptional situations” (Williams 2015). However, closer analysis reveals many clinical practices that place the interests of other patients, individuals or society above the interests of a given present patient (Wendler 2010). Consider, for example, that clinicians routinely give vaccinations not to protect the present patient, but to maintain herd immunity; that clinicians transplant kidneys from healthy individuals to patients with renal failure in order to enable these patients to live a better and longer life; and that clinicians allow younger colleagues to gain experience and perform procedures for the first time, even though they know that this will result in higher complication rates for the present patient. Importantly, these practices are not only frequent, but ethically justified provided that a number of conditions are met (Wendler 2010). This finding raises the possibility of evaluating the ethical acceptability of rational use programmes for antibiotics in comparison to other practices that involve compromising the interests of individual patients for the benefit of others.

For such comparisons to be both sound and useful, it is essential to identify comparator practices that are relevantly similar to rational use programmes for antibiotics and widely considered to be acceptable (Sunstein 1993). Moreover, detailed ethical guidance on how to evaluate the given comparator practice should ideally exist. Clinical research fulfils all three desiderata, as the following paragraphs demonstrate.

Clinical research is a subset of research with human participants that focuses on evaluating methods to prevent, treat or cure illness and disease, or on generating the knowledge necessary to develop such methods. One of the key characteristics of clinical research is that investigators expose participants to some risks for the potential benefits of unknown patients in the potentially distant future. Clinician-investigators routinely perform research procedures that do not promote the best clinical interests of the present patient-participant, but serve solely to address research questions. In fact, the majority of clinical trials involve procedures that have no prospect of clinical benefit for participants but help to answer important research questions. For example, most clinical trials involve additional blood draws, biopsies or imaging procedures that would not be performed as part of routine clinical care, but serve to test the safety and efficacy of an investigational drug. Moreover, phase 1 trials with healthy individuals by definition have no prospect of clinical benefit. It is widely considered acceptable for clinician-investigators to perform these “net risk” procedures and trials—and thereby deviate from the general clinical norm of promoting the present patient’s best clinical interests—because clinical research generates knowledge with the potential to improve the health or care of future patients (WMA 2013; CIOMS 2016). That is, the potential benefits of clinical research for future populations are widely considered to justify that investigators expose patient-participants to some level of net research risk.

Importantly, the potential benefits for future patients—or the so-called “social value” of the research—is generally seen as the fundamental justification for exposing participants to net research risks, not the participants’ informed consent. Participants’ consent is, of course, relevant for determining what levels of net research risk are acceptable in socially valuable research. Only small levels of net risk are acceptable when participants cannot give their own informed consent, for example in research involving children, or when obtaining participants’ consent is not feasible, for example in research on large datasets that were originally collected for clinical (and not for research) purposes. By contrast, greater net risks are acceptable when participants give their voluntary and informed consent. However, it is generally accepted—and for good reason in our view—that research without social value is not justified even when the research poses low net risks and participants consent (Wendler and Rid 2017). This underscores that the fundamental justification for exposing participants to net research risks lies in the potential benefits of clinical research for future patients.

Restricting antibiotic use with the goal of curbing AMR is relevantly similar to clinical research in this important respect. Just like in research, clinicians expose the present patient to some risks of harm for the potential benefits of unknown patients in the potentially distant future. Moreover, what level of risk is acceptable arguably depends on whether the present patient consents to delaying or foregoing antibiotic treatment that would be in her best clinical interests. In clinical research, it is not certain that any given clinical trial generates findings that make a significant contribution to improving the health or care of future patients and populations. Moreover, the enrolment of any single patient-participant is uncertain to make a significant difference to the value of any given trial. Yet without sufficient overall participation, trials—and clinical research more generally—would fail. Rational use programmes that involve delaying or withholding antibiotics are similar in this respect. It is not only uncertain that such programmes will make a significant contribution to addressing AMR, but restricting a single patient’s access to antibiotics may not lead to a measurable effect on the overall level of AMR. Yet reducing antibiotic prescriptions overall is a key component of strategies for curbing AMR and ensuring that bacterial infections can still be treated effectively in the future (Costelloe et al. 2010; Davies et al. 2013). Given this relevant similarity between clinical research and rational use programmes that delay or withhold antibiotics, and considering that it is widely accepted for clinician-investigators to impose some level of net risk on patient-participants for the potential benefit of future patients (Selgelid 2007; Kollef and Micek 2014), it should also be acceptable to impose comparable levels of risks on patients in order to preserve antibiotic effectiveness for future patients.Footnote 2

