Soy misconceptions: hormones, GMOs and the environment

Of the three main charges brought against soy, two collapse under the evidence and one needs rewording. A meta-analysis of 41 clinical trials found no effect of soy or isoflavones on testosterone, measured in 1753 men. GM soy cannot end up in your Swiss tofu, because the law forbids it and the supply chain does not import it. Soy does weigh on deforestation, but 77 % of world production goes into animal troughs rather than into blocks of tofu.

One genuine precaution remains, and most articles about soy skip it: if you take levothyroxine, soy can reduce how much of your medication you absorb. This is not a matter of hormonal disruption but a matter of timing. What follows sets out the evidence behind each of these points, with the figures and the sources.

What soy is, in three orders of magnitude

Soy is a legume native to East Asia, farmed for several thousand years and, over the past sixty, turned into one of the most traded crops on the planet. Three numbers frame the whole debate.

First, where it goes. According to data compiled by Our World in Data, more than three-quarters of global soy, 77 %, feeds farm animals, mostly as processed cake. A small share is eaten directly by humans as tofu, tempeh, soy drink or edamame.

Second, genetic modification. Roughly 80 % of the soy grown worldwide is genetically modified, and the share approaches 100 % in Brazil, Argentina and the United States, the three leading producers. The figure alarms people, but it describes a market Switzerland does not take part in, for reasons set out below.

Third, the doses. In Western cohorts, the highest isoflavone intake category typically sits at 3 mg per day or less, about two tablespoons of soy drink. In Asian populations where soy is an everyday food, the same category exceeds 50 mg per day. Clinical trials test between 40 and 100 mg. That fifteen-fold gap explains why most Western observational findings on isoflavones say very little: exposure there is too low to produce a physiological effect at all.

Hormones: what the trials actually measured

Isoflavones are not oestrogens

Soy contains isoflavones, chiefly genistein, daidzein and glycitein. These molecules belong to the phytoestrogen family and are structurally similar to human oestradiol, which lets them bind to oestrogen receptors.

The resemblance stops there, and the detail matters. Isoflavones bind preferentially to the beta oestrogen receptor, whereas human oestradiol binds both subtypes with comparable affinity. The two receptors are not distributed the same way through the body and do not trigger the same responses. Depending on the tissue, an isoflavone can therefore behave as a weak agonist or as an antagonist. That is the profile of a modulator, not of a replacement hormone, and it explains why the intuition “it looks like an oestrogen, so it acts like one” turns out to be wrong once anyone measures it.

Testosterone: 41 trials, no effect

This is where the demonstration is cleanest. A meta-analysis published in Reproductive Toxicology in 2021 pooled 41 clinical trials in adult men consuming soy foods, soy protein or isoflavone extracts. Total testosterone was measured in 1753 men, free testosterone in 752, oestradiol in 1000, oestrone in 239 and SHBG in 967.

The result leaves little room for interpretation: no significant effect on any of these measures, whatever statistical model was applied. The authors then split the data by isoflavone dose and by study duration, and no effect appeared there either. Fears of feminisation rest on a handful of heavily reported individual cases and on extrapolations from animal studies, not on human trials.

Breast cancer: the most counter-intuitive reversal

This is where most articles stop short, simply pointing readers towards a health professional. The data deserve better.

The Shanghai Breast Cancer Survival Study, published in JAMA in 2009, followed 5042 women who had survived breast cancer. Compared with the lowest quartile of soy protein intake, the highest quartile was associated with a hazard ratio of 0.71 (95 % confidence interval 0.54 to 0.92) for all-cause mortality and 0.68 (0.54 to 0.87) for recurrence. In absolute terms, four-year mortality fell from 10.3 % to 7.4 %. The association held in women whose tumours expressed oestrogen receptors and in those whose tumours did not, and in tamoxifen users as well as non-users.

A meta-analysis published in 2013 in the Asian Pacific Journal of Cancer Prevention pooled five cohorts and 11 206 patients, with the same direction of effect: a hazard ratio of 0.85 (0.77 to 0.93) for mortality and 0.79 (0.72 to 0.87) for recurrence.

Two caveats apply. These are observational studies, so associations rather than proof that adding soy improves prognosis. And much of the data comes from Asian populations whose soy intakes are far higher and often begin in childhood, which limits direct transfer. What the data do establish is the absence of any danger signal, which is enough to retire the idea that a woman treated for breast cancer should avoid tofu.

