The Twilight of Antibiotics: we publish a model that estimates how much time they have left

A threat that does not grow: it accelerates

Problems are usually pictured as straight lines. Antimicrobial resistance does not behave that way. The most troubling figure is not how much it grows, but how: it does not rise at a constant rate, it raises the rate at which it rises.

That distinction changes everything. A phenomenon that is merely exponential is already a concern. One in which the rate itself accelerates, known as super-exponential growth, reaches the point where first-line antibiotics become ineffective sooner. The study The Twilight of Antibiotics was built on that observation.

The scale of the problem

The available figures are already severe, and probably conservative:

• Antimicrobial resistance caused around 1.27 million attributable deaths in 2019 (Murray et al., The Lancet).
• In many developing countries those deaths go unrecorded or are classified differently, so the real figure is underestimated.
• It is not confined to countries with high antibiotic consumption: it is global. From Mycobacterium tuberculosis to Neisseria gonorrhoeae, virtually every microbe under selective pressure evolves.
• Published projections reach some 1.91 million annual deaths by 2040 (Tai et al., 2025) and tens of millions by 2050 (O'Neill Review).
• Meanwhile, the discovery of genuinely new antibiotic classes has stalled: none has appeared since 1987.

What is at stake

Losing antibiotic effectiveness does not only complicate the treatment of pneumonia or a urinary tract infection. It puts at risk every medical practice that depends on controlling opportunistic infections.

Oncology and transplant medicine work with immunosuppressed patients, often because of the myelotoxic drugs used in treatment itself. Major surgery, intensive care and dentistry equally depend on being able to prevent and treat infection. Without antibiotics that work, much of modern medicine regresses. The study raises the risk of a reverse epidemiological transition: a partial return to pre-antibiotic conditions.

The main actors: the ESKAPEE pathogens

Much of the threat is concentrated in a group of pathogens particularly adept at developing resistance, known by the acronym ESKAPEE: Enterococcus faecium, Staphylococcus aureus, Klebsiella pneumoniae, Acinetobacter baumannii, Pseudomonas aeruginosa, Enterobacter and Escherichia coli.

Several of their resistant phenotypes are now routine in hospitals: MRSA (methicillin-resistant S. aureus), VRE (vancomycin-resistant enterococcus), extensively resistant K. pneumoniae and carbapenem-resistant P. aeruginosa. Resistance to carbapenems and third-generation cephalosporins has been widely documented since 2020, and is accelerated by horizontal gene transfer through plasmids (KPC, NDM, OXA-48), which moves resistance from one bacterium to another without each having to develop it independently.

How the model was built

At the core of the study is a closed-form mathematical model. In simple terms, it is a generalised logistic curve with a quadratic term added to the exponent. That term is the simplest way to express that the rate of increase is not fixed, but grows steadily over time.

The contrast is worth noting. Resistance is usually modelled as discrete geometric growth or as a continuous exponential, both at a constant rate. This model starts from a different premise: that rate is itself growing. When more weight is given to that acceleration, the time remaining until critical ineffectiveness shortens. To allow comparison, the platform also plots a constant-rate logistic curve as a reference.

Rigour and reproducibility

A forecast of this magnitude is only useful if it can be audited. The model was therefore built on three deliberate decisions:

• Deterministic, with no randomness. No parameter fitting and no random sampling. The output is identical on any computer, and every figure reproduces exactly.
• Anchored to evidence. Coefficients are calibrated to published milestones, drawing on more than ten peer-reviewed sources (Lancet/GBD, O'Neill, GRAM, Tai 2025, Magiorakos 2012, WHO GLASS).
• Back-tested. The model was retro-validated: calibrated with data up to 2015, its prediction was then compared blind against observations from 2019 to 2025. For the short term it also incorporates a time-series analysis (SARIMA) with its diagnostics.

That commitment to verification extends to the software: it is open, MIT-licensed and deposited in Zenodo with its own citable identifier, following the FAIR principles for research software and the Force11 citation guidelines.

What the model projects

The resistance pressure index, a composite metric from 0 to 100 that condenses different resistance signals into a single trajectory, follows a clear path: it starts at around 70 in 2025, sharpens towards 90 around 2040 and approaches values of 95 to 98 between 2040 and 2047. The study interprets that range as the threshold of critical ineffectiveness against gram-negative nosocomial infections.