Moreover, there is a long-standing and nuanced debate about acceptable risk in clinical research. This does not only mean that our judgments about acceptable risks are relatively considered in the research context, but also that detailed ethical guidance in this area exists (WMA 2013; CIOMS 2016; Emanuel et al. 2008). Clinical research therefore fulfils the three desiderata for sound and helpful comparisons formulated above: it is relevantly similar to rational use programmes that delay or withhold antibiotics for the sake of curbing AMR; it is widely considered an acceptable exception from the general principle that clinicians should act in the best interests of the present patient; and detailed ethical guidance on how to evaluate research risks, including net research risks, exists. This suggests that judgments about acceptable risk in clinical research can inform the ethical evaluation of rational use programmes that impose risks on present patients with the goal of curbing AMR and safeguarding antibiotic effectiveness for future patients (Rid et al. 2019).

5 An Ethical Framework for Evaluating Rational Antibiotic Use

For clinician-researchers to justifiably impose risks on patient-participants in clinical research, a number of conditions must be met. Specifically, the research must be scientifically valid and socially valuable; the risks to participants have to be minimised and reasonable in relation to the potential clinical benefits for them and/or the social value of the research; and, when participants cannot or do not consent, any net risks to them should be no greater than minimal or, in cases of compelling social value, no greater than a minor increase over minimal risk (WMA 2013; CIOMS 2016).

To operationalise these conditions, one of the authors of this paper (AR) has developed a systematic framework for evaluating research risks and potential benefits based on prominent research ethics guidelines and literature (Rid and Wendler 2011). This framework can be usefully adapted to evaluate the risks of rational use programmes that involve delaying or withholding antibiotics from patients who could clinically benefit (Appendix). The framework is most useful for evaluating rational use programmes that do not require clinicians to obtain the patient’s informed consent for restricting antibiotics. In particular, the minimal risk threshold for research without informed consent (Rid 2014) provides much needed orientation regarding acceptable levels of unconsented to net risks in the public health context. However, the framework could also help to evaluate rational use programmes that would require the patient’s informed consent. Here, the debate about upper net risk limits in research with competent consenting adults is particularly informative (London 2006; Miller and Joffe 2009).

The framework sets out six steps for evaluating rational use programmes that involve delaying or withholding antibiotics. A hypothetical programme to restrict antibiotic use for lower respiratory tract infections (LRTIs) helps to illustrate how these steps would work in practice (Table 20.1

Table 20.1 Ethical framework for rational antibiotic use
Full size table

, Ethical framework for rational antibiotic use).

The first step—ensure and enhance the programme’s social value—requires decision-makers to develop rational use programmes based on sound evidence and policy methods. This is to ensure that all programmes are rigorously developed and judged to be both feasible and effective. For example, the available data should suggest that delaying or withholding the given antibiotics for LRTIs has the potential to help curb AMR. Data should also be sufficient to estimate the risks of restricting the antibiotics for patients; if there is too much uncertainty regarding how likely patients will suffer harm, relevant data should first be gathered in a research study. For programmes that meet these conditions, step one also requires conducting (and/or reviewing) observational research on their implementation. For example, how does restricted antibiotics use for LRTIs affect prescription rates, clinical outcomes, and antimicrobial resistance in the community? Such research serves to confirm that the given programme is indeed effective, while also expanding the existing knowledge of AMR. Evidence from accompanying observational research should inform regular re-evaluations of the programme.

