Same Diet, Two Different Brains: The Alzheimer's Trial That Split People by Their Metabolism First

A six-month randomized trial in Istanbul put 60 people with early to mid stage Alzheimer's on a plant-rich ketogenic diet with intermittent fasting. The cognitive gain sat with the 19 participants whose glucose handling was intact, and much less with the 11 whose metabolism was already struggling. The subgroup is tiny. The question it raises is not.

A tape measure and a fasting blood draw decided something in this trial. Before anyone discussed olive oil or a twelve-hour eating window, the researchers weighed people, calculated BMI, and tested for insulin resistance, and those numbers sorted participants into metabolic boxes. Then came the finding that makes the whole thing interesting: within the group that got the diet, the people without a glycotoxic profile (n = 19) showed a higher change in their Mini Mental State Examination score than the people who had one (n = 11).

Read that twice. Insulin resistance and excess glucose doing damage to the brain is precisely the problem a ketogenic diet is supposed to answer. In this trial, over six months, the people carrying that problem appeared to move the least.

The study itself: an assessor-blinded randomized controlled trial run by Beyza Tagraf and colleagues at Istanbul Medipol University, published online on July 24, 2026 in The Journal of Nutrition, Health & Aging, registered as NCT06898424. Sixty people with a clinical NIA-AA diagnosis of Alzheimer's, aged 55 to 85, Clinical Dementia Rating of 1 or 2. Thirty received the diet plus standard care; thirty received standard care alone. Cognition and brain imaging were assessed at baseline and at six months.

The group-level numbers are modest and real. Mean MMSE rose by 0.9 points in the diet group and fell by 1.5 points in the control group, with the between-group difference in change scores reported at p < 0.001. Baseline scores were close: 19 (SD 5.1) in the diet arm, 20 (SD 5.2) in control.

The authors then did something more careful than a mean. They applied the Reliable Change Index, a statistic that asks whether one individual's change exceeds the test's own noise. By that measure, 93 percent of the diet group counted as stable against 70 percent of controls, with improvement in 7 percent and 3.3 percent respectively. So the headline here is holding ground, which for a family watching a slow slide is its own kind of news, and also a reminder of how small the arithmetic is.

The subtype scheme comes out of the precision-medicine tradition around the KetoFLEX 12/3 protocol: inflammatory, glycotoxic, toxic, vascular, atrophic, categories the paper describes as overlapping rather than exclusive. Glycotoxic was defined concretely: BMI above 30 kg/m2 together with insulin resistance. Eleven of the thirty diet participants (37 percent) met it, as did seven of thirty controls (23 percent), a difference that did not reach significance (p = 0.40).

Eleven people. That is the entire glycotoxic diet subgroup, inside an arm of thirty, in a six-month single-center trial. A comparison that small can be moved by one person with a bad flu week. Treat it as a hypothesis with a number attached, which is what it is.

Still, it points at something the rest of the literature has been muttering about for years: ketones and fasting may speak different dialects in different bodies. In amyloid-beta infused rats studied in 2020, intermittent fasting improved memory on passive avoidance, Y maze, and water maze tests while a ketogenic diet did not, even though both raised serum ketones, and the ketogenic diet worsened gut dysbiosis and glucose tolerance. Those were rats, in an eight-week experiment, and rat hippocampi are not human lives. But it is a second, independent hint that raising ketones and fixing metabolism are two separate events that happen to travel together.

One more detail deserves attention from clinicians. The trial also compared responders against non-responders on the Modified Visual Magnetic Resonance Rating Scale, a seven-parameter visual score read from MRI acquired during routine care, and found a statistically significant difference between them. Structure at baseline, in other words, may condition who can respond at all. Metabolism is one filter. The brain's remaining tissue is another.

What holds this trial back is worth saying plainly. Participants knew what they were eating, so only the assessors were blinded; the control arm received no comparable attention or placebo diet; diagnosis rested on clinical criteria; follow-up stopped at six months; and the registry lists the study as running from March 2024 to October 2024, with the record first posted in March 2025 and the paper citing ethics approval dated March 6, 2025. Those dates sit oddly together, and a reader weighing the evidence is entitled to notice.

For the daughter reading this with a parent's clinic appointment in two weeks, the usable part is a question, not a menu. If a metabolic intervention is being discussed, what measurements would define her parent's metabolic picture, who would take them, and what would count as a response at six months. Eleven people in a subgroup cannot answer that for anyone's mother. Her clinician, holding her mother's actual bloodwork and actual MRI, is the one who can start.

Metabolism has grammar. A ketogenic diet is a sentence spoken into a body that already has rules about how it handles fuel, and this trial is an early, small piece of evidence that the same sentence lands differently depending on those rules. The next trial worth running is the one that sorts people first and powers each box properly.