Moderate Diet

Cruciferous Vegetables and the Thyroid: Real Chemistry, Wrong Dose in Both Directions

Summary

Cruciferous vegetables really do contain compounds that can interfere with the thyroid — thiocyanate competes with iodide for uptake and goitrin interferes with hormone synthesis — but the human evidence puts the effect where almost nobody eats: the one documented case of thyroid failure severe enough to reach intensive care involved one to one and a half kilograms of raw bok choy daily for months in an 88-year-old, the only human experiments that ever produced a measurable antithyroid effect used a few hundred grams of raw turnip or rutabaga and lost the effect entirely once the vegetables we

Why Moderate

Moderate Evidence overall, because the entry's verdict is a synthesis whose parts sit at very different strengths and whose load-bearing human evidence is a set of small studies rather than one good one.

NOT Foundational, even though two of its components are. The underlying chemistry — glucosinolate hydrolysis by myrosinase, and the identity of thiocyanate and goitrin as goitrogenic species — is genuinely Foundational and undisputed. But the entry's actual claim is not the chemistry; it is a clinical verdict about who is and is not affected, and that verdict rests on several small human studies, one severe case report, an old uptake experiment and a systematic review, carries a live both-ways commercial tension, and requires judgement about dose thresholds only partly established. A Foundational label would let the certainty of the chemistry lift a conclusion the chemistry does not by itself supply.

NOT Strong, because the safety base is thin in exactly the way that should prevent an upgrade: ten volunteers over four weeks in 1986 with no iodine status reported, forty-eight over four weeks, forty-five women over twelve weeks in a population whose iodine status and pollutant exposure differ from most readers' and published by researchers with a direct stake in the answer, and one twelve-person trial with a non-significant upward drift in thyroid-stimulating hormone. Convergent nulls from several decades, several preparations and at least one disinterested reviewing group are genuinely reassuring — reassuring enough that the practical confidence is high — but they are not the powered, long, sex-inclusive, iodine-characterised trial that Strong would require, and none of them loaded goitrin specifically.

NOT Emerging, because the harm claim has been directly tested at ordinary intakes and failed to appear, the dose boundary is documented at both ends including one experimental human threshold, and the mechanism explains the pattern without special pleading. This is better established than a compound with mechanism and no human data.

The per-sub-area split (read this, not the headline):
• Glucosinolate → myrosinase → isothiocyanate chemistry, and the identity of the goitrogenic species: Foundational (characterised enzymology, uncontested by any camp).
• Cooking deactivates myrosinase and reduces isothiocyanate yield: Strong (mechanistic, human pharmacokinetic data, and a cooked-versus-raw contrast demonstrated in human iodide-uptake experiments; the "cooking method matters" nuance is where it gets sloppy in practice).
• Ordinary cruciferous intake does not measurably affect thyroid function: Moderate (several small human feeding studies across eighty years, cooked and raw, all null, plus an independent systematic review).
• Severe harm is possible at extreme raw intake: Emerging (one case report published as correspondence, with unestablished baseline thyroid status, plus one confounded milder case — real, and n=1 each).
• Raw root brassicas at a few hundred grams can inhibit human iodide uptake, and cooking abolishes it: Moderate (old human experimental work with an acute uptake endpoint, reached through a systematic review rather than the primaries).
• Iodine status and existing thyroid disease are the effect modifiers: Moderate (physiologically compelled, review-level, with mixed signals even inside the vulnerable group).
• Gut microbiota partially convert glucoraphanin, with high interindividual variation: Moderate (narrative review, not re-verified in this audit; the bacterial-myrosinase attribution carries the authors' own hedge).
• Active myrosinase raises sulforaphane bioavailability roughly three to fourfold: Moderate for direction, and explicitly not a population estimate (a series of very small cohorts).
• Sulforaphane increases urinary pollutant-conjugate excretion in humans: Moderate (a good randomised trial, on a biomarker endpoint, in a high-exposure population, by an interested group).
• Sulforaphane improves any clinical outcome: Emerging at best (eight randomised trials, surrogate endpoints dominant, short trials rarely showing efficacy, no incidence or survival result).
• Cruciferous intake lowers cancer risk in populations: Emerging (case-control-versus-cohort divergence signals bias; heavy lifestyle confounding; no trial).

Practical takeaway

If you have no thyroid diagnosis and are not iodine-deficient, this is a non-issue and you can stop here. Eat cruciferous vegetables in any normal pattern, raw or cooked, without thinking about goitrogens. The direct human evidence tested this repeatedly — cooked Brussels sprouts, raw broccoli sprouts, radish sprouts, a concentrated sulforaphane beverage, and radioiodine-uptake experiments with cooked broccoli, raw cabbage and cooked cauliflower — and found nothing.

If you have hypothyroidism, nodular thyroid disease, or Hashimoto's, do three cheap things and nothing more. Cook the bulk of your cruciferous intake — steaming is fine and is not the enemy of benefit. Do not build a daily habit around large volumes of raw brassica, especially raw turnip or swede, which are the forms that actually inhibited iodide uptake in human experiments, and raw kale or bok choy at the volumes implicated in the documented harm. Make sure iodine intake is adequate, because iodine sufficiency is what makes the competitive part of the mechanism irrelevant. Do not eliminate the food group; the twelve-week concentrated-beverage trial found no change in autoantibodies, and elimination has real nutritional costs. Management of the condition itself belongs with your clinician and with thyroid_dysfunction.

If you might be iodine-deficient, treat iodine as the variable, not broccoli. Populations without iodised salt, and pregnancy in particular, are where iodine supply is genuinely at stake. The honest caveat is that iodine status is hard to establish at the individual level — a spot urinary iodine test is unreliable for one person — so the group most affected by this cannot easily self-identify. That argues for securing iodine intake rather than for restricting vegetables.

If you drink raw green smoothies daily, know what the ceiling looks like. The documented severe harm was 1.0–1.5 kg of raw brassica every day for months, and the experimental inhibition needed a few hundred grams of raw root brassica — which is not how anyone eats turnip or swede. A daily handful of raw kale is nowhere near either. A regimen built around a kilogram of raw cruciferous material daily is a genuinely different thing, and if that describes you and you have any thyroid history, cook some of it and get thyroid function checked rather than reasoning from first principles.

Preparation, if you actually want the sulforaphane. Chop or chew thoroughly and let it stand briefly before heat, so hydrolysis begins while the enzyme is alive. Prefer steaming to boiling; if you boil, you are discarding a fraction of the compounds with the water. Raw sprouts and raw cruciferous vegetables retain the most active enzyme. Adding a small amount of a raw mustard-family food (mustard powder, raw rocket, radish) to a cooked cruciferous dish reintroduces myrosinase, which follows directly from the bioavailability finding — plausible and mechanistically sound rather than trial-proven.

On buying an extract. The honest position is that concentrated sulforaphane is an unproven bet, not a proven intervention: real human target engagement, no clinical outcome evidence. If you buy anyway, the one thing that materially changes what you receive is whether active myrosinase is present or co-supplied — a glucoraphanin figure alone tells you little. Do not pay for it on the strength of the pollutant-excretion trial, which measured urine, not health.