The second step—identify the programme interventions—requires considering how the care of patients under the given rational use programme would differ from recommended standard care. Of course, the key difference would be that antibiotics are delayed or withheld under clearly specified circumstances. However, programmes might also include supplementary, non-routine interventions that aim to reduce the risks of restricting antibiotic use. Importantly, these non-routine interventions can carry their own risks. For example, if antibiotics for LRTIs were delayed or withheld, patients might be informed about signs of pneumonia or empyema (i.e. potential more serious complications of an initial infection) in writing, rather than verbally as in standard practice, so as to ensure that they know when to seek expert advice and thereby reduce risks to them. However, receiving written information can itself cause concern among patients. It is therefore essential to identify all programme interventions in order to comprehensively evaluate the risks that the given rational use programme poses to patients. As part of this step, the need for any supplementary interventions should also be scrutinized with a view to avoiding unnecessary risks. For example, do patients really require written information?

Step three—evaluate and reduce the risks to patients—requires evaluating the risks of all programme interventions, including those of any supplementary interventions, based on the available evidence. Moreover, all reasonable measures must be taken to reduce the risks to patients. For example, if written information about pneumonia and/or empyema is considered necessary, how can this information be conveyed in ways that reduce anxiety and stress for patients? Similarly, rational use programmes that involve delaying or withholding antibiotics need to ensure that patients who suffer serious harms are adequately treated at no financial cost. In the case of restricting antibiotics for LRTIs, for instance, it must be guaranteed that any collections of pus in the pleural cavity (i.e. empyemas) are promptly treated for free. Patients with any lasting serious harms should also receive compensation. Furthermore, patient groups at increased risk of experiencing serious harm—for example, patients with immune deficiency—should either be actively monitored or excluded from the programme. Guidelines should also instruct clinicians to exercise their judgment in individual cases.

Step four—evaluate and enhance the potential benefits for patients—requires evaluating based on the available evidence to what extent patients could benefit clinically from the given rational use programme. For example, not receiving antibiotics for LRTIs means avoiding their side effects and reducing the risk of infection with Clostridium difficile—which is associated with antibiotic use and sometimes leads to a life-threatening infection of the colon—or resistant infection in future. Moreover, non-routine supplementary interventions can have potential clinical benefits for patients. For instance, if a one-time respiratory therapy session is introduced to help manage patients’ symptoms, this should be considered when evaluating the risks of the overall programme. Any supplementary interventions with potential clinical benefits for patients should also be targeted at groups who would benefit most, especially when they cannot be provided to all patients. For example, the one-time respiratory therapy session might be focused on patients who suffer from chronic pulmonary problems (provided they are not excluded from the programme for being at increased risk).

Step five—evaluate whether the interventions pose net risks to patients—requires judging, again based on the available evidence, whether the risks of each individual intervention exceed the intervention’s potential clinical benefits. If the answer to this question is yes, then the given intervention is said to pose “net risks” to the patient. One way of making this determination is to ask whether an “informed clinician” who is committed solely to the promoting the patients’ best clinical interests would recommend that they undergo the intervention in question (Rid and Wendler 2011). If the clinician would recommend the intervention, it promotes the patients’ clinical interests and thus does not pose net risks. This implies that the intervention’s risk-benefit profile is acceptable and—assuming the requirements of the previous steps are satisfied—needs no further evaluation. If the clinician would be indifferent, then undergoing the intervention neither undermines nor promotes the patients’ clinical interests. Provided that including the intervention in the rational use programme is necessary to ensure its safety (step 2), this suggests that it has an acceptable risk-benefit profile. If the clinician would advise against the intervention, then it poses net risks that require further evaluation (step 6). For example, an informed clinician who is committed solely to promoting the patients’ best clinical interests would not recommend delaying or withholding antibiotics for LRTIs (at least where the expected benefit of antibiotic treatment outweighs the expected harms of adverse effects of antibiotics outlined above). This means that this intervention poses net risks whose acceptability needs to be considered further.

Finally, step six—evaluate whether the net risks to patients are justified by the programme’s social value—serves precisely this purpose. It first requires determining the cumulative level of net risk of the given rational use programme by adding the net risks of all its interventions. For example, restricting antibiotics for LRTIs likely poses a relatively small level of cumulative net risk from delaying or withholding antibiotic treatment and providing patients with written information in order to manage the resulting risks. Evaluators must then judge whether the programme’s cumulative net risks fall within the general range of acceptable net risk to patients and, if the answer is yes, whether the risks are proportionate to the social value of implementing the programme.

