A Stanford Analysis Ties LDL Increases on Low-Carb Diets to a Genetic Score

August 26, 2026 · Science · Biomarkers · Nutrition

What the Analysis Reports

People with a stronger genetic tendency toward high LDL cholesterol saw larger LDL increases on a low-carb diet than people without it. The analysis was presented by Alexa Barad, PhD, RDN, a postdoctoral scholar at Stanford University School of Medicine. She presented it at NUTRITION 2026, the annual meeting of the American Society for Nutrition, held July 25 to 28, 2026 in National Harbor, Maryland.

It is a secondary analysis of DIETFITS, a randomized controlled trial that assigned more than 600 adults to either a healthy low-carbohydrate or a healthy low-fat diet for one year. Genetic data was available for 431 of those participants, and the outcome examined was the change in LDL cholesterol from baseline to six months. The researchers combined thousands of variants from across the genome into a polygenic score estimating each participant's overall genetic tendency toward higher or lower LDL cholesterol.

Among participants in the low-carb arm, those with a higher polygenic score were more likely to see LDL rise. The release describes that pattern as driven by greater sensitivity to saturated fat, with participants at higher genetic risk showing the largest LDL increases when they consumed more saturated fat. Participants following the low-fat diet did not show the same pattern.

No effect sizes were published. The American Society for Nutrition release reports the direction of the finding without a single figure in milligrams per deciliter or in percentage terms. None appears in the ScienceDaily write-up that carried it into wider circulation on August 23 either. The society's own disclaimer states that NUTRITION 2026 abstracts have not generally undergone the peer review required for journal publication and should be considered preliminary.

Why an Unpublished Effect Size Matters

A conference abstract with no effect size cannot support a dosing or dietary decision. The claim reported here has a direction and a proposed mechanism, and it has no magnitude. That distinction decides how a reader should use it. If people in the top genetic group rose by a few milligrams per deciliter more than the rest, the result is a research curiosity. If they rose by thirty or forty, it would describe a real difference in who a low-carb diet suits.

The design does carry weight. DIETFITS randomized participants to their diets, so the comparison between the low-carb and low-fat arms is not subject to the self-selection that weakens observational diet research. The genetic layer sits on top of that randomization as a secondary analysis. That is a weaker form of evidence than the trial's primary comparison, and it is best treated as generating a hypothesis. The researchers themselves note that additional research is needed to confirm the results and to determine whether they apply across more diverse populations.

What It Means for You

This finding lands on top of something our own guides already say. Our page on ApoB records that dietary saturated fat moves some people 20 to 30 milligrams per deciliter on the same change, while others barely move. It tells readers to test rather than assume. The Stanford analysis proposes part of the reason for that spread. It does not change the instruction, and if anything it strengthens it, because a genetic tendency you cannot see makes the measured response more informative rather than less.

The practical sequence costs the price of two lipid panels. Get a baseline before changing how you eat, then repeat at eight to twelve weeks, which is long enough for a dietary change to show up in circulating lipids. ApoB is the better marker to follow than LDL cholesterol alone, because it counts atherogenic particles directly, and our guide to what an ApoB test costs covers the standalone and panel-included routes. Numbers can move sharply on a low-carb diet. Our page on normal versus optimal lab ranges covers why a result inside a lab's reference range can still warrant a conversation.

One thing this does not justify is buying a genetic test to decide what to eat. The score used here was constructed for research from thousands of variants, no consumer equivalent has been validated for diet selection, and the measured response is both cheaper and more direct. Readers who already have genetic data from another source have no reason to reinterpret it against an abstract carrying no numbers. Lipoprotein(a) is the one inherited lipid marker with a clear case for a single lifetime measurement, and it is unrelated to the polygenic score in this analysis.

Who Should Ignore This Finding

Anyone already taking a statin or another lipid-lowering therapy should treat this as background. Their LDL response is dominated by the medication, and a prescriber has already set their monitoring schedule. Anyone with diagnosed familial hypercholesterolemia should do the same. That condition is driven by single high-impact variants rather than by the polygenic tendency this score measures, and it carries its own management pathway. Anyone eating a low-carb diet whose lipid panels have been stable for years has their answer already, and it is a better answer than a genetic score would give them.

What Would Change This Read

A peer-reviewed publication with effect sizes would change it immediately, because the magnitude is the missing piece and everything else is in place. A result showing a large separation between genetic groups would make individual response prediction a genuine clinical question. A small separation would confirm that measuring beats predicting, which is the position our ApoB guide already takes. Replication in a more diverse cohort would address the limit the researchers name themselves. DIETFITS was a single trial in one population, and polygenic scores built in one ancestry group often transfer poorly to others.

Related Coverage

A recurring theme in these briefings is the distance between a measurement and a decision. An earlier one covered a Finnish cohort where conventional risk factors outpredicted epigenetic clocks. Another covered a federal guideline that reviewed the body fat measurements and endorsed none of them. This analysis points the same way from the genetic side. A score that predicts a tendency is worth less than a panel that reports the response, and the panel is the cheaper of the two. The next step for anyone changing their diet is to book the baseline draw before the change rather than after it.

Sources

• American Society for Nutrition, "Is your diet driving up your cholesterol? That may depend on your genes," news release for NUTRITION 2026, eurekalert.org (accessed August 26, 2026). Source of the 431-participant analysis figure, the six-month timepoint, the presenter attribution, the conference dates, and the peer-review disclaimer.
• ScienceDaily, "Your genes may decide whether a low-carb diet sends cholesterol soaring," August 23, 2026, sciencedaily.com (accessed August 26, 2026). Source for the polygenic score description and the saturated fat interaction as reported.