The Resistance Paradox: Why Reasonable Dosing is Riskiest

A Question Never Answered by Evolutionary Biology

Andrew Read had spent his career thinking about evolution and infectious disease. It was therefore unsurprising when a gap in the field's reasoning began to trouble him. Medicine had developed an intricate body of rules for fighting infection. But when Read, an evolutionary biologist at Penn State University, examined those rules through the lens of evolutionary theory, the foundations did not always hold.

The core principle of "hit hard, hit fast, use the highest safe dose" rested on one assumption: eliminating microbes quickly would leave no survivors. For certain infections, this logic is sound. HIV treated with a modern drug cocktail is the clearest example. If a sufficiently powerful combination of antiretrovirals can kill every viral particle in the body, the evolutionary game ends before resistance has a chance to emerge.

But Read recognized that this thinking accounted for only one of two evolutionary forces operating inside an infected patient. It addressed how resistance arises; it said almost nothing about what happens to resistant microbes once they exist.

Working with Troy Day, a mathematician, statistician, and biologist at Queen's University in Ontario, Canada, Read set out to model both forces simultaneously. Their framework, published in 2016 in the journal PLOS Computational Biology, applies across a wide range of pathogens and dosing scenarios. Its conclusion is both precise and unsettling: the highest dose is not always the best option for preventing resistance. Neither, it turns out, is the instinctive middle ground.

"There is nothing in evolutionary theory that says that the dogma of hitting infections hard with high doses of medication should be the best rule of thumb to prevent drug resistance," Read said at publication. "Our analysis demonstrates that although the traditional 'hit hard' approach often works, in some cases it also can be the very worst thing to do."

The Competitor You Never Knew You Needed

To grasp why, it helps to think about what an infection actually looks like at the level of individual microbes, and why a moderate antibiotic dose can sometimes be more dangerous than a lower one.

Most bacterial infections are not composed of a genetically uniform population. Inside a patient's body, trillions of bacteria compete for space, nutrients, and resources. The overwhelming majority of those microbes are drug-sensitive and will die when exposed to antibiotics. But buried within that population, usually in tiny numbers, are mutations. Mutated microbes carry genetic variations that give them some capacity to tolerate the drug. Under ordinary circumstances, those resistant microbes are boxed in. The competition for resources is intense, and their drug-sensitive neighbors vastly outnumber them. The resistant strains cannot expand because there is no room.

Now introduce an antibiotic dose in a moderate range. That dose kills a large proportion of the drug-sensitive majority, but the resistant minority survives, and suddenly the landscape has changed. The competition that was holding resistant microbes in check has been dramatically thinned. Where there was once fierce biological pressure from millions of drug-sensitive competitors, there is now open territory: space to colonize, resources to consume, and no rivals to block the way.

Ecologists call this "competitive release." The principle is well-established in nature: remove a dominant species from an ecosystem and the populations it suppressed will surge. The same mechanism, Read and Day demonstrated, operates inside the human body. And moderate antibiotic doses may be unintentionally calibrated to trigger it.

Dr. Jose Vazquez, an infectious disease researcher and public health consultant who treats patients with resistant bacterial infections at Augusta University, put the mechanism plainly in a February 2026 interview. "When you wipe out the weak bacteria," he said, "you remove the competition. That gives the resistant microbes a clear avenue to multiply."

Vazquez was already familiar with the Day-Read model before the interview and said it reframes assumptions his field had operated under for decades. "For decades, we assumed stronger doses were met with less resistance," he said. "But evolution isn't always intuitive to us."

A Hard Pill to Swallow

What makes the model genuinely difficult to communicate, and genuinely dangerous if communicated carelessly, is where the risk peaks.

Read and Day's framework traces the probability of resistance evolving as a function of dose strength. The resulting curve does not slope smoothly downward as doses rise, in the way the "hit hard" doctrine would predict. Instead, it rises as dosing increases from the minimum effective level, crests somewhere in the moderate range, and then falls as doses approach the maximum a patient can safely tolerate. Resistance risk, in other words, is highest not at the extremes but in the middle. "The surprising finding is that intermediate dosing may backfire the most," Vazquez confirms.

The model does not conclude that low doses are always safer than high ones, or that the answer is simply to prescribe less. It makes a more specific and in some ways more demanding claim: the optimal dose for preventing resistance will fall at one of the two ends of what clinicians consider acceptable, either as high as the patient can safely tolerate, or as low as is needed to reliably clear the infection. The middle, where medicinal instinct most naturally settles, is where the danger is concentrated.

