Sulforaphane benefits

Sulforaphane has been given to people in trials for type 2 diabetes, prediabetes, Helicobacter pylori gastritis, autism, air pollution exposure, tobacco smoke exposure, inflammation, pneumonia and bleeding around the brain. That is a wider range of conditions than most dietary compounds reach, and it makes sulforaphane unusual. The list also hides a pattern that matters more than its length: in nine of the eleven trials the thing measured at the end was not a disease. Eight finished with a marker in blood or urine, and one with behavior scales filled in by caregivers.

That distinction decides what this section can tell you. A trial showing that urinary excretion of a benzene conjugate rose 61 percent is a real result about detoxification chemistry. It is not evidence that anybody avoided a cancer. Both sentences are true and only the first one is usually reported. Every claim below names its trial, its sample size and its outcome type, and says when the honest answer is that the question has not been asked.

What has sulforaphane been tested for in people?

Eleven trials carry the human evidence. Sample sizes run from 12 to 291, durations from one week to eighteen, and the doses are expressed in different units because the preparations differ. Nine of the eleven were randomized with a comparison group; two had no control arm at all, and that is marked in the table rather than left for you to work out.

Read the right-hand column before the left one. Eight of the eleven trials ended in a laboratory measurement. One measured behavior, in 44 people. Two measured a clinical endpoint, the kind a doctor would act on, and those two are the largest trials on the list at 105 and 133 patients. Neither found an effect. Both used stabilized sulforaphane, a synthetic preparation given as a drug at a dose well above a food portion, set out with its pharmacokinetics on that page.

So the trials that measured a marker in blood or urine mostly found something, and the two that followed patients to recovery, length of stay and death did not. Those two were in acutely ill patients over 14 and 28 days, which is a hard place to show a benefit, so this is not proof that the marker results lead nowhere. It does mean that the strongest claim the human evidence supports is about chemistry measured in a sample, and that every time anyone has asked the harder question, the answer has come back flat.

Does sulforaphane lower blood sugar?

A little, in the trials built to find out, and the most recent and most careful of them fell short of its own target. Dwibedi and colleagues randomized drug-naive adults with prediabetes to broccoli sprout extract (n = 35) or placebo (n = 39) once daily for 12 weeks, with a prespecified primary outcome of a 0.3 mmol/l reduction in fasting blood glucose against placebo. The reduction they measured was 0.2 mmol/l, with a confidence interval of -0.44 to -0.01 and P = 0.04. The trial reports that result as not meeting its primary outcome, and that is the correct way to describe it.

The earlier work pointed higher. Bahadoran's team gave 10 g or 5 g a day of broccoli sprout powder, or placebo, to 81 patients with type 2 diabetes for four weeks, analyzed 63 of them, and found lower fasting serum insulin and HOMA-IR on the 10 g dose at P = 0.05 for the treatment effect. Four weeks is short, HOMA-IR is a calculation from fasting insulin and glucose rather than a measurement of insulin action, and the 5 g arm showed nothing.

Between those two sits Axelsson 2017, the paper that sent sulforaphane into diabetes research in the first place. It was mostly a computational and laboratory study, matching a liver gene-expression signature for type 2 diabetes against 3,800 drug signatures, confirming the effect on hepatic glucose production in cells and animals, and finishing with a clinical arm in obese patients with poorly controlled type 2 diabetes. Its abstract states neither the number of patients nor the dose, and the full text is behind a subscription, so no figure from that human arm appears on this site. The animal comparison in it, an effect on glucose production of a size similar to metformin, is a result in diabetic animals and nothing more.

Put the three together and the picture is consistent and modest: a real but small effect on fasting glucose in people whose glucose is mildly raised, measured over weeks, with the best-designed trial landing below the threshold its own authors set. Doses used across all three, with the assay for each, are set out on how much sulforaphane per day.

Can sulforaphane prevent cancer?

Nobody has tested that. The human trials in this area measure how fast the body conjugates and excretes a carcinogen, which is a step on the way to a cancer risk argument and is not the same as a cancer avoided. Two trials in Qidong, a region with high aflatoxin exposure and high rates of liver cancer, are the clearest case.

Kensler's trial gave 200 adults nightly infusions of three-day-old broccoli sprouts containing either 400 or under 3 µmol glucoraphanin for two weeks and measured urinary aflatoxin-N7-guanine, a DNA adduct. There was no difference between the two arms (P = 0.68). Phenanthrene tetraol, the second marker, showed no difference either (P = 0.29). The trial's positive finding was of a different kind: within the group receiving glucosinolates, the more sulforaphane metabolite a person excreted, the lower their adduct levels, at P = 0.002 for aflatoxin and P = 0.0001 for phenanthrene. That is an association across individuals, not an effect of the assigned dose, and the authors said plainly that understanding what drives hydrolysis and bioavailability would be needed before broccoli sprouts could be used well in human studies.

