Soybeans get more of their calories from protein and fat than from carbohydrate. Cooked soybeans get about 42% of their calories from protein (18.2 g per 100 g) and 46% from fat. They have only 2.4 g of net carbs per 100 g.
They score 100 out of 100 for essential amino acids on this site, so they are a complete protein. A cup (172 g) of cooked soybeans has 31.3 g of protein.
Soy is also the center of the isoflavone debate. Isoflavones, such as genistein and daidzein, are plant compounds with a chemical shape similar to human estrogen. That fed early worries, but AICR (the American Institute for Cancer Research) says they work differently in the body.
A 2025 review of 40 trials in postmenopausal women found no estrogen-like effect on four measures. Reviews of men and of thyroid hormones found little or no effect, as the studies below show.
Soybeans must be cooked. Raw ones contain protease inhibitors and lectins. These interfere with digestion, and heat destroys them.
Most research below tested soy protein isolate, soy milk, tofu or isoflavone pills instead of cooked beans, so each card says what was tested. The roasted soy nut trials come closest to eating whole soybeans.
Health benefits of soybeans
The top findings from 30 studies, strongest evidence first
What the stars mean
- Strong. Several randomized trials, pooled in a meta-analysis, point the same way, or regulators (FDA or EFSA) have approved the claim.
- Moderate. Good randomized trials support it, but there are fewer of them, or their results vary.
- Limited. Small, short or single trials, or mixed results. Promising, but it needs confirmation.
Population and lab studies get no stars: they show a link or a mechanism, not proof.
How we rate evidence →- ★★★Slightly lower “bad” LDL cholesterolSoy protein (a median of 25 g a day) cut LDL about 3 to 4% versus other proteins (4.76 mg/dL across 41 trials). These were soy protein studies, not whole beans.
- ★★★No change in estrogen measures from isoflavonesIn 40 trials (3,285 postmenopausal women), soy isoflavones, median 75 mg a day, did not change womb lining thickness, vaginal maturation or hormone levels. The doses came from isoflavones, not necessarily from cooked beans.
- ★★Lower LDL and blood pressure with soymilkSwapping cow’s milk for soymilk (about 500 mL a day) lowered LDL (by 0.19 mmol/L) and the top blood pressure number (by 8.00 mm Hg). This was across 17 trials of soymilk, not whole beans.
- ★★No change in men’s testosterone or estrogenIn 41 studies, soy foods, soy protein or isoflavone extracts did not change men’s testosterone or estradiol, the main estrogen. Dose and study length made no difference.
- ★★Mixed evidence on hot flashesPooled trials of isoflavone extracts, not soy foods, found hot flashes 20.6% less frequent than with placebo. A Cochrane review found no conclusive evidence that soy or similar plant estrogens ease flushes.
All 30 studies, strongest evidence first
Soy protein cut LDL cholesterol about 3 to 4% versus other proteins
This meta-analysis (a study that pools many trials) used the 46 trials the FDA weighed when it considered revoking its soy heart claim. Of these, 43 had usable data and 41 measured LDL. On a median of 25 g of soy protein a day for 6 weeks, LDL fell 4.76 mg/dL and total cholesterol 6.41 mg/dL. About 75% of the trials showed a drop, but only a minority were significant (unlikely to be chance) on their own. These were soy protein studies, not whole beans, and one author works for the Soy Nutrition Institute.
Soy’s cholesterol effect stayed significant for 14 years after 1999
This cumulative meta-analysis added the same 46 trials the FDA reviewed in the order they came out, as in the study above. By 1999, when the FDA approved its heart claim, LDL had fallen 6.3 mg/dL. In later years the drop stayed between 4.2 and 6.7 mg/dL and never lost significance. The authors say this supports the original claim; the trials tested soy protein, not whole beans.
Soy isoflavones changed no measure of estrogen activity after menopause
This meta-analysis pooled 40 randomized trials (people assigned to groups by chance) in 3,285 postmenopausal women. Isoflavones, at a median of 75 mg a day for a median of 24 weeks, were compared with non-isoflavone controls. Endometrial thickness (the womb lining), vaginal maturation, FSH and estradiol (hormone levels) did not change significantly, and the certainty of evidence was high to moderate. The doses came from soy isoflavones, not necessarily from cooked beans.