A note on why to eat these vegetables at all. The population evidence is confounded and no trial exists, so the reason to eat cruciferous vegetables is the same as for vegetables generally — fibre, potassium, folate, food variety, satiety, displacement of worse options — and belongs with whole_food_emphasis. That reason survives whether or not the sulforaphane story ever produces an outcome trial, which is exactly why it is the better reason.

Evidence detail

Why This Entry Exists

This is one of the cleanest examples in nutrition of two commercial camps inflating the same real chemistry in opposite directions. The chemistry is not in dispute by anyone: cruciferous vegetables store glucosinolates, damaging the tissue by chopping or chewing releases the plant enzyme myrosinase, and myrosinase hydrolyses those glucosinolates into isothiocyanates — sulforaphane from the precursor glucoraphanin, and, from other precursors, the goitrogenic species thiocyanate ion and goitrin. Both camps cite that chemistry accurately. Then one camp deletes the dose, the preparation and the iodine conditions the goitrogen effect requires, and sells fear of vegetables. The other camp deletes the gap between a biomarker and a disease outcome, and sells a capsule.

What makes this question unusually answerable is that the human evidence brackets it from both ends, and there is more of it than either camp lets on. The harm end has a single case report in the New England Journal of Medicine whose dose is roughly an order of magnitude above anything anyone has fed deliberately, in raw form, in an elderly patient over months. It also has something better than a case report and almost never quoted: an old series of human radioactive-iodine-uptake experiments in 100 subjects that fed seven different brassicas raw and cooked, found nothing from cooked broccoli, raw cabbage, cooked cauliflower or raw radish, and found genuine uptake inhibition only from raw turnip and raw rutabaga at a few hundred grams — which vanished on cooking. The safety end has several small feeding studies spread over eighty years, using cooked Brussels sprouts, raw broccoli sprouts, radish sprouts and a concentrated sprout beverage, all null on thyroid hormones. That structure is rare in nutrition and it is worth stating plainly, because neither of the loud positions will state it: the mechanism is real, the everyday risk is not, the exceptions are narrow enough to name, and the vegetables that actually did something to human thyroids in an experiment are not the ones anyone is arguing about.

The second job of this entry is to be equally unimpressed by the pro side. Sulforaphane is the most commercially amplified plant compound of the last decade, and its human evidence is a detoxification-flux biomarker — a rise in the urinary conjugates of benzene and acrolein — plus a large body of rodent and cell-culture work on the NRF2 pathway. That is real target engagement and genuinely interesting science. It is not evidence that anyone lived longer, avoided a cancer, or breathed better. Selling the extract on the mechanism is precisely the move this corpus exists to catch. And one honest limitation cuts against the reassuring reading of that same programme's thyroid-safety trial: a broccoli sprout beverage is a worst case for isothiocyanate exposure, not for goitrin, whose precursor progoitrin is concentrated in root brassicas rather than broccoli. That trial clears a sulforaphane-rich preparation. It does not clear a goitrin-rich one.

What bad advice this protects against, in all directions:
• "Cruciferous vegetables destroy your thyroid — cut out broccoli, kale and cabbage" → the mechanism is real and the dose and preparation conditions are silently dropped. Feeding 150 g of cooked Brussels sprouts daily for four weeks — vegetables carrying 220 mg of glucosinolates per 100 g, so the substrate was abundantly present — produced no change in thyroid-stimulating hormone, thyroxine or tri-iodothyronine in the ten volunteers tested. In the human uptake experiments, cooked broccoli, raw cabbage and cooked cauliflower did nothing at all.
• "There is a case in the New England Journal of Medicine linking cabbage to coma, so this is proven" → the case is real and it is n=1, and the dose is the finding: one to one and a half kilograms of raw bok choy every day for several months, in an 88-year-old woman with diabetes whose thyroid status before that regimen was never established. Any retelling that omits "raw", "1.0–1.5 kg daily", "for months" and "aged 88" is technically sourced and substantively dishonest.
• "Goitrogens are a myth invented by vegans" → the opposite over-correction, and it is also wrong. Thiocyanate ion and goitrin are genuine goitrogenic species, human iodide uptake really was inhibited by a few hundred grams of raw turnip or rutabaga in a controlled experiment, iodine status is a real effect modifier, and people with existing thyroid disease, iodine deficiency, or a daily raw kale or bok choy habit are a legitimate exception rather than a rhetorical concession.
• "Sulforaphane is clinically proven to detoxify pollutants" → the best human trial in the field, in 291 people, measured how much conjugated benzene and acrolein appeared in urine. Excretion of the benzene conjugate rose 61% and acrolein 23%; crotonaldehyde did not move at all. That is target engagement, not a health outcome, and no trial anywhere shows sulforaphane reduces cancer incidence.
• "Broccoli extract is the same as broccoli, just concentrated" → sulforaphane from preparations that retain active plant myrosinase is roughly three to four times more bioavailable than glucoraphanin delivered without it, in a series of very small pharmacokinetic cohorts. A label reading "contains X mg glucoraphanin" tells you very little about delivered sulforaphane.
• "Cooking destroys all the benefit, so eat your cruciferous vegetables raw" → half true and dangerously simplified. Boiling both denatures myrosinase and leaches glucosinolates into water you throw away; light steaming can preserve much of the enzyme. And raw is the exact form implicated in the one documented severe human harm, and the only form that inhibited iodide uptake in the human uptake experiments, so "always raw" is the worst advice available to the small group who actually needs to be careful.
• "If you have Hashimoto's you must eliminate all cruciferous vegetables" → in the twelve-week randomised trial of a concentrated sprout beverage, thyroid autoimmunity status did not change. Elimination costs fibre, potassium, folate and food variety for a benefit that has not been demonstrated even in the concentrated case.

This entry owns two questions and only two: whether cruciferous vegetables harm the thyroid, and what sulforaphane actually does in humans. It defers thyroid disease itself — diagnosis, hypothyroidism management, autoimmunity, iodine repletion as clinical care — to thyroid_dysfunction. It defers the wider plant anti-nutrient argument (phytate, lectins, oxalate, tannins) to antinutrients_evidence_and_context. It defers the general "mild stressors produce adaptive benefit" framing that sulforaphane marketing leans on to hormesis_and_adaptive_stress and plant_defence_compounds_and_xenohormesis. When a query resolves to one of those, it should route there rather than being re-argued here.

Evidence

Read the tiers, not the thesis. This entry's chemistry is as solid as nutrition biochemistry gets, its human safety data are a scatter of small studies spread over eighty years, its human harm data are one severe case report plus an old uptake experiment with a genuine positive result in raw root brassicas, and its benefit data stop at a urinary biomarker. Those things are not the same strength and must never be surfaced as though they were. Several of the older human studies below were reached through a 2024 systematic review rather than opened as primary papers, and that is stated on each one rather than hidden.

Sub-area 1: the chemistry (Foundational, and not disputed by either camp).

