The claim
In late October a study in the Journal of Internal Medicine made headlines around the world. Researchers followed nearly half a million Chinese adults for about nine years and found that people who drank tea regularly had an 11% lower risk of heart and blood-vessel disease than people who did not. The sweet spot seemed to be six to eight cups a day. Black tea did better than green. And tea drunk “burning hot” was linked to higher risk.
The headline version is easy to remember: drink tea, protect your heart.
As in earlier posts on this blog, I am not going to try to prove the claim. I am going to try to break it. First I ask why it feels so convincing. Then, following Karl Popper, I set out in advance what evidence would refute it. Only then do I look at the data.
What the study actually found
Before testing the claim, it helps to state it accurately. According to the publisher’s summary and the news coverage:
- Who: 485,934 adults aged 30–79 in China.
- How long: median follow-up of 9.05 years.
- Outcome: 14,863 cases of cardiovascular, cerebrovascular, pulmonary heart or peripheral artery disease.
- Exposure: “regular” tea drinking, which one account defines as at least three days a week, compared with less frequent drinking.
- Main result: about 11% lower relative risk for regular drinkers, after adjusting for age, smoking, physical activity, body weight and existing illness.
- Secondary results: lowest estimated risk at about 6–8 cups a day among weekly drinkers, with risk rising again above that. Black tea showed more consistent associations than green tea. Moderate-strength tea did best. Self-described “burning hot” tea was linked to about 16% higher risk.
Two independent experts quoted in the coverage made the key caveats. The study is observational. Tea intake was measured only once, at the start. Cup size was not standardised. The model was not fully adjusted for diet and income. And many comparisons were run, so the striking sub-findings about black tea and burning-hot tea should be treated as exploratory. The authors themselves cautioned against reading “6–8 cups” as a recommended intake.
That is a careful study, honestly reported. The question is what it can and cannot tell us.
Part 1: Why the claim feels true (the Kahneman–Tversky audit)
1. A big number feels like strong evidence
Half a million people sounds conclusive. But sample size fixes only one problem: random noise. It does nothing about bias. If tea drinkers differ from non-drinkers in ways the researchers did not measure, a bigger sample just gives a more precise estimate of the wrong number. Kahneman called the tendency to trust a big, vivid data set “What You See Is All There Is”: we judge the evidence we have and forget to ask what is missing.
2. A plausible mechanism makes the story feel causal
Tea contains polyphenols, including flavonoids, which have antioxidant and anti-inflammatory effects in the lab. They may help blood vessels relax. A mechanism like that makes the association look like cause and effect. But as the B12 post on this blog showed, plausible mechanisms that pass through biomarkers fail in trials often. Antioxidant vitamins are the classic case: a convincing story, followed by randomised trials that found no benefit.
3. Relative risk sounds bigger than absolute risk
“11% lower risk” is a relative figure. A rough calculation from the reported totals shows why that matters. About 14,863 events among 485,934 people is roughly 3% over nine years. An 11% relative reduction on a baseline of about 3% is a difference of around one-third of a percentage point over nine years. That is a rough illustration, not a figure from the paper, since baseline risk differs between groups. But it shows the scale: even if the effect were entirely causal, it would be modest for any one person. One of the experts quoted in the coverage made the same point, that the reduction is relative.
4. The healthy-user effect
In many populations, the people who make one “healthy” choice tend to make others. The study itself found that the lowest risk was in people who drank tea, did not smoke and were physically active. That joint pattern is exactly what you would expect if tea partly acts as a marker of an overall lifestyle, not just as a cause in its own right. Adjusting for smoking and exercise helps, but adjustment can only remove the confounders that were measured, and measured well. Diet and income were not fully adjusted for.
Verdict of the bias audit: the claim feels stronger than the evidence alone justifies, because it combines a huge sample, a neat mechanism and a relative-risk headline. None of that is evidence of causation.
Part 2: Making the claim falsifiable (the Popperian setup)
For Popper, a claim is useful only if we can say in advance what would prove it wrong. So here is the causal version, and its predictions:
H: Regular tea drinking causally lowers the risk of cardiovascular disease.
If H is true, we should see:
- P1: Tea drinkers have fewer cardiovascular events than non-drinkers in cohort studies, across different countries and cultures.
- P2: In randomised trials, tea improves the risk factors through which it is supposed to work (blood pressure, LDL cholesterol).
- P3: In randomised trials with hard outcomes, people assigned to drink tea have fewer heart attacks and strokes.
- P4: The sub-findings hang together in a coherent way, rather than looking like noise from many comparisons.
P1 is the easy test, because confounding can produce it even if H is false. P3 is the severe test. P2 and P4 sit in between.
Part 3: The evidence
P1 — Observational associations: ✅ Passed (but this is the easy test)
The new study is not alone. A UK Biobank analysis of about half a million British adults, published in Annals of Internal Medicine in 2022, found that drinking two or more cups of (mostly black) tea a day was associated with roughly 9–13% lower all-cause mortality over about 11 years, with inverse associations for deaths from cardiovascular disease, ischaemic heart disease and stroke. Its authors, too, said the results did not mean people should start drinking tea to gain the benefit. That is a different country, a different tea culture and a different way of drinking tea, often with milk. The association turns up repeatedly.
Consistency across populations is worth something. It makes a pure fluke unlikely, and it makes confounding by one country’s specific habits less likely. But it does not rule out the healthy-user effect, which can appear in every country.
P2 — Trials on risk factors: ⚠️ Passed, weakly
Here we do have randomised evidence. A 2014 meta-analysis in PLoS ONE (Greyling and colleagues) pooled eleven placebo-controlled trials of black tea, with 378 participants drinking about four to five cups a day. Systolic blood pressure fell by about 1.8 mmHg and diastolic by about 1.3 mmHg. That is small for an individual, though the authors argued it could matter across a population.
A 2013 Cochrane review (Hartley and colleagues) of green and black tea for primary prevention found 11 trials with 821 participants that lasted at least three months. Tea produced small reductions in blood pressure, and black tea lowered LDL cholesterol. The reviewers stressed that there were very few long-term studies.
So the prediction survives, but only just. The effects on risk factors are real but small. They are roughly in line with a modest benefit, and nowhere near large enough to carry the whole story by themselves.
P3 — Trials on heart attacks and strokes: ❓ Not tested
This is the decisive point. The Cochrane review found no trials that reported cardiovascular events at all. As far as I could find, no one has randomised large numbers of people to drink tea or not for years and counted heart attacks and strokes.
In Popper’s terms, the severe test has never been run. The hypothesis has not failed it, but it has not passed it either. That is a very different position from “proven”.
There are good practical reasons for the gap. Such a trial would need tens of thousands of people, years of follow-up and strict adherence to a daily habit, and no company has a strong financial reason to fund it. But the reason a test was not run does not change what we know without it.
P4 — Internal coherence: ⚠️ Mixed
Some of the sub-findings fit a causal story. A dose-response pattern (more tea, lower risk, up to a point) is one of the classic signs that an association might be causal.
Others are harder to read:
- The U-shape. