Has the blood type diet ever been tested?

Once, and indirectly. A 16-week randomised trial of a low-fat vegan diet did ABO typing on 68 participants in its intervention arm and compared outcomes by blood group. Weight fell 5.7 kg in type A against 7.0 kg in everyone else (p = 0.09) and 7.1 kg in type O against 6.2 kg in everyone else (p = 0.33), with total cholesterol and every other outcome also null. Nobody has ever randomised people to their matched diet against their mismatched one, which is the design the hypothesis actually requires.

Why do people say it worked for them?

Because three of the four diets are recognisable healthy eating patterns under different names, and the largest observational study is fairly clear about it. Across 1,455 people, adherence to the Type A pattern was associated with lower BMI, waist, blood pressure, cholesterol, triglycerides and insulin measures, the Type AB pattern with nearly the same, and the Type O pattern with lower triglycerides. Matching the diet to the blood group did not change any of those associations. The food was doing the work.

Does blood type matter for health at all?

Yes, and not for this. ABO genotype is genuinely associated with disease: in twelve prospective cohorts, A1 alleles carried a pancreatic cancer odds ratio of 1.48 to 1.71 against O/O while A2 alleles carried none, and in a Norwegian cohort the highest quartile of plasma ABO transferase carried a venous thromboembolism hazard ratio of 2.12. Those associations run through glycosyltransferase activity and no trial has tested whether any diet modifies them.

What about the idea that blood groups evolved at different stages of human diet?

The genetics runs against the chronology. The ABO polymorphism is a trans-species polymorphism shared with the great apes and maintained by long-term balancing selection that appears to act on immune response to pathogens, which places its origin far before agriculture. That does not disprove the dietary claim, which would need its own test; it removes the historical argument usually offered in its support.

Is there any harm in trying it?

It depends which of the four you are given. The Type A and Type AB patterns are plant-forward and were associated with better markers in the observational data. The Type O pattern excludes grains and dairy, which is the fibre and calcium question that also applies to paleo. The Type B pattern excludes chicken, corn, lentils, peanuts, sesame and wheat, was associated with no cardiometabolic benefit in that same study, and is a broad elimination with nothing behind it.

Elimination6 human studies cited

Blood type diet

Written by Aaron CuhaReviewed Sep 2026

Also called: Eat Right 4 Your Type, ABO diet, D'Adamo diet

One randomised test exists, a secondary analysis inside a 16-week vegan trial with blood typing on 68 people. Blood type A lost 5.7 kg and everyone else lost 7.0 kg, p = 0.09, running opposite to the prediction. Nobody has ever randomised anyone to their matched diet against their mismatched one.

What it actually is

Four prescribed eating patterns assigned by ABO blood group, from a 1996 popular book rather than from a research programme. Type O is directed toward a high-protein, meat-forward pattern avoiding grains and dairy; type A toward a largely vegetarian one; type B toward a pattern that allows dairy while excluding chicken, corn and several other foods; type AB toward a mixture. The four differ from each other in ways that go far beyond ABO: the type A diet is close to a low-fat vegetarian pattern and the type O diet is close to a paleo one, which is the reason the observational data is more interesting than it first appears.

The mechanism its advocates propose

Stated at its strongest before it is tested. The proposal is that dietary lectins interact with ABO antigens on gut and blood cells in a blood-group-specific way, that the four blood groups emerged at different stages of human dietary history, and that eating outside your type causes agglutination, inflammation and chronic disease. There is a genuine substrate underneath this, which is why it is not simply dismissible. ABO is a real functional polymorphism: the antigens are sugars placed on cell surfaces by a glycosyltransferase, and ABO genotype is genuinely associated with disease. In a consortium of twelve prospective cohorts with 1,533 pancreatic cancer cases and 1,582 controls, A1 alleles carried an odds ratio of 1.48 for A1/O and 1.71 for A1/A1 against O/O while A2 alleles carried no excess, tracking glycosyltransferase activity. In a Norwegian cohort with 294 incident events, the highest quartile of plasma ABO transferase carried a venous thromboembolism hazard ratio of 2.12 (1.39 to 3.22). So ABO matters biologically. The chronology half of the claim runs into a different problem: the ABO polymorphism is shared with the great apes as a trans-species polymorphism maintained by balancing selection acting on immune function, which places its origin far before agriculture. The question the diet asks is narrower than any of this, and it is the only one that can be tested directly: does matching a diet to blood group change what that diet does to you.