Concentrated supplements: a different question

EFSA reviewed the safety of isolated isoflavones in food supplements for post-menopausal women in 2015. Its conclusion: no indication of harmful effects on the mammary gland, the uterus or the thyroid at the doses found in supplements sold in Europe.

Two limits come attached to that conclusion, and quoting one without the other is dishonest. The studies reviewed covered durations from three months to just under a year, which says nothing about taking a supplement for ten years. And the agency found no data allowing it to assess safety in women during the transition to menopause or in those with a personal or family history of cancer. Men and infant formula were outside the scope of the assessment.

So keep the distinction in mind: eating tofu and swallowing a capsule of concentrated isoflavones are not the same act. The second is a medical decision, the first is cooking.

Soy infant formula

Soy-based infant formula attracts a large share of the worry, and it has been studied. A Korean cohort published in the Journal of Nutrition in 2021 compared 11 535 children fed soy formula with 142 864 fed cow’s milk formula, out of 153 841 children in total, using propensity score matching. No significant association was found with epilepsy (adjusted hazard ratio 1.318; 0.825 to 2.106), attention deficit hyperactivity disorder (1.26; 1.00 to 1.60), autism spectrum disorder (0.99; 0.54 to 1.83) or developmental delay.

On the skeletal side, a longitudinal study published in the American Journal of Clinical Nutrition in 2023 followed 433 children from 3 months to 6 years and found no significant difference in bone mineral density between breast milk, dairy formula and soy formula.

None of this makes soy formula a default choice. The NHS is explicit: only give a baby soy formula if a health professional has advised it. The reason is not hormonal. It is that soy formula does not suit every situation, not least because cow’s milk protein allergy is frequently accompanied by a reaction to soy.

The one interaction that genuinely calls for care

Here is the point missing from the vast majority of articles about soy.

A review published in Thyroid in 2006, covering 14 trials, concludes that in euthyroid, iodine-replete people, soy foods and isoflavones do not adversely affect thyroid function, or only very modestly. The same review, however, reports evidence that soy, by inhibiting absorption, may increase the dose of synthetic thyroid hormone that hypothyroid patients require. The authors add that hypothyroid adults need not avoid soy foods, and stress one condition: regular soy consumers should make sure their iodine intake is adequate.

In practice the answer is a schedule, not a ban. The NHS advises taking levothyroxine on an empty stomach, 30 to 60 minutes before the first meal of the day, before any caffeinated drink and before any other medicine. Soy joins a list that already includes calcium and iron supplements, antacids and several other treatments. If you take levothyroxine and you change your soy intake substantially, tell your doctor: it is the dose that gets reviewed, not the tofu that gets dropped.

Iodine is the critical nutrient in this chapter, and our article on avoiding iodine deficiency without fish sets out the reliable sources and the Swiss reference figures.

GMOs: what Swiss law actually says

A moratorium, not a preference

The usual advice is to check labels to avoid transgenic soy. In Switzerland that precaution is largely beside the point, and the reason is legal.

The Federal Act on Non-Human Gene Technology (SR 814.91) states in article 37a that no authorisation may be granted, up to 31 December 2030, for placing genetically modified plants, plant parts, seeds or other propagating material on the market for agricultural, horticultural or forestry purposes. The current wording of that article comes from an amendment of 20 June 2025, in force since 1 August 2026. Nobody grows transgenic soy in Switzerland, and it is not a question of labelling.

The 0.9 % threshold and what it means

For imports, the FDHA Ordinance on Genetically Modified Foodstuffs (SR 817.022.51) requires, in article 8 paragraph 1, the statement “produced from genetically modified X”. Paragraph 7 sets out the only tolerance: the statement may be omitted if no ingredient contains more than 0.9 % by mass of such material, and if it can be shown that appropriate measures were taken to prevent its presence.

This threshold is not permission to put 0.9 % GMO into a product. It covers unintentional mixing during transport, storage or processing, and it demands proof of a preventive effort. Tofu sold in Switzerland with 0.9 % of deliberately added transgenic soy would be illegal.

The blind spot nobody mentions

Here is a paradox worth knowing. The labelling obligation applies to foods that are GMOs, contain them or are produced from them. It does not apply to milk, meat or eggs from animals fed on genetically modified feed, because the animal product itself is not a GMO product within the meaning of the ordinance.

