Resistance is evolution, not malice
Get the subject of the sentence right first. It is not the patient who becomes resistant — a person does not become resistant to amoxicillin, and a candidate who implies otherwise has lost the station in one line. It is the bacterial population, and the mechanism is ordinary Darwinian selection running at a speed we can watch.
In any large population of bacteria there is variation. A few cells already carry a mutation, or a gene acquired from elsewhere, that lets them survive a particular drug. Before the antibiotic arrives those cells have no special advantage — they compete with billions of neighbours for the same space. Introduce the antibiotic and you remove the competition: the susceptible cells die, the survivors inherit an empty niche, and because a bacterial generation can be twenty minutes, the population that regrows is the resistant one.
So the antibiotic did not create resistance. It selected for it. That distinction is the whole topic compressed into a sentence, and it is why the language of bacteria “learning” or “fighting back” is worth dropping: it implies intent where there is only differential survival. Many of the genes involved are older than clinical medicine, because the compounds we turned into drugs were made by soil organisms competing with each other long before we arrived.
- Enzymatic destruction. The bacterium makes an enzyme that breaks the drug apart — beta-lactamases cleave the ring at the centre of penicillins and cephalosporins, and extended-spectrum and carbapenemase versions widen that to the drugs kept in reserve.
- Target modification. The molecule the drug binds to changes shape just enough that the drug no longer fits. Methicillin-resistant Staphylococcus aureus works this way, via an altered penicillin-binding protein.
- Efflux pumps. Membrane proteins that pump the drug back out faster than it accumulates, often with broad specificity, so one pump can reduce susceptibility to several unrelated classes at once.
- Reduced permeability. Changes to the outer membrane and its porin channels so less drug gets in — one reason Gram-negative organisms are intrinsically harder to treat than Gram-positive ones.
The step that turns this from a patient-level problem into a public health one is horizontal gene transfer. Bacteria do not only pass genes down to daughter cells; they exchange them sideways, across species. Plasmids — small circular pieces of DNA carrying resistance genes, sometimes several at once — move directly between cells by conjugation, bacteria take up free DNA from their surroundings, and bacteriophages carry genetic material between hosts. So a resistance gene selected in one organism, setting or country can appear in a completely unrelated pathogen somewhere else.
That is why exposure matters even when it is not therapeutic. An unnecessary course applies selection pressure for no clinical gain, and sub-therapeutic exposure — enough drug to kill the most susceptible cells but not enough to clear the infection — is a particularly efficient way to leave the hardier ones behind. One nuance is worth holding: the familiar advice to always finish the course was challenged in a 2017 BMJ analysis, on the grounds that its evidence base is weaker than its confidence suggests, and the argument is still live. The safe position is that patients should take a course as prescribed, and that trial evidence has been moving towards prescribing shorter courses in the first place.
- Selection pressure
- Any force that makes survival differ between variants in a population. An antibiotic is a very strong one: it kills the susceptible majority and leaves the resistant minority the whole environment.
- Horizontal gene transfer
- Movement of genetic material between bacteria other than by inheritance — conjugation, transformation or transduction. It lets resistance cross species boundaries.
- Plasmid
- A small circular DNA element separate from the chromosome, often carrying several resistance genes together, and readily transferred between cells.
- Broad and narrow spectrum
- How many bacterial types a drug covers. Broad-spectrum agents are useful when the organism is unknown, but they select across a far wider range of the patient’s own flora.
- Antimicrobial stewardship
- The coordinated effort to make prescribing more selective — right drug, dose, route and duration — in order to preserve the drugs that still work.
- AWaRe
- The World Health Organization’s classification of antibiotics into Access, Watch and Reserve categories, used to steer prescribing towards first-line agents and protect the last-line ones.
The antibiotic selects, it does not create
If you retain one line from this article, retain that one. Resistance is not a punishment for misuse and bacteria are not adapting on demand; every antibiotic course, appropriate or not, applies pressure that favours the cells already able to survive it. The clinical question is therefore not only “is this drug safe to give this patient” but “is the benefit to them worth the pressure it applies”.
Explain antibiotic resistance to a patient in thirty seconds.
This is a standard explanation station, and it is marked on clarity and on the absence of jargon, not on depth. Something like: “Antibiotics kill bacteria, but in any large group of bacteria a few happen to survive. When we use an antibiotic, we clear out all the ones it works on and leave the survivors room to multiply — so over time the drug stops working, not just for you but for everyone. It is not that your body gets used to it; it is that the bacteria that are left are the tough ones.” Then check understanding rather than moving on. Avoid “superbug”, avoid “immune”, and do not open with a global statistic — you are talking to one person about their own illness.
