Imagine having a genetic copy of yourself. You share essentially the same DNA, the same birthday, and many early experiences. Then researchers let you each follow two different dietary plans for eight weeks.
One twin eats a healthy vegan diet. The other eats a healthy omnivore diet with both plant and animal foods. Researchers track LDL cholesterol, fasting insulin, and body weight, among other things.
Who "wins"? It is a tempting question, but the study's most useful answer is more nuanced. The vegan group had larger short-term changes in certain cardiometabolic markers. The study cannot, however, name a universally best diet or say who thrived in any broad sense.
Key takeaways
- The study randomised 22 adult identical twin pairs to a healthy vegan or healthy omnivore diet for eight weeks.
- After eight weeks, the change in LDL cholesterol was larger in the vegan group than in the omnivore group.
- Changes in fasting insulin and body weight also differed between groups, but the study was not designed to separate the dietary pattern from energy intake and weight loss.
- Both plans were intended to be healthy; this was not a comparison of plant-based food against fast food.
- The study was short, small, open-label, and had no follow-up. It does not show long-term disease outcomes, longevity, or which diet suits each individual.
Why twins are researchers' dream participants
When two unrelated people respond differently to a dietary change, it is hard to know how much depends on genetic differences. Identical twins share essentially the same DNA sequence, which makes the comparison within each pair unusually informative.
But the twin design does not remove all differences. Participants may differ in everyday habits, adherence, prior diet, gut microbiome, and other factors. Randomisation strengthens the study, but does not make every result a guarantee for all twins or all adults.
How the Stanford study worked
Researchers recruited 22 pairs of adult identical twins, 44 participants in total. Within each pair, one twin was randomly assigned to a vegan diet and the other to an omnivore diet. Twenty-one pairs completed both dietary plans.
The study ran for eight weeks. During the first four weeks, participants received delivered meals. During the last four weeks, they bought and prepared food themselves according to the study's guidelines.
The design was open-label: participants knew which diet they were eating. This makes practical dietary research possible, but may affect behaviour, reporting, and expectations.
Healthy vegan versus healthy omnivore
This was not a test of lentils against burgers. Both groups were encouraged to eat plenty of vegetables, choose minimally processed foods, and limit added sugar and refined grains.
The dietary patterns differed in several ways nonetheless. The vegan group ate more vegetables and less cholesterol, while protein intake, energy intake, and other nutritional components could also differ.
The result therefore cannot be reduced to a single ingredient or to the label "vegan". It is about complete dietary patterns followed in an everyday-life study.
Finding 1: LDL cholesterol
The study's primary outcome was LDL cholesterol. After eight weeks, the estimated difference in change between groups was 13.9 mg/dL in favour of the vegan group.
That is a clear short-term result in this study. It does not mean that an omnivore diet is automatically harmful, that a vegan diet always produces the same change, or that the study measured future cardiovascular events.
LDL-C is a cardiometabolic risk marker. The article therefore does not translate the laboratory value into a promise of a longer life or absence of disease.
Finding 2: body weight
Both groups lost weight during the study. The estimated difference in change was 1.9 kg more in the vegan group after eight weeks.
The study was not isocaloric. Participants were therefore not given identical amounts of energy, and researchers noted differences in reported energy intake.
That makes weight important for interpretation: the LDL change cannot be fully separated from the weight loss or from other differences between the dietary plans.
Finding 3: fasting insulin
The vegan group also had a larger reduction in fasting insulin. The estimated difference between groups was 2.9 μIU/mL after eight weeks.
Fasting insulin is not the same as a complete diagnosis or a direct measure of how each participant's health develops over time.
The article therefore does not say the study proved improved insulin sensitivity for everyone. It describes the outcome that was actually measured: fasting insulin.
How should the results be interpreted?
The results suggest that the vegan dietary plan in this short study produced larger changes in LDL-C, fasting insulin, and body weight than the healthy omnivore plan.
But the groups also differed simultaneously in energy intake, weight change, and several aspects of dietary composition. Researchers could not isolate whether the difference was driven by the absence of animal foods, more plant foods, less saturated fat, more fibre, lower energy intake, or a combination.
The most scientifically defensible answer to "who won?" is therefore: the vegan group had an advantage in certain measured markers over eight weeks. The study does not determine which diet is best in all respects or for all people.
What the study does not show
The study does not show what happens after a year, five years, or the rest of life. It did not measure heart attacks, stroke, longevity, or other long-term clinical outcomes.
It does not show that all vegan diets are healthy or that all omnivore diets are equivalent. Food quality, nutritional balance, access, and adherence matter in both dietary patterns.
Nor does it show who subjectively "thrived most". Blood markers and body weight are not the same as energy levels, enjoyment of food, wellbeing, or a diet that is practically sustainable for a given person.
And what happened afterwards?
Here the story must stop at the study's boundary. The researchers had no follow-up period after the eight weeks and could therefore not assess how stable the changes were or how participants' dietary habits developed.
The study later received considerable attention in popular culture. Such media coverage can spark curiosity, but it is not evidence and is not used as support for the article's health claims.
A separate epigenetics publication appeared in 2024. It is not included in this article's conclusions because it addresses different questions and requires its own source review.
