Bottom line
Eating a wide mix of plant foods, in many colors, tracks consistently with lower risk of death and of chronic disease. Once a single compound is purified into a supplement, large randomized trials have mostly failed to deliver on the promise 283031.
One class of evidence is the firmest, and it is bad news. People who smoke and take beta-carotene for a long time get more lung cancer, not less 2223. The US Preventive Services Task Force therefore opposes using it to prevent cardiovascular disease and cancer 32.
The large cocoa-extract trial COSMOS did not lower total cardiovascular events. It saw fewer cardiovascular deaths only on a secondary outcome 191.
Sulforaphane is the compound in this group with the clearest mechanism. Small human trials show a signal. There is not yet hard evidence on disease outcomes 102527.
The views split clearly. One is the most enthusiastic. One is the most skeptical. Another is practical: food first, a supplement optional 926.
What it is
Plant compounds are chemicals plants make to resist ultraviolet light, insect bites, and microbes. They are not vitamins. Missing one does not produce a "deficiency disease."
Polyphenols are the largest family. Flavonoids are a large class inside polyphenols. The flavanols in cocoa, the quercetin in onions and apples, and the resveratrol in grapes all belong to it. Tea, berries, cocoa, beans, and nuts are rich in polyphenols.
Sulforaphane comes from cruciferous vegetables such as broccoli, cauliflower, and cabbage. What the vegetable stores is the precursor, glucoraphanin. It becomes sulforaphane only with the help of the enzyme myrosinase 10.
That conversion depends on chopping and chewing. The enzyme is afraid of heat, and a long boil inactivates it 10. Broccoli sprouts contain ten to one hundred times more of the precursor than mature broccoli 10.
Carotenoids are the pigments that make carrots, tomatoes, squash, and egg yolks orange, yellow, and red. They include beta-carotene, lycopene, lutein, and zeaxanthin. They dissolve in fat and are absorbed better with fatty food 12.
Nrf2 is the cell's master defense switch. Activated, it turns on many antioxidant and detoxification genes 915.
Hormesis means a small stress makes the body stronger. One view holds that many plant compounds work as that kind of small challenge 8. Bioavailability is how much of what was eaten enters the blood. Quercetin's oral absorption is only about 1% 13.
Why it matters
Cardiovascular disease, diabetes, and brain decline are all tied to chronic inflammation and oxidative damage, and plant compounds are thought to modulate those processes.
People who eat more vegetables and fruit do live longer. In two large Shanghai cohorts, 134,796 people altogether, the fifth of the population eating the most cruciferous vegetables had a total death risk about 22% lower than the fifth eating the least 28. The association showed up mainly in cardiovascular death 28.
Among 124,805 people in the UK Biobank, those with the most diverse flavonoid sources had a 6% to 20% lower risk of all-cause death and of several chronic diseases 30.
There is a historical lesson here. Researchers once found that people with high blood beta-carotene had less lung cancer, so they ran supplement trials. People who smoked and then took the supplement got more lung cancer, and total mortality was higher too 2223.
So the real question is not "are plants good." It is whether the benefit comes from a particular molecule, from the whole food, or from the fact that people who like vegetables and fruit were already healthier.
Different perspectives
Main positions
One view is the skeptic. It weights randomized trials and absolute risk, and it holds that most plant-compound benefits have been overstated.
On COSMOS, the primary outcome did not differ. Cardiovascular death had a 27% lower relative risk, and the absolute risk fell only about 0.26 percentage points, from 0.97% to 0.71% 1. The speculation is that cocoa may have lessened severity, and the benefit sat in people who had smoked, not in never-smokers 1. Reaching the trial's flavanol level with dark chocolate would put about 57% of the day's calories in that chocolate. The slogan that chocolate protects the heart does not hold, in that view 1.
On flavonoid diversity, that account names healthy-user bias 2. The most varied eaters also tend to move more and smoke less. There is no staircase of "more diverse, better." The result looks like noise 2. It still argues for colorful vegetables, fruit, and tea, not as a silver bullet 2.
Resveratrol draws the hardest criticism. A meta-analysis of randomized trials showed no improvement in lipids, glucose, or liver enzymes, and the logical chain is held to have failed from the start 3. In an interview, the same view noted the mouse Interventions Testing Program found no lifespan effect 5. The red-wine headache study, the note says, was a test-tube experiment 4.
Another account is practical. What it discusses most is sulforaphane and lutein: food first, a supplement optional.
On a microplastics episode, sulforaphane is treated as support for the liver's phase II detoxification, hanging a handle on a harmful substance so it can leave in urine 6. Animal-study amounts, the account says, are roughly a large serving of raw broccoli 6. A light steam or a short sauté can still keep sulforaphane 6.
