Sulforaphane side effects

What have the trials actually recorded?

Digestive complaints, counted and graded. The most detailed account comes from a randomized crossover trial in 49 current smokers at the University of Arizona, who took a broccoli seed and sprout extract for two weeks at each of two doses. Twenty-nine of the 49 reported at least one adverse event of grade 1 or 2, and all 29 of them reported a gastrointestinal complaint. Nothing reached grade 3 or higher. Loose stool that did not meet the formal criteria for diarrhea affected 24 percent, and diarrhea itself 22 percent. Grading followed the National Cancer Institute's Common Terminology Criteria for Adverse Events, version 4.0, which matters because it is the difference between a counted observation and a reader's impression.

The only formal phase I safety study of broccoli sprout preparations is older and much smaller. Johns Hopkins ran three cohorts, each of three treated volunteers and one placebo recipient, as in-patients on a crucifer-free diet. They took sprout extract orally every eight hours for seven days, 21 doses in total, at 25 µmol of glucosinolate, 100 µmol of glucosinolate, or 25 µmol of isothiocyanate. Thirty-two types of hematology and chemistry test were run before, during and after. Liver transaminases and thyroid function were examined in detail. The authors reported no significant or consistent abnormal events of any kind. Nine treated people is a small number, and the study's own framing is that no concerns had been reported and a formal check was overdue.

The longest trial to report on tolerability is a 2025 placebo-controlled trial in Swedish adults with prediabetes: 35 people on broccoli sprout extract against 39 on placebo, daily for 12 weeks. It recorded gastrointestinal side effects and no severe adverse events.

The cleanest placebo comparison comes from a pharmaceutical dose rather than a food one. In the phase 1 study of stabilized sulforaphane, 174 adverse events were recorded in 17 of the 18 healthy men who took the drug for seven days, with abdominal discomfort in 94 percent, nausea in 56 percent and vomiting in 28 percent. The 6 men on placebo reported nothing at all. Two events reached moderate severity and none was serious. That dose was about six times the preformed sulforaphane in the Qidong beverage and arrived in one enteric-coated tablet, so the rate does not transfer to a meal or a capsule. What it does establish, better than any food study, is that the digestive effects belong to the compound.

Do the side effects get worse at a higher dose?

In the one trial built to answer that, yes. The Arizona crossover gave every participant 148 µmol of glucoraphanin daily for two weeks and 296 µmol daily for two weeks, in a randomized order with a washout between, so each person acted as their own control. Mild abdominal pain and diarrhea were significantly more frequent during the higher-dose period. Ten of the 49 had abdominal pain or cramping on the higher dose, typically alongside loose stool, and six of those ten events were graded 2 rather than 1.

That is a narrow finding and worth stating precisely. It compares two doses of one extract over two weeks in people who all smoked. It does not locate a threshold, and no trial has escalated the dose until something went wrong, so there is no published upper limit. What it does establish is direction: the digestive effects are dose-related rather than idiosyncratic. For what the trials have used and how the units convert, see how much sulforaphane per day.

What about the study that found sulforaphane harmed heart tissue?

It exists, it is real, and it was done entirely outside the body. A group at University Medical Center Hamburg-Eppendorf published it in Redox Biology in 2021, and it is the source behind most of what circulates online about sulforaphane and the heart. Google's AI Overview for this query currently cites a Reddit thread discussing it, which is a fair sign that nobody has explained the paper properly.

What they did: grew three-dimensional engineered heart tissue from neonatal rat cardiomyocytes and, separately, from human induced pluripotent stem cell-derived cardiomyocytes, then exposed it to sulforaphane. Contraction force fell. In chronically treated rat tissue the length-tension response was markedly worse than vehicle control, 191 ± 16 µm against 353 ± 25 µm on the DMSO vehicle and 337 ± 38 µm untreated, with seven to eight tissues per group. The human tissue showed larger force differences when paced at higher frequencies. Sulforaphane raised lactate formation, increased mitochondrial production of reactive oxygen species and lowered mitochondrial membrane potential. Electron microscopy showed disturbed sarcomere organization and swollen mitochondria with whorled membranes. Adding l-glutamine as an alternative fuel partly restored force, which is what pointed the authors at mitochondrial dysfunction as the cause. Their closing recommendation is that cardiac function should be monitored in patients with cardiovascular disease who are taking sulforaphane.

The concentrations are the part to hold on to. The chronic experiments used 1 µM for up to three weeks of continuous culture. The acute dose-response ran from 0.1 to 30 µM, and the authors treated 30 µM as a high bolus. Set that against the only published human pharmacokinetics: four volunteers given about 200 µmol of broccoli sprout isothiocyanates peaked between 0.943 and 2.27 µmol per liter in plasma at one hour, measured by cyclocondensation assay, then cleared it with a half-life of 1.77 hours. So the acute 30 µM bolus sits roughly 13 to 32 times above a real human peak, and no plausible oral dose reaches it.