If TwinPare ran the same experiment
Imagine two twins following different, well-designed dietary plans while documenting sleep, activity, food experience, and health markers. That would create many fascinating questions — but also a great responsibility.
A TwinPare concept must not become a competition about who eats "cleanest" or loses the most weight. Medical safety, nutritional adequacy, consent, and professional study design would need to come before entertainment value.
The real gain would not be crowning one twin. It would be understanding variation better without turning individual measurements into universal dietary advice.
The TwinPare perspective
The twin study shows why identical genetic starting points can make lifestyle questions unusually clear — and why the answers are rarely simple even then.
The genes were very similar. The dietary plans differed. Several markers changed differently. But even a randomised twin trial needs to be interpreted within its timeframe, its sample, and its method.
TwinPare's conclusion is therefore not "only eat this way". It is that well-designed twin studies can help us ask better questions about food, health, and individual variation.
Source and limitations
This article is based on a peer-reviewed open-label randomised clinical trial published in JAMA Network Open 2023. The study recruited 22 adult identical twin pairs, ran for eight weeks, and had no follow-up period.
The sample was small and consisted mainly of women. The study lacked formal power calculations, was not isocaloric, and could not blind participants to the dietary plan.
The primary article received a published correction in December 2023. The correction addressed an erroneous sentence about non-significant secondary outcomes; the corrected PMC version and the reported tables are used for this draft.
The source supports short-term group differences in LDL-C, fasting insulin, and body weight. It does not support a universally best diet, long-term disease outcomes, automatic weight loss, or the claim that one group thrived most in all respects.
Source notes
The source has been verified and editorially reviewed for this article. The limitations below show which level of conclusion the sources support.
- [landry-2023] Cardiometabolic Effects of Omnivorous vs Vegan Diets in Identical Twins: A Randomized Clinical Trial. Matthew J. Landry; Catherine P. Ward; Kristen M. Cunanan; Lindsay R. Durand; Dalia Perelman; Jennifer L. Robinson; Tayler Hennings; Linda Koh; Christopher Dant; Amanda Zeitlin; Emily R. Ebel; Erica D. Sonnenburg; Justin L. Sonnenburg; Christopher D. Gardner. JAMA Network Open, 2023. Evidence type: Eight-week open-label randomised clinical trial of 22 adult identical twin pairs assigned to a healthy vegan or healthy omnivore diet Limitation: The study was small, short, open-label, and not isocaloric. The first four weeks used delivered meals and the last four used self-prepared food. Differences in energy intake, weight, fibre, saturated fat, and other dietary components mean the results cannot be attributed to a single property of a vegan diet. No follow-up was conducted. A published correction (DOI 10.1001/jamanetworkopen.2023.50422) corrected an erroneous sentence about non-significant secondary outcomes; the corrected PMC version is used here. PubMed PMC DOI
Editorial source review
This section shows how the article's key factual claims are linked to the source.
Phrasings that require caution
- Write that the vegan group had larger short-term changes in certain markers, not that a vegan diet won or is superior for everyone.
- Describe LDL-C, fasting insulin, and body weight as studied outcomes; do not translate them into guaranteed health, a longer life, or absence of disease.
- Use fasting insulin, not proven insulin sensitivity, unless additional direct source support is added.
- Remind readers that both dietary plans were designed as healthy and that the study did not compare a vegan diet with fast food.
- Keep dietary pattern, energy intake, weight change, fibre, saturated fat, cholesterol, and overall food quality separate.
- Note that the first four weeks used delivered meals and the last four used self-prepared food.
- Do not use Netflix or the separate epigenetics study as evidence for this article's health claims.
- Give no individual dietary, weight-loss, or treatment advice.
| ID | Claim | Source support | Caution |
|---|---|---|---|
| C1 | The study randomised 22 identical twin pairs to a vegan or omnivore diet for eight weeks; 21 pairs completed both plans. | 2023 | Distinguish the number enrolled from the number who completed, and describe the study as open-label. |
| C2 | Both intervention diets were designed as healthy dietary patterns, but they differed in several food and nutritional components. | 2023 | Do not claim that a single component explained the results. |
| C3 | After eight weeks, the estimated difference in LDL-C change between groups was −13.9 mg/dL for vegan compared with omnivore diet. | 2023 | Describe this as a short-term group result for a risk marker, not as evidence of clinical events or universal effect. |
| C4 | Weight decreased in both groups, with an estimated difference of −1.9 kg for vegan compared with omnivore diet after eight weeks. | 2023 | The study was not isocaloric; do not call the difference a guaranteed effect of veganism. |
| C5 | The change in fasting insulin differed by −2.9 μIU/mL for vegan compared with omnivore diet after eight weeks. | 2023 | Use the term fasting insulin and avoid making the result a universal diagnosis or proven mechanism. |
| C6 | The short study cannot separate the dietary pattern from concurrent weight loss, energy intake, and multiple nutritional differences. | 2023 | Avoid a single mechanism and a universal winner claim. |
| C7 | This article is published with a clear boundary because the short study and its correction do not support universal dietary advice or long-term health promises. | 2023 | Keep the correction and the boundary between short-term markers and subjective wellbeing visible. |