1. Glucosinolates are hydrolysed by the plant enzyme myrosinase — released when tissue is damaged by chopping or chewing — into isothiocyanates, notably sulforaphane from the precursor glucoraphanin. (Foundational.) Enzyme and substrate are physically compartmentalised in intact plant tissue and only meet when cells rupture. One nuance carries real practical weight: hydrolysis does not always yield the isothiocyanate. Depending on pH and the presence of epithiospecifier protein, the same substrate can yield sulforaphane-nitrile instead, which is not bioactive in the same way. Glucosinolate content is therefore not a linear predictor of sulforaphane delivered. (Fahey JW, Holtzclaw WD, Wehage SL, Wade KL, Stephenson KK, Talalay P. PLoS One. 2015;10(11):e0140963. PMID 26524341. Characterised enzymology reproduced in vitro and in vivo, not an epidemiological association. Cui bono, both ways: supplement labels cite this chemistry accurately and then let the reader assume "contains glucoraphanin" means "delivers sulforaphane", omitting both the myrosinase requirement and the nitrile branch; veg-fear content cites the identical chemistry to establish that cruciferous vegetables "release compounds that attack the thyroid", borrowing the accuracy of the mechanism to underwrite a dose claim the mechanism does not support. Neither side disputes this step, which is exactly why it is the rhetorical anchor for both.)

2. The goitrogen mechanism is real: glucosinolate breakdown yields thiocyanate ion and goitrin (5-vinyloxazolidine-2-thione), which can interfere with thyroid function — and the precursors are not evenly distributed across the food group. (Foundational for the compounds; the two-route mechanistic detail is stated carefully below.) The 1986 Brussels sprouts paper states the identification directly in a primary source: certain glucosinolates are readily converted into goitrogenic species, notably 5-vinyloxazolidine-2-thione and thiocyanate ion. Goitrin comes from progoitrin, which is concentrated in root brassicas — turnip, rutabaga, swede — rather than in broccoli and its sprouts, and total glucosinolate content differs by the same axis: a 2024 systematic review reports roughly 28–330 mg per 100 g fresh weight in root brassicas against 10–110 mg per 100 g in leafy ones. That distribution is the single most under-reported fact in this debate, because it means the vegetables people argue about are not the vegetables carrying the most goitrin. The commonly stated mechanistic split — thiocyanate competing with iodide at the sodium-iodide symporter, goitrin inhibiting thyroid peroxidase downstream — is uncontroversial and appears consistently across sources, but we reached it through secondary review summaries rather than a primary endocrinology paper we opened, and we flag that rather than weld it onto a citation that does not carry it. (McMillan M, Spinks EA, Fenwick GR. Hum Toxicol. 1986;5(1):15-19. PMID 2419242 — for the identification of the goitrogenic species; content ranges from "Do Brassica Vegetables Affect Thyroid Function?—A Comprehensive Systematic Review", Int J Mol Sci. 2024;25(7):3988, PMID 38612798. The symporter-versus-peroxidase split requires a separate primary citation we did not verify. Cui bono: this is the entire foundation of the veg-fear claim, and it is genuinely true — the anti-plant camp is not fabricating chemistry, it is citing real chemistry, deleting the conditions, and smearing a root-vegetable compound across the whole food group, which is what makes the claim hard to rebut in a soundbite and profitable to sell.)

Sub-area 2: what actually happens in humans at real intakes (Moderate — several small studies, almost all null, with one instructive positive).

3. Ordinary cooked cruciferous intake did not measurably affect thyroid function. (Moderate.) Ten volunteers ate 150 g of cooked Brussels sprouts daily for four weeks on an otherwise normal diet. No effect on thyroid-stimulating hormone, thyroxine or tri-iodothyronine — despite the sprouts containing 220 mg of glucosinolates per 100 g. The authors attribute the null to inactivation of myrosinase during cooking: the substrate was there in quantity, the enzyme to convert it was not. The design caveats are serious and must travel with the finding — n=10, four weeks, and the authors titled it "preliminary observations". A null in ten people over four weeks is weak evidence of no effect on its own, tests only the cooked form, reports no iodine status, and gives no dose-response information; it carries weight here because it is one of several convergent human nulls, not because it is strong alone. (McMillan M, Spinks EA, Fenwick GR. Hum Toxicol. 1986;5(1):15-19. PMID 2419242. Cui bono: this study is essentially never cited in veg-fear content, despite directly testing and refuting the everyday version of the claim using the exact vegetable most associated with it. Its obscurity relative to the case report is a clean demonstration of citation selection — the n=1 extreme travels, the n=10 null does not. Note also that it comes from the same research group that characterised the goitrogenic compounds, so it is not a null published by advocates.)

4. The one human experiment that ever produced a real antithyroid effect used raw root brassicas at a few hundred grams — and cooking abolished it. (Moderate for direction; old work, acute uptake endpoint, reached through a systematic review.) In a classic radioactive-iodine-uptake series in 100 subjects (technicians, medical students and physicians), cooked broccoli, raw cabbage, cooked cauliflower and raw radish all showed no effect on iodide uptake. Raw rutabaga at 280–363 g showed significant antithyroid activity, with complete inhibition at the higher doses; cooked rutabaga at 500–686 g showed none. Raw turnip inhibited uptake at 441 g but not at 220 g; cooked turnip was inconsistent. This is the closest thing to a human dose-response that exists in this literature, and it says three things at once: the effect is real, it is confined to raw root brassicas at substantial volume, and heat removes it. The caveats are heavy — the work is old, the endpoint is acute radioiodine uptake rather than circulating hormones over time, individual study details come to us through a 2024 systematic review rather than the primary papers, and 220 g of raw turnip is already an unusual daily intake. (Greer et al., as described in "Do Brassica Vegetables Affect Thyroid Function?—A Comprehensive Systematic Review", Int J Mol Sci. 2024;25(7):3988, PMID 38612798, PMC11012840. Primary papers not opened in this audit. Cui bono: neither camp quotes this, and both should. The fear camp loses because the vegetables it names showed nothing and cooking cancelled what did show. The dismissal camp loses because a genuine, dose-dependent human inhibition exists and "myth" cannot survive it.)

5. The rest of the small human feeding literature is null, including in raw form — with one non-significant blip that we quote rather than bury. (Moderate.) Forty-eight healthy volunteers took 20 g of raw broccoli sprouts daily for four weeks with no significant influence on thyroid-stimulating hormone, free tri-iodothyronine or free thyroxine and no hypothyroid symptoms. Thirty women with obesity took 100 g of red radish sprouts daily for eight weeks within a low-calorie diet, with no significant change in thyroid-stimulating hormone. Healthy iodine-sufficient adults given 8 mg of thiocyanate daily in milk for twelve weeks kept thyroxine, tri-iodothyronine and thyroid-stimulating hormone within normal ranges despite measurably raised serum and urinary thiocyanate — a direct test of the competitive mechanism in people with enough iodine. Against those, one small double-blind placebo-controlled trial of broccoli sprout extracts (25–100 µmol of glucosinolates or isothiocyanates) in twelve healthy volunteers recorded thyroid-stimulating hormone values above the normal limit (4.4–7.6 µIU/mL), though the changes were not statistically significant. We state that blip explicitly, because an entry that demands the harm camp quote its dose cannot quietly drop the one result pointing the other way; with twelve subjects and no significance it is a hint to be tested, not a finding. (All four described in "Do Brassica Vegetables Affect Thyroid Function?—A Comprehensive Systematic Review", Int J Mol Sci. 2024;25(7):3988, PMID 38612798; the raw-sprout study attributed there to Yanaka et al. Primary papers not opened in this audit — treat participant numbers and doses as the review reports them.)*

