Soy, Estrogen, and Equol: What Your Gut Microbes Change

Soy does not appear to act like estrogen therapy in postmenopausal women. A 2025 meta-analysis of 40 randomized controlled trials involving 3,285 participants found no statistically significant pooled effect of soy isoflavones on endometrial thickness, vaginal maturation index, follicle-stimulating hormone, or circulating estradiol. That is the best direct answer to the common fear that dietary phytoestrogens broadly disrupt female hormones.

It is not the whole answer. Gut bacteria convert one soy isoflavone, daidzein, into S-equol in some people but not others. This difference may help explain why an average result can look neutral while some individuals show measurable biological responses. The evidence supports caution with sweeping claims in either direction: soy is neither a hormonal wrecking ball nor a personalized estrogen prescription.

What the best trial evidence says

Viscardi et al.'s 2025 systematic review and meta-analysis pooled 40 randomized trials of soy isoflavones in postmenopausal women. The authors used a registered review protocol, PRISMA methods, a random-effects model, and the Cochrane RoB 2.0 tool to assess risk of bias.

The review examined four prespecified markers of clinical estrogenicity: endometrial thickness, vaginal maturation index, follicle-stimulating hormone, and estradiol. None showed a statistically significant pooled difference between isoflavone and control groups. These are surrogate outcomes, not breast-cancer incidence, fracture rates, or quality of life. Still, they are the right markers to examine when the concern is whether soy creates a classical estrogen-like effect on reproductive tissues or circulating hormones.

The available report summary does not provide the pooled mean differences, confidence intervals, or heterogeneity estimates for each endpoint, so those numbers should not be guessed. The conclusion is narrower and more useful: across the trial populations studied, soy isoflavone supplementation did not produce a statistically significant average change in these four markers.

Soy is not estradiol in a bean costume

Genistein and daidzein resemble 17β-estradiol enough to bind estrogen receptors, but their pharmacology differs from estradiol or prescribed menopausal hormone therapy. Reviews by Sekikawa et al. and Intharuksa et al. describe preferential binding to estrogen receptor beta, or ERβ, rather than estrogen receptor alpha, ERα.

ERα has a prominent role in proliferative signaling in breast and uterine tissue. ERβ is expressed in bone, blood vessels, and the central nervous system. This receptor preference is why researchers describe isoflavones as selective estrogen receptor modulators, or SERMs. That label does not mean every effect is beneficial, nor does it prove safety at every dose in every clinical setting. It means the compounds cannot be treated as interchangeable with estrogen therapy.

The microbiome changes the exposure

Daidzein becomes S-equol only when the right intestinal microbes are present. Leonard et al.'s 2022 review estimates that about 25% to 30% of Western women and 50% to 60% of Asian women produce equol. Adlercreutzia equolifaciens is one bacterium with the necessary pathway; a 2021 molecular biology study identified and expressed its equol-biosynthesis genes in other bacterial hosts.

S-equol has greater estrogenic and antioxidant activity than the parent isoflavones and shows ERβ-selective activity. Therefore, two people can eat similar amounts of soy yet produce different circulating metabolites. Many of the randomized trials in the meta-analysis did not measure or stratify participants by equol-producer status. If producers and non-producers respond differently, pooling them could dilute a subgroup effect.

That explanation is biologically plausible, but human producer-stratified trial evidence remains limited. A plausible explanation is not a substitute for the trial that would test it.

The opposing signal deserves attention

Fujitani et al. in 2021 found that higher urinary equol was positively associated with higher urinary estradiol excretion in women. This observational finding does not establish that equol raises estradiol. Equol could alter estrogen metabolism or clearance, or equol and estradiol excretion could track with shared dietary or microbiome factors. But it does mean that the population-average null result cannot prove estrogenic neutrality for every equol producer.

A 2024 gnotobiotic mouse study adds mechanistic support for treating producer status as meaningful. Researchers created equol-producing and non-producing mouse phenotypes under controlled conditions and observed divergent outcome trajectories after daidzein exposure. Mice are useful for isolating a biological variable; they cannot tell us the size of an effect on human symptoms, cancer risk, or fractures.

Premenopausal evidence is also less tidy than a social-media warning label. In a cross-sectional study of 348 Japanese women with regular menstrual cycles, Tsuji et al. found no significant association between urinary equol level and measured endogenous sex-hormone levels after adjustment for covariates. The study examined associations, not supplementation effects, and its Japanese population has a higher expected prevalence of equol production than Western populations.

Possible health effects are not proven outcomes

Equol production has been associated with favorable metabolic and obesity-related measures. In a cross-sectional study of 386 postmenopausal women, Takahashi et al. compared equol producers and non-producers using urinary equol, body mass index, waist circumference, and lifestyle data. The researchers reported lower obesity and metabolic-risk markers among producers. Cross-sectional data cannot separate an effect of equol from the dietary habits and other characteristics of people who produce it.

Frailty evidence has the same limitation. A 2026 cross-sectional analysis of 651 community-dwelling adults aged 70 to 85 found that higher urinary equol concentration was associated with lower frailty status. The study classified producers at urinary equol concentrations of at least 1,000 nmol/L. Kojima et al. provide an interesting association, not evidence that increasing equol will prevent frailty.

For a more ordinary dietary decision, a 2024 meta-analysis of randomized trials found that replacing cow's milk with soymilk did not adversely affect intermediate cardiometabolic markers. Erlich et al. tested substitution, not whether soymilk outperforms cow's milk. It supports soy as a compatible option, not as a cardiometabolic cure in a carton.

What to do with this information

Whole-food soy does not need to be avoided because of a general fear that it will raise estrogen. The strongest randomized evidence in postmenopausal women found no statistically significant average change in four clinical estrogenicity markers. Yet equol production is a real source of biological variation, and the current trials cannot provide a clean producer-specific estimate because most did not classify participants that way.

For routine dietary soy, the evidence supports a measured view: no population-level signal of classical estrogenicity, with unresolved individual variation driven in part by the gut microbiome.

Women considering long-term, high-dose isoflavone supplements—especially those with hormone-sensitive conditions or treatment histories—should discuss the decision with their clinician. A urinary equol test after a soy challenge can identify producer status, but no controlled human evidence yet shows that changing producer status with probiotics produces meaningful clinical benefits. The next useful trials will stratify by metabotype before asking soy to answer questions it was never designed to answer alone.