The chronic concentration is the harder one, and glossing over it would be dishonest. One micromolar falls inside that measured human plasma range. The difference is exposure time rather than concentration: a human peak decays within hours on a 1.77 hour half-life, while the engineered tissue sat in 1 µM without interruption for up to three weeks. A tissue bathed continuously and a person who swallows a capsule each morning are not the same exposure, and that is the comparison the paper cannot make for you.

Two things follow. No trial in the table above reported a cardiac adverse event, and not one of the four looked for one with cardiac imaging or function testing, so their silence is weak evidence. And the paper's own authors wrote their caution about patients with existing cardiovascular disease, which is a reasonable thing to raise with a cardiologist and not a reason for anyone to avoid broccoli.

Does sulforaphane affect the thyroid?

The human measurement that exists is reassuring and small. The Johns Hopkins phase I study examined thyroid-stimulating hormone, total T3 and free T4 in detail across its treatment period, in volunteers taking sprout extract three times daily for seven days, and found no consistent abnormality. Nine treated people over one week is not enough to detect a slow drift, and nobody has repeated the measurement over months.

The worry itself is often misattributed. The goitrogenic compound in brassica vegetables is goitrin, and goitrin comes from progoitrin, a different glucosinolate from glucoraphanin. A supplement standardized to glucoraphanin is not a goitrin source, and the two travel together only in whole vegetables. A review of the brassica data reported that radioiodine uptake into the thyroid was inhibited by 194 µmol of goitrin but not by 77 µmol, and that commercial broccoli, broccoli rabe and kale of the Brassica oleracea group hold under 10 µmol of goitrin per 100 gram serving, while collards, Brussels sprouts and some Russian kale hold enough to matter. That is a review rather than a trial, and the thiocyanate figure in it is an estimate the authors derived rather than a measurement in people taking sulforaphane. Anyone treated for a thyroid condition, or eating little iodine, has a sensible question for their doctor before starting a concentrated product.

Who should not take sulforaphane?

Four groups have a real reason to ask first, and in every case the reason is missing evidence rather than demonstrated harm.

Anyone on an anticoagulant or antiplatelet drug. Sulforaphane blocks aggregation in isolated human platelets and has an antithrombotic effect in animals, both measured in the laboratory. No human trial has tested whether it changes bleeding time or interacts with warfarin, clopidogrel or a direct oral anticoagulant. The laboratory signal is consistent enough to raise and far too thin to quantify.

Anyone pregnant or breastfeeding. No trial has dosed either group, so there is nothing to report in either direction.

Anyone with established cardiovascular disease, on the strength of the Hamburg authors' own recommendation above, and with the concentration argument in hand so the conversation starts from what the study actually showed.

Anyone being treated for a thyroid condition, for the reasons in the section above.

Two more groups are worth naming for a different reason. People taking a concentrated extract rather than eating sprouts are the ones in whom every adverse event above was recorded, and the digestive effects scale with dose. And how much sulforaphane a given person produces from a glucoraphanin supplement depends partly on their gut bacteria: the 2025 Gothenburg trial found that the people whose blood glucose responded carried more of a Bacteroides gene needed to convert the inactive precursor, and that the abundance of that gene operon tracked serum sulforaphane concentration. The same variability that makes a dose unpredictable makes a side effect unpredictable. What a label can and cannot tell you about this is covered on the sulforaphane supplement page.

None of this applies to eating broccoli sprouts or broccoli. No adverse event in any trial above came from a vegetable.

What the trials cannot tell you

Every number on this page comes from a study of between 9 and 49 treated people, running from one week to twelve. A trial that size over that span cannot detect an effect that appears in one person in a thousand, or after three years of daily use. Nor can it find harm it did not measure: none of these trials ran cardiac function testing, and the sprout-extract phase I study that did run broad laboratory panels had nine people on treatment.

So the accurate summary is narrow. At doses up to 296 µmol of glucoraphanin daily for up to twelve weeks, and up to 92.4 mg of preformed sulforaphane daily for a week, the recorded side effects are digestive, dose-related and mild by formal grading, and the broad laboratory panels that have been run came back clean. Beyond that, the honest answer is that nobody has looked.

The same constraint applies at the other end of the ledger. These are the trials the benefit claims come from too, and most of them finished by measuring a marker in urine rather than a disease. What each one gave people and what it found is set out claim by claim under sulforaphane benefits.