Put differently, the model places the critical point around 2040, that is, some fourteen years away, plus or minus three. Between 2045 and 2060 the curve slowly approaches near-total ineffectiveness across all classes. And this is not an abstract forecast: in parallel, mortality from carbapenem-resistant pathogens, grouping Enterobacterales, A. baumannii and P. aeruginosa, climbs steadily through 2035.

A possible escape route: the metabolic hypothesis

Faced with that scenario, the study does not stop at diagnosis. It describes a possible metabolic escape route, the line of research of Ernesto Prieto Gratacós.

The logic is a change of terrain. Instead of attacking the same molecular target that bacteria learn to evade time and again, the proposal is to interfere with the metabolism they depend on to grow. Many ESKAPE pathogens behave as facultative anaerobes, with a marked dependence on glycolytic pathways (GLUT-type transporters, hexokinase-2, lactate dehydrogenase A). That dependence makes them vulnerable to enzyme inhibition by structural analogues of glucose, an approach made viable by the marked functional asymmetry between normal eukaryotic cells and bacteria, and one that repurposes inhibitors already characterised in oncology.

Among the candidates the paper discusses are 2-deoxy-D-glucose, a non-degradable pseudo-sugar with an antiproliferative effect, and sodium ascorbate, an analogue of glucose at its carbonyls, whose antimicrobial and antineoplastic effect by intravenous route is widely recognised. The effect may be enhanced with autophagy inhibitors such as hydroxychloroquine. The underlying reasoning is that these antimetabolic interventions have retained their effectiveness across decades: by restricting the energy and biosynthetic substrate, fast-growing bacteria do not tolerate prolonged starvation, unlike antibiotic pressure, which selects for resistance.

It bears repeating: this is a direction of research, presented as such in the paper, not an available treatment or medical advice.

The platform is open and auditable

None of this asks for an act of faith. The entire body of work lives on an open, interactive platform, resistome.imhoit.com, organised into sections that mirror the structure of the paper:

• An overview with the super-exponential curve, its constant-rate reference and the critical window 2040-2047.
• A map of ESKAPEE pathogens by antibiotic class, with World Health Organization priorities and multidrug-resistance categories.
• A time-series section with SARIMA forecasts and their diagnostics.
• A section on industry and bibliometrics: the growth of publications on resistance and the divergence between awareness of the problem and available effectiveness.
• A section devoted to the metabolic line, presented as a working hypothesis.

Anyone can move the assumptions, export the figures at publication quality and audit every data point.

Limitations and scientific honesty

The study is explicit about its limits. The pressure index is a conceptual composite, not a single measured quantity, so it cannot be validated against one observable. Global aggregation masks considerable regional heterogeneity: an index of 90 on one continent may coexist with 40 on another. And the metabolic route, however promising its biochemical basis, is a line that warrants prospective controlled trials to consolidate the evidence.

Publication and presentation

The paper was published in the British Journal of Medical and Health Research (2026, volume 13, issue 8, pages 43 to 56) and will be presented at the II International Congress of Biological, Functional and Regenerative Medicine, and Biological Dentistry (BIOMINDS 2026), in Buenos Aires.

In the authors' words

We did not write this to frighten anyone, but to put a date on the table. If the forecast is right, the window to change strategy is shorter than we tend to assume.

Ernesto Prieto Gratacós

We wanted no one to have to take our word for it. The model is deterministic and open: anyone can run it and arrive at exactly the same result.

Julio Botto, IMHOIT

Frequently asked questions

Is this alarmism?

No. It is the trend shown by published data, modelled reproducibly. It is published precisely so that action can be taken in time.

Is the metabolic route already a treatment that can be requested?

No. It is a line of research described in the paper, a direction to explore. It is not an available treatment or medical advice.

Is the data invented or simulated?

No. Mortality and surveillance values come from cited sources, and the model is deterministic: every figure reproduces from the open code.

Why «super-exponential» and not simply «exponential»?

Because it is not only resistance that increases, it is the speed at which it increases. That detail brings the critical point forward.

What margin of uncertainty does the forecast carry?

The model assumes an uncertainty of plus or minus three years and was tested against observations from 2019 to 2025.

How to explore and cite it

• Interactive platform: resistome.imhoit.com
• Paper (open access): 10.5281/zenodo.21898960
• Software (code and data): 10.5281/zenodo.22117640

Suggested citation, Vancouver style:

Prieto Gratacós E, Botto J. The Twilight of Antibiotics: a predictive mathematical model of declining antimicrobial effectiveness, and a possible metabolic escape route. Br J Med Health Res. 2026;13(8):43-56. doi:10.5281/zenodo.21898960

The authors