6. A twelve-week randomised trial of a sulforaphane-rich broccoli sprout beverage found no effect on thyroid hormonal or autoimmune status. (Moderate — the best-designed direct test of a concentrated preparation.) Forty-five female participants, 84 days, 100 mL daily of the formulated broccoli sprout beverage delivering 600 µmol glucoraphanin and 40 µmol sulforaphane — the same formulation as the Qidong parent trial. Serum thyroid-stimulating hormone, free thyroxine and thyroglobulin were unaffected, and so was thyroid autoimmunity status. This is a stronger safety signal than the 1986 study on design — randomised, longer, with autoimmunity markers as well as hormones, and using a concentrated preparation rather than ordinary food. It remains modest and its scope is narrower than it looks: n=45, all female, so no data in men; twelve weeks is short for thyroid autoimmunity, which evolves over years; the population is a rural Chinese cohort whose baseline iodine status and pollutant exposure differ from Western populations, with iodine status being the single most important effect modifier for this question; and, decisively, a broccoli sprout beverage is a concentrated source of sulforaphane, not of goitrin, whose precursor sits mainly in root brassicas — so this trial clears the isothiocyanate load, not the goitrin load. (Chartoumpekis DV, Ziros PG, Chen JG, Groopman JD, Kensler TW, Sykiotis GP. Food Chem Toxicol. 2019;126:1-6. PMID 30735751. Funding and interest note, stated rather than hidden: the authors include the principal investigators of the broccoli-sprout chemoprevention programme — this is a safety null published by the people whose intervention is being cleared. The result is plausible and consistent with the other human data and it is not independent adjudication. Discrepancy disclosed: the primary abstract reports 45 female participants, while the 2024 systematic review describes the same twelve-week beverage thyroid study as 267 subjects; we use the primary figure and have not reconciled the difference.)

7. There is one documented human case of thyroid failure severe enough to reach intensive care from extreme raw cruciferous intake, and the dose is the finding. (Emerging — a single case report, published as correspondence.) An 88-year-old Chinese woman consumed 1.0–1.5 kg of raw bok choy daily for several months in an attempt to improve her diabetes control. She was intubated for hypoxaemic and hypercarbic respiratory failure, admitted to intensive care with severe hypothyroidism and myxedema coma, and treated with intravenous methylprednisolone and levothyroxine. The proposed mechanism is that raw brassica releases myrosinase on chewing, accelerating hydrolysis of glucosinolates to thiocyanates, nitriles and oxazolidines, whereas cooking deactivates the enzyme. This is the weakest design that still counts as evidence: baseline thyroid status before the regimen is not established, so pre-existing subclinical hypothyroidism unmasked by an added iodide-uptake inhibitor cannot be excluded and is arguably the more likely full explanation; iodine status is not reported; causation is inferred from timing and mechanistic plausibility. A second, much milder case report exists — a patient with Hashimoto's disease eating large servings of raw broccoli whose thyroid parameters normalised after both levothyroxine treatment and dietary change, which leaves causality unresolvable. Together they establish that the effect is possible at extreme raw intake and nothing about frequency or threshold at ordinary intake. (Chu M, Seltzer TF. N Engl J Med. 2010;362(20):1945-1946. PMID 20484407; the Hashimoto's case as described in Int J Mol Sci. 2024;25(7):3988. Cui bono: the most-cited and most-stripped item in the whole debate — an NEJM byline confers authority, and the three qualifiers that neutralise it get dropped. The pro-vegetable side's error is the mirror image: dismissing the case as an outlier. It is a real documented harm and the correct response is to quote the dose, not to bury the case.)

8. An independent 2024 systematic review of the whole literature reaches the same practical conclusion. (Moderate — a systematic review, and the only substantial adjudication here with no stake in a sulforaphane product.) The review followed PRISMA, searched to January 2024 and analysed 123 in vitro, animal and human studies of brassica plants and extracts against thyroid mass and histology, thyroid-stimulating hormone, tri-iodothyronine, thyroxine, iodine uptake and thyroid cancer cells. Its conclusion is that the majority of results cast doubt on the assumption that brassica plants have antithyroid effects in humans, and that including brassica vegetables in the daily diet — particularly alongside adequate iodine intake — poses no adverse effect on thyroid function; it notes that the longstanding advice to avoid them rests largely on older animal work. This matters structurally as much as substantively: the entry's other reassuring trial was published by parties with a commercial stake, and this one was not. Its limits are the ordinary limits of a narrative-leaning systematic review — no pooled effect estimate, heterogeneous designs, and it inherits the weaknesses of the small studies it summarises. ("Do Brassica Vegetables Affect Thyroid Function?—A Comprehensive Systematic Review." Int J Mol Sci. 2024;25(7):3988. PMID 38612798. PMC11012840. Cui bono: nobody sells anything on the back of this paper, which is roughly why it is invisible in both camps' content.)

Sub-area 3: delivery — why the same vegetable behaves differently depending on preparation (Strong to Moderate).

9. Cooking deactivates myrosinase and substantially reduces both isothiocyanate yield and goitrogen formation. (Strong.) High-temperature cooking such as boiling causes significant loss of myrosinase activity and a corresponding drop in isothiocyanate formation; a review reports human recovery in blood and urine at around 3.4% after consumption of microwaved broccoli. This is also the mechanism the 1986 investigators invoked to explain their thyroid null, and it is directly visible in the human uptake experiments, where raw rutabaga and raw turnip inhibited iodide uptake and their cooked equivalents did not. The important honesty caveat is that "cooking" is not one thing: boiling is worse than steaming because it both denatures the enzyme and leaches water-soluble glucosinolates into discarded water, whereas light steaming to around 60 °C can inactivate epithiospecifier protein while sparing much myrosinase, which can actually raise sulforaphane yield. Blanket claims in either direction are wrong, and the 3.4% figure belongs to microwaved broccoli specifically and should not be generalised. (Dmytriv TR, Lushchak O, Lushchak VI. Front Physiol. 2025;16:1497566 — not re-verified in this audit; corroborated on the thyroid side by McMillan et al. 1986 and by the cooked-versus-raw contrast in the human uptake series. Cui bono: sellers weaponise this correctly but selectively — "your cooked broccoli is worthless, buy the extract" — while omitting that many commercial extracts have also had myrosinase inactivated in processing and are subject to the same problem. The veg-fear camp must suppress this finding entirely, because it is the single fact that most cleanly defuses the goitrogen scare for the majority of people who cook their vegetables. Its absence from that content is conspicuous.)

