Soy milk: half a litre, and the bacteria in between

One study in this set actually poured soy drink into people. A network meta-analysis of 14 trials on 543 adults found that replacing 500 mL of cow milk a day with soy drink may lower LDL cholesterol by 0.47 mmol/L, graded low certainty, with no effect on HDL. It also reports falls of 8.23 mmHg systolic and 7.82 diastolic, which is a large swing for a food and is graded low certainty for that kind of reason. Half a litre a day is a lot of drink. In our own FNDDS measures a glass of soy drink is 8 fl oz, about 244 g, so the trial dose is roughly two glasses every day for at least three weeks. No trial there measured a heart attack. Everything else on this page was measured on something other than a carton: equol capsules, isolated soybean protein, fermented soy powder. The most useful fact about soy is that it may work only on some people. Daidzein from soy becomes equol only when the right gut bacteria are present, and in the pooled trials the benefit for hot flashes came from equol. The reviewers call that ability the likely main determinant of whether soy works, which may be why soy trials contradict each other without anyone being wrong.

The numbers, per 100 g

What your body absorbs

• Whether soy does anything for menopausal symptoms may depend mainly on a gut bacterium you either have or do not. — peri- and postmenopausal women, split by whether they produce equol from soy isoflavones (n = 6 studies (779 participants) reviewed, 5 (728 participants) meta-analysed) Evidence A source
  The meta-analysis found a significant benefit of equol for lowering hot flash scores at generally low risk of bias, and supplementing equol to non-producers lowered the incidence or severity of hot flashes; two of the five pooled studies individually found no significant benefit

What cooking and storage change

• In a 12-week trial of 61 adults aged 65 and over, a fermented soy powder improved memory scores against placebo, and fermented soy is a different food rather than the same one processed. — adults aged 65 and over, mean age 74 (n = 61 randomised, 47 completers) Evidence B source
  Memory scores improved in the fermented soy group against placebo (p=0.041); in a post hoc analysis of 29 women aged 70 and over, global cognition (p=0.028) and memory (p=0.049) improved, with adjusted mean change in global cognition 2.86% against 0.06% for placebo

What it does to your health

• Sixty trials on six thousand women found no bone benefit from soy as a food, and a small one from soy protein as an ingredient. — healthy women, mean age 54.2 years (n = 60 randomized trials, 6284 participants) Evidence A source
  No dairy or soy intervention at food level significantly improved outcomes against control; at component level soybean protein raised total body bone mineral density by 0.03 g/cm2 (95% CI 0.01 to 0.05) and casein plus whey by 0.04 g/cm2 (0.01 to 0.06)
• Half a litre of soy drink a day in place of cow milk may lower LDL cholesterol by 0.47 mmol/L over trials of at least three weeks, and the people who pooled the trials call the evidence low certainty. — adults in randomized trials (n = 14 trials, 543 participants) Evidence A source
  Replacing 500 mL/day of cow milk with soy drink may reduce LDL cholesterol by 0.47 mmol/L (95% CI -0.85 to -0.10, low certainty) with no effect on HDL cholesterol (moderate certainty), and may reduce systolic blood pressure by 8.23 mmHg and diastolic by 7.82 mmHg, both low certainty

What people believe that the data does not support

• Europe has refused the soy and cholesterol claim, and has authorised nothing at all about soy protein. — — Evidence E source
  The wording that soy protein has been shown to lower blood cholesterol and that lowering cholesterol may reduce the risk of coronary heart disease carries the status Non-authorised; four soy protein submissions were refused, including one built on 25 g of soy protein a day, and the number of authorised soy protein claims in the register is zero
• America allows on a package the exact sentence Europe refused, and prints the wording out for manufacturers to copy. — — Evidence E source
  The model health claim reads that 25 grams of soy protein a day, as part of a diet low in saturated fat and cholesterol, may reduce the risk of heart disease, with the serving's soy protein content to be filled in by the manufacturer

What is still unknown

• Observational data link higher soy intake in adolescence with about a third less breast cancer in adulthood, and isoflavones on their own showed nothing. — adolescent girls followed into adulthood, studies mostly from the United States (n = 51 studies, of which 20 were rated high quality) Evidence C source
  Higher adolescent soy intake was associated with lower adult breast cancer risk (RR 0.67; 95% CI 0.55 to 0.82; 3 studies), alongside fruits and vegetables (RR 0.90), fibre (RR 0.78) and vegetable fat (RR 0.76); no significant association appeared for meat, fish, eggs, dairy, milk, grains, alcohol, total fat, animal fat or isoflavone