The scale, and how to state it honestly
The World Health Organization consistently identifies antimicrobial resistance as one of the leading threats to global health, and it has been the subject of a United Nations General Assembly high-level meeting more than once, which is unusual for a clinical problem. The organisation maintains a priority pathogens list to steer research funding, including carbapenem-resistant Gram-negative bacteria, drug-resistant tuberculosis and increasingly resistant gonorrhoea.
On mortality, describe magnitude rather than reciting a figure you cannot defend. A large international modelling study published in 2022 estimated that bacterial resistance was directly responsible for over a million deaths in 2019, and was a contributing factor in several million more, placing it among the largest causes of death worldwide. Both are estimates built from surveillance data of very variable quality, with wide uncertainty intervals. Describing the order of magnitude is accurate; quoting a number to three significant figures is not.
The number candidates reach for comes from the O’Neill Review, commissioned by the UK government in 2014, chaired by the economist Jim O’Neill and reporting in 2016. It projected that on current trajectories resistance could cause around ten million deaths a year by 2050, alongside an enormous cumulative economic cost. That is the most influential number in the field and it is entirely reasonable to cite, provided you frame it as what it is: a modelled projection, extrapolated from a limited set of organisms and built on assumptions about how resistance rates will rise. Researchers have criticised it as an upper-bound scenario, and the review was commissioned partly to make the political case for action. Its value was as a policy lever, and by that measure it worked.
Do not present a projection as a fact
“Ten million people will die a year by 2050” is the single most over-quoted line in medical interviews, and it is usually delivered as though it were a measurement. An interviewer who knows the field will ask where the number comes from, and a candidate who cannot say “it is a 2016 modelled projection from the O’Neill Review, and it has been criticised for its assumptions” looks like someone reciting rather than reasoning. Saying the number and then qualifying it yourself is a considerably stronger move than either quoting it flat or avoiding it.
What actually drives it
Over-prescribing in primary care is the driver candidates name first, and the interesting part is why it happens, because “lazy GPs” is both wrong and unattractive. Work commissioned by public health bodies in England and published in 2018 concluded that a substantial minority of primary care prescriptions — on the order of a fifth — were not clinically justified. But a GP has ten minutes, no imaging and no blood results, and is separating a self-limiting viral illness from early bacterial infection on clinical judgement alone. The asymmetry is the point: the cost of not prescribing when you should have is immediate, visible and personal, while the cost of prescribing when you should not is diffuse and lands on someone else years later. Add patient expectation, and the fact that saying no properly takes longer than writing the prescription, and the behaviour is explicable.
Agriculture accounts for a large share of global antibiotic use by volume, historically as growth promoters — sub-therapeutic doses that make animals put on weight faster — and as routine prophylaxis for intensively farmed livestock rather than treatment of diagnosed disease. Growth promoters have been banned across the European Union since 2006, the UK retains that position, and UK veterinary antibiotic sales have fallen substantially since 2014, but practice varies worldwide. The illustration worth having is colistin, an old and toxic drug held in reserve as a last resort in humans yet for years used heavily in livestock production in some countries: a plasmid-borne colistin resistance gene was identified in China in 2015 and reported from multiple countries within months — horizontal gene transfer crossing the animal-to-human boundary, documented close to real time.
Two further drivers matter and are often missed. In many countries antibiotics are available over the counter without a prescription, so courses are self-selected, frequently wrong for the organism and often stopped when the money runs out. And substandard or falsified antimicrobials, which are a serious problem in weakly regulated markets, deliver exactly the sub-therapeutic exposure that selects resistance most efficiently.
Infection prevention sits underneath all of it. Poor sanitation, unsafe water, overcrowding and weak hospital infection control produce more infections, and therefore more antibiotic courses, while spreading resistant organisms between patients. There is a counterweight a strong candidate raises unprompted: on widely cited estimates, lack of access to effective antibiotics kills at least as many people as resistance to them does, particularly children with pneumonia and newborns with sepsis in low-income countries. Framing this purely as “use less” is a high-income framing of a problem whose worst burden falls on people who cannot reliably get the drugs at all. Health inequalities sets up the same pattern domestically.
The broken market, and the UK’s answer
Most recent antibiotic approvals are refinements of existing classes rather than genuinely new ones, and the WHO has repeatedly described the clinical pipeline — particularly against resistant Gram-negative organisms — as insufficient to meet the need. A handful of novel agents have reached approval recently, but the decades-long trend is a pipeline that thinned while resistance did not.