What happened when calories were controlled

Strongest design available: not applicable, observational

Nobody assigned anything. People who choose a diet differ from people who do not in ways that are not fully adjustable, so this shows association and not effect.

No calorie-controlled trial of this pattern exists, and neither does a dedicated randomised trial of any design. The only randomised evidence is a secondary analysis inside a trial run for another purpose, and it found no interaction with blood group on any outcome it measured.

The randomised test, Wang 2021, is a secondary analysis among the intervention-group participants of a 16-week randomised trial of a low-fat vegan diet against making no dietary change. ABO typing was done on 68 intervention participants and outcomes compared between blood type A and everyone else, and between blood type O and everyone else. Mean weight change was minus 5.7 kg for type A against minus 7.0 kg for all others (p = 0.09), and minus 7.1 kg for type O against minus 6.2 kg for all others (p = 0.33). Total cholesterol fell 17.2 mg/dL in type A against 18.3 mg/dL in all others (p = 0.90), and 17.4 against 18.4 mg/dL for type O against all others (p = 0.89). There were no significant differences in any outcome, including body fat, visceral fat, blood lipids, fasting glucose or HbA1c.
Read the direction of those point estimates before reading the p values. The blood type hypothesis predicts that type A should do best on a plant-based diet and type O worst. On weight, the point estimates ran the other way in both comparisons, and neither reached significance. Sixty-eight people in one arm of one trial is a small test and the uncertainty around those differences is wide. It is also the only randomised evidence that exists.
Wang 2014 is the large observational test and its result is more interesting than its reputation. 1,455 participants of the Toronto Nutrigenomics and Health study completed a 196-item food frequency questionnaire covering one month, an adherence score was computed for each of the four blood type diets, and ABO blood group was determined by genotyping rs8176719 and rs8176746. Adherence to the Type A diet was associated with lower BMI, waist circumference, blood pressure, serum cholesterol, triglycerides, insulin, HOMA-IR and HOMA-Beta (p < 0.05). Adherence to the Type AB diet was associated with the same set except BMI and waist. Adherence to the Type O diet was associated with lower triglycerides (p < 0.0001). The Type B diet was associated with nothing.
And then the sentence the study was built to produce: matching the blood type diets with the corresponding blood group did not change the effect size of any of those associations. Three of the four diets look good on cardiometabolic markers, for everyone, regardless of blood group, which is what you would expect if the diets were working through their food content and not through ABO matching.
Cusack 2013 is the systematic review, conducted with GRADE across the Cochrane Library, MEDLINE and Embase. 1,415 references were screened and 16 articles identified, of which one was eligible, and that one examined LDL cholesterol response to a low-fat diet by MNS blood type rather than by ABO, so it did not answer the question. No studies showing the health effects of ABO blood type diets were identified. The review's own statement of what would be needed is the design nobody has run: a comparison of health outcomes between people adhering to a particular blood type diet and people continuing a standard diet, within a particular blood type population.

What happened when people just ate it

The observational layer is the Toronto study and it is worth reading as a positive finding rather than a negative one. People eating more like the Type A or Type AB pattern had better cardiometabolic markers than people who did not, and their blood group made no difference to that. Both of those patterns are, in practice, plant-forward and lower in processed food, and the Type O pattern, which is closest to paleo, was associated with lower triglycerides in the same direction the carbohydrate restriction literature predicts. The Type B pattern, the one with the least overlap with any conventional healthy eating pattern, was associated with nothing at all. That distribution of results is what a food-content explanation looks like. Nobody has published adherence, dropout or long-term follow-up data for this pattern in any blood group. There is a useful parallel from a different personalisation claim: the Food4Me trial randomised adults across seven European countries to conventional dietary advice or to personalised advice based on individual diet alone, on diet plus phenotype, or on diet plus phenotype plus five diet-responsive genetic variants. Personalised advice produced larger and more appropriate changes than conventional advice, cutting red meat, salt and saturated fat and raising Healthy Eating Index scores, and there was no evidence that including phenotypic or genotypic information enhanced the effect. Different genes and a different claim, and the same structural result: the personalisation framing did the motivating, and the genetic layer added nothing measurable on top.

