Low-carbohydrate and low-fat diets, genetic susceptibility, and long-term risk of dementia: a prospective cohort study

What APOE and polygenic risk scores is

APOE is a gene whose variants raise or lower Alzheimer risk, and a polygenic risk score adds up many small genetic effects into a single number. Both are used here to ask whether diet matters more, or less, for people who start with higher inherited risk.

Drawn from background genetics, not from this paper.

Why they ran it

The authors note that low-carbohydrate and low-fat diets are both promoted for cardiometabolic prevention, and set out to examine how each relates to incident dementia.

They also wanted to test whether genetic susceptibility changes that relationship, which is why APOE genotype and a polygenic risk score were built into the analysis rather than treated as confounders.

Drawn from the paper's stated objective.

The Health and Retirement Study supplied 5,301 dementia free adults aged 55 and over, with diet captured by food frequency questionnaire in 2013 to 2014. Overall low-carbohydrate and low-fat indices were built from macronutrient composition rankings, then split into plant based, animal based, healthy and unhealthy sub scores. Incident dementia through 2022 was defined by the Langa-Weir algorithm. Genetic susceptibility was captured by APOE genotype and an Alzheimer disease polygenic risk score.

Over nine years, 506 people developed dementia. A higher overall low-carbohydrate score was associated with lower risk, hazard ratio 0.90 per standard deviation. The overall low-fat score was not associated at all. The sub scores are where it gets interesting: plant based and healthy low-carbohydrate patterns showed stronger inverse associations, while animal based low-carbohydrate and unhealthy low-fat patterns were associated with higher risk.

Associations were consistent across APOE genotype and across strata of the polygenic risk score. Higher plant based and healthy low-carbohydrate scores also tracked with better global and domain specific cognitive performance. The authors read the whole pattern as evidence that macronutrient quality matters alongside quantity.

The numbers

undefined	undefined
undefined	undefined
undefined	undefined
undefined	undefined
undefined	undefined
undefined	undefined
undefined	undefined

Why this might happen

Open question The authors do not claim to know why.

No biological mechanism is offered. As an observational study the authors stop at the pattern, and the conclusion they draw is a claim about what the data distinguishes rather than about physiology.

What they underline is that macronutrient quality and source, not just how much carbohydrate or fat a diet contains, is what tracked with cognitive outcomes here.

Drawn from the paper's abstract and conclusion.

What this does not show

This cannot show that changing what you eat changes dementia risk. It is observational. Diet was measured once and people were followed, so the association may reflect what kind of person eats which way rather than what the eating does.
This is not a ketogenic or very low carbohydrate diet. The low-carbohydrate index is a relative ranking of ordinary eating patterns against each other. Nobody in this cohort was necessarily eating what a clinician would call low carb.
Consistency across genetic strata is not proof of no interaction. The associations looked similar across APOE genotype and polygenic risk score. Detecting a genuine gene by diet interaction takes far more power than detecting a main effect.
Diet was captured once, nine years before the last outcome. A single food frequency questionnaire in 2013 to 2014 stands in for the whole follow up, and people change how they eat, particularly as cognition declines.

Where this leaves us

It refines the low-carb versus low-fat argument by showing the labels are the wrong unit. Within each pattern, quality and food source separated lower risk from higher risk, and that separation held regardless of inherited risk.

US adults aged 55 and over, followed about nine years, with diet measured once by questionnaire.

A trial or a repeated measures cohort that changes diet quality and follows cognition, which would test whether the association survives being intervened on.

Caveats worth holding

Observational design with a single dietary assessment, so reverse causation and residual confounding are both live.
Dementia was ascertained by algorithm rather than clinical diagnosis.
The macronutrient indices are relative rankings within this cohort, not absolute intakes.
Sourced from the abstract and record; the full discussion was not reachable.