A study of 10,754 Americans puts a number on flavonoids: 219 mg a day. That is also, almost to the milligram, where the top quarter of its sample begins

In 2023 the American Heart Association gave a name to something doctors see every day: heart disease, kidney disease, diabetes and obesity turning up in the same people and feeding one another. It called the cluster cardiovascular-kidney-metabolic syndrome, CKM for short, and set out five stages, from 0 (no risk factors) through excess weight or prediabetes (1), further metabolic risk factors or kidney disease (2) and high predicted risk (3) to established heart disease or stroke (4). A team of cardiologists at Anqing Municipal Hospital in China has now asked whether flavonoids, the plant compounds in tea, berries, citrus, onions and soy, track with those stages and with who dies. Their paper came out in the journal Medicine on 2 October 2026.

They did not recruit anyone. They used the US government's public health survey, NHANES, taking 10,754 adults from the 2007 to 2010 and 2017 to 2018 rounds. Each person had described everything they ate and drank on up to two separate days, and those records were matched to a US Department of Agriculture database of 29 flavonoid compounds. The survey is linked to national death records up to the end of 2019, and 1,500 of the participants had died by then: 368 from cancer and 350 from heart and circulatory disease.

People were split into four equal groups by flavonoid intake. Among those at stages 1 and 2, the top quarter had a 40% lower risk of dying from cancer than the bottom quarter after adjustment for age, income, smoking, weight and a dozen other things. Among those at stages 3 and 4, the top quarter had a 20% lower risk of dying from any cause. The curve flattened at 218.8 mg a day; below that, every extra 50 mg went with a 9% lower risk of death. The authors call this figure a potential target for future research, and to their credit they add that it is not meant for advising patients and needs confirming in trials.

Now the caveats, and from here the reading is ours. The 40% figure has a range running from 63% lower to 4% lower, and the 20% figure from 35% lower to 1% lower. Those are the two results that passed the usual test, out of three causes of death examined in several groupings of stages, and the main text mentions no correction for making that many comparisons. Deaths from heart disease, the outcome this syndrome is named for, showed no clear link in any group. The crude death rates were 11.1% in the bottom quarter and 8.7% in the top, a gap that was not statistically significant. And the top quarter was a different crowd: 63% had been to college against 47% of the bottom quarter, and 48% were in the highest income band against 31%. A model can adjust for the differences it measured, not the ones it didn't.

Then there is the 219 mg. The cut-off for the top quarter of the sample was 218.47 mg a day, and the threshold found by the curve-fitting was 218.83 mg, so the 'target' is in effect the line between the top quarter and everyone else. Total flavonoids in this data moved almost in lockstep with one subclass, the flavan-3-ols (a correlation of 0.99), and an earlier Department of Agriculture analysis of the same survey found that tea supplies about 80% of the flavonoids American adults get. So 'high flavonoid intake' here mostly means 'drinks tea'. The paper also credits isoflavones, the soy compounds, with half of the benefit on overall deaths, yet 38% of participants ate none and the typical intake among the rest was 0.01 mg a day. That part came from a method which, the authors say, was run without the survey's weights and was set to look for protective effects only.

A few figures in the first table look like slips. It lists 53.5% of participants as current smokers and 18.6% as never having smoked, which looks like the two labels were swapped. It gives average carbohydrate intake as 18 g a day beside 2,076 calories, which cannot both be right. A result described as a trend is printed with P = .51. And the 9.6-year median follow-up covers people surveyed in 2017 and 2018, who could have been followed for three years at most. None of this decides whether tea or berries help; plenty of earlier studies point the same way. It is a reason to read '219 mg a day' as a feature of one dataset and not a dose. The work was funded by the Anqing Science and Technology Bureau, and the authors report no conflicts of interest.