The bottom line

Two regulators read the same trials and reached opposite conclusions, and you should see that rather than be handed one of them. In the United States, 21 CFR 101.82 lets a package say that 25 grams of soy protein a day, as part of a diet low in saturated fat and cholesterol, may reduce the risk of heart disease. The regulation prints that wording out for manufacturers to copy. Europe refused that same sentence, turned down four soy protein submissions, and has authorised nothing about soy protein at all. Twenty-five grams a day is roughly three glasses of soy drink, which is more than most people who buy it are drinking. On bone, the split between a food and an ingredient is sharp. Across 60 randomised trials on 6284 women, no dairy or soy intervention at food level beat the control, while isolated soybean protein raised total body bone mineral density by 0.03 g per square centimetre. That is a small number and it is a long way from a fracture prevented. Soy is also a food people are told to fear, and the evidence does not support the fear. In three studies of what girls ate as teenagers, higher soy intake was associated with about a third less breast cancer in adulthood, while isoflavone intake on its own showed nothing. Association is all this is. Nobody fed anyone soy and waited. Those are adults recalling adolescence, in a literature where only 20 of 51 studies were rated high quality. Read it as a reason to stop worrying rather than a reason to start drinking. One warning about labels. Fermented soy is a different food, because fermentation produces peptides and aglycone isoflavones that unfermented soybeans do not contain. The cognition trial you may have seen used a fermented powder in 61 older adults, lost a quarter of them along the way, and its headline finding came from a post hoc subgroup. An ordinary carton of soy milk is not a fermented product. This page also has no evidence on who should be careful with soy. Nothing in this set of studies looked at soy as an allergen or at soy alongside medication, and that is an absence of evidence here, not a finding that there is nothing to know.

Sources

1. sym-r1-01 · systematic review and meta-analysis of randomized trials
   Soy isoflavones may benefit some, but not all, menopausal women, and the ability of the women to produce equol may be the major determinant of effectiveness.
   J Med Food, 2019 · checked 2026-09-22
2. sym-r1-07 · randomized controlled trial
   Fermented soy contains bioactive compounds not found in unfermented soybeans, such as peptides and aglycone isoflavones, which may support cognition by reducing neuroinflammation and oxidative stress.
   Nutrients, 2025 · checked 2026-09-22
3. sym-r1-02 · network meta-analysis of randomized trials
   At the food level, no dairy or soy interventions significantly improved outcomes versus control, although milk + yogurt ranked numerically highest based on SUCRA values.
   Nutrients, 2025 · checked 2026-09-22
4. sym-r1-04 · network meta-analysis of randomized trials
   Replacing 500 mL/d of cow's milk with soy drink may reduce low-density lipoprotein (LDL) cholesterol [MD: -0.47 mmol/L (-0.85; -0.10); low certainty], but showed no effect for high-density lipoprotein (HDL) cholesterol [MD: 0.01 mmol/L (-0.03; 0.05); moderate certainty].
   Adv Nutr, 2026 · checked 2026-09-22
5. sym-r1-05 · regulatory register entry
   POL-HC-6260 Soy protein Soy protein has been shown to lower/reduce blood cholesterol; blood cholesterol lowering may reduce the risk of (coronary) heart disease -/- Non-authorised
   EU Register on nutrition and health claims, snapshot of 2026-09-21 · checked 2026-09-22
6. sym-r1-06 · regulation
   (1) 25 grams of soy protein a day, as part of a diet low in saturated fat and cholesterol, may reduce the risk of heart disease.
   21 CFR 101.82, Health claims: Soy protein and risk of coronary heart disease · checked 2026-09-22
7. sym-r1-03 · systematic review and meta-analysis of observational studies
   Higher adolescent intakes of fruits and vegetables [relative risk (RR): 0.90; 95% confidence interval (CI): 0.82, 0.99; n = 3 studies], soy (RR: 0.67; 95% CI: 0.55, 0.82; n = 3), dietary fiber (RR: 0.78; 95% CI: 0.67, 0.92; n = 3), and vegetable fat (RR: 0.76; 95% CI: 0.66, 0.88; n = 2) were associated with lower risks of breast cancer in adulthood.
   Adv Nutr, 2025 · checked 2026-09-22