The reason is commercial rather than scientific, and explaining it well is the most distinctive thing you can do with this topic. Every other drug class rewards volume: a statin taken daily for twenty years generates revenue proportional to how widely it is used. A new antibiotic that works against a resistant organism will be deliberately locked away and used as little as possible, precisely because it works. The better the drug, the more determined stewardship will be to reserve it. Development costs match any other drug; the revenue does not. Large companies have largely withdrawn from the field, and small biotechnology firms have gone bankrupt shortly after bringing a new antibiotic to market — the American company Achaogen did exactly that in 2019, a year after its drug was approved.
Policy responses divide into two kinds, and the vocabulary is worth a lot in the room. Push incentives subsidise discovery up front — public grants and international partnerships that fund early development so a company is not risking its own capital. Pull incentives guarantee a reward at the end, decoupled from how much is sold: market entry rewards, extended exclusivity, and subscription payments.
The UK-specific answer is the last of those, and it is the detail that makes an examiner sit up. NHS England and NICE developed a subscription model — sometimes described as paying like a streaming service — in which a company receives a fixed annual fee for making an antimicrobial available to the NHS, valued by NICE according to the drug’s importance rather than the number of packs dispensed. It ran as a world-first pilot covering two antimicrobials before expanding into a broader scheme from 2024. The logic is that it decouples the two: the manufacturer is paid the same whether the drug is used once or never, so within the NHS the commercial and stewardship incentives no longer pull against each other. If the commissioning structures are unfamiliar, how the NHS works covers who NICE and NHS England are.
Is the subscription model the right fix?
The case for it
- It solves the specific market failure: revenue is decoupled from volume, so a company can profit from a drug that stewardship keeps on the shelf.
- Payment tracks an assessment of clinical value rather than sales, which is a more defensible basis for reward.
- It gives the NHS guaranteed access to reserve antimicrobials rather than depending on a manufacturer finding the market worth serving.
- It is a concrete, testable policy in a field dominated by declarations of intent, and other countries have watched it closely.
The case against
- The sums are small relative to the cost of developing a drug, so one country’s scheme is unlikely to revive a global pipeline alone.
- A pull incentive only works at scale if several large markets run parallel schemes, and coordinating that has proved slow.
- Valuing a drug you hope never to use is methodologically hard, and the assessment rests on modelled future benefit.
- Within a fixed NHS budget, money spent here is money not spent on services with demonstrable present-day benefit.
- It does nothing about access in the countries where untreated, rather than untreatable, infection kills the most people.
Why have so few new antibiotics been developed?
Answer the economics, not the science, because that is where the actual answer lies. Say that discovery is genuinely hard — the easy compounds were found decades ago and Gram-negative bacteria are difficult to penetrate — but that the binding constraint is commercial. Then deliver the core line: antibiotics are the only drug class where success means using the product as little as possible, so the better a new agent is, the less it will be sold. Name the consequence (companies exiting the field, firms going bankrupt shortly after approval) and then show you know there is a response: push incentives that fund the research, pull incentives that reward the outcome, and the NHS subscription model as a UK example of the second. Finishing with a criticism of that model is what turns a good answer into a distinctive one.
Stewardship, concretely
Stewardship is where the topic becomes something a doctor does on a Tuesday rather than something a committee announces. The UK framework taught to foundation doctors is usually summarised as starting smart and then focusing, and it maps onto a sequence you can walk an interviewer through.
- Decide whether this is bacterial at all, and whether it needs an antibiotic if it is. Colds are viral, most coughs and sore throats are too, and no antibiotic touches a virus; other common presentations such as earache are usually self-limiting even when bacteria are involved. Tools such as FeverPAIN and Centor make that judgement in a sore throat less arbitrary than a hunch.
- Take samples before starting where you reasonably can. Cultures taken after the first dose may grow nothing, and you have lost the information that would have let you narrow treatment later.
- Start smart: the narrowest agent covering the likely organism, at a dose high enough to clear rather than merely select, by the appropriate route, following local guidelines because local resistance patterns differ.
- Document the indication, dose, route and a review or stop date at the moment of prescribing. An antibiotic with no stop date continues by inertia.
- Review at 48 to 72 hours, once cultures and the clinical response are back, recording one of a small set of outcomes: stop, switch intravenous to oral, change the agent, continue with a further review date, or move to outpatient parenteral therapy. This is the step most often skipped, and where most unnecessary days of treatment are removed.