In an eye-health newsletter, lutein takes part in retinal light sensing, comes from egg yolk and dark leaves, and has research tied to age-related macular degeneration 7. The principle is still more unprocessed vegetables and fruit 7. Detail in that material about polyphenols often comes from a guest, including mustard-seed powder on cooked broccoli raising sulforaphane about fourfold 8.
The most enthusiastic account treats sulforaphane as the strongest natural Nrf2 activator and polyphenols as antioxidants with a dual action: clearing free radicals directly, and activating Nrf2 indirectly. Gut bacteria split polyphenols into absorbable pieces. A poor microbial community discounts the effect 9. It holds sulforaphane can mobilize more than 200 protective genes 1015, and it often cites the Shanghai result, a 22% lower total death risk in the highest cruciferous fifth 15.
Trials cited there: a broccoli-sprout drink in Qidong raised urinary benzene 61% and acrolein 23% 10. In men with rising PSA after prostate-cancer surgery, sulforaphane put doubling time at about 29 months, against about 16 on placebo 10.
Cocoa, that account says, is among the highest-flavonoid foods, and processing loses more than 60%. A 2022 consumer survey found heavy metals above California limits in most cocoa products tested 11. Eight men who drank a cocoa smoothie with banana had a flavanol-metabolite peak about 84% lower, because banana carries an enzyme that breaks polyphenols down 16. The stress is that it is a small study, and banana itself is still nutritious 17. Blueberry trials are mixed 17.
High blood carotenoids track with lower risk of obesity, insulin resistance, inflammation, and early death. With fat, as in eggs, absorption can rise about 3.8-fold 12. In 4,173 postmenopausal women, higher blood carotenoids tracked with slower epigenetic aging 18. On quercetin the interest is in pairing it with another drug to clear senescent cells. Median lifespan in mice rose 36%. The anti-inflammatory effect in humans has not been reproduced reliably 13.
Where views agree and differ
Food first, and single molecules get over-promoted. They differ on how far observational evidence can be trusted, and on supplements.
Colorful vegetables and fruit, tea, and cruciferous vegetables are a reasonable daily habit 279. Absorption matters: heat destroys the enzyme, carotenoids need fat, and polyphenols need the gut microbiota 101296. Not every star molecule survives. One view rejects resveratrol 3. Another grants the human anti-inflammatory evidence on quercetin is inconsistent 13.
On cohorts, one account pairs them with mechanism as evidence the food helps 15. Another holds this is where healthy-user bias misleads most easily 2. One wants two lines of clues. The other wants a hard outcome from a randomized trial.
On supplements, one view holds that a sulforaphane supplement can make up for few vegetables 6. Another weights sprouts and food 15. A verdict on cocoa extract is a possible very small benefit, far behind exercise 1.
On COSMOS, one reading stresses a negative primary outcome and a tiny absolute benefit 1. A cocoa page talks more about blood pressure, endothelium, and small trials 11. One asks whether fewer people die. The other asks whether the body changed.
PearHealth's synthesis
A varied plant diet is the adopted center. Single-compound supplements are held back. People who smoke, or who used to, are not well served by beta-carotene supplementation. The observational evidence points one way 2830, the foods are low risk, and they also bring fiber.
COSMOS's primary outcome was negative 19. A clinic substudy of 573 people saw no overall cognitive improvement over two years 21. An online study of 3,562 saw a memory improvement only where diet quality was poor and flavanol intake was low 20. That looks like filling a gap, not "more is better."
"More antioxidant, better" is in doubt. Cochrane pooled 78 randomized trials and nearly 300,000 people. In the low-bias trials, antioxidant supplements slightly raised the risk of death 31.
Sulforaphane work in autism and prostate cancer is interesting and still research. In children, the primary outcome was not statistically significant 26. That use belongs with a clinician, inside a proper trial, and cannot stand in for treatment.
We do not sell any supplement, and an ordinary person does not need an extract to get these benefits.
How strong is the evidence
The evidence of harm is the firmest. Food benefits come mainly from observational research. Supplement trials are mostly negative, or show a signal only on secondary outcomes and in subgroups. A randomized trial can speak to cause. Observational research is easily disturbed by confounders.
The firmer part:
- Beta-carotene: in the Finnish trial of 29,133 male smokers, lung-cancer incidence rose 18% 22. In the US trial of 18,314 smokers and people exposed to asbestos, the lung-cancer hazard ratio was 1.28 and the all-cause death hazard ratio was 1.17, and the trial was stopped 21 months early 23.
- Lutein and zeaxanthin: in an eye-disease trial of 4,203 people, adding them did not further slow late macular degeneration 24. The group that included beta-carotene had more lung cancer, seen mainly in former smokers 24.