10. Gut bacteria partially compensate for inactivated plant myrosinase, inefficiently and with large variation between people. (Moderate.) Mammalian gut microbiota can convert glucoraphanin into sulforaphane available to the host — the reviewing authors' own word is "probably", via bacterial myrosinase. The route yields a mixture rather than sulforaphane alone: glucoraphanin can be metabolised to sulforaphane, sulforaphane-nitrile, glucoerucin, erucin and erucin-nitrile. Interindividual variability is substantial and plausibly microbiota-driven, with Dorea, Bifidobacterium and Ruminococcus torques positively associated with sulforaphane metabolite excretion — associations, not demonstrated causal converters. The same review concedes that plant myrosinase probably does not deliver therapeutically significant sulforaphane concentrations and that insufficient data exist directly comparing bioavailability between active-myrosinase and processed sources: a candid admission of a gap, in a review otherwise favourable to the compound. The practical consequence is that any individual's actual sulforaphane exposure from cooked cruciferous vegetables is close to unpredictable. (Dmytriv TR, Lushchak O, Lushchak VI. Front Physiol. 2025;16:1497566 — a narrative review, not a pooled analysis, and not re-verified in this audit. Cui bono: variability is sold as a reason to buy a standardised extract, a real argument that collapses if the extract lacks active myrosinase. It also undercuts the fear direction, since it means goitrin and thiocyanate exposure from cooked intake is likewise low and erratic rather than reliably damaging. Nobody sells "it depends on your microbiome and we cannot predict it", which is why that honest framing is rare.)

11. Sulforaphane from preparations retaining active plant myrosinase is roughly three to four times more bioavailable than from glucoraphanin without it. (Moderate — direction and rough magnitude, in very few people.) The paper's own summary is that sulforaphane in preparations with active endogenous myrosinase is three- to four-fold more bioavailable than sulforaphane from glucoraphanin delivered without it, with roughly 10–18% conversion to sulforaphane metabolites in the no-myrosinase condition. The design caveat is decisive for how this may be quoted: the work is a series of very small human cohorts rather than one powered trial, and we could not confirm the exact per-arm participant numbers from the abstract in this audit. This is a mechanistic pharmacokinetic study, not a population estimate, and the three-to-four-fold figure must never be presented as a precise or generalisable multiplier. The endpoint is urinary metabolite recovery — higher plasma exposure is not itself a health outcome. (Fahey JW et al. PLoS One. 2015;10(11):e0140963. PMID 26524341. Authors have long-standing involvement in the broccoli-sprout research programme and associated intellectual-property interests. Cui bono, and note that it cuts within the industry: this finding is a direct commercial attack on the large, cheaper category of glucoraphanin-only extracts and a boost for sprout-based or myrosinase-added products — a recent paper on adding exogenous mustard-seed myrosinase to a broccoli seed extract is a product-shaped solution to a product-shaped problem. For the buyer, the consequence is simple: a milligram figure for glucoraphanin on a label is close to uninformative absent active myrosinase.)

Sub-area 4: what sulforaphane actually does in people (Moderate for target engagement, Emerging for anything clinical).

12. In humans, a broccoli sprout beverage rapidly and durably increased urinary excretion of airborne-pollutant conjugates — a detoxification-flux biomarker, not a disease outcome. (Moderate.) A twelve-week randomised clinical trial in 291 participants recruited from the rural He-He Township, Qidong, in the Yangtze River delta region of China — an area with substantial airborne pollutant exposure — used a beverage delivering 600 µmol glucoraphanin and 40 µmol sulforaphane daily. Excretion of the glutathione-derived conjugate of benzene rose 61% and acrolein 23% versus placebo (P ≤ 0.01), sustained across the trial with no decline in bioavailability over twelve weeks of daily dosing; crotonaldehyde did not move, meaning the response is pollutant-specific rather than a general "detox" effect. Benzene-derived mercapturic acid excretion was higher in GSTT1-positive participants irrespective of study arm. What this shows is real, statistically robust human target engagement — the pathway moves. What it does not show is reduced cancer, reduced respiratory disease, or reduced mortality. Species note, stated explicitly: the NRF2 mechanism literature underpinning the rationale is overwhelmingly rodent and cell-culture, and must be labelled mechanism, never human proof. (Egner PA, Chen JG, Zarth AT, et al. Cancer Prev Res. 2014;7(8):813-823. PMID 24913818. Conducted by the Johns Hopkins broccoli-sprout programme, which has long-standing institutional and intellectual-property investment in this compound class; the high-exposure rural setting also limits generalisability to low-exposure Western populations, where the effect could plausibly be smaller. Cui bono: this is the single most misrepresented study in the category — "clinically proven to detoxify pollutants" is built on a 61% rise in one urinary metabolite. The conversion of a conjugation-flux biomarker into an implied disease-prevention claim is where the margin lives.)

13. Hard clinical-outcome evidence for sulforaphane is thin: trials are short, small, and dominated by surrogate endpoints. (Emerging as a benefit claim.) A systematic review of randomised controlled trials in cancer management, searching to May 2023, found eight trials in prostate, breast and pancreatic cancer and melanoma, with variable and inconsistent dosing. The main efficacy endpoints explored were overall survival, tumour biomarkers, and histological and gene-expression results — a predominance of surrogate markers — and shorter durations of a few weeks or a month rarely led to significant efficacy outcomes; an overall-survival improvement in pancreatic cancer did not reach statistical significance. Later commentary in the field concedes that surrogate molecular changes function as the principal indicators of biological activity, and argues that long-term endpoints such as cancer incidence or survival are impractical within feasible study periods. That is an honest methodological constraint rather than an excuse, but the consequence is the same: the outcome evidence does not exist and is not imminently coming. Note the scope limit: this review covers cancer therapeutics, and separate small trials in autism, schizophrenia, asthma and metabolic endpoints exist and should not be swept in under this citation. Absence of outcome trials is not evidence of absence of effect — it is genuinely hard to run incidence trials on an unpatentable food-derived compound — but the correct statement is that the evidence has not been generated. (ElKhalifa D, Al-Ziftawi N, Awaisu A, Alali F, Khalil A. Front Oncol. 2023;13:1251895. PMCID PMC10710291. Cui bono: the entire supplement category is priced and marketed as though outcome trials existed. The gap between "increases a urinary conjugate" and "prevents cancer" is the product. Running the other way, skeptic and carnivore content uses the thin outcome evidence to dismiss cruciferous vegetables wholesale, which is a separate error — the absence of evidence for a concentrated extract says nothing about the food.)

14. Population evidence linking cruciferous intake to lower cancer risk is suggestive, observational, weaker in cohorts than in case-control studies, and untested by any randomised trial. (Emerging.) The pattern is consistent and instructive: retrospective case-control studies report inverse associations, while prospective cohorts find them weaker and less consistent. For ovarian cancer, intake was associated with reduced risk in case-control studies but not in cohorts; for gastric cancer, case-control results were significant while prospective results were only borderline; a bladder-cancer meta-analysis in the same literature did not support an association with lower risk. That design divergence is the classic signature of recall bias, not of a robust causal effect. The confounding is major and unresolved: cruciferous intake is a marker for an entire lifestyle cluster — higher total vegetable intake, less smoking, higher socioeconomic status, more physical activity, less ultra-processed food — and no adjustment fully removes it. Food-frequency measurement adds substantial misclassification, and publication bias favours positive findings. Cardiovascular meta-analyses of prospective cohorts show the same structural gap: association present, intervention evidence limited. (Guo et al. "Effect of cruciferous vegetable intake on cancer: An umbrella review of meta-analysis." J Food Sci. 2024;89:5230-5244. doi:10.1111/1750-3841.17300 — 22 meta-analyses covering 175 independent cancer studies, quality-assessed with AMSTAR-2 and PRISMA 2020. Underlying design divergence documented in the ovarian (PMC3937240) and gastric (PMID 23679348) meta-analyses; the bladder meta-analysis identifier (PMC9434151) was not re-opened in this audit. Cui bono: "studies link broccoli to lower cancer risk" underwrites both supplement marketing and produce-industry messaging, and the case-control-versus-cohort divergence is never mentioned. The carnivore camp correctly points out that this literature is confounded, then illegitimately extends "confounded" into "harmful", which requires evidence it does not have either. Both sides overclaim from the same weak dataset.)