Soy protein lowered LDL cholesterol by 5.5% in the better-designed trials
This meta-analysis pooled 20 parallel and 23 crossover randomized trials. People took a median of 30 g of soy protein a day for 4 to 18 weeks, not cooked soybeans. In the parallel trials, which scored higher on quality, HDL (the “good” cholesterol) rose 3.2% and triglycerides fell 10.7%. In the crossover trials, where people tried both diets, LDL fell 4.2%.
American Heart Association: soy protein lowered LDL by only about 3%
This science advisory reviewed 22 randomized trials of isolated soy protein with isoflavones. The soy protein averaged 50 g a day, about half of a usual day’s protein. Across 19 studies of isoflavones alone, the average effect on LDL was nil. The authors did not recommend isoflavone supplements, but noted soy foods are high in polyunsaturated fat, fiber, vitamins and minerals and low in saturated fat.
Soymilk in place of cow’s milk lowered LDL and blood pressure
This meta-analysis pooled 17 randomized trials in 504 adults who replaced a median of 500 mL of cow’s milk a day with soymilk. LDL fell 0.19 mmol/L (millimoles per liter), systolic pressure (the top number) 8.00 mm Hg and CRP, an inflammation marker, 0.82 mg/L. Added sugar in sweetened soymilk did not change the results. Certainty was high for LDL and moderate for blood pressure; this tested soymilk, not whole beans.
Unfermented soy did not significantly lower LDL cholesterol in postmenopausal women
This meta-analysis of 29 randomized trials (2,457 women) found total cholesterol fell 9.46 mg/dL and triglycerides 10.86 mg/dL, and HDL rose 2.32 mg/dL. LDL fell 4.55 mg/dL, a drop that was not significant, so it could be chance. LDL did fall in women with lipid disorders and with soy protein that contained isoflavones. Subgroups covered soy isolate protein and soy protein with isoflavones, not cooked beans.
Soy isoflavone extracts cut hot flash frequency by 20.6% compared with placebo
This meta-analysis pooled 17 double-blind randomized trials: 13 on frequency and 9 on severity. They tested extracted or synthesized isoflavones (median 54 mg a day, for 6 weeks to 12 months), not soy foods, against a placebo (dummy pill). Severity fell 26.2%. Results differed widely between trials, and supplements with more than 18.8 mg of genistein worked more than twice as well as lower doses.
A Cochrane review: no conclusive evidence phytoestrogens ease flushes or night sweats
Cochrane, a group known for careful reviews, looked at 43 randomized trials (4,364 women) of soy foods, soy extracts, red clover and other phytoestrogens. Most trials were too different to pool, and many were small with a high risk of bias. Placebo cut flush frequency by 1% to 59%. Four trials of extracts with more than 30 mg a day of genistein consistently found fewer flushes.
Soy supplements did not change free thyroid hormones, and TSH rose slightly
This meta-analysis pooled 18 studies: randomized trials that measured free T3, free T4 and TSH after soy or soy products. TSH, the signal that tells the thyroid to work, rose by 0.248 mIU/L on average, a borderline result. The authors say its clinical meaning, if any, is unclear. There was no sign of publication bias, and results varied a lot between trials.
Soy protein and isoflavones did not change men’s testosterone or estrogen
This updated meta-analysis pooled 41 clinical studies of soy foods, soy protein or isoflavone extracts. Total testosterone was measured in 1,753 men and estradiol (the main estrogen) in 1,000. Dose and study length made no difference. It extends a 2010 meta-analysis that reached the same result.
Soy nuts cut systolic pressure 9.9% in women with high blood pressure
In this crossover trial, 60 postmenopausal women ate soy nuts (roasted whole soybeans) for 8 weeks. The nuts supplied 25 g of soy protein and 101 mg of isoflavones a day, replacing other protein in a cholesterol-lowering diet. In women without high blood pressure, systolic pressure fell 5.2%; LDL fell 11% in the women with it and did not change in the others. It was small and short.
Soy nuts lowered LDL and improved blood vessel function in older adults
In this randomized crossover trial, 23 healthy adults aged 60 to 70 ate 67 g of soy nuts a day for 16 weeks. The nuts gave 25.5 g of protein and 174 mg of isoflavones, and the comparison was a period with no nuts. LDL fell 0.17 mmol/L, flow-mediated dilation (how well arteries widen) rose 1.48 percentage points and mean arterial pressure fell 3 mm Hg. Arterial stiffness did not change, and the trial was small and single-blinded.
