An Arms Race Against Evolution: Why the Mathematics of Antibiotic Resistance May Determine Medicine's Future

In 2023, the Centers for Disease Control and Prevention estimated that antibiotic-resistant infections cause more than 35,000 deaths in the United States each year, with millions more experiencing infections that require more toxic, more expensive, or less effective treatments as a result of resistance. The global toll is far larger. And unlike many public health crises, this one is not the result of a novel pathogen arriving from outside the existing medical system. It is, at its root, a mathematical problem — a consequence of evolutionary dynamics playing out across billions of bacterial cells in hospitals, farms, and communities worldwide.

Applied mathematicians have been building models of this process for decades. The question increasingly being asked within both the infectious disease research community and mathematics departments is not whether the models are sophisticated enough, but why the gap between mathematical insight and clinical practice remains so stubbornly wide.

Evolution in Real Time

Understanding why antibiotic resistance is a mathematical problem requires a brief account of the biology. Bacteria reproduce rapidly — a single Escherichia coli cell can divide every twenty minutes under favorable conditions — and each replication carries a small probability of mutation. In a bacterial population numbering in the billions, even rare mutations occur with statistical regularity. When a mutation confers resistance to an antibiotic, natural selection acts swiftly and decisively: resistant cells survive and reproduce while susceptible ones are killed. Within a matter of days, a population that was once overwhelmingly susceptible can become dominated by resistant strains.

This is Darwinian evolution operating on a timescale humans can observe directly, and it is precisely the kind of process that population dynamics mathematics was developed to describe. The foundational tools are systems of ordinary differential equations — the same mathematical framework used to model predator-prey interactions, tumor growth, and epidemic spread. In the antibiotic resistance context, these equations track the sizes of susceptible and resistant bacterial subpopulations over time, incorporating parameters for growth rates, death rates under antibiotic exposure, mutation rates, and the fitness costs that resistance mutations sometimes impose on bacterial cells.

What the Models Reveal

The insights generated by mathematical models of resistance are not merely descriptive. They have produced genuinely counterintuitive predictions that have reshaped how researchers think about treatment strategy.

One of the most significant involves the concept of competitive release. When a patient is treated with an antibiotic, the drug preferentially kills susceptible bacteria, which in a mixed population had previously been outcompeting resistant strains for nutrients and space. By eliminating the susceptible competitors, antibiotic treatment can inadvertently accelerate the growth of the resistant subpopulation — a phenomenon that standard clinical intuition, focused on killing pathogens rather than managing ecological dynamics, tends to overlook.

Mathematical models have also formalized the concept of adaptive therapy — a treatment strategy in which antibiotic doses are calibrated not to eradicate all bacteria as rapidly as possible but to maintain a controlled population of susceptible cells that suppresses resistant ones through competition. Originally developed in the context of cancer treatment by researchers at the Moffitt Cancer Center in Tampa, Florida, adaptive therapy frameworks have been extended to bacterial infections with promising early results in both computational and animal models. The core mathematical principle is a form of evolutionary game theory: the goal is not to win every battle against the pathogen but to manage the competitive dynamics between bacterial subpopulations in a way that preserves treatment options over time.

Differential equation models have further illuminated the population-level dynamics of resistance spread — how resistant strains move between patients in hospital settings, how antibiotic use in agriculture contributes to the community-level reservoir of resistance genes, and how treatment guidelines in one region affect resistance patterns in another. These are fundamentally mathematical questions about coupled dynamical systems, and the answers have direct implications for public health policy.

Where the Models Break Down

Despite these advances, the translation of mathematical insights into clinical practice has been slow and frequently disappointing. The reasons are multiple and instructive.

First, the biological systems being modeled are substantially more complex than the equations used to describe them. Real bacterial populations are not well-mixed homogeneous communities; they form spatially structured biofilms, colonize heterogeneous tissue environments, and interact with immune system dynamics that are themselves extraordinarily difficult to model. Parameters estimated from laboratory cultures frequently fail to capture behavior in living patients, and the gap between in vitro model calibration and in vivo clinical reality has undermined the predictive accuracy of even sophisticated differential equation frameworks.

Second, clinical medicine operates under constraints that mathematical models rarely incorporate. Physicians must make treatment decisions in real time, without access to the detailed population-level data that adaptive therapy protocols require. A strategy that is mathematically optimal under the assumption of perfect monitoring may be operationally impossible in a busy hospital setting. The feedback loop between model prediction and bedside application is far slower and noisier than the models assume.

Third, and perhaps most importantly, the collaboration between applied mathematicians and infectious disease clinicians has historically been limited by cultural and institutional barriers. Mathematicians are trained to value generality and formal rigor; clinicians are trained to value practical applicability and empirical validation. These are not incompatible values, but bridging them requires sustained interdisciplinary engagement that the structure of American academic medicine does not consistently support. Research funding, publication incentives, and career advancement pathways in both mathematics departments and medical schools tend to reward work within disciplinary boundaries rather than across them.

Building the Collaboration

There are signs of progress. The National Institutes of Health has increasingly emphasized quantitative approaches to infectious disease research, and programs such as the Models of Infectious Disease Agent Study network have funded mathematical modeling work with direct clinical relevance. Several academic medical centers, including those affiliated with Johns Hopkins University and the University of Michigan, have established formal collaborations between mathematics faculty and infectious disease divisions.

For students and early-career researchers, the antibiotic resistance problem represents a field where mathematical training translates directly into work of urgent public health significance. The core mathematical skills required — differential equations, dynamical systems theory, stochastic modeling, parameter estimation, and sensitivity analysis — are standard components of applied mathematics curricula. What is less standard, and what the field genuinely needs, is training that also encompasses the biological and clinical context in which those tools must operate.

Graduate programs that combine rigorous mathematical training with substantive engagement with microbiology and infectious disease epidemiology are still relatively rare in the United States, but their graduates are in high demand. Positions at the CDC, the NIH, academic research centers, and pharmaceutical companies developing novel antimicrobial therapies all draw on precisely this combination of skills.

The Stakes of Getting It Right

The World Health Organization has identified antimicrobial resistance as one of the top global public health threats of the twenty-first century. Some projections suggest that without effective intervention, resistant infections could cause more deaths annually than cancer within several decades. Those projections are themselves the products of mathematical models — and they are contested, uncertain, and dependent on assumptions that the research community continues to debate.

What is not contested is that the biological arms race between human medicine and bacterial evolution is accelerating, and that the window for developing mathematically informed treatment strategies that can stay ahead of resistance is narrowing. The differential equations exist. The dynamical systems frameworks are being refined. The question is whether enough mathematically trained researchers will choose to engage with the full complexity of the clinical problem — not just the elegance of the model — to make those tools count when it matters most.