Protein, and whether it confounds the result

The four diets differ enormously in protein, which is the largest unexamined variable in the whole pattern. Type O is steered toward a meat-forward, grain-excluding pattern that in practice resembles paleo, and type A toward a largely vegetarian one. No study of this pattern has reported protein intake, and none has measured lean mass in any blood group, so the difference in protein between the four arms of the hypothesis has never been quantified even descriptively. The measured protein intakes and what each produced are in the protein section of this site and in the high protein entry.

The measured intakes behind the protein figures are on the protein page.

What reliably moves

Marker	Direction	From
Body weight on a plant-based diet, by blood group	No difference	Minus 5.7 kg in blood type A against minus 7.0 kg in everyone else (p = 0.09), and minus 7.1 kg in type O against minus 6.2 kg in everyone else (p = 0.33), among 68 typed participants in one arm of a 16-week randomised trial. The point estimates ran opposite to the prediction.
Total cholesterol on a plant-based diet, by blood group	No difference	Down 17.2 mg/dL in type A against 18.3 in all others (p = 0.90), and 17.4 against 18.4 for type O against all others (p = 0.89). Body fat, visceral fat, fasting glucose and HbA1c were also null.
Cardiometabolic markers by adherence to the Type A and Type AB patterns	Better, in everyone	Lower BMI, waist, blood pressure, cholesterol, triglycerides, insulin, HOMA-IR and HOMA-Beta for Type A adherence, and the same except BMI and waist for Type AB, across 1,455 participants. Matching the diet to the corresponding blood group did not change any effect size.
Markers by adherence to the Type B pattern	Nothing	No significant association with any cardiometabolic biomarker in the same study. The Type B pattern is the one with the least overlap with any conventional healthy eating pattern.
Venous thromboembolism by ABO	Up in non-O, and nothing to do with diet	Highest quartile of plasma ABO transferase carried a hazard ratio of 2.12 (1.39 to 3.22) against the lowest, across 294 incident events, rising to 3.71 for unprovoked events. Attenuated only moderately by von Willebrand factor and factor VIII.
Pancreatic cancer by ABO allele	Up for A1, not for A2, and nothing to do with diet	Odds ratios of 1.48 for A1/O and 1.71 for A1/A1 against O/O, with A2 alleles carrying no excess (A1 against A2, p = 0.004), across twelve prospective cohorts. Tracks glycosyltransferase activity, which is the biology the diet borrows and does not test.

Long term, and hard outcomes

Nothing. No cohort has followed people eating a matched blood type diet, no trial has run longer than the 16 weeks of the study the secondary analysis sits inside, and no outcome of any kind, event or death, has been measured in any blood group on any of these four patterns. The 2013 systematic review found one eligible article out of 1,415 screened and it did not address ABO. In the years since, the evidence base has grown by one secondary analysis in 68 people. That is the complete record.