Whole-bean soy milk cut high LDL about 5% versus dairy milk
In this crossover trial, 28 adults with LDL of 160 to 220 mg/dL drank each milk for 4 weeks. Each milk gave them 25 g of protein a day. LDL ended at 161 mg/dL on whole-bean soy milk and on soy protein isolate milk, against 170 mg/dL on dairy milk. HDL, triglycerides, insulin and glucose did not differ.
Frozen tofu did not beat whey protein on cholesterol or blood pressure
In this double-blind crossover trial, 45 adults aged 40 to 70 with mildly raised cholesterol ate kori-tofu, a frozen tofu, for 4 weeks. Total, LDL and HDL cholesterol, HbA1c (a long-term blood sugar marker) and systolic pressure fell during the tofu period. Most of those changes also appeared with whey, so the authors say there is no proof of a tofu effect. A Japanese food company, Asahimatsu Foods, funded the study.
Five years of isoflavone tablets went with more endometrial hyperplasia
In this double-blind trial, 376 postmenopausal women with a uterus took 150 mg of isoflavones a day (179 women) or placebo (197 women). Biopsies found endometrial hyperplasia, an overgrowth of the womb lining, in 3.37% on isoflavones against 0% on placebo, and no cancers. The authors say the result calls long-term safety of phytoestrogens for the endometrium into question. These were tablets, not soy foods, and the next trial found no such effect.
Three years of isoflavone soy protein did not thicken the womb lining
In this double-blind trial, 350 postmenopausal women took 154 mg of isoflavones in soy protein, or a milk protein placebo, each day. Of them, 224 who still had a uterus were analyzed. No soy-group biopsy showed hyperplasia or cancer, against 14.3% in the placebo group. That difference was not significant, so it could be chance.
Breast cancer patients eating most soy protein had 29% lower death risk
The Shanghai Breast Cancer Survival Study followed 5,033 women treated with surgery for a median of 3.9 years. There were 444 deaths and 534 recurrences or breast cancer deaths. Women in the top quarter of soy protein intake had 0.71 times the risk of death and 0.68 times the risk of recurrence. The link held in ER-positive and ER-negative cancers and in tamoxifen users and nonusers. It is observational, so soy eaters may differ in other ways.
Breast cancer survivors eating more isoflavones had 25% lower recurrence risk
This pooled analysis followed 9,514 survivors in 2 US cohorts and 1 Chinese cohort (groups followed over time) for 7.4 years on average. Those who ate at least 10 mg of isoflavones a day had 0.75 times the risk of recurrence. Risk of death from any cause (0.87) and from breast cancer (0.83) was lower too, but not statistically significant. Results were similar in US and Chinese women despite large differences in intake.
Soy food after a breast cancer diagnosis was linked to better survival
This meta-analysis pooled 5 cohort studies with 11,206 patients. Soy food was linked to 15% lower mortality (0.85 times the risk) and 21% lower recurrence (0.79 times). The link appeared in ER-positive and ER-negative disease. The authors say soy food intake might be associated with better survival; cohort studies cannot prove cause.
Isoflavones linked to lower breast cancer risk in Asian, not Western, women
This meta-analysis pooled 14 prospective studies (which follow people forward in time) of breast cancer incidence and 4 of recurrence. Higher intake went with 0.89 times the risk of breast cancer overall, 0.76 in Asian populations and 0.97 in Western ones. For recurrence the figure was 0.84. No dose-response link was found for incidence.
In Japan, fermented soy, not total soy, went with lower death risk
This cohort followed 92,915 adults aged 45 to 74 for 14.8 years, with 13,303 deaths. For the top against the bottom fifth of total soy, the risk of death was 0.98 times as high in men and in women. For fermented soy, such as natto and miso, it was 0.90 in men and 0.89 in women, about 10% lower. Natto also went with fewer cardiovascular deaths. The authors warn that unadjusted confounding (other differences between people) could explain the fermented soy link.
Across 23 cohorts, soy was linked to fewer cancer and cardiovascular deaths
This dose-response meta-analysis pooled 23 prospective studies with 330,826 participants. Soy was linked to 12% lower cancer deaths (0.88 times the risk) and 15% lower cardiovascular deaths (0.85). Soy protein alone was not linked to all-cause or cardiovascular death. But each 5 g a day was linked to 12% fewer breast cancer deaths.
