I treat obesity. I do not treat cancer. So I owe you an explanation for why a cervical cancer paper is the most stirring thing I read this month, and why I am handing it to a newsletter built around metabolic medicine.

It comes down to one number. Zero.

In England, across the five most recent years of national data, not a single woman between the ages of 20 and 24 died of cervical cancer. Not one. Based on the death rates that held steady for the previous two decades, you would have expected around 23 deaths in that window. The expected number was 23.1. The observed number was 0.

That gap is not a rounding quirk or a screening fluke. It is what happens when you vaccinate a generation of 12-year-olds against the virus that causes the disease, and then wait.

England began offering the HPV vaccine to schoolgirls aged 12 to 13 in September 2008, with a catch-up campaign for teenagers up to 18, and added boys in 2019. Uptake was extraordinary by the standards of any vaccine program. Around 80% to 90% of girls received at least one dose before the pandemic. Most high-income countries would envy those numbers.

Peter Sasieni and Milena Falcaro at Queen Mary University of London tracked cervical cancer deaths in women aged 20 to 34 from 2001 through 2024, and lined those deaths up against vaccination coverage by birth cohort. The result, published in The Lancet, is the first robust national evidence that the HPV vaccine does not just prevent cervical cancer. It prevents dying from it.

We already knew the vaccine slashed cervical cancer incidence. Earlier work from this same group showed diagnoses falling by up to 87% in vaccinated cohorts. But incidence and mortality are different animals, and a careful skeptic could have held a real objection, which I will come back to. Mortality is the number that counts. Mortality is the one nobody had shown until now.

Look at the shape of it. The youngest women, the ones vaccinated at 12 or 13 before they were ever exposed to HPV, fell off a cliff to zero. The 25 to 29 group, vaccinated a little older and a little later, fell hard but not all the way. And the 30 to 34 group, fewer than half of whom were ever vaccinated, and most of those as older teenagers who may already have been infected, barely budged.

That staircase is the tell. When the size of an effect tracks the dose of an intervention this cleanly, across age groups, with each decline arriving exactly when that cohort's vaccinated women aged into the data, you are no longer looking at coincidence.

The headline numbers are almost hard to write down. In vaccinated women, the model estimated a 100% reduction in cervical cancer death for the 20 to 24 group (95% CI 81 to 100), and 100% for the 25 to 29 group (89 to 100). For 30 to 34, where vaccination came late and sparse, the estimate dropped to 63%, with a confidence interval wide enough to include no benefit at all. The vaccine works best precisely where the biology says it should: in girls protected before exposure.

Now I slow down, because this is observational data and you deserve the caveats with their boots on.

This is an ecological study. The authors did not have individual vaccination records linked to individual deaths. They worked from population-level coverage and population-level mortality, in five-year age bands, and they had to assume no herd protection among the unvaccinated to model the effect in the vaccinated. Those are real limitations, and the authors state them plainly.

The zero itself is the most seductive and the most fragile figure in the paper. The authors are refreshingly upfront that it is likely a chance finding sitting on top of a very low underlying rate, not proof that cervical cancer has been eradicated as a cause of death in young women. Cervical cancer death under age 25 was always rare. Push a rare event toward an effective intervention and you can land on zero by good fortune as much as by mechanism. The wide confidence intervals are the math admitting exactly this.

So why am I confident this is causal and not a mirage? Because Sasieni and Falcaro do the thing good epidemiologists do with observational data. They walk it through the Bradford Hill criteria, and it passes on every count that matters.

Temporality. Rates were flat for years, then dropped sharply the moment vaccinated cohorts entered each age band, five years later in each successive group. Dose response. The staircase above. Biological plausibility. Randomized trials already proved the vaccine prevents persistent HPV infection and the precancerous lesions that lead to cervical cancer, and a UK series found 93% of cervical cancers in women under 30 were driven by the two HPV types the original vaccine targets. Coherence. The competing explanations do not hold. Screening changes in this period would, if anything, have pushed mortality up. Sexual behavior barely shifted. Better cancer treatment would have helped all ages at once, not in a tidy five-year cascade by birth cohort.

Put together, the ecological data stop being a weak argument and become a strong one.

Here is the objection I promised. A thoughtful critic could have said: sure, the vaccine prevents cancers, but maybe it only prevents the early, screen-detectable ones we would have caught and cured anyway. The cancers that actually kill are the advanced ones in women who slip through the screening net. Population vaccination might leave those untouched, in which case incidence falls but deaths do not.

This study is the answer to that objection. Deaths dropped, not just diagnoses. The cancers being prevented include the lethal ones. The fall in mortality was, in the authors' words, as good as or better than predicted. That closes a loop that had stayed open for a decade.

One detail will interest you for a different reason. Read the acknowledgments and you find that the authors used ChatGPT to scan the literature for prior reports they might have missed, and Microsoft Copilot to draft the first version of their summary, then reviewed and edited everything and took full responsibility for the content.

A landmark Lancet paper, disclosing routine large language model use in its workflow, stated as plainly as the funding source. That is where serious medicine now sits. Not AI writing the science, but AI compressing the grunt work of search and drafting while humans keep the judgment. I have argued in this newsletter that this is the only defensible version of the technology in clinical work. It is quietly becoming the norm in the pages of the best journals in the world.

This proof arrives exactly as vaccine confidence is sliding. HPV uptake is falling in many places and hesitancy is rising, often loudest among parents weighing a shot for a healthy 12-year-old who feels invincible. I have more sympathy for that hesitation than the internet tends to reward. It is genuinely hard to vaccinate a child against a threat that is decades away and abstract.

But this is the rare case where the abstraction just became concrete. A generation of English girls got a shot, and a cancer that killed their mothers and grandmothers is, for them, approaching zero. Not managed. Not merely caught earlier. Closer to gone.

In obesity medicine, I spend my days fighting diseases that are already entrenched, with tools that work but demand a lifetime of use. It is humbling to look across the hall at a field that prevented the disease outright with a childhood vaccine. That is the cleanest win in modern medicine, and we should be shouting about it.

Zero. I had to read it three times too.

REFERENCES