Citations

Human2021
Blood Type Is Not Associated with Changes in Cardiometabolic Outcomes in Response to a Plant-Based Dietary Intervention
Journal of the Academy of Nutrition and Dietetics
The only randomised evidence. Secondary analysis of 68 typed participants in the intervention arm of a 16-week randomised low-fat vegan trial. Weight minus 5.7 kg in type A against minus 7.0 in all others (p = 0.09); minus 7.1 in type O against minus 6.2 (p = 0.33). Total cholesterol p = 0.90 and p = 0.89. No significant difference in any outcome.
Human2014
ABO genotype, 'blood-type' diet and cardiometabolic risk factors
PLoS One
1,455 participants, 196-item food frequency questionnaire, ABO genotype from rs8176719 and rs8176746. Adherence to the Type A diet associated with lower BMI, waist, blood pressure, cholesterol, triglycerides, insulin, HOMA-IR and HOMA-Beta; Type AB with the same except BMI and waist; Type O with lower triglycerides; Type B with nothing. Matching the diet to the blood group did not change any effect size.
Review2013
Blood type diets lack supporting evidence: a systematic review
American Journal of Clinical Nutrition
1,415 references screened across the Cochrane Library, MEDLINE and Embase, 16 articles identified, one eligible, and that one examined LDL response to a low-fat diet by MNS blood type rather than ABO. No studies showing health effects of ABO blood type diets were identified. GRADE approach.
Human2017
Effect of personalized nutrition on health-related behaviour change: evidence from the Food4Me European randomized controlled trial
International Journal of Epidemiology
A different personalisation claim with a structurally similar result. 1,269 adults across seven countries randomised to conventional advice or to personalised advice based on diet, diet plus phenotype, or diet plus phenotype plus five genetic variants. Personalised advice cut red meat, salt and saturated fat and raised diet quality, and there was no evidence that phenotypic or genotypic information enhanced the effect.
Human2025
Histo-blood group ABO system transferase plasma levels and risk of future venous thromboembolism: the HUNT study
Blood
A real ABO disease association, unrelated to diet. 294 incident venous thromboembolism cases and a 1,066-person subcohort. Highest quartile of plasma ABO transferase carried a hazard ratio of 2.12 (1.39 to 3.22) against the lowest, 3.71 for unprovoked events, attenuating only moderately on adjustment for von Willebrand factor and factor VIII.
Human2010
Variant ABO blood group alleles, secretor status, and risk of pancreatic cancer: results from the pancreatic cancer cohort consortium
Cancer Epidemiology, Biomarkers and Prevention
Twelve prospective cohorts, 1,533 cases and 1,582 controls. Against O/O, odds ratios were 1.48 for A1/O and 1.71 for A1/A1, while A2 alleles carried no excess (A1 against A2, p = 0.004). Risk tracked glycosyltransferase activity. Secretor phenotype was not an effect modifier.
Human2009
Genome-wide association study identifies variants in the ABO locus associated with susceptibility to pancreatic cancer
Nature Genetics
The genome-wide scan that established the ABO locus association, cited here to show that the ABO biology the diet borrows from is real and was discovered independently of any dietary hypothesis. It says nothing about diet.
Review2015
Trans-species polymorphism in humans and the great apes is generally maintained by balancing selection that modulates the host immune response
Human Genomics
Bears on the chronology half of the claim. The histo-blood ABO group is one of a small number of well-characterised trans-species polymorphisms shared between humans and other primates and maintained by long-term balancing selection acting on immune function, which places its origin far earlier than agriculture.
Review2001
ABH and Lewis histo-blood group antigens, a model for the meaning of oligosaccharide diversity in the face of a changing world
Biochimie
Argues that the polymorphism of the ABH and Lewis antigen gene families provides intraspecies diversity for coping with diverse and rapidly evolving pathogens, and that the role of these antigens in cancer and cardiovascular disease is incidental rather than of evolutionary significance. Background on what the antigens are believed to be for.

What people report

These are uncontrolled self-reports, not evidence. They are here because they tell you what to expect and what to watch for, which the trial literature does not. They cannot tell you whether anything works.

Feeling better within weeks, the most common report, and the expected result whenever anyone replaces a haphazard diet with any structured one.
Weight loss on the Type O and Type A patterns, reported frequently, and consistent with those being recognisable low-carbohydrate and plant-forward patterns under different names.
No effect at all, reported commonly and rarely written up, which is the finding the randomised secondary analysis supports.
Symptoms returning on reintroducing a forbidden food, reported without blinding of any kind and indistinguishable from expectation.
Frustration at the Type B list in particular, which excludes chicken, corn, lentils, peanuts, sesame and wheat and is the hardest of the four to run.
Confusion when family members with different blood groups are told to eat opposite diets, the most commonly cited practical reason for stopping.
Cost of the type-specific supplement lines sold alongside the framework.