The consequence is counter-intuitive: in a country where GMO labelling applied without exception, the consumer most exposed to transgenic soy would be the meat eater, not the tofu eater. In Switzerland that gap is closed in practice by private standards. According to Soja Netzwerk Schweiz, the Swiss label programmes (IP-Suisse, Naturafarm, Bio Suisse Bud, Suisse Garantie and others) all require GM-free animal feed, so imported feed soy is GM-free too.

The soy that actually enters Switzerland

Figures from Soja Netzwerk Schweiz give the measure of the subject. Switzerland imported 266 557 tonnes of feed soy in 2025. Those imports are certified against sustainability standards, GM-free and deforestation-free, and 97.8 % of them come from Europe, notably Italy, Ukraine, Germany and Austria. The network estimates that associated CO2 emissions have fallen by 85 % compared with 2010.

Domestic production exists but stays modest: around 3300 hectares yielding roughly 7000 tonnes of beans a year, of which some 5000 tonnes go to animal feed. Swiss tofu exists; it is simply made in large part from European beans. Our article on the local tofu economy in Switzerland covers the supply chain in detail.

Environment: the number that changes the question

Where the soy really goes

The reasoning “I eat tofu, therefore I help destroy the Amazon” rests on a confusion of flows. More than three-quarters of world soy, 77 %, is destined for animal feed. About 7 % is fed directly to animals as whole beans; the rest goes through crushing first, with the oil extracted and the protein-rich cake sent to fatten livestock, mainly pigs and poultry. A further share feeds biofuels and industry. Direct human consumption, meaning tofu, tempeh, soy drink, miso and edamame, takes up a minority slice of the total.

What actually drives deforestation

The data assembled by Our World in Data from the work of Pendrill and colleagues break tropical deforestation down by type of production. Pasture expansion for cattle comes first at 41 % of the total, and Brazilian beef alone accounts for 24 % of global tropical deforestation, which works out at roughly 72 % of Brazil’s deforestation. Oilseeds, mainly soy and palm oil, account for 18 %.

Soy is not innocent, and it would be wrong to pretend otherwise. But it sits in second place, well behind cattle ranching, and most of what it produces feeds precisely that ranching.

The moratorium and its leak

The story of the Brazilian soy moratorium deserves telling in full, because it contains both a success and its downside.

In 2006, under pressure from retailers and NGOs, the major soy traders signed the Soy Moratorium, the world’s first voluntary zero-deforestation agreement. They undertook not to buy soy grown on land in the Brazilian Amazon deforested after July 2006. The result is measurable: in the two years before the agreement, 30 % of soy expansion in the region came at the expense of forest; by 2014 that share had fallen to 1 %.

The downside has a name, the Cerrado. Production shifted to that wooded savannah south of the Amazon, often at the expense of its native vegetation, which enjoys neither the same protection nor the same media attention. By 2015 only 13 % of Brazilian soy came from the Amazon against 48 % from the Cerrado. It is the textbook case of leakage: an agreement that works within its perimeter can displace the problem instead of solving it, which is why several researchers have called for the moratorium to be extended beyond the Amazon.

Soja Netzwerk Schweiz also reports that the agreement is now under pressure: in early 2026, leading Brazilian industry associations including Abiove announced their withdrawal from the moratorium under political influence. If that withdrawal takes effect, the sector’s main safeguard for twenty years disappears.

What a Swiss consumer can actually do

The practical conclusion is less glorious than “boycott soy”, but it is more accurate. Since most of the pressure on land comes from meat and from the cake that feeds it, the main lever is the quantity of animal products on the plate, not the brand of tofu. For food-grade soy itself, the European origin that dominates the Swiss market settles the deforestation and GMO questions in practice.

The same logic extends across the aisle: our comparison of the carbon impact of plant-based drinks versus cow’s milk shows the same gap between intuition and data.

What soy delivers, and what it does not

Protein

Soy is one of the few plant sources whose indispensable amino acid profile covers requirements without particular complementation, including lysine, the amino acid most often limiting in cereals. That is what earns it the “complete protein” label.

One point of vocabulary deserves care, though. Protein quality measures such as DIAAS put isolated soy protein above soy flour, which in turn sits above wholemeal wheat but below dairy proteins. Soy is therefore excellent among plants without being equivalent to whey. This has no practical importance in a varied diet, where the range of sources evens out the differences. Our complete guide to plant proteins gives the portions and combinations that matter.

LDL cholesterol: a real, modest, correctly attributed effect

The cholesterol reduction is often credited to soy’s soluble fibre. That is inaccurate, and the mistake is a common one.