One number we deliberately do not print. A specific quantitative reassurance circulates widely in this debate — a stated goitrin content per 100 g of commercial broccoli and kale set against a stated amount required to inhibit thyroid iodide uptake, implying roughly a twentyfold safety margin. We could not trace it to a primary source. It appears to originate in a secondary reference work whose pages we could not open, the underlying study is unidentified, and the experimental conditions behind the inhibitory threshold — species, route, acute or chronic, and what degree of inhibition counts as the endpoint — are unknown. The comparison is directionally consistent with everything verified above, but directional consistency is not verification, and this is exactly the shape of number that gets welded to a plausible-looking citation and propagates uncorrected. Two traceable statements do the same job better: the glucosinolate content of leafy brassicas runs roughly 10–110 mg per 100 g fresh weight against 28–330 mg per 100 g in root brassicas, and in the human uptake experiments the leafy and floret vegetables produced no measurable inhibition at all while a few hundred grams of raw root brassica did.

Mechanism

The enzyme and the substrate live apart until you break the cell. Glucosinolates sit in one compartment of intact plant tissue and myrosinase in another. Chopping, blending or chewing ruptures the cells and lets them meet, and only then does hydrolysis begin. This single fact explains most of the apparent contradictions in the popular literature: the same vegetable is a potent sulforaphane source, a modest goitrogen source, or neither, depending entirely on whether the enzyme is alive when the tissue is damaged.

Hydrolysis branches, and only one branch is the one being sold. The reaction does not reliably yield the isothiocyanate. Depending on pH and the presence of epithiospecifier protein, the same glucosinolate can yield a nitrile instead, which is not bioactive in the same way. A glucosinolate number on a label or in a food composition table is therefore an upper bound on possibility, not a measure of delivery.

Two different goitrogenic routes, not one. Thiocyanate ion acts upstream, competing with iodide for uptake into the thyroid — the same general route by which tobacco smoke, a far larger thiocyanate source in smokers, affects thyroid iodine handling. Goitrin acts downstream, interfering with hormone synthesis by inhibiting thyroid peroxidase. Because the upstream route is competitive, iodine supply is the decisive variable: with ample iodide available, competition at the transporter is a much smaller problem than it is when iodide is already scarce. That is the mechanistic reason iodine status, not vegetable intake, is the real risk determinant — and it is why twelve weeks of daily thiocyanate in milk changed nothing measurable in iodine-sufficient adults.

The goitrin is not where the argument is. Goitrin derives from progoitrin, which is concentrated in root brassicas — turnip, rutabaga, swede — rather than in broccoli, kale or broccoli sprouts, and total glucosinolate content follows the same split. This is why the human uptake experiments found real inhibition from raw turnip and raw rutabaga and none from cooked broccoli, raw cabbage, cooked cauliflower or raw radish. Any version of the argument that treats "cruciferous" as one uniform goitrogen source has already lost contact with the chemistry.

Heat is the switch, and it works in both directions at once. Boiling denatures myrosinase and leaches water-soluble glucosinolates into water that is usually discarded, which reduces both the sulforaphane you might have wanted and the goitrogenic species you did not. Light steaming can inactivate epithiospecifier protein while sparing much myrosinase, raising isothiocyanate yield. Raw maximises both. In the human uptake series the same root vegetables that inhibited iodide uptake raw did not do so cooked, which is the mechanism demonstrated in people rather than inferred.

The microbiome is a partial and unreliable backup converter. When plant myrosinase is gone, gut bacteria can perform some of the conversion, producing a mixture that includes sulforaphane and several related and less-studied compounds. The efficiency varies substantially between individuals in ways that track gut community composition. This is why two people eating identical cooked broccoli can have very different circulating sulforaphane, and why the practical exposure — of benefit and of goitrogen — is far less predictable than either marketing or fear-content implies.

NRF2 is a mechanism, not an outcome. Sulforaphane activates the NRF2 transcription factor, which upregulates a battery of phase II conjugating and antioxidant enzymes. In humans this produces measurable pathway output — more conjugated benzene and acrolein leaving in urine. Whether accelerating conjugation of a pollutant reduces the disease that pollutant causes is a plausible but entirely untested inferential step, and almost all of the NRF2 literature that makes it feel inevitable was done in rodents and cell culture.

Dose is the whole argument. One kilogram of raw bok choy daily supplies an enzymatically active, uncooked, unleached glucosinolate load roughly an order of magnitude beyond anything the feeding studies delivered, over months rather than weeks, in a body with less thyroid reserve. Even the experimental positive required several hundred grams of the most progoitrin-dense vegetables, raw. Nothing in the mechanism is surprising once the dose and the preparation are stated; nothing in the mechanism predicts harm once the dose is ordinary.

Risks And Contraindications

• The one documented severe harm requires all of: raw, extreme volume, sustained for months. One to one and a half kilograms of raw bok choy daily, over months, in an 88-year-old, produced myxedema coma requiring intensive care. This is the boundary case and must be quoted with its dose every single time. It is not a reason for anyone eating ordinary quantities to change anything.
• The only measured human threshold sits in raw root brassicas, not in the vegetables people argue about. Raw turnip inhibited iodide uptake at 441 g but not 220 g, and raw rutabaga at 280–363 g showed significant inhibition; cooked forms of both did not. That is a real number and it belongs in the record — and it is a volume of raw swede or turnip that essentially nobody eats.
• Existing thyroid disease is the real risk group, and the risk is modest and manageable. Someone with reduced thyroid reserve has less capacity to absorb an added iodide-uptake inhibitor. That argues for cooking most cruciferous intake and avoiding high-volume raw habits, not for elimination.
• Iodine deficiency is the effect modifier that matters most, and the evidence within that group is itself mixed. One study of pregnant women with low urinary iodine reported an association between higher thyroid-stimulating hormone and thiocyanate; another found no association between low-level thiocyanate exposure and thyroid function in mildly iodine-deficient pregnant women. We flag that these two studies come to us through secondary summaries and were not traced to primary sources, so they are stated as a mixed signal rather than cited by name. The practical implication is unchanged: secure iodine.
• Smoking is a far larger thiocyanate exposure than food. In smokers, tobacco-derived thiocyanate plausibly dwarfs any dietary cruciferous contribution, which is both a large uncontrolled confound in observational thyroid analyses and a much better target than the vegetable drawer.
• Nobody has tested high raw cruciferous intake in iodine-deficient people, and nobody will. That is the exact scenario of concern and it is ethically untestable. The safe-side evidence is a handful of small human feeding studies plus one independent systematic review; the harm-side evidence is one severe case, one confounded milder case, and an old uptake experiment in raw root vegetables. Everything between 150 g cooked daily and 1 kg raw daily in an iodine-deficient person is inference, and should be presented as inference.
• Concentrated sulforaphane supplements carry the ordinary unknowns of an unregulated category. Content varies, myrosinase status is often undeclared, long-term human safety data beyond twelve weeks are absent, and no interaction profile has been characterised in people on thyroid replacement or other medication. The twelve-week thyroid null is reassuring for twelve weeks in 45 women taking a sulforaphane-rich rather than goitrin-rich preparation, and says nothing beyond that.
• Do not use this entry to reassure someone with untreated thyroid symptoms. Fatigue, cold intolerance, weight change and low mood warrant thyroid testing and clinical assessment, not a dietary debate — see thyroid_dysfunction.
• Raw kidney beans are a different and far more acute plant hazard and are not covered here — that is genuine acute lectin poisoning and belongs to antinutrients_evidence_and_context. Do not let a "are plants dangerous" query bundle the two.

