The Headline Said You Could Slash Dementia Risk. The Study Did Not.

Michael Beiter

Jul 31

8 min read

A lesson in how modest research becomes a miracle before breakfast

I went looking for one of the week’s most widely circulated nutrition stories. I found a small study about meal timing and cognitive tests that had been picked up by The Guardian, The Times, MarketWatch, the New York Post, and The Sun.

By the time it reached The Sun, the headline read:

“The four-hour eating rule to ‘slash dementia risk’ — AND lose 7kg in 6 months.”

Well, shit. Who wouldn’t click that?

One simple rule. Less dementia. Fifteen pounds gone. Apparently we’ve been one early dinner away from solving two of aging’s largest problems.

Except that isn’t what the study found.

The study did not measure dementia. It did not establish a reduction in dementia risk. It did not show that time-restricted eating caused the weight loss. It wasn’t even a published, peer-reviewed paper yet. It was a conference abstract describing a pilot trial of 47 women.

There may be something interesting in the research. There is also a canyon between something interesting and the promise made in that headline.

Let’s walk across it.

What the researchers actually did

The study included 47 women between 50 and 79 years old who had overweight or obesity. Everyone received counseling to reduce daily food intake by 500 calories for six months.

Twenty-six of the women were also asked to eat within a window of less than nine hours per day, generally around 10 a.m. to 6 p.m. Their average eating window was 8.2 hours. The comparison group ate across an average of 12.3 hours.

Both groups lost about 15 pounds. There was no meaningful difference in weight loss between them.

At the end of six months, the time-restricted group performed better on tests of spatial planning and problem-solving. They also appeared to make fewer errors on memory and learning tests, but those differences did not reach statistical significance. There were no significant differences in reaction time or multitasking.

The researchers described the effect as modest. They called the study preliminary and said larger trials were needed. The results were presented at the American Society for Nutrition’s 2026 conference and had not completed the peer-review process required for journal publication.

That is the study.

Forty-seven women. Six months. Similar weight loss in both groups. Improvement in some cognitive tests, not others. No dementia diagnoses. No measured change in dementia incidence. No proof that anyone’s future dementia risk had been cut, slashed, chopped, diced, or otherwise attacked with a kitchen knife.

Even the “four-hour eating rule” is slippery. The participants did not eat within a four-hour window. They ate within an eight- to nine-hour window and were asked to stop eating four hours before bed. The article explains this eventually. The headline lets you misunderstand it first.

Watch what happens to the verbs

This is where science communication often goes sideways. The numbers may remain technically recognizable while the verbs quietly mutate.

“Performed better” becomes “protects.”

“Some cognitive test results” becomes “brain health.”

“Brain health” becomes “dementia risk.”

“May” gets paired with “slash,” allowing the sentence to sound certain and uncertain at the same time.

That last move is especially useful. May provides legal and scientific cover. Slash does the emotional work. Your cautious brain receives a qualifier while your frightened brain receives a promise.

The headline never gives a number for the alleged reduction in dementia risk because the researchers never measured one. “Slash” implies a large effect without having to name it. It is statistical cosplay.

A cognitive test is not dementia

The largest distortion is the jump from performance on a few cognitive tasks to future dementia risk.

Those ideas are related, but they are not interchangeable.

Dementia is a clinical outcome involving progressive impairment that affects daily life. A short-term change on a test of spatial planning is a surrogate outcome: something researchers can measure now that may or may not predict what actually matters years later.

Surrogate outcomes are useful. We cannot run every study for 20 years and wait to see who develops dementia. But a surrogate is a clue, not a conviction.

This trial did not follow participants long enough to observe dementia. It did not report dementia diagnoses. It did not establish that the particular test improvements translate into fewer cases later. Yet the headline promotes the unmeasured outcome to center stage because “some women improved on selected cognitive tests” has all the click appeal of a municipal drainage report.

The fair conclusion is that time-restricted eating may deserve more study as one possible influence on certain aspects of cognition in this specific population.

The unfair conclusion is that a four-hour rule slashes dementia risk.

The weight-loss promise is worse

The second half of the headline promises that the rule will help you lose 7 kilograms in six months.

But every participant was counseled to eat 500 fewer calories per day, and both groups lost about the same amount of weight. The women eating across 12 hours lost just as much as the women eating across eight or nine.

That means the trial did not show that the shorter window produced additional weight loss.

The 15-pound loss is still a useful result. It just belongs to the calorie-reduction program shared by both groups, not uniquely to the timing rule featured in the headline.

