Soy: 41 studies on testosterone, and a label claim FDA tried to withdraw

The question most people bring to soy has a clear answer. Pooling 41 studies, soy protein and isoflavones left men's hormones where they were. Total testosterone was measured in 1753 men, free testosterone in 752 and estradiol in 1000, and nothing moved under any statistical model the authors tried, at any dose, over any duration. Read what went into those studies, though. They fed soyfoods, soy protein and isoflavone extracts, and some of the extracts came from red clover rather than from soy. The heart claim American labels may still carry has the same shape. It is written for 25 grams of soy protein a day, the protein pulled out of the bean and sold as a powder, and in 2017 FDA proposed to revoke it because the totality of the evidence had stopped holding together. That proposal was never finalised, so the claim stays legal. Europe read the same case for isolated soy protein and refused it.

The numbers, per 100 g

What cooking and storage change

• Fermented soy shifted waistlines, visceral fat, glucose and cholesterol in pooled trials. — participants in 15 eligible randomised controlled trials out of 2205 records screened (n = 15 trials) Evidence A source
  Body mass index fell by 0.14 kg/m2 (95% CI -0.28 to -0.01, P = 0.039), waist circumference by 1.50 cm (-2.94 to -0.07, P = 0.04), visceral fat by 692.17 mm2 (-1011.58 to -372.77, P < 0.001), fasting plasma glucose by 6.39 mg/dL (-10.38 to -2.40, P = 0.002) and total cholesterol by 5.0 mg/dL (-6.60 to -3.39, P < 0.001); other parameters did not change significantly

What it does to your health

• The link between soy isoflavones and lower breast cancer risk shows up when women are asked to remember, and fades when they are followed forward. — participants in 7 cohort studies and 17 case-control studies (n = 24 studies) Evidence C source
  The summary odds ratio for breast cancer comparing highest against lowest isoflavone intake was 0.71; the inverse association was significant in the case-control studies but not in the cohort studies; below 10 mg per day of isoflavones no effect on risk was detected; in the cohort dose-response models a 10 mg per day increase gave reductions of 6.8% and 3.2% depending on the model
• Across every meta-analysis anyone has done, soy foods come out more helpful than harmful, except the fermented ones. — 23 meta-analyses of observational research covering 39 unique health outcomes, plus one meta-analysis of randomised controlled trials covering four outcomes (n = 24 meta-analyses) Evidence C source
  Dietary soy food consumption was generally more beneficial than harmful, particularly for cancer and cardiometabolic disease, but the benefits of fermented soy products were inconclusive and some studies indicated possible adverse effects; the authors call for more studies on biological indicators

What people believe that the data does not support

• American labels may still say 25 grams of soy protein a day helps the heart. — — Evidence E source
  The model claim permitted on a label states 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
• FDA asked to withdraw its own soy and heart claim, saying the evidence no longer holds together. — — Evidence E source
  FDA stated that although some evidence suggests a relationship between soy protein intake and reduced risk of coronary heart disease, the totality of the evidence is inconsistent and not conclusive, and therefore does not support the agency's earlier finding of significant scientific agreement; the rule was proposed and has not been finalised, so the claim in 21 CFR 101.82 remains in force
• Europe looked at the same soy and cholesterol claim and refused it. — — Evidence E source
  The single register entry for isolated soy protein, claiming that the protein-rich soybean component lowers blood cholesterol and that lowering cholesterol may reduce the risk of coronary heart disease, carries the status Non-authorised; authorised claims for soy protein in this snapshot number zero

Who this works differently for

• Soy does not lower men's testosterone or raise their oestrogen, and 41 studies say so. — adult men across 41 studies; total testosterone measured in 1753 men, free testosterone in 752, estradiol in 1000, estrone in 239 and sex hormone binding globulin in 967 (n = 1753 for total testosterone) Evidence A source
  No significant effects of soy protein or isoflavone intake were found on total testosterone, free testosterone, estradiol, estrone or sex hormone binding globulin under any statistical model, and sub-analysis by dose and duration also showed no effect
• Women who already had breast cancer and ate more isoflavones did not do worse, and pooled data suggest they did better. — 592 breast cancer survivors followed prospectively, plus 14 observational studies of isoflavones eaten as food (n = 592 plus 14 studies) Evidence C source
  In the prospective cohort, 47 recurrences occurred over a median 4.3 years and the hazard ratio for the highest against lowest tertile of isoflavone intake was 1.29 (0.60 to 2.78); in the pooled analysis the hazard ratios comparing extreme categories were 0.81 (0.67 to 0.98) for recurrence and 0.85 (0.76 to 0.96) for all-cause mortality, with a nonlinear inverse association

