Introduction

Antibiotic resistance is a burgeoning problem and there are few new antibiotics on the immediate horizon.1 Currently it is estimated that in the European Union antibiotic-resistant bacteria cause 25 000 deaths a year, and contribute a yearly cost in the US of $21–34 billion.1 Each time an antibiotic is used there is some impact on the burden of antibiotic-resistant microbes either through direct effects on the individual recipient2–4 or arising through contamination of the environment with antibiotics, or resistance genes, or resistant microbes.5 Despite widely publicised concerns about the risks of antibiotics, there is evidence for continuing over-prescribing.6 ,7 Antibiotics are like a leaky rescue craft which becomes less effective the more it is used. If the supply of new antibiotics is drying up, then we should ensure that we have good reasons to justify the use of those antibiotics that are still effective. This is not just a question of evidence. There are also ethical considerations, such as reasons for not using antibiotics, the justification of ‘acceptable’ levels of risk, the valuations of adverse outcomes (such as death) and of (not) selecting antibioticresistance, and the extent to which we should use aggregate costs and benefits to determine antibiotic policies.

The WHO chose combating antimicrobial resistance as the theme of this year's World Health Day (7 April 2011). ‘Inappropriate and irrational use of medicines’ is given by the WHO as a major contributor to the burden of antibiotic-resistant infections.8 The control of ‘inappropriate’ use is a common theme in strategies for the control of antibioticresistance.9 The term ‘inappropriate use’ is not consistently defined in the medical literature. Information on cost and effectiveness is frequently used to differentiate ‘appropriate’ from ‘inappropriate use.’2–9 It is the criteria that we can use to reject the use of antibiotics, that is, the focus of this paper.

Antibiotic treatment guidelines can be controversial

A potential antibiotic prescriber has to decide not only whether to prescribe an antibiotic (or not), but also what threshold level of antibiotic resistance (potential ineffectiveness) precludes use of a particular antibiotic for a given condition. Medical guidelines for the use of antibiotics for specific conditions may specify antibiotic resistance thresholds above which more active alternatives should be used, for example, ‘10%–20%’ for the antibiotic trimethoprim for uncomplicated cystitis,10 and 5% for quinolones for gonorrhoea.11 The % figure is the proportion of patients with gonorrhoea who have infection with a strain considered (usually by laboratory testing) to be resistant to the antibiotic (quinolone in this case). This means that 5% of patients treated with quinolones for gonorrhoea would be expected to show a sub-optimal response by comparison with an alternative.

A paper by Daneman et al (2008)12 illustrates controversy over what are ‘acceptable’ resistance thresholds. Infectious Diseases Society of America (IDSA) treatment guidelines for community-acquired pneumonia advocate use of macrolide antibiotics as monotherapy in situations where up to 25% of clinical isolates of Streptococcus pneumoniae (the predominant cause of pneumonia) are resistant by laboratory tests. Daneman et al suggest that this guideline is endorsing ‘therapy-attributable mortality in ∼ 1 of every 100 persons treated with empirical macrolide therapy’. The suggestion is that the IDSA guidelines are implicitly accepting a 1% mortality rate for community-acquired pneumonia, even though this rate of death could be reduced by an alternative antibiotic treatment regimen.

Daneman et al (2008)12 go on to suggest that “It might be argued that a more rational and reasonable threshold for use of alternate antimicrobial therapy would be the resistance prevalence at which the hidden cost of mortality exceed the cost differential between macrolide agents and other therapeutic modalities”. Daneman et al are suggesting that the aggregated treatment costs should be used to establish acceptable levels of death attributable to sub-optimally treated infection. The individual outcome is being weighed against the overall costs of a particular strategy, and aggregated small benefits (treatment costs) are being weighed against the cost of a human life. This approach reflects the widespread use of cost effectiveness as a substantial consideration in determination of treatment guidelines.13 Use of cost effectiveness as a criterion to determine ‘appropriate’ antibiotic policy is problematic for many reasons. These include empirical14 and philosophical challenges15 with the placing of a monetary value on a human life and the attribution of future costs consequent on infection with antibiotic-resistant bacteria. These costs may be substantial (particularly with the prospect of the spread of untreatable infection) but are unpredictable and almost impossible to accurately quantify.16

An even more important problem with the use of cost-effectiveness criteria to determine antibiotic policies is to explain why doctors should not prescribe antibiotics to gain small potential benefits. Most antibiotics are relatively inexpensive (because most antibiotic patents have long expired).