Eating tofu weekly was linked to 18% lower coronary heart disease risk
These 3 US cohorts were the Nurses’ Health Studies and the Health Professionals study, with 8,359 coronary events. Tofu once a week or more, against less than once a month, went with 0.82 times the risk. The top fifth of isoflavone intake had 0.87 times the risk, and soy milk 0.87, which was not significant. In women the tofu link was stronger before menopause and in postmenopausal women not using hormones.
Across 27 cohorts, soy intake was not linked to type 2 diabetes
This federated meta-analysis combined cohorts in the Americas, Europe, the Eastern Mediterranean and the Western Pacific. It covered up to 42,473 cases among 807,785 adults. Pulses and soy showed no link. Total legumes had a weak link, a rate ratio of 1.02 per 20 g a day, driven by some European cohorts.
Soy foods linked to lower prostate cancer risk; fermented soy was not
This meta-analysis pooled 30 articles from observational studies. Total soy food, genistein, daidzein and unfermented soy intakes were each linked to lower risk; fermented soy and total isoflavones were not. Neither soy food nor isoflavones in the blood were linked to advanced prostate cancer, though very few studies exist.
About half of children with soy allergy outgrew it by age 7
This review of records covered 133 children seen at a referral clinic, followed for a median of 5 years. Soy allergy was predicted to resolve in 25% by age 4, 45% by age 6 and 69% by age 10. Higher soy IgE (allergy antibody) levels meant slower recovery, and 88% also had peanut allergy. Referral patients may be more severe than average.
In mice, genistein-rich soy isolate grew breast tumors more at higher doses
Study in mice, not people. The mice had implanted estrogen-dependent human breast cancer cells and ate soy isolates with 15, 150 or 300 ppm (parts per million) of genistein. Tumor growth rose with the dose, and cell division was greatest at 150 and 300 ppm. AICR notes that rodents break down isoflavones differently from humans, and the same effect has not been seen in people.
In mice, soy flour left breast tumors unchanged; purified isoflavones grew them
Study in mice, not people. Each diet gave the same amount of genistein (750 ppm) as soy flour, soy molasses, an extract, mixed isoflavones or genistin. Tumors on a soy-free diet shrank, tumors on soy flour stayed the same size, and tumors on the extracts, mixed isoflavones or genistin grew. The authors advise soy flour over purified isoflavones for postmenopausal women with estrogen-dependent breast cancer.
Homemade soy foods held far fewer isoflavones than commercial ones
These were lab measurements of isoflavones, not a study in people. Homemade soy milk held 20 times fewer isoflavones than commercial soy milk, tofu 2.6 times, tempeh 4.5 times and miso 9.8 times fewer. Soaking cut isoflavones in textured soy protein by more than 70%. The authors argue that intake estimates from commercial products may overstate what home cooks and traditional diets supply.
Who should be careful with soybeans
Most people can eat soybeans freely. Take a little care if one of these applies:
- Soy allergy. Soy is one of the nine major food allergens that US labels must declare. Many children outgrow it, about half by age 7 in a clinic study, though 88% of those children also had peanut allergy. If you react to soy, tofu, soy milk and edamame count too.
- Thyroid medicine. Soy does not appear to harm a healthy thyroid, as the studies above show. A review found some evidence that soy foods can reduce absorption of thyroid hormone tablets, so the dose may need adjusting. A later review found mostly case reports and one randomized trial with no difference. Adults with an underactive thyroid need not avoid soy; tell your clinic, keep your intake steady and make sure you get enough iodine.
- Raw or undercooked soybeans. Raw soybeans contain protease inhibitors and lectins that heat destroys, so cook dried soybeans thoroughly instead of eating them raw or lightly cooked.
How to choose, store and eat soybeans
- Cooked from dry: soak for 6 to 8 hours or overnight, because a quick soak works poorly for soybeans. Simmer for about 5 to 6 hours, skimming the foam, until a bean is soft when pressed. The nutrition numbers on this page are for cooked beans.