The article also converts 6.8 kilograms into “just over two stone.” It is just over one stone. That mistake has nothing to do with interpreting complex statistics. It is arithmetic.

This is a common media trick: take an outcome that occurred during the study and attach it to the most novel part of the study, even when the comparison group achieved the same outcome.

If two groups take the same bus downtown, but one group wears blue hats, we have not discovered that blue hats are a transportation system.

Small studies produce unstable answers

A trial with 47 participants is not worthless. Pilot studies are supposed to be small. They help researchers test procedures, estimate possible effects, and decide whether a larger, more expensive trial is justified.

But small samples produce less precise estimates. A few unusually strong or weak performances can move the average. Results from 47 older women with overweight or obesity do not automatically apply to men, younger adults, people at a lower body weight, or anyone already experiencing cognitive impairment.

There is another issue hiding in plain sight: the researchers tested several cognitive outcomes. The restricted group did better on spatial planning and problem-solving, showed only a nonsignificant trend on memory and learning, and did not differ on reaction time or multitasking.

When several outcomes are tested, the chance of finding at least one apparently positive result increases. That does not prove the significant findings were random. It does mean we need the full paper to see the prespecified outcomes, analysis plan, effect sizes, uncertainty ranges, missing data, adherence, and any corrections for multiple comparisons.

We do not have those details yet. That is precisely why “preliminary” matters.

Peer review is not a holy purification ceremony. Published studies can be wrong too. But a conference abstract gives us much less information to interrogate. Treating it as settled guidance is buying a house after seeing one photo of the kitchen.

The distortion didn’t begin with the tabloid

It would be convenient to blame the entire mess on one overheated headline writer. The real pipeline is more complicated.

The original research team appears appropriately cautious in places, describing modest effects and calling for larger studies. But the institutional news release itself framed the work around dementia risk, even though the trial measured cognitive-test performance rather than dementia. That framing made the next leap easier.

The coverage then spread across major outlets with different levels of restraint.

The Guardian called it a preliminary study and told readers exactly who participated, which tests changed, which did not, and why clinical meaning remained uncertain. Its headline still leaned toward reduced cognitive decline, which the study did not measure directly, but the article gave readers enough information to judge the claim.

The New York Post offered “the secret to a sharper brain.” The Sun escalated to slashing dementia risk and losing 7 kilograms.

Same pilot study. Increasingly muscular verbs.

This matters because the public rarely reads the abstract. Most people do not even read the full news story. The headline becomes the finding, and the finding becomes another rule they think they should follow.

Then a client sits across from me worried that eating at 7:30 p.m. is damaging her brain, despite the fact that she worked all day, trained after work, and would like to eat dinner with her family like a normal person.

The cost of bad science communication is not merely that people know one fact incorrectly. It adds another demand to a health culture already stuffed with them.

How to read the next miracle headline

You do not need a research degree to catch most of this. Ask five questions:

What did the researchers actually measure?

Cognitive tests are not dementia diagnoses. Blood markers are not heart attacks. Muscle-protein synthesis is not long-term muscle gain.

Who was studied?

Forty-seven older women with overweight or obesity is a legitimate study population, not a substitute for humanity.

What was the comparison?

Both groups reduced calories and lost the same amount. The novel timing rule did not produce extra weight loss.

How much of the result was positive?

Some tests improved. Some showed only a trend. Others did not differ. “Mixed but interesting” is often more accurate than “breakthrough.”

What kind of evidence is this?

A small conference presentation should change your confidence a little. It should not reorganize your life by Monday.

Also watch the verbs. Associated with, suggests, and may belong to uncertainty. Prevents, protects, slashes, and reverses make much larger causal promises. If the headline’s verb is stronger than the study’s design, somebody is selling confidence the evidence did not earn.

What I would actually tell a client

An eight- or nine-hour eating window may be a perfectly reasonable tool if it fits your life. Finishing dinner a few hours before bed may help some people with reflux, sleep, food structure, or late-night grazing. The broader evidence on early time-restricted eating suggests possible metabolic benefits, although results vary and long-term adherence still matters.

But this pilot study is not a reason to believe that eating after 6 p.m. causes dementia. It is not proof that time-restricted eating prevents it. And it does not show that an eating window is superior for weight loss when calories are reduced similarly.

Try the schedule if it makes your life easier or solves a problem you actually have. Do not adopt it because a headline turned a tentative cognitive finding into a neurological protection plan.

The science is allowed to be interesting before it becomes useful. It is allowed to be promising before it becomes advice. It is allowed to leave us uncertain.

Headlines are not paid to tolerate that uncertainty. We should be.