"It's going to be one of those two, but you can't just toss a coin," Read said. Which extreme is best depends on the specific pathogen, the patient, and the proportion of resistant microbes already present, all variables that must be tested case by case in clinical trials and not inferred from a general rule.

When Medicinal Guidelines are Wrong

Fleming's warning was about ignorance: patients who did not understand the drug misusing it. The Day-Read model raises a quieter and harder problem, the possibility that even careful and well-intentioned clinical practice may sometimes be producing the very outcome it was designed to prevent.

That possibility carries ethical weight. The World Health Organization estimates antimicrobial resistance already contributes to hundreds of thousands of deaths each year globally. If current trends continue, that burden could grow into one of the defining public health crises of the coming decades. In that context, evidence suggesting that standard dosing guidelines may sometimes accelerate resistance is not a theoretical curiosity. It is a potential contributor to ongoing harm.

But the ethical calculus does not run in only one direction. Undertreated infections kill patients directly and quickly: the harm is attributable and measurable. On the other hand, the harm of microbial resistance is distributed across populations and time, harder to trace to any individual prescription. Changing clinical guidelines on the basis of a mathematical model without the clinical trial evidence to validate it risks trading one source of harm for another.

Vazquez is acutely aware of this tension. He emphasized that the research should not be read as an instruction for clinicians to begin experimenting with dosing on their own judgment, nor should patients interpret it as permission to self-adjust their prescriptions. "The lesson isn't to under-treat patients," he said. "It's to recognize that there is no one-size-fits-all solution in dosing." Clinical trials, he stressed, remain necessary before any guidelines change.

He was equally pointed about the responsibilities of those who communicate the research to public audiences. "When reporting on this," Vazquez said, "it's crucial not to oversimplify it as 'take less medicine.'" In a media environment where health headlines routinely shed their qualifications in the passage from journal to social feed, the distance between a carefully worded research finding and a dangerous misunderstanding can be measured in patient outcomes.

When Moderation Becomes the Enemy

The philosophical discomfort the Day-Read model produces runs deeper than clinical practice. For millennia, Western medicine has prized moderation, and the "reasonable" dose is the moderate one. Caution counsels against extremes.

But the model performs a quiet inversion of this logic. In the evolutionary contest unfolding inside a human body, moderation does not represent careful balance: it may represent the worst of both strategies simultaneously. A moderate dose is not high enough to eliminate all competing microbes and end the evolutionary game. It is not low enough to leave the microbial ecosystem intact, with resistant strains still held in check by their drug-sensitive neighbors. Instead, it strips away the biological pressure that was suppressing resistant strains without delivering enough force to destroy them.

Fleming feared underdosing. Contemporary medicine fears overdosing. Read and Day's mathematics suggests that the place where both fears converge in the middle may be the place where resistance is most reliably cultivated.

"In evolution," Vazquez said, "balance isn't always best."

Before Clinical Trials and News Stories

There is a practical implication buried in the Day-Read framework that points toward a more immediate contribution. Rather than testing the full spectrum of doses in incremental steps, a process that is lengthy and expensive, the model suggests that well-designed clinical trials might need to compare only two conditions: the highest tolerable dose and the lowest effective one. That narrower research question is potentially faster, cheaper, and more actionable than the sprawling trials the field has historically run.

That shift in trial design, using mathematical optimization to prioritize which experiments are worth running, is part of a broader movement in medicine. Researchers in a multitude of clinical specialties are increasingly turning to mathematics to replace guesswork with measurable predictions, particularly in areas where the number of possible variables outruns the capacity of conventional trials to test them.

The Day-Read paper is, in this sense, an argument about method as much as it is about dosing. It does not hand clinicians a new prescription to follow. It gives researchers a smarter framework for designing the experiments that might eventually yield one.

For patients navigating this uncertain landscape in the present, the guidance is both simpler and more demanding than the research itself suggests. Take antibiotics exactly as prescribed. Do not stop early because symptoms improve: that risks leaving a residual population of microbes, potentially a resistant one, in an environment where the competition has already been eliminated. Patients must recognize that the prescription they receive today reflects both the best available evidence and the honest limits of it. Fleming's warning in Stockholm was about what happens when powerful medicine is misunderstood and misused. The warning in the Day-Read model is subtler: even rigorous, well-intentioned application of medical logic can be undermined by biology it does not yet fully account for.

Resistance, it turns out, is not only a microbial problem. It is an evolutionary one, and evolution does not reliably reward the reasonable, the moderate, or the instinctively safe. It rewards whatever survives. Science is slowly learning to think the same way.