Controversy

Nature: a genuine, well-characterised biochemical mechanism whose dose, preparation and iodine-status conditions are deleted by one commercial camp to sell fear of vegetables, and whose refutation is over-claimed by the opposite commercial camp to sell a capsule on a biomarker. Error is available at both poles, and both camps cite the same true chemistry.

Position A — "Cruciferous vegetables damage the thyroid; restrict or avoid them." The veg-fear and carnivore-adjacent take.
• Best evidence: the mechanism is real and Foundational. Thiocyanate and goitrin are genuine goitrogenic species, iodide-uptake competition is real physiology, human iodide uptake really was inhibited by a few hundred grams of raw turnip or rutabaga in a controlled experiment, and there is a documented human case of myxedema coma attributed to raw brassica intake, published in a top-tier journal.
• Where it overreaches: it drops every condition the mechanism requires. The documented case was raw, at 1.0–1.5 kg daily, over months, at age 88, without baseline thyroid status. The experimental inhibition was raw root brassica only and disappeared on cooking, while cooked broccoli, raw cabbage and cooked cauliflower did nothing in the same experiment. It ignores the feeding studies — cooked and raw — that tested the everyday claim directly and found nothing, and the randomised trial that tested a concentrated preparation and found nothing including on autoantibodies. It generalises one extreme case and one root vegetable into a dietary rule for everyone.

Position B — "Goitrogens are a debunked myth; the whole concern is nonsense." The over-correction, common in pro-plant and supplement-adjacent content.
• Best evidence: several small human feeding studies over eighty years, cooked and raw, all null on thyroid hormones; an independent 2024 systematic review of 123 studies concluding no adverse effect on thyroid function with adequate iodine; and a mechanistic argument that ordinary goitrin exposure is far below inhibitory amounts.
• Where it overreaches: the individual studies are ten people over four weeks, forty-eight over four weeks, forty-five women over twelve weeks, with iodine status unreported or unconfirmed and one twelve-person trial showing a non-significant drift upward in thyroid-stimulating hormone — that cannot carry "debunked". Human iodide uptake was inhibited experimentally by raw root brassicas, which "myth" cannot accommodate. It waves away a real documented harm rather than quoting its dose. And it erases a genuine, identifiable at-risk group: people with existing thyroid disease, people who are iodine-deficient, and people consuming very large raw quantities daily. "Myth" is as sloppy as "toxin".

Position C, running alongside both — "Sulforaphane is a proven detoxifier and cancer-preventive." The supplement pitch.
• Best evidence: a genuine randomised trial in 291 people showing rapid, sustained, statistically significant increases in urinary excretion of benzene and acrolein conjugates, plus a large and coherent NRF2 mechanism literature.
• Where it overreaches: the endpoint is urinary excretion of conjugates, not disease. Crotonaldehyde did not move, so this is not a general "detox" effect. The mechanism literature is overwhelmingly rodent and cell-culture and is routinely presented as human proof. A systematic review of eight randomised trials found surrogate endpoints dominate and short trials rarely showed efficacy, with even the pancreatic-cancer survival signal falling short of significance. No trial anywhere shows reduced cancer incidence.

The funding/bias dimension — cui bono, both ways. Toward inflating the harm: veg-fear and carnivore content, meat-adjacent commerce, and elimination-protocol sellers monetise the goitrogen mechanism precisely by leaving out the raw-and-extreme dose condition, the root-versus-leaf distinction, and the iodine condition. The case report is their best asset and they strip it. Toward inflating the benefit: the sulforaphane supplement category is priced as though outcome trials existed, and converts a conjugation-flux biomarker plus a rodent mechanism into implied disease prevention. There is a third, sharper irony worth stating: the research programme with the largest institutional and intellectual-property stake in sulforaphane is also the group that published the thyroid-safety null — a result that serves its product, produced by people with an obvious interest, and probably also correct. Interest does not make a finding wrong; it makes it non-independent, and the honest move is to say so, note that the 1986 null came from the group that characterised the goitrogens in the first place, and note that a 2024 systematic review with no product to sell reached the same conclusion. Meanwhile the honest middle position — a real, conditional interaction relevant to a small identifiable group — sells nothing to anyone, which is a large part of why it is so rarely published.

Realised Position: the chemistry is real; the everyday risk is not; the exceptions are narrow enough to name. For people with adequate iodine and no thyroid disease, ordinary cruciferous intake in any preparation has no demonstrated thyroid effect, and multiple direct human studies looked for one. For people with existing thyroid disease, iodine deficiency, or a sustained high-volume raw habit, the interaction is genuine and the response is small: cook most of it, avoid daily raw kilogram-scale intake and raw root brassica in volume, secure iodine, and do not eliminate the food group. On sulforaphane, we hold that it produces real, measurable human target engagement and that this is not a health outcome; concentrated supplements are an unproven bet sold at the price of a proven one, and the reason to eat cruciferous vegetables is the ordinary whole-food reason, which survives whatever the sulforaphane literature eventually shows.