What is still unknown

• Most of the soy and breast cancer evidence comes from one part of the world and one study design. — the 24 included observational studies, published between 2001 and 2021 (n = 24 studies) Evidence C source
  17 of the 24 included studies were case-control and 7 were cohort; the average Newcastle-Ottawa score was 7.6 with a range of 6 to 9, with eight studies scoring 7, nine scoring 8 and four scoring 9; the studies were carried out mainly in Asia, with a minority in America

The bottom line

If you have had breast cancer, the honest summary is that eating soy has not been shown to make things worse, and the pooled numbers lean the other way. Across 14 studies, women eating the most isoflavones had a recurrence hazard of 0.81 and an all-cause mortality hazard of 0.85. Inside that same paper sits a single prospective cohort of 592 survivors with 47 recurrences over about four years, and there the number ran the wrong way at 1.29, with a confidence interval from 0.60 to 2.78. That is a small study saying it does not know. The prevention picture is weaker again. The odds ratio of 0.71 for breast cancer comes out significant when women are asked to remember what they ate, and stops being significant when they are followed forward. Seventeen of the 24 studies were case-control, and most were run in Asia, where soy is eaten from childhood. Below 10 milligrams of isoflavones a day nothing was detectable at all. Fermented soy is where our own sources disagree in public. Fifteen randomised trials found small shifts in body mass index, waist, visceral fat, fasting glucose and total cholesterol, the largest of them 1.50 centimetres off the waist. The umbrella review that read every meta-analysis on soy calls the fermented results inconclusive and says some studies point to possible harm. We would rather hand you that disagreement than pick a winner. Two things belong on a soy page and are not here. We hold no evidence card on soy as an allergen, and none that separates soy eaten as food from isoflavones taken as a capsule, even though the trials above fed both and some of the capsules were not made from soy at all. If you are avoiding soy because of an allergy, or thinking about an isoflavone supplement, this page has nothing for you, and that is our gap rather than a verdict.

Sources

1. soy-dr-05 · systematic review and meta-analysis of randomised controlled trials
   Results demonstrated significant beneficial effects of fermented soy on body mass index (WMD = -0.14 kg m-2, 95% CI: -0.28, -0.01, P = 0.039), waist circumference (WMD = -1.50 cm, 95% CI: -2.94, -0.07, P = 0.04), visceral fat (WMD = -692.17 mm2, 95% CI: -1011.58, -372.77, P < 0.001), fasting plasma glucose (WMD = -6.39 mg dL-1, 95% CI: -10.38, -2.40, P = 0.002), and total cholesterol (WMD = -5.0 mg dL-1, 95% CI: -6.60, -3.39, P < 0.001) compared with controls.
   Food Funct, 2025 · checked 2026-09-22
2. soy-dr-02 · meta-analysis of observational studies
   The inverse association was significant in the case-control studies but not in the cohort studies.
   Nutrients, 2023 · checked 2026-09-22
3. soy-dr-06 · umbrella review of meta-analyses
   However, it is important to note that the benefits of fermented soy products were inconclusive, and even some studies indicated their possible adverse effects.
   Mol Nutr Food Res, 2026 · checked 2026-09-22
4. soy-dr-07 · regulatory authorisation
   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, заява про здоров'я: соєвий білок і ризик ішемічної хвороби серця · checked 2026-09-22
5. soy-dr-08 · proposed regulatory action
   Our reevaluation of the totality of the publicly available scientific evidence indicates that, although some evidence suggests a relationship between soy protein intake and reduced risk of CHD, the totality of the evidence is inconsistent and not conclusive.
   Federal Register, 82 FR 50324, пропозиція FDA скасувати заяву про соєвий білок і ІХС, 31 жовтня 2017 · checked 2026-09-22
6. soy-dr-09 · regulatory register entry
   POL-HC-6237 Isolated soy protein Protein-rich soybean component 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, заява POL-HC-6237, знімок 2026-09-21 · checked 2026-09-22
7. soy-dr-01 · meta-analysis of clinical studies
   Regardless of the statistical model, no significant effects of soy protein or isoflavone intake on any of the outcomes measured were found.
   Reprod Toxicol, 2021 · checked 2026-09-22
8. soy-dr-04 · prospective cohort analysis with meta-analysis
   The combined HRs (95% CIs) comparing the extreme categories were 0.81 (0.67-0.98) for recurrence and 0.85 (0.76-0.96) for all-cause mortality.
   Nutr Cancer, 2024 · checked 2026-09-22
9. soy-dr-03 · study characteristics within a meta-analysis
   Of the 24 included studies, 17 were case-control studies [5,6,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36] and 7 were cohort studies [7,8,9,37,38,39,40].
   Nutrients, 2023 · checked 2026-09-22