I suggest that the contractualist approach of Thomas Scanlon (1998)17 has the potential to justify antibiotic prescribing decisions and explain why we should not use antibiotics to gain small benefits. I start from the premise that each use of antibiotics diminishes the availability of effective antibiotics by some amount and that there are few new antibiotics likely to be available for many years—making effective antibiotics a limited resource at least in the short to medium term. The specific aim of this paper is to articulate an ethically justified principle for regulating the use of antibiotics based on Scanlon's contractualist approach.

A contractualist approach to the use of antibiotics

Scanlon (1998)17 grounds his approach to moral reasoning on the idea that actions are right only if there is a principle permitting them that no one could reasonably reject. Scanlon emphasises the role of reasons for rejecting principles. “An act is wrong if and only if any principle that permitted it would be one that could reasonably be rejected by people” (Scanlon, p4)17 motivated to find generalisable principles. For Scanlon, principles that could reasonably be rejected are those that would cause a person serious hardship and where there are feasible alternatives that would not impose such burdens (Scanlon, p196)17 – this rejection comes before bargaining for mutual advantage. Moral principles must be justifiable to each person. A general feature of a contractualist framework is the rejection of the idea that multiple aggregated smaller benefits can justify serious harm to an individual (Scanlon, p238).17 When comparable harms and benefits are distributed between different people, then numbers count as tie-breakers.

It is also important to recognise that Scanlon advocates generalisable principles (Scanlon, p204).17 Particular individual (situational) preferences provide insufficient justification for a particular principle (or antibiotic treatment policy)—“our assessment cannot be based on the particular aims, preferences, and other characteristics of specific individuals. We must rely instead on commonly available information about what people have reason to want”. On the other hand “the question is whether the fact that a principle would help or hurt specific individuals can be a grounds for preferring it, and for reasonably rejecting alternatives that would not have this effect” (Scanlon, p211)17.

I propose that the dominant justification for initiating or continuing the use of antibiotics is to prevent an irretrievable adverse effect on the capabilities of individuals (usually directly related to health effects) or others (perhaps close contacts). This consideration outweighs (most) other considerations. For Scanlon (Scanlon, p224)17 “all we need take into account in deciding whether a principle could reasonably be rejected are such things as the following (my additions in italics): (a) the importance of being able to get aid (effective antibiotics) should one need it; (b) the degree of inconvenience involved in giving it, should one be called upon to do so (diagnosis and prescription); (c) the generic costs of having a standing policy of giving aid in the way this principle requires (potential reduction in antibiotic effectiveness); and (d) the generic benefits of having others have this policy (control of infectious diseases)”. I am suggesting that the principle of using antibiotics to prevent irretrievable injury in patients or their contacts cannot be reasonably rejected. This can be expressed as a principle: antibiotics should be used to prevent some substantial risk of irretrievable harm in patients or their contacts (P1). In the case of antibiotics the same individual generally gains from the control of antibiotic resistance as from the use of antibiotics. An individual alive today may not benefit today but may well benefit in the future from a policy that restricts use of antibiotics to those at significant level of risk of irretrievable adverse consequences. “What is at issue here is not the likelihood that any particular individual, given all, that is known about him or her, will be burdened, or benefited, or both by the principle, but rather the likelihood that anyone who is burdened by the principle will also benefit from it” (Scanlon, p224).17

On the other hand, a principle that we should use a limited resource to treat short-term reductions in capability (P2), or conditions with inevitable outcomes (that will not be substantially ameliorated by antibiotics) (P3), or to satisfy preferences (without evidence that use will prevent harm) (P4), or for conditions that don't respond to antibiotics, such as viral coughs and colds (P5), could be rejected. This rejection comes from the perspective of anyone who does or will need effective antibiotics to prevent irretrievable adverse consequences, and whose outcomes may be prejudiced as a consequence of antibiotic- (treatment-) resistant forms of infection. The burdens associated with not using antibiotics in each of these cases (P2-5) are not comparable with the harm consequent on treatment-resistant infection.

How much risk of an irretrievable adverse consequence justifies use of an effective antibiotic?

The principle that antibiotics should be used to prevent some substantial risk of irretrievable harm in patients or their contacts (P1) is consistent in some cases with a category to which Scanlon would apply ‘the rescue principle’. “The cases in which it would most clearly be wrong not to give aid—and most clearly unreasonable to reject a principle requiring that aid be given—are cases in which those in need of aid are in dire straits; their lives are immediately threatened, for example,…” (Scanlon, p224).17 Community-acquired pneumonia is an example of a condition in which the lives of individuals are immediately threatened and 1% mortality from community-acquired pneumonia12 is almost certainly outside of an acceptable (or tolerable) level of risk for those who will carry the more severe adverse consequences (death) of the IDSA policy, particularly while feasible more effective treatment strategies are available.