- Soy nuts, edamame and tofu: the soy nut trial that lowered LDL by 0.17 mmol/L used 67 g a day for 16 weeks. AICR counts 1/4 cup of soy nuts, 1/2 cup of shelled edamame or 1/3 cup of tofu as one serving. Each has about 7 g of protein and 25 mg of isoflavones.
- Tofu and soy milk for a sensitive gut: Monash rates firm tofu low in FODMAPs in a standard serve, and silken tofu high. FODMAPs are carbs that can upset a sensitive gut. Monash says soy milk made from whole soy beans tends to be high in GOS, a FODMAP, and soy milk from soy protein extract low.
Pick whole, clean dried beans and discard any that are dark or discolored. Freeze cooked soybeans in bags with a little of their cooking liquid. Residues are no reason to skip soybeans. In USDA tests of 608 samples of soybean grain in 2022, every residue found was below its legal limit.
Sources
- Cooking dried soybeans. University of Arkansas System Division of Agriculture.
- Soy and breast cancer: myths and misconceptions. American Institute for Cancer Research. 2024.
- Tofu and FODMAPs. Monash FODMAP blog. 2015.
- Milk alternatives on a low FODMAP diet. Monash FODMAP blog. 2016.
- Food allergies. US Food and Drug Administration.
- Savage JH, Kaeding AJ, Matsui EC, Wood RA. The natural history of soy allergy. J Allergy Clin Immunol. 2010.
- Messina M, Redmond G. Effects of soy protein and soybean isoflavones on thyroid function in healthy adults and hypothyroid patients: a review of the relevant literature. Thyroid. 2006.
- Gatta E, Maltese V, Cavadini M, et al. Interference or noninterference between soy and levothyroxine: that is the question. A narrative review of literature. Endocr Pract. 2023.
- Liener IE. Implications of antinutritional components in soybean foods. Crit Rev Food Sci Nutr. 1994.
- Tischmann L, Adam TC, Mensink RP, Joris PJ. Longer-term soy nut consumption improves vascular function and cardiometabolic risk markers in older adults: results of a randomized, controlled cross-over trial. Clin Nutr. 2022.
- Pesticide Data Program annual summary, calendar year 2022. USDA Agricultural Marketing Service.
History of soybeans
Soybeans were first domesticated in China. Wikipedia places the change from wild plant to crop in north-central China, probably 6,000 to 9,000 years ago. The earliest known remains of the wild ancestor, Glycine soja, are charred seeds from the Neolithic (late Stone Age) site of Jiahu in Henan province. They are about 9,000 to 7,800 years old.
The story is still being rewritten. A 2023 review notes seeds from Japan’s Jōmon period, 6,000 to 4,000 years old. They suggest soybean use and larger seeds began there earlier than in China and Korea. A 2025 study of the genomes of 8,105 soybean accessions (stored seed samples) found that black soybeans were a key stepping stone in domestication.
Tofu came later. Wikipedia says tofu making was first recorded in China during the Han dynasty (202 BC to AD 220), with a legend crediting Prince Liu An. It became popular under the Song dynasty (960 to 1279).
It appears in Japanese records by AD 965, brought by Zen Buddhist monks. The English word “soy” comes from the Japanese soi, a form of shōyu. That word comes from the Chinese jiangyou, meaning soy sauce.
Soybeans reached the West slowly. Samuel Bowen brought them from China to Georgia in 1765. That year Henry Yonge grew the first crop in North America on Skidaway Island. For the next 155 years soybeans were grown mainly as forage (animal feed).
In 1770 Benjamin Franklin, the first American to mention tofu, wrote about it to the botanist John Bartram. He called it cheese from China.
In 1931 Henry Ford hired chemists to make a soy protein fiber called Azlon, which never reached the market. He also used soybean oil in car paint. By 2000, US soybean plantings covered more than 70 million acres, second only to corn.
The heart claim has its own history. In 1999 the FDA allowed foods to state that 25 g of soy protein a day may reduce the risk of heart disease. The claim applied within a diet low in saturated fat and cholesterol. In October 2017 the FDA proposed revoking it after reviewing 709 publications, because newer trials gave inconsistent results.
It said it might allow a qualified (more limited) claim instead. The rule, 21 CFR 101.82, is still in the Code of Federal Regulations.
Sources
- Sedivy EJ, Wu F, Hanzawa Y. Soybean domestication: the origin, genetic architecture and molecular bases. New Phytol. 2017.