Egner's trial nine years later answered part of that by raising the dose and the duration: 291 participants, 600 µmol glucoraphanin plus 40 µmol sulforaphane daily for 12 weeks, with a beverage rather than an infusion. Urinary excretion of the benzene conjugate rose 61 percent and the acrolein conjugate 23 percent against placebo, both at P ≤ 0.01, and crotonaldehyde did not move. Bauman's crossover in 49 smokers found the same direction at a smaller dose, with the higher of two doses raising all three conjugates.

So the detoxification claim has three independent human trials behind it, and it is a claim about pollutant handling. Sulforaphane's anticancer literature is enormous and is almost entirely cultured cells and animal models, where it does things that cannot be read across to a person at these doses. The clinical trials that exist have measured urine.

What happened in the autism trial?

This is the one human result where the measured outcome was behavior rather than chemistry, and it deserves care in both directions. Singh and colleagues randomized 29 young men aged 13 to 27 with moderate to severe autism spectrum disorder to 50 to 150 µmol of sulforaphane daily and 15 to an indistinguishable placebo, for 18 weeks, followed by four weeks with no treatment. Scores came from caregivers and physicians on three established scales.

Placebo scores barely moved, under 3.3 percent. The treated group improved 34 percent on the Aberrant Behavior Checklist (P < 0.001 between treatments) and 17 percent on the Social Responsiveness Scale (P = 0.017), and more of them improved on the Clinical Global Impression scale for social interaction, abnormal behavior and verbal communication. When treatment stopped, scores on all scales rose back toward pretreatment levels.

The withdrawal result is the part that makes the trial hard to dismiss, because a placebo response would not usually reverse on schedule. Against that: 44 participants, one site, all male, caregiver-rated outcomes in a trial where caregivers knew a trial was happening, and no replication at this size in the decade since. One trial of 29 treated people is a reason to run a larger one.

Does it help the stomach, and does it reduce inflammation?

For H. pylori, one trial says yes on markers. Yanaka's team randomized 48 infected patients to 70 g a day of broccoli sprouts, carrying 420 µmol of the sulforaphane precursor, or an equal weight of alfalfa sprouts, for eight weeks. Urease measured by the urea breath test fell, so did H. pylori stool antigen, and so did serum pepsinogens I and II, which track gastric inflammation. Alfalfa did nothing. Two months after the sprouts stopped, every value had returned to where it started, which says the sprouts suppressed colonization without eradicating the organism. Standard antibiotic therapy clears H. pylori; this did not.

For inflammation the evidence is weaker than its citation count suggests. López-Chillón's study fed 30 g of broccoli sprouts a day to 40 overweight adults for ten weeks and found mean IL-6 down from 4.76 to 2.11 pg/ml at P < 0.001, with C-reactive protein falling as well. It had no placebo arm. Ten weeks of attending a nutrition study changes what people eat and how much they move, and inflammatory markers drift on their own, so the design cannot separate the sprouts from everything else. The authors call for further studies, and they are right to.

Why does the same dose do different things to different people?

Because the conversion step happens partly outside your control. Sulforaphane is not in the food; the food holds glucoraphanin, and something has to cut it. The plant's own enzyme does it when you chew fresh tissue, and when that enzyme has been destroyed by cooking or by commercial freezing the job falls to gut bacteria, which is where the variation comes from. That mechanism is set out in foods high in sulforaphane.

Kensler's 2005 trial recorded striking differences between individuals in metabolite excretion on an identical dose and flagged them as the obstacle. Twenty years later the prediabetes trial identified part of the cause. The adults who responded to the extract carried a higher abundance of a Bacteroides-encoded transcriptional regulator required to convert the inactive precursor to sulforaphane, and the abundance of that gene operon correlated with serum sulforaphane concentration. The responders also shared a clinical profile: mild obesity, low insulin resistance and reduced insulin secretion, with a 0.4 mmol/l fall in fasting glucose rather than 0.2.

Those were exploratory analyses in a trial that missed its primary outcome, and the authors say they need confirming in future trials. Treat them as the best available explanation of why your result may differ from the trial average, not as a way to predict it. Nothing accumulates in the meantime: plasma half-life was 1.77 ± 0.13 hours in four subjects dosed once, by cyclocondensation assay, so each day's dose is handled and cleared.