A meta-analysis published in the Journal of Nutrition in 2019 revisited the 46 controlled trials identified by the US FDA when it was considering revoking the health claim granted to soy protein. Across the 43 usable trials, a median dose of 25 g of soy protein per day over a median six weeks reduced LDL by 4.76 mg/dl (95 % confidence interval -6.71 to -2.80), roughly 3 to 4 %.

Three things are worth noting. The effect is attributed to soy protein, compared with non-soy protein, not to fibre. It is real and statistically solid. And it is modest: 3 to 4 % less LDL replaces no treatment, but it adds to the other dietary levers.

Minerals, and the phytate nuance

Tofu set with calcium sulphate is a worthwhile source of calcium; tofu set with nigari (magnesium chloride) is not. The information sits in the ingredient list, and the difference is far from trivial for anyone relying on tofu for calcium. Our article on calcium without dairy covers the alternatives.

Soy also supplies iron and zinc, but in non-haem form, whose absorption is reduced by phytates. Fermentation, soaking and sprouting all lower phytate content, which is why tempeh and miso behave differently from plainly cooked beans. The practical techniques are set out in our article on optimising plant iron absorption.

What soy does not fix

No unfortified soy food supplies vitamin B12. Tempeh and miso sometimes contain traces of microbial origin, in quantities too variable to count as a source. The Swiss FSVO notes that a vegan diet requires B12 supplementation through tablets or fortified products, and that even a lacto-vegetarian diet without eggs can fall below the recommendation. Our reference points are in the article on vitamin B12 needs and reliable sources.

Nor does soy supply iodine in useful amounts, which matters all the more given that the 2006 Thyroid review insists on adequate iodine intake for regular soy consumers.

How much to eat, in practice

The Swiss FSVO sets no soy-specific quota and treats it as one protein source among others. In the Swiss recommendations, a protein portion corresponds to 120 g of tofu or 60 g of dry pulses, at roughly one portion a day for an adult, on top of dairy for those who eat it.

On that basis, one to three portions of soy products a day fits comfortably into a balanced vegetarian diet. Two practical pointers:

  • Spread it rather than concentrating it. A 200 g block of tofu split across two meals serves your intake balance better, and that is all it does.
  • Do not make soy your only pulse of the week. Lentils, chickpeas, dry beans and split peas bring different fibre and micronutrient profiles. Our article on varying plant protein sources sets out simple rotations.

And if you want or need to do without it, cooking without soy: protein-rich alternatives covers substitutions with no nutritional loss.

Soy allergy: the one genuine reason to avoid it

Soy is a declarable allergen in Switzerland as in the European Union, meaning it must be highlighted in the ingredient list of any prepacked product containing it. That matters for the people concerned, because soy hides in many processed products as lecithin, oil or textured vegetable protein.

Soy allergy is more common in young children and often resolves with age, though not always. It is also why soy formula does not always suit an infant allergic to cow’s milk protein: the two allergies frequently coexist. Unlike the hormonal worries, this is a documented and sometimes serious risk, one that belongs to medical diagnosis rather than personal choice. Our allergen-free cooking guide offers complete menus for these situations.

For infants and young children generally, the NHS notes that unsweetened calcium-fortified plant drinks, including soy drink, can be offered as a main drink from the age of 1 but not before, and that rice drinks should be avoided under the age of 5 because of their arsenic content.

The products, and what really sets them apart

The word “soy” covers foods whose properties differ markedly, something product lists tend to obscure.

Tofu. Made by coagulating soy drink and pressing the curd. Its firmness depends on pressing time, and its nutritional interest depends on the coagulant, as above. Firm tofu suits grilling and stir-frying, silken tofu suits creams and desserts. Our marinade techniques for tofu and tempeh address the main complaint levelled at this food, its blandness.

Tempeh. Whole soybeans fermented by Rhizopus and moulded into a firm cake. Fermentation lowers phytates and changes both texture and flavour, which is more pronounced than tofu’s. Making it at home is covered in our homemade tempeh guide.

Miso and soy sauce. Long-fermented products used as condiments rather than protein sources: their salt content confines them to small quantities. Miso is a remarkable flavour enhancer, as our article on miso in Western cuisine shows.

Edamame. Beans harvested immature and cooked in the pod. They are eaten as a vegetable rather than as a dry pulse.

Soy drink. Soaked beans ground and filtered. Always check two things on the label: fortification with calcium and vitamins B12 and D, which is not automatic, and the absence of added sugars in flavoured versions.