With regard to the general range of acceptable net risk, there is a long-standing debate about this issue in clinical research that can inform judgements about the acceptability of rational use programmes. Debate has been most intense in the context of research involving participants who cannot give their own informed consent, such as children or patients with dementia. Many ethical guidelines and regulations, as well as many research ethicists, endorse a “minimal risk” threshold for such research in order to protect participants from excessive net research risks (Emanuel et al. 2008; WMA 2013; CIOMS 2016). The precise interpretation of this threshold remains contested. However, a recent analysis of the existing literature identifies several “lessons learned” that are equally pertinent for the evaluation of rational use programmes (Rid 2014). Specifically, the most convincing definitions of minimal risk refer to the two basic components of research risk, likelihood and magnitude of research harm. Further, these definitions distinguish different magnitudes of harm (e.g. small, moderate, serious) and set approximate likelihood thresholds for each magnitude of harm, where the likelihood thresholds are anchored with numeric information (e.g. 1 per 100,000) derived from risk comparisons. For example, if the risks of participating in a charity soccer game are considered acceptable, and this activity is thought to be relevantly similar to clinical research, data on the risks of playing soccer can be used to anchor likelihood thresholds for different magnitudes of research harm. At present, comparator risks are best specified as the risks that average, healthy, normal individuals in different age groups face in riskier but still acceptable activities that are directed at benefiting others, such as playing soccer for charitable purposes in children. In practice, this implies that the most convincing definitions of the minimal risk threshold equate minimal net research risks with very low likelihoods of serious and moderate harm and modest likelihoods of small harm (Rid 2014). Using this threshold as a preliminary guide for evaluating the acceptability of rational use programmes, net risks such as those reported for delaying antibiotics for LRTIs would seem to qualify as minimal. Specifically, if the two main complications—empyema and pneumonia—are classified as moderate harms, the up to 10% likelihood that they will occur (Schuetz et al. 2009) is broadly consistent with the likelihood of moderate harms observed in appropriate comparator risks (Rid et al. 2010).

Similar arguments have been advanced in the debate about upper risk limits in research with competent consenting adults. Specifically, several authors suggest that these upper risk limits should be delineated in comparison to appropriate comparators risks for competent adults, such as emergency medical assistance (London 2006; Miller and Joffe 2009). In rational use programmes that require clinicians to obtain informed consent for delaying or withholding antibiotics, the range of acceptable net risks could be evaluated based on what levels of risk are considered acceptable in research involving competent consenting adults. However, all current rational use programmes that we are aware of proceed without the patients’ informed consent (though individual doctors may discuss prescribing decisions with their patients and seek to understand patients’ preferences), so that the minimal net risk threshold is more pertinent. Moreover, although some ethical guidelines allow a “minor increase” above minimal risk in research without consent when it is not possible to gather the necessary data in a less risky manner and the social and scientific value of the research is compelling (CIOMS 2016), the minor increase over minimal risk threshold remains underspecified in the research ethics literature. We therefore would not recommend applying it systematically in the context of rational use programmes.

With regard to judging whether the net risks of a given rational use programme are proportionate to the social value of implementing it, one way of making this judgment is to ask whether an “informed and impartial social arbiter” would recommend the programme in question after carefully considering the risks and potential benefits for all affected parties and giving everyone’s claims fair consideration, while treating like cases alike in similar areas of policy (Rid and Wendler 2011). This test provides no more than a heuristic to guide evaluators’ judgment; however, the idea of a social arbiter helps to ensure that rational use programmes are evaluated just like other programmes or policies that impose some level of risk on individuals for the sake of realising an important social good—within clinical medicine, but also beyond. In the case of delaying or withholding antibiotics for LRTIs, the (arguably) minimal net risks that this programme involves for patients would seem proportionate to the importance of preserving antibiotic effectiveness and the potential of this programme to help curb AMR. Moreover, given that net risks are minimal, the programme could proceed without obtaining the patients’ informed consent.