Does sulforaphane activate Nrf2 in people?

One trial has checked, at a drug dose, for two weeks, in patients, and the answer was no. Almost every benefit claimed for sulforaphane is explained through Nrf2, a transcription factor that switches on a set of antioxidant and detoxification genes. The cell and animal evidence for that mechanism is large and consistent. The human evidence for it is one measurement.

Long and colleagues gave 300 mg of a stabilized synthetic sulforaphane daily for 14 days to hospitalized patients with community-acquired pneumonia, 65 of them, against 68 on placebo, in Dundee between November 2020 and May 2021. Alongside the clinical outcomes they sequenced messenger RNA from blood leukocytes and ran a 45-cytokine serum panel. Expression of Nrf2 target genes was not significantly changed. Neither were interleukin-6, interleukin-1 beta or tumor necrosis factor alpha. The authors' own conclusion is that the drug "did not improve clinical status or modulate key Nrf2 targets".

The sequencing still returned findings, which is what makes the result worth reading rather than filing as a failed trial. It identified 310 genes differentially expressed between the groups at day 8 and 42 at day 15, at a false discovery rate below 0.05, so the treatment was changing gene expression. What it did not change was the specific pathway the whole story rests on, at a dose roughly six times the preformed sulforaphane in the Qidong beverage. SARS-CoV-2 caused the pneumonia in 103 of the 133 patients, so this is mostly a trial in COVID-19.

One trial in one patient group cannot settle a mechanism. But it is the only direct human test that exists, and anyone citing Nrf2 activation as the reason to take sulforaphane should know that when somebody went looking for it in people, they did not find it. The claim turns up most often on a label, and what a sulforaphane supplement can support is set out in the supplements section.

Does sulforaphane reach the brain?

Less than the field assumed, and that was measured for the first time in 2024. Sulforaphane is often discussed as though brain effects follow from the animal work, because a small fat-soluble molecule was expected to cross the blood-brain barrier without difficulty. The SAS study tested that expectation by sampling spinal fluid.

Zolnourian and colleagues randomized 105 patients after spontaneous subarachnoid hemorrhage, 54 to 300 mg of stabilized sulforaphane twice daily for up to 28 days and 51 to placebo, and made plasma and spinal fluid concentrations a primary outcome rather than an afterthought. The drug reached the blood: areas under the curve were 16.2 for sulforaphane, 277 for its glutathione conjugate and 415 h × ng/ml for the N-acetylcysteine conjugate. It did not reach the spinal fluid to the same degree. Concentrations there were low, many samples fell below the limit of quantification, and what was measurable was predicted by the spinal-fluid-to-serum albumin ratio (R² = 0.182, p = 0.039), meaning the compound appeared where the barrier was already damaged. The authors write that sulforaphane "penetrated CSF less than expected".

They also say something blunt about the literature they were building on. Of the large body of animal work on sulforaphane in neurological disease, only one study had ever quantified brain or spinal fluid levels, and none had done it in humans. The trial's clinical outcomes matched the pharmacology: no difference in vasospasm, in spinal fluid haptoglobin or malondialdehyde, or in functional recovery on either scale.

Two limits belong with that. These were patients whose blood-brain barrier had been disrupted by a hemorrhage, which is not a healthy brain, and the only sampling site was spinal fluid, which tracks brain tissue imperfectly. The honest summary is that the one human measurement of brain delivery came in lower than the animal literature had assumed, and that no trial has tested sulforaphane for memory, mood, focus or cognitive decline in people at all.

What would change this picture?

Trials in people who are well, followed long enough to matter. The two trials here with hard clinical endpoints both recruited acutely ill hospital patients, where a 14 or 28 day course of anything has little room to show an effect, and both were negative. The prediabetes trial is the closest thing to the other design, a registered placebo-controlled trial of a standardized extract in people who were not ill, and over 12 weeks it produced a small effect that missed its own target. Nobody has run a long trial in healthy adults with an outcome a person would notice.

There is a reason those are scarce. Sulforaphane cannot be patented as a compound, so the academic trials are small and the funded ones belong to a company developing a drug version. The preparations also differ enough that a dose in one trial is not a dose in another, which is why the doses in the table above are quoted in whatever unit each paper used.

Until more of those exist, the useful questions are narrower ones: what a trial actually gave people, how much of it was absorbed, and what the preparation on a label does and does not contain. Those are covered on sulforaphane supplements and broccoli sprouts. What the trials recorded going wrong is on sulforaphane side effects, where the gastrointestinal events in these same studies are set out by dose.

No pages in this section yet.