Cross-Pillar Connections

• Conditions (thyroid_dysfunction): owns the thyroid itself — diagnosis, hypothyroidism and Hashimoto's management, replacement therapy, iodine repletion as clinical care, and when to test. This entry owns only whether cruciferous vegetables affect thyroid function and defers everything about the condition there. Any query with symptoms in it routes there, not here.
• Diet (antinutrients_evidence_and_context): owns the broad plant anti-nutrient question — phytate and mineral absorption, lectins and the acute raw-kidney-bean hazard, oxalate and stones, tannins. This entry is the deep-dive on the goitrogen strand only, and defers the rest of the class, including the whole "are plants toxic" framing, there.
• Diet (whole_food_emphasis): owns the actual reason to eat cruciferous vegetables — fibre, micronutrients, satiety, food variety, displacement of worse options — which does not depend on the confounded cancer epidemiology or on any sulforaphane claim. This entry defers the positive case for eating them there, and holds only that the population cancer evidence is too confounded to carry it.
• Cross-pillar (hormesis_and_adaptive_stress): owns the general framing that a mild stressor provokes an adaptive response — the conceptual scaffolding sulforaphane marketing borrows when it invokes NRF2. This entry defers that framing entirely and holds only the specific human data on what sulforaphane does and does not do.
• Diet (plant_defence_compounds_and_xenohormesis): owns the class-level story of plant defence chemistry and the xenohormesis hypothesis. This entry is a single worked example within it — one compound family, one organ, one both-ways commercial distortion — and defers the general theory there.
• Diet (capsaicin_and_spicy_food_evidence): the closest structural parallel in this batch — another plant irritant sold on mechanism and biomarker endpoints in one direction and feared as damage in the other. Cited for the shared pattern of evaluation, not for shared biochemistry; the capsaicin evidence itself lives there.
• Cross-pillar (immune_function_cross_pillar_optimisation): owns immune-system claims, including the redox and inflammatory framing that sulforaphane marketing frequently reaches for after the detoxification claim. This entry holds only the pollutant-conjugate excretion result and defers any immune interpretation there.

What would change our mind

Falsifiability: explicit upgrade/downgrade criteria from source

• We would raise the harm side above its current narrow framing if a controlled trial or a well-designed prospective cohort with measured iodine status showed thyroid-function changes at ordinary cruciferous intakes in iodine-sufficient people. At present every human study at ordinary intake is null, and the one experimental positive required raw root brassica at several hundred grams.
• We would tighten the caution for the at-risk group considerably if a case series or cohort documented thyroid decompensation in people with existing hypothyroidism at moderate raw intakes — say, a few hundred grams daily rather than a kilogram. The uptake experiments make that a live possibility rather than a fantasy, which is why it is the first thing we would want tested.
• We would strengthen the safety claim from Moderate toward Strong if a properly powered randomised trial — several hundred participants, both sexes, at least six months, with iodine status measured and reported, and using a goitrin-rich rather than sulforaphane-rich preparation — confirmed no effect on thyroid hormones and autoantibodies. The current nulls are numerous and directionally consistent but individually underpowered, and none of them loaded goitrin specifically.
• We would raise sulforaphane's benefit claim off Emerging if a randomised trial with a clinical endpoint — cancer incidence, respiratory outcomes, a hard cardiometabolic event, not a biomarker — showed benefit. Target engagement is established; efficacy is not.
• We would revise the "extracts are an unproven bet" line if head-to-head human trials established a reliable dose-to-effect relationship for a specific standardised preparation with declared myrosinase status, on an outcome that matters to a person rather than to an assay.
• We would print the quantitative safety margin if the goitrin content and inhibitory-threshold figures that circulate in this debate were traced to an identifiable primary source with stated experimental conditions. Until then we state the comparison generically and name the gap.
• What would NOT move us: the existence of the goitrogen mechanism (it is real and we already grant it in full); the New England Journal case report re-cited without its dose (we already grant the case and the dose is the finding); another rodent or cell-culture NRF2 paper, however elegant (mechanism is not outcome, and this literature is already large); another case-control study reporting lower cancer risk with higher cruciferous intake (the divergence from cohort results is itself the evidence that bias is operating); another urinary-biomarker trial showing increased conjugate excretion (that question is answered, and it is not the question that matters); or a supplement company's internal or sponsored data on a proprietary preparation. Absent the trials named above, the conditional Moderate position stands.

Industry bias note

Structural incentives the evidence base may reflect

• The fear end has sellers, and they are selling by omission rather than fabrication. Carnivore and veg-fear content, elimination-protocol practitioners, and thyroid-diet merchants monetise a mechanism that is entirely real by removing the conditions it requires: raw intake at extreme volume, root brassicas rather than broccoli, and iodine deficiency. The tell is citation shape — the New England Journal case appears constantly and its dose almost never; the 1986 Brussels sprouts feeding study, which tests the everyday claim directly with the most emblematic vegetable, is effectively invisible in that content, as is the raw broccoli sprout study that tested the raw form and found nothing. An n=1 extreme travels and a null does not, and that asymmetry is chosen, not accidental.
• The benefit end has larger sellers, and they are selling by category error. The sulforaphane supplement market runs on an NRF2 mechanism established mostly in rodents and cell culture, plus one human trial whose endpoint is how much conjugated benzene appears in urine. "Clinically proven to detoxify" is the phrase where a conjugation-flux biomarker becomes an implied disease claim, and that phrase is worth a great deal of money. The systematic review of randomised trials found exactly what you would expect underneath it: surrogate endpoints, short durations, and efficacy rarely demonstrated.
• There is a commercial war inside the supplement category too. The bioavailability finding — that active myrosinase roughly triples to quadruples delivered sulforaphane — is a direct attack on the cheaper, shelf-stable, glucoraphanin-only extract segment and a boost for sprout-based and myrosinase-added products. Papers on adding exogenous mustard-seed myrosinase to broccoli seed extract are product-shaped solutions to a product-shaped problem. Read bioavailability claims with an eye to which segment benefits.
• The most-cited safety result is not independent, and we say so rather than hiding it. The twelve-week thyroid-safety null was published by the principal investigators of the broccoli-sprout chemoprevention programme, whose intervention it clears. That does not make it wrong — it is consistent with the other human data and with the mechanism — but it is not disinterested adjudication and should never be presented as such. It also clears the wrong compound: a sulforaphane-rich beverage is not a goitrin load.
• Even the reassuring number is unowned. The widely-repeated goitrin content-versus-threshold comparison is rhetorically perfect for the pro-vegetable side, gets repeated uncritically, and has apparently never been challenged by the fear side either — which suggests nobody has checked it. Both camps are content with an untraced figure when it is convenient or when contesting it is not worth the trouble. We decline to print it.
• The clean signal. Three independent legs converge here. The group that characterised the goitrogenic compounds in 1986 also fed them to people and found nothing, four decades before any sulforaphane market existed. A differently-designed randomised trial with the opposite commercial interest found nothing either. And a 2024 systematic review of 123 studies, by a group with nothing to sell in this category, concluded that brassica vegetables in the daily diet pose no adverse effect on thyroid function when iodine intake is adequate. Convergence across a hostile-interest pair plus a disinterested reviewer is about as good as this literature gets. And the honest middle position it produces — a real conditional interaction affecting a small, identifiable group — has no seller on either side, which is the usual signature of a position worth trusting. See plant_defence_compounds_and_xenohormesis for the same distortion pattern across the wider class of plant compounds.

Sources (13)

HIGH on the practical verdict (the harm evidence and the safety evidence converge on the same dose boundary from opposite sides, and they now converge from more than one research group); MODERATE on the individual safety studies (each small and short, several reached through a systematic review rather than opened as primaries); LOW/EMERGING on every clinical-outcome claim made for concentrated sulforaphaneLast updated July 2026
Open in the Library: search, filter, every entry →

We set no cookies and run no ad trackers. We count visits with Cloudflare's cookieless, privacy-first analytics. The only thing stored on your device is which example you last viewed.