Sources: Blood type diet forums, practitioner write-ups from the framework's own network, and the popular books that carry it. Note the commercial structure: the pattern is attached to a branded supplement and testing line, which is a disclosure that belongs next to the community reports. Uncontrolled self-report throughout, and in this pattern it is close to the entire body of affirmative material, because the peer-reviewed record contains one systematic review that found nothing eligible, one observational study that found the effects independent of blood group, and one secondary analysis that found no interaction.

Who this is wrong for

Anyone with blood type O being steered away from grains and legumes without a plan to replace the fibre, and anyone with blood type B excluding chicken, corn, lentils, peanuts, sesame and wheat. Those are broad eliminations with no trial behind them, and the Type B pattern was the one associated with no cardiometabolic benefit at all in the largest observational study.
Anyone who will accept their blood group as the explanation for a result. In the only randomised test, the point estimates ran opposite to the prediction and no comparison reached significance.
Anyone paying for blood-group-specific supplements, testing or meal plans. A blood type test tells you your blood group, which is worth knowing for transfusion and for the genuine ABO disease associations, and it does not tell you what to eat.
Anyone with a history of disordered eating, for whom four lists of forbidden foods assigned by an immutable personal characteristic is a poor structural fit.
Anyone in a high-risk group for the real ABO-associated conditions who adopts the diet as a response to that risk. The venous thromboembolism and pancreatic cancer associations run through glycosyltransferase activity, and no trial has tested whether any diet modifies them.
Anyone treating this as a personalised nutrition method. The one randomised trial of genotype-based personalisation in a different gene set found that adding genetic information on top of individualised dietary advice produced no further benefit.

Questions

Has the blood type diet ever been tested?: Once, and indirectly. A 16-week randomised trial of a low-fat vegan diet did ABO typing on 68 participants in its intervention arm and compared outcomes by blood group. Weight fell 5.7 kg in type A against 7.0 kg in everyone else (p = 0.09) and 7.1 kg in type O against 6.2 kg in everyone else (p = 0.33), with total cholesterol and every other outcome also null. Nobody has ever randomised people to their matched diet against their mismatched one, which is the design the hypothesis actually requires.
Why do people say it worked for them?: Because three of the four diets are recognisable healthy eating patterns under different names, and the largest observational study is fairly clear about it. Across 1,455 people, adherence to the Type A pattern was associated with lower BMI, waist, blood pressure, cholesterol, triglycerides and insulin measures, the Type AB pattern with nearly the same, and the Type O pattern with lower triglycerides. Matching the diet to the blood group did not change any of those associations. The food was doing the work.
Does blood type matter for health at all?: Yes, and not for this. ABO genotype is genuinely associated with disease: in twelve prospective cohorts, A1 alleles carried a pancreatic cancer odds ratio of 1.48 to 1.71 against O/O while A2 alleles carried none, and in a Norwegian cohort the highest quartile of plasma ABO transferase carried a venous thromboembolism hazard ratio of 2.12. Those associations run through glycosyltransferase activity and no trial has tested whether any diet modifies them.
What about the idea that blood groups evolved at different stages of human diet?: The genetics runs against the chronology. The ABO polymorphism is a trans-species polymorphism shared with the great apes and maintained by long-term balancing selection that appears to act on immune response to pathogens, which places its origin far before agriculture. That does not disprove the dietary claim, which would need its own test; it removes the historical argument usually offered in its support.
Is there any harm in trying it?: It depends which of the four you are given. The Type A and Type AB patterns are plant-forward and were associated with better markers in the observational data. The Type O pattern excludes grains and dairy, which is the fibre and calcium question that also applies to paleo. The Type B pattern excludes chicken, corn, lentils, peanuts, sesame and wheat, was associated with no cardiometabolic benefit in that same study, and is a broad elimination with nothing behind it.