6 Potential Objections

Critics might argue that there are less invasive strategies for curbing the spread of AMR that we should exhaust before delaying or withholding antibiotics in clinical medicine, such as eliminating unnecessary antibiotic treatments, reducing antibiotic use in farming, and investing more, and more effectively, into research on new antibiotics. These other strategies should indeed be pursued with urgency, and there is little reason to believe that adopting rational use programmes would undermine them. However, given that AMR is now recognized as one of the major public health threats of the twenty-first century, it is essential to consider all approaches that could delay or prevent AMR in an ethically justifiable way. As this paper argues, rational use programmes that involve delaying or withholding antibiotics should become part of the conversation.

It might also be objected that few such programmes would satisfy the minimal risk threshold known from the context of clinical research. At this point, it is difficult to respond to this objection because few reliable data exist on patient outcomes when antibiotics are delayed or withdrawn. However, we have identified at least one case—restricting antibiotic use for LTRIs—where reasonable estimates suggest that the minimal risk threshold would be respected. As more data become available, more cases could follow. Moreover, our proposed ethical framework would also be helpful for evaluating rational use programmes that pose more than minimal risks to patients and therefore require obtaining their informed consent to participate. However, as mentioned, all current rational use programmes we are aware of proceed without obtaining the patients’ informed consent, and we suspect this will stay the same in future. To be effective, rational use programmes need to be implemented as widely as possible—and in practice this is likely to be feasible only when patients are not asked to give their informed consent to having antibiotic treatment withheld or delayed. Given the considerable uncertainty about the social value of rational use programmes in terms of curbing AMR, and considering that such programmes are generally rolled out without requiring that clinicians obtain patients’ informed consent, we would argue that rational use programmes should pose no more than limited net risks to patients until more evidence supports their social value.

Finally, some might argue that the present argument rests on an ethical framework for risk-benefit evaluations in clinical research that is not universally accepted. In particular, prominent research ethicists hold that clinician-investigators should generally not delay or withhold established effective treatments from patient-participants in clinical trials (Weijer and Miller 2004). This argument is based on the idea that clinician-investigators continue to have clinical obligations in the research context, which make it unacceptable for them to act against the best clinical interests of patient-participants by depriving them of proven effective interventions. And, it might be argued, if it is not acceptable to delay or withhold proven interventions in the research context, then it is not acceptable to do the same in rational use programmes. However, it is important to see that this latter position is not universally accepted in research ethics. This is because it cannot explain why it is justified for clinician-investigators to pose any net research risks to patient-participants—for example, by performing research-specific blood draws or biopsies that have no prospect of clinical benefit. After all, these interventions by definition do not promote the best clinical interests of patient-participants (Miller and Brody 2003). The ethical framework for risk-benefit evaluations that forms the basis of the present paper avoids this problem. It justifies why some level of carefully evaluated net risk can be imposed to patient-participants, namely because clinical research has important social value. Moreover, the framework applies the same standards for acceptable net risk to all research interventions, whether or not net risks result from delaying or withholding effective treatments or from performing research procedures without a prospect of clinical benefit (Rid and Wendler 2011). Thus, the ethical framework for risk-benefit evaluations in clinical research used here, while not universally accepted, is arguably more defensible than the existing alternatives.

7 Conclusion

For clinicians, rational use programmes can pose ethical conflicts when they involve delaying or withholding antibiotics and thereby compromise the clinical interests of their present patients for the sake of preserving antibiotic effectiveness for future patient populations. Such exceptions from the general norm that clinicians should always act in the best interests of the present patient need to be carefully developed and managed. In this paper, based on a comparison to clinical research, we presented an ethical framework that enables an explicit and transparent evaluation of the net risks that rational use programmes can pose to patients in order to address AMR. Because these evaluations require judgments about complex empirical facts and normative questions, it is essential that rational use programmes be developed transparently and with the involvement of patients, clinicians and other relevant stakeholders. This involvement would not only improve the quality of deliberations, but also help to ensure the fairness and legitimacy of the resulting policies and their successful implementation.

Finally, the rational use of antibiotics can only be one part of a comprehensive strategy to address the danger of AMR; other measures have to be pursued with equal urgency, including better infection control, comprehensive strategies to reduce antibiotics use in animals, and the development of new antibiotics.