In many other cases we need to decide on the level of risk (or uncertainty) of an adverse irretrievable outcome below which an antibiotic should not be prescribed. Antibiotic treatment decisions would seem to provide a relatively straightforward case for Scanlon's contractualism (see Lenman, p10918) in that those who are burdened at any one time will also most likely be benefited at another, so we are not talking (generally) about redistributing risks. Even so Lenman (Lenman, p11418) suggests we still need to be able to take account of degrees of risk when those who carry benefits and burdens cannot be clearly distinguished. This seems to be a relevant consideration when it comes to the use of antibiotics. There is always some possibility (perhaps very small) that a patient will suffer a very serious consequence because he or she did not receive an antibiotic. This (even remote) possibility could give reason to reject almost any principle constraining the use of antibiotics. Many physicians are reluctant to impose even small avoidable risks of harm on patients.19

The United Kingdom Health and Safety Executive refer to broadly acceptable, tolerable and unacceptable risks.20 Tolerable risks are those that we are prepared to accept in order that we might accrue some benefit, for example, we are (generally) prepared to tolerate the risks associated with working in food production or with travel because of the associated benefits. The HSE offers a 1/10 000 risk of death as a guideline for a tolerable risk, that is, imposed. By accepting some small risk of an adverse outcome from infection (by constraining the use of antibiotics) we gain the benefit of sustaining effective treatment(s) when the risks are more substantial.

I argue that the level of risk at which we consider the use of an antibiotic to be justified should be above a level which approximates to the levels of risk which we tolerate in our daily lives. Use of antibiotics when risks of irretrievable harm are comparable with those that we tolerate every day can be reasonably rejected. The risks that we should prioritise for our concern are those which exceed the levels of risk associated with cooking, driving, playing sports, unprotected sex, and other activities which entail avoidable risks to oneself or others and which are a part of an everyday human life (Lenman, p10918). The principle that antibiotics should be used to prevent some substantial risk of irretrievable harm in patients or their contacts (P1) can be augmented by specifying ‘substantial risk’. This level of risk is the additional risk to individual(s), that is, potentially ameliorated by antibiotics and that falls outside the range of everyday risks that we tolerate, such as the risk of accidents.

P1 can be further developed to P6: antibiotics should be used to prevent some substantial risk of irretrievable harm in patients or their contacts, where a substantial risk is a level of risk that can be reduced by the use antibiotics, and which exceeds the range of risks of irretrievable harm that we tolerate in our day--day lives (P6).

This additional burden of risk also has to take in to account the potential risks of antibiotics themselves, such as those arising from allergy and supra-infection, and the confidence that we have in the estimate of risk. The HSE guidelines give a tolerable risk of 1 in 10 000 for an imposed risk and 1 in 1000 for a consented risk20 of irretrievable injury (death). It is intuitively plausible to suggest that the levels of risk below which use of antibiotics could be rejected falls within this range of tolerable risk. It is not the purpose of this paper to describe exactly what levels of risk these might be. These can be determined by empirical research, evidence synthesis, statistical analysis and public debate. It is important though that these levels of risk are made explicit in antibiotic policy guidelines, and are open to review.

The argument for reasonable rejection of the use of antibiotics for any condition other than those conditions associated with a significant risk of irretrievable adverse consequences is founded on the dire consequences of selecting antibiotic resistance for some. If the use of an antibiotic is unlikely to contribute to the burden of antibiotic-resistant infections, then the grounds for reasonable rejection of use would be undercut. It may be that there are some categories of patient or conditions that provide no contribution to the burden of antibiotic resistance, but it is hard to identify any current human treatment condition in which this would generally apply.

Implications of a restricted range of indications for antibiotic prescribing

If we accept the principle that antibiotics should be used to prevent some substantial risk of irretrievable harm in patients or their contacts, where a substantial risk is a level of risk that can be reduced by the use antibiotics, and which exceeds the range of risks of irretrievable harm that we tolerate in our day-to-day lives (P6) and reject P2–5, then what are the implications?

For many conditions a contractualist position would seem to be broadly aligned with current medical practice and would accept many of the current medical indications for use of antibiotics and policy guidelines based on cost-effectiveness considerations. Most current guidelines (see eg,13) exclude the use of antibiotics for conditions such as viral coughs and colds, for which antibiotics are ineffective (P5). Conditions for which there is a significant probability that some will suffer an irretrievable harm (foreseeably) preventable by use of specific antibiotic(s) include conditions such as complicated cystitis, community-acquired pneumonia, suspected neonatal sepsis, gonorrhoea, and prophylaxis for many types of surgical procedure. There is also a category of infections that, if left effectively untreated, will lead to irretrievable adverse consequences but for which there is time to change the empirical treatment options if appropriate to do so, for example, uncomplicated cystitis. Treatment of these infections empirically with antibiotics to which there are significant associated levels of antibiotic resistance could not be reasonably rejected as long as persistent infection can be effectively treated should the patient fail to respond to antibiotics.