Two other tools belong in the answer. The first is the delayed or back-up prescription — one the patient holds and collects only if things have not settled after a few days — which NICE recommends for several self-limiting infections and which reduces how many courses are actually taken while still giving the patient a plan. The second is better diagnostics: point-of-care C-reactive protein testing in lower respiratory tract infection, rapid tests for influenza and group A streptococcus, and molecular tests that name an organism and its resistance genes hours rather than days ahead of culture. Much of over-prescribing is a diagnostic uncertainty problem, so faster tests are as much of the solution as restraint.
The angle most candidates miss is vaccination. An infection that never happens needs no antibiotic, so vaccination reduces resistant disease directly and reduces the prescribing that follows infection: pneumococcal conjugate vaccination cuts invasive disease including resistant strains, and influenza vaccination cuts the secondary bacterial infections and inappropriate prescriptions that come with them. Alongside it sit the unglamorous infection control basics — hand hygiene, catheter and cannula care, isolating colonised patients, clean water and sanitation. Note too that stewardship is not a doctors-only activity: antimicrobial pharmacists, microbiologists and infection control nurses lead most of it in practice.
The patient in front of you, and the patient who does not exist yet
This is where the marks are, and it is why the topic sits alongside assisted dying and gene editing rather than in a facts-and-figures section. An antibiotic prescription is one of the few decisions in medicine where the harm of over-treatment falls mainly on somebody other than the person receiving it. That makes it a clean, examinable collision, and running it through the four pillars takes fifteen seconds.
Beneficence: an antibiotic for a viral illness offers this patient no meaningful benefit, so the pillar people assume is on the patient’s side is largely silent. Non-maleficence: it carries real individual harms — rash, diarrhoea, Clostridioides difficile infection after the gut flora is disrupted, occasionally anaphylaxis — plus the subtler harm of medicalising a self-limiting illness, because a prescription now teaches the patient to return for one next time. Autonomy: a patient with capacity may refuse any treatment, but autonomy is not a right to demand one that is not indicated, and GMC guidance is clear that a doctor is not obliged to provide a treatment they judge to be of no clinical benefit. Justice: the person who bears the cost is a future patient with an infection that no longer responds, and they are not in the room.
Say the asymmetry out loud, because it is the honest part. The patient is real, present, unwell and unhappy; the future patient is statistical and absent. The doctor personally carries the consequence of missing a bacterial infection, and almost none of the consequence of one more unnecessary course. Recognising that the incentives push against the right decision distinguishes a candidate who has thought about this from one who has read about it.
Two ways to fail this station
The first is to cave — to prescribe in order to end an uncomfortable consultation, which is exactly the behaviour the station is testing for. The second, and more common among well-prepared candidates, is to lecture: to answer a person who feels genuinely ill with a speech about global resistance and ten million deaths. Both score badly. The mark goes to the candidate who takes the illness seriously, explains once and briefly why an antibiotic will not help this particular illness, and offers something real instead.
A patient comes in with a cold and insists on antibiotics. What do you do?
Do not lead with the refusal. Start by finding out what is driving the request: how long they have been unwell, what they are worried it might be, and whether something specific hangs on being better by a particular day — a flight, an exam, a child they cannot afford to be too ill to look after. Legitimise the illness, because feeling dismissed is what turns a request into a demand. Then explain once, briefly and without jargon, that this is viral and an antibiotic will not shorten it or make them feel better. Offer what you can do: a realistic timeline (a cough after a viral illness commonly lasts three weeks or more), symptomatic advice, and clear safety-netting — the specific things that would mean coming back sooner. A delayed prescription they collect only if things have not improved is a genuine middle path, not a fudge. Say explicitly what you would not do: prescribe to end the consultation. Concede the boundary — if they were immunosuppressed, or there were red flags, the answer changes. Then name the tension: beneficence towards one patient against justice towards patients who are not in the room, and it is hard because only one of them can argue with you. Role-play stations covers the delivery, which is marked separately from the reasoning.
The UK’s current five-year national action plan on antimicrobial resistance covers 2024 to 2029, and sits under a longer-term ambition to contain and control resistance by 2040, and it is a One Health plan: human medicine, veterinary medicine and the environment addressed together rather than in isolation. That framing is worth borrowing for a closing line, because it captures what makes this topic different — no single prescriber, hospital or country can solve it alone, and every one of them can make it worse. For breadth across the remaining current issues, the hot topics guide is the last stop on this path.