Six common mistakes

  1. Confusing food with supplement. A block of tofu and a capsule of concentrated isoflavones raise different questions. EFSA assessed only the second, and only in post-menopausal women.
  2. Reading the 0.9 % threshold as permission. It covers proven unintentional mixing, not deliberate addition.
  3. Taking levothyroxine with a soy latte. The problem is timing, not soy. Empty stomach, 30 to 60 minutes before any food or caffeinated drink.
  4. Crediting the LDL reduction to soy fibre. The 2019 meta-analysis attributes the effect to the protein, with a fall of about 3 to 4 %.
  5. Assuming all tofu is high in calcium. Only calcium-sulphate-set tofu is. The coagulant appears in the ingredient list.
  6. Boycotting tofu for the Amazon while leaving meat consumption untouched. The 77 % of world soy that feeds livestock makes that arithmetic very unfavourable.

Limits of the available data

Honesty requires marking the limits of this dossier.

Most of the reassuring data on high soy intakes comes from East Asian populations, where consumption often begins in childhood and takes largely fermented forms. Transferring those results to a European adult discovering tofu at thirty calls for caution.

The data on isoflavone supplements cover durations of three months to a year. Nobody has ten-year trials, and EFSA explicitly recorded the absence of data for perimenopause and for a history of cancer.

Finally, these data do not replace medical advice. If you are on thyroid treatment, under follow-up for a hormone-dependent cancer, or wondering about feeding an infant, the useful conversation is with your doctor, with these elements in hand.

Frequently asked questions

Does soy give men breasts? No. The 2021 meta-analysis of 41 trials found no effect on total testosterone, free testosterone, oestradiol, oestrone or SHBG, at any dose or duration. The cases reported in the press involved extreme and isolated intakes.

Can you eat soy after breast cancer? The available observational data show no danger signal and point the other way, with lower mortality and recurrence in the highest intake quartiles. These remain associations, and the decision belongs with the team treating you, particularly during active treatment.

Does soy disrupt the thyroid? In someone with no thyroid disorder and adequate iodine intake, the 2006 review of 14 trials found no worrying effect. In someone treated with levothyroxine, soy can reduce absorption of the medication: the answer is to keep to the empty-stomach rule, 30 to 60 minutes before the first meal, and to report any dietary change.

Could my Swiss tofu contain GM soy? It is highly unlikely. Growing genetically modified plants is prohibited in Switzerland until 31 December 2030, labelling is mandatory above 0.9 % per ingredient, and 97.8 % of the soy imported for the Swiss market comes from Europe under GM-free certification.

Does eating tofu contribute to Amazon deforestation? Marginally, and far less than meat does. 77 % of world soy feeds livestock, cattle ranching drives 41 % of tropical deforestation against 18 % for oilseeds, and food-grade soy sold in Switzerland is overwhelmingly European.

Should children avoid soy? Ordinary soy foods have no reason to be excluded from a varied child’s diet. Soy infant formula is the special case: according to the NHS it should only be used on the advice of a health professional. Fortified plant drinks are suitable as a main drink from the age of 1.

Is fermented soy better than unfermented soy? Better on one specific point: fermentation lowers phytates, which improves iron and zinc availability, and it changes the isoflavone profile. It confers no particular hormonal benefit, and miso and soy sauce remain salty condiments rather than protein sources.

Buying soy in Switzerland

Two pieces of label information are worth checking in a Swiss shop, because they vary widely between products.

On tofu, the coagulant appears in the ingredient list, usually as calcium sulphate or as magnesium chloride, sometimes named nigari. Only the first contributes meaningfully to calcium intake, and the difference between two blocks sitting side by side on the same shelf can be substantial. On soy drink, fortification is what decides its usefulness as a milk substitute: organic products sold in Switzerland are frequently not fortified with calcium and vitamins, because organic rules restrict such additions, whereas conventional drinks usually are. If you use soy drink to replace milk, check rather than assume.

On origin, the situation is more relaxed than soy’s reputation suggests. A growing share of the food-grade soy sold in Switzerland comes from Swiss or European fields, and the country of production stated on the pack makes that traceable. At a farm shop or a market, asking directly about the growing region is worth the trouble, since processors working with domestic beans are generally happy to say so.

One last point on price. Plain tofu, tempeh and unflavoured soy drink sit among the cheapest protein sources per 100 g of protein, well below meat and below most nut-based products. The expensive items in this aisle are the flavoured, marinated and ready-cooked versions, where you are largely paying for a marinade you can make at home in five minutes for a few centimes.

Sources