Soybean benefits: the isoflavones did not act like oestrogen

Key benefits

1. 01Strong evidence

   Soy isoflavones did not act like oestrogen

   A 2025 systematic review and meta-analysis in Advances in Nutrition. MEDLINE, Embase and the Cochrane Library were searched to August 2024 for randomised trials of at least three months comparing isoflavones with non-isoflavone controls.

   Forty trials, 52 comparisons, 3,285 participants. Median dose 75 mg/day of isoflavones, median duration 24 weeks, in postmenopausal women.

   All four measures of oestrogenicity were unchanged.

   • Endometrial thickness −0.22 mm (95% CI −0.45 to 0.01), P = 0.059
   • Vaginal maturation index 2.31 (−2.14 to 6.75), P = 0.310
   • FSH −0.02 IU/L (−2.39 to 2.35), P = 0.987
   • Oestradiol 1.61 pmol/L (−1.17 to 4.38), P = 0.256

   Certainty was high to moderate for all four. This is not an absence of evidence; it is many trials consistently pointing at no change.

   The authors read this as isoflavones acting like selective oestrogen receptor modulators (SERMs), behaving clinically differently from the hormone itself.

   The breast cancer survival data point the same way. A 2019 meta-analysis in the European Journal of Nutrition pooled 12 articles covering 37,275 women with breast cancer. Highest versus lowest pre-diagnosis soy and isoflavone intake gave overall survival HR 0.84 (0.71–0.98). But breast-cancer-specific survival was not significant at HR 0.89 (0.74–1.07), and post-diagnosis intake gave an overall survival HR of 0.80 that was also not significant (0.62–1.04). These are observational studies pooled; they cannot be read causally.

   Now the interest disclosure. One co-author of the 2025 meta-analysis (M. Messina) was employed by the Soy Nutrition Institute Global, which is funded by the United Soybean Board. The paper disclosed this itself.

   That does not overturn the result. Forty randomised trials assessed with GRADE exist independently of who wrote them up. But the same author appears on both entries below, and a reader should know that.

   [1][2]
2. 02Moderate evidence

   LDL falls by roughly 3–4%

   A 2019 meta-analysis in the Journal of Nutrition, with an unusual backstory: the US FDA was considering revoking soy protein's heart health claim, and this paper re-analysed the exact 46 trials the FDA had identified as its basis.

   Of the 46, 43 supplied data; 41 supplied LDL. Baseline LDL across participants ran from 110 to 201 mg/dL.

   Comparing soy protein at a median of 25 g/day over a median of 6 weeks against non-soy protein:

   • LDL −4.76 mg/dL (95% CI −6.71 to −2.80, P < 0.0001, I² = 55%)
   • Total cholesterol −6.41 mg/dL (95% CI −9.30 to −3.52, P < 0.0001, I² = 74%)

   There was no dose-response effect and no evidence of publication bias.

   The look at individual trials is the interesting part. About 75% of them showed LDL falling, across a range from −0.77 to −58.60 mg/dL — but only a minority were individually significant. Each was too small to detect on its own; pooled, they pointed one way.

   The authors' summary is a reduction of about 3–4%.

   Take the size honestly. Someone at 130 mg/dL would move to about 125. That is the scale of one component of a diet, not a medicine.

   The same M. Messina is among the authors here as above.

   [3]
3. 03Moderate evidence

   Thyroid: fine if yours is normal, but space it from your medication

   "Soy is bad for the thyroid" is another old claim. A 2006 review in Thyroid went through it.

   It identified 14 trials that measured at least one thyroid function marker — eight in women only, four in men, two in both.

   With one exception, there were either no effects or very modest changes.

   The authors' summary splits three ways.

   • In euthyroid, iodine-replete people, there is little evidence that soy foods or isoflavones harm thyroid function.
   • In hypothyroid patients on thyroid hormone, there is evidence that soy inhibits absorption and may raise the dose required. That is not a reason to avoid soy foods.
   • In people with already compromised thyroid function or marginal iodine intake, a theoretical concern remains, based on in vitro and animal data.

   So their practical advice is one line: if you eat soy foods, make sure your iodine intake is adequate. In a Korean diet, sea mustard, kelp and gim do that work.

   An important limit — thyroid function was not the primary outcome in any of the 14 trials. This is a review of values measured on the side.

   The first author here is also M. Messina. The same person appears on all three.

   [4]
4. 04Nutrition data

   The highest protein and fat of any bean

   Per 100 g of boiled soybeans:

   • Protein 18.2 g — the highest among cooked pulses (lentils about 9 g, chickpeas 8.9 g, black beans 8.9 g)
   • Iron 5.14 mg — 64% of the adult requirement
   • Fibre 6.0 g — 24%
   • Potassium 515 mg, phosphorus 245 mg, magnesium 86 mg, calcium 102 mg
   • Manganese 0.824 mg, copper 0.407 mg, folate 54 µg, vitamin K 19.2 µg
   • 172 kcal

   What sets soy apart from other beans is the fat: 8.97 g per 100 g, against 0.4 g in lentils and 2.6 g in chickpeas. That is why soybeans are pressed for oil and other pulses are not.

   The fat breaks down as 1.3 g saturated, 1.98 g monounsaturated, 5.06 g polyunsaturated — over half polyunsaturated.

   The protein quality is high too. Soy is one of the few plant proteins carrying all nine essential amino acids in sufficient amounts, with an amino acid score close to animal protein.

   But the iron is non-haem, absorbed at only 2–10%. Eating it with vitamin C raises that.

   Treat it as a different food from tofu or soy milk. Tofu is the protein and fat strained out of the bean, so it has almost no fibre; soy milk is mostly water. Only eating the bean itself brings the 6.0 g of fibre with it.

How to eat it well

What to watch out for