- Takahashi Y, Nasu H, Nakayama S, Tomooka N. Domestication of azuki bean and soybean in Japan: from the insight of archeological and molecular evidence. Breed Sci. 2023.
- Zhu Z, Wang Y, Liu S, et al. Genomic atlas of 8,105 accessions reveals stepwise domestication, global dissemination, and improvement trajectories in soybean. Cell. 2025.
- Soybean. Wikipedia.
- Tofu. Wikipedia.
- Food labeling: health claims; soy protein and coronary heart disease. Proposed rule, 82 FR 50324. US Food and Drug Administration. 2017.
- Food labeling: health claims; soy protein and coronary heart disease. Final rule, 64 FR 57700. US Food and Drug Administration. 1999.
- 21 CFR 101.82: health claims: soy protein and risk of coronary heart disease (CHD). Legal Information Institute, Cornell Law School.
Frequently asked questions about soybeans
Are soybeans good for you?
For most people, yes: a cup (172 g) of cooked soybeans has 296 calories, 31.3 g of protein and 10.3 g of fiber. It is a complete protein, and soy protein lowered LDL cholesterol by 4.76 mg/dL in the trials the FDA reviewed. In Japan, total soy intake was not linked to the risk of dying. The evidence is modest, and most trials tested soy protein, soy milk or isoflavone pills rather than cooked beans.
Does soy lower cholesterol?
A little: soy protein at a median of 25 g a day lowered LDL by 4.76 mg/dL, about 3 to 4%. An American Heart Association advisory put it at about 3%. The FDA proposed revoking its 1999 soy heart claim in 2017 because trial results varied, yet a cumulative analysis found the effect never lost significance. In the soy nut trial, 67 g a day for 16 weeks lowered LDL by 0.17 mmol/L.
Does soy act like estrogen, and does it help hot flashes?
Not in the way people fear. In a 40-trial review of postmenopausal women, isoflavones did not change estradiol, FSH, endometrial thickness or vaginal maturation. Isoflavone extracts cut hot flash frequency by 20.6% against placebo in 13 trials, but a Cochrane review of 43 trials found no conclusive evidence. A 5-year trial of 150 mg isoflavone tablets found more endometrial hyperplasia (3.37% against 0%); a 3-year trial of 154 mg found none.
Is soy safe if you have had breast cancer?
Soy foods appear to be, and AICR says it is safe to include them after a diagnosis. In the Shanghai study of 5,033 survivors, the top quarter of soy protein intake had 0.71 times the risk of death. Five pooled cohorts found 0.85 times the risk for mortality and 0.79 for recurrence, but all are observational, rated limited suggestive by AICR. The old worry came from mice fed genistein-rich soy isolates, and AICR notes rodents handle isoflavones differently; ask your cancer team about supplements.
Does soy lower testosterone in men?
No. A 2021 meta-analysis of 41 clinical studies found no significant effect of soy protein or isoflavones on total or free testosterone, estradiol or estrone. That held whatever the dose or study length, and total testosterone was measured in 1,753 men. In observational studies, higher soy food intake was linked to a lower risk of prostate cancer.
Is soy bad for your thyroid?
Not in the trials: a meta-analysis of 18 articles found no change in free T3 or free T4. TSH rose a little, by 0.248 mIU/L, which the authors say may not matter clinically. If you take thyroid hormone tablets, one review found some evidence that soy foods can reduce absorption. So tell your clinic about regular soy and keep your intake steady.
How much soy is okay per day?
No trial here set an upper limit. AICR describes studies using 1 to 2 servings a day, each about 7 g of protein and 25 mg of isoflavones. Servings include 1/3 cup of tofu, 1/2 cup of shelled edamame, 1/4 cup of soy nuts and 1 cup of soy milk. In Japan the top fifth of total soy intake had 0.98 times the risk of death. The soy nut trial used 67 g a day for 16 weeks without serious adverse events.
Are soybeans GMO?
Most US soybeans are: USDA data show 96% of US soybeans planted in 2025 were genetically engineered, mainly to tolerate herbicides. In USDA’s 2022 tests of 608 soybean grain samples, glyphosate was the most common residue, found in 558 samples. Every detection was below its legal limit. No trial here compared GMO with other soybeans, and organic and non-GMO soybeans are available if you prefer them.