Fleming's Warning

Alexander Fleming received the Nobel Prize in 1945. In his lecture he described a scenario.

A man with a sore throat buys penicillin and takes too little of it. The organisms in his throat are exposed to a dose that kills the susceptible ones and leaves the hardier ones alive. He infects his wife, who develops pneumonia from bacteria that penicillin no longer touches, and dies.

Fleming's conclusion: the thoughtless person playing with penicillin treatment is morally responsible for the death of the man who succumbs to infection with the penicillin-resistant organism.

This was said at the moment of triumph, by the person with the most reason to celebrate. The subsequent eighty years are therefore harder to excuse as ignorance.

The reason antibiotics work at all is selective toxicity: they attack something bacteria have and we do not. Penicillin blocks the construction of the peptidoglycan cell wall, which human cells do not possess. Others target bacterial ribosomes, which differ from ours enough to be distinguished. The bigger the difference between the target and anything of yours, the safer the drug.

This also explains why antivirals and antifungals are so much harder. A virus uses your own cellular machinery, so there is little that belongs to it alone to attack. A fungus is a eukaryote with cells uncomfortably similar to yours, which is why antifungal drugs have such unpleasant side effects.

Resistance is not caused by the drug, and this is the most commonly misunderstood point.

Bacteria do not mutate in response to antibiotics. Variation in susceptibility already exists in a population of billions, and the drug sorts it, killing the susceptible and leaving the rest to repopulate. Resistance genes have been recovered from bacteria in permafrost thousands of years old and from isolated cave systems never touched by human medicine, because these are ancient compounds in an ancient war between microbes.

What makes it fast is that bacteria trade. Horizontal gene transfer lets them exchange plasmids carrying resistance genes across species boundaries, so resistance arising in one organism can spread to unrelated ones without any reproduction at all. A resistance gene acquired in a farm animal's gut can end up in a human pathogen.

Which brings us to where most antibiotics actually go.

In many countries, the majority of antibiotic use by volume is agricultural rather than medical, given to healthy animals at low doses, because it promotes growth and prevents disease in crowded conditions. Low-dose continuous exposure across enormous populations is close to a purpose-built apparatus for selecting resistance, and the resulting organisms do not stay on the farm.

Several countries have restricted growth-promotion use, and measurable falls in resistance followed within a few years. The mechanism is that resistance usually carries a fitness cost: maintaining an efflux pump, an altered target or a degrading enzyme takes energy, so resistant strains are at a disadvantage when the drug is absent. Remove the pressure and susceptible strains recover ground.

That is slow and partial, and compensatory mutations can reduce the cost over time, and it is genuinely reversible, which is the basis for hospital stewardship programmes and antibiotic cycling.

The hardest part of the problem is not biological.

A new antibiotic is taken for a week, cures the patient, and is then deliberately held in reserve for resistant cases in order to preserve its usefulness. Every incentive the pharmaceutical market provides points the other way: the profitable drug is one taken daily for decades by a large population.

So the economics of antibiotics are close to inverted. Most large companies have left the field. Several small firms that successfully brought new antibiotics to market went bankrupt doing so: Achaogen developed a drug for multi-drug-resistant infections, got it approved, and filed for bankruptcy within a year.

This is a clean case of a market failing to supply something that is badly needed, because the thing's value lies mostly in its not being used. The proposed remedies are all attempts to break the link between volume and revenue: subscription models where a health service pays an annual fee for access regardless of usage, prizes for successful development, and extended exclusivity on other products. The UK has trialled the subscription approach.

None of it has yet produced a pipeline resembling what the problem requires, and the resistance continues to accumulate on the schedule Fleming described in 1945.