On the other hand a contractualist position would seem to oppose the tolerance of some current antibiotic usage scenarios (P2–5 above). Many drugs, including antibiotics, are available over the counter and in countries where antibiotics are more heavily regulated are still accessible for self-administration through the internet.21 Unregulated access to antibiotics may be an important factor in the selection and dissemination of bacterial strains resistant to our remaining effective antibiotics (carbapenems).22 Unregulated use of antibiotics motivated by advertising, profit, preference or ignorance could reasonably be rejected. Sandel23 has argued that it is immoral to buy (or be entitled to buy) the right to pollute, because it allows the wealthy to evade obligations; it turns pollutants into commodities (removing the moral stigma, and it undermines a sense of shared responsibility. These arguments can equally be applied to the generation of antibiotic-resistant microbes. A principle (P2 or P4) allowing patients to pay a premium to obtain a real or perceived short-term benefit from antibiotics could reasonably be rejected if we take account of those who will suffer an irretrievable loss associated with the consequences of an increasing burden of antibiotic-resistant infections.

Preferences (P4) of relatives and healthcare workers may also sometimes drive treatment decisions to use antibiotics24 even when patient benefits are marginal or non-existent. A large proportion of patients in the terminal stages of dementia are treated with antibiotics,25 in many cases without much evidence of patient benefit (P5).26 Schwaber & Carmeli in an editorial,27 while recognising the tension between antibiotic use and resistance, state that “The solution is not to categorically deny antibiotics to the severely demented elderly, or even to impose limits on their use …Such decisions, in addition to being ethically untenable, would run counter to the expressed wishes of patients and their families”. Unfortunately, the ethical justification for this statement is unclear. A contractualist might insist that the interests of those who carry the consequences of infection with treatment-resistant infection should also be considered. Patients dying in Intensive Care Units frequently receive antibiotics right up to the time of death even when treatment is futile (P3). ‘Withdrawal of treatment’ orders may reduce the risk that patients will harbour (and potentially provide a reservoir for) antibiotic-resistant bacteria (such as MRSA) at the end of their lives.28 I am suggesting that a contractualist approach would reasonably reject unregulated access to antibiotics, use in response to the preferences of relatives or healthcare workers, or the ‘routine’ use of antibiotics for terminally ill patients—“Physicians may feel more comfortable in continuing to try to correct a theoretically reversible condition by use of antibiotics even in the face of an irreversible dying process”.24 Instead, a contractualist could insist on reasons for believing that use of antibiotics would ameliorate some substantial risk of irretrievable harm to the patient as justification in each case.

Use of antibiotics in animals

It is hard to see how animals might be given an equal status with humans such that we would use antibiotics whenever it is possible to prevent an irretrievable harm (to an animal) through the use of antibiotics. Attempting to use antibiotics to prevent irretrievable harm to any threatened animal would be both extremely (probably impossibly) demanding, and would almost certainly promote the spread and prevalence of resistant microbes and therefore be an entirely counter-productive strategy. Application of infection control precautions (see below) to wild animals (in order to control the spread of antibiotic resistance) is also implausible. Even if we were to give animals equal moral status with humans, the use of antibiotics in animal husbandry to promote the growth rate of an animal (as a food source) is hard to justify (to those who will suffer the consequences of infection with antibiotic-resistant microbes), or to compensate for managing animals in degrading and overcrowded conditions,29 or the use of antibiotics when the animal is imminently to be slaughtered for food. On the other hand, if uncontrolled infection in animals will put many others at risk of irretrievable harm (perhaps animals and humans), then use of antibiotics to prevent some from dying from animal-derived infectious diseases would be more difficult to reasonably reject.

The control of antibiotic-resistant microbes

Scanlon17 argues that probabilities should determine the extent of the precautions that we should take. As the probability of untreatable infection increases, so we should do more to control antibiotic-resistant bacteria. Currently infection control precautions to control reservoirs or the transmission of antibiotic-resistant bacteria are not routinely advocated or implemented when antibiotics are prescribed, and there are continuing concerns about the priority given to control of the spread of antibiotic-resistant microbes in some healthcare institutions even in countries which have given great emphasis to standards in infection control.30 Those who suffer adversely as a consequence of antibiotic-resistant bacteria have reason to reject tolerance of reasonably avoidable spread of antibiotic-resistant bacteria.