The middle part:
- Cocoa: COSMOS enrolled 21,442 older adults, with a median follow-up of 3.6 years. The hazard ratio for total cardiovascular events was 0.90, and the confidence interval crossed 1, so it does not count as significant. The hazard ratio for cardiovascular death was 0.73 19.
- A high-polyphenol Mediterranean diet: the DIRECT PLUS trial, 284 people over 18 months, showed slower age-related brain shrinkage 29. People aged 50 and older had a clearer reduction in brain shrinkage, and younger participants saw a similar direction 29. But 88% of the participants were men, and the diet changed several things at once, so the credit cannot all be given to polyphenols 29.
- Sulforaphane: the Qidong trial, 291 people over 12 weeks, increased benzene excretion by 61% 27. That is an intermediate marker, not a disease outcome.
- Autism: in a trial of 44 young men, one behavior scale improved 34% 25. In a later trial of 57 children, the primary outcome did not differ significantly 26.
The weaker part:
- The links between cruciferous vegetables, flavonoids, and longevity come from cohort studies 2830.
- The US National Cancer Institute's summary holds that human studies of the relationship between cruciferous vegetables and cancer are not consistent 33.
What you can do
Directions many people take:
- Rotate the cruciferous vegetables: broccoli, cauliflower, cabbage, Chinese broccoli, and, where it is available, broccoli sprouts 10.
- Steam lightly or heat briefly. Avoid a long boil. A little mustard-seed powder on cooked broccoli can put the converting enzyme back 815.
- Pair orange, yellow, red, and dark-green vegetables with a little fat or with eggs, which helps carotenoid absorption 12.
- Rotate polyphenol sources: tea, berries, apples, oranges, grapes, beans, and nuts 30.
- For cocoa, a less-processed cocoa powder is one option. Treat chocolate as a snack, and keep an eye on sugar and calories 111.
- If flavanol absorption is a concern, a smoothie can be made without banana. That comes from only one very small study 1617.
Questions worth raising with a clinician:
- I smoke, or I used to smoke. Does the multivitamin or the eye supplement I take contain beta-carotene 2432?
- I have a thyroid problem, or my iodine intake may be low. Does eating a large amount of cruciferous vegetables, or using a sulforaphane supplement, need attention 15?
- Could a plant-extract supplement interact with a medicine I am taking?
- I have a history of prostate cancer and my PSA is rising. How much does the sulforaphane research have to do with my situation, and how much does it not 10?
- A child has autism. What do the current sulforaphane trial results show, and what do they not show 2526?
Common misconceptions
Misconception one: the more antioxidant eaten, the better. A pooling of randomized trials showed that antioxidant supplements did not lower death, and in the low-bias trials the risk rose slightly instead 31.
Misconception two: eating dark chocolate is the same as protecting the heart. The trial used a concentrated extract. Reaching the same level by eating chocolate is nearly impossible 1. Some chocolate products also have a heavy-metal problem 11.
Misconception three: the resveratrol in red wine slows aging. Randomized trials and a mouse lifespan test both fail to support that 35.
Misconception four: a supplement is concentrated vegetables. Sulforaphane-precursor supplements often lack the enzyme the conversion needs 15. Vegetables also contain fiber and other components. A single molecule cannot stand in for the whole food.
How it connects to your blood work
A routine blood panel does not measure polyphenols, sulforaphane, or carotenoids. Related research does often use the markers below.
- High-sensitivity C-reactive protein (hs-crp): a reflection of chronic inflammation throughout the body. A meta-analysis one review cites shows that quercetin lowered it in people over 40 or in people who already had a disease, and the human results are not consistent 13.
- Glycated hemoglobin (hba1c) and fasting glucose (fasting-glucose): in a sulforaphane study of 97 people with type 2 diabetes, the improvement showed up mainly in people with obesity and poor glucose control 10. A higher share of the high-polyphenol diet group also improved glucose control 14.
- Insulin-resistance index (HOMA-IR) (homa-ir) and fasting insulin (fasting-insulin): in DIRECT PLUS, improvement in insulin sensitivity was the marker most closely tied to slower brain shrinkage 29.
- Triglycerides (triglycerides) and LDL cholesterol (ldl-c): one view holds that cocoa may improve lipids 11. The resveratrol meta-analysis saw no change in lipids 3.
- Thyroid-stimulating hormone (tsh) and free thyroxine (free_t4): cruciferous vegetables are worried over for an effect on the thyroid. One view holds that this needs attention mainly when iodine is severely lacking 15.
- Prostate-specific antigen (psa): one sulforaphane trial used PSA doubling time as the outcome, in men after prostate-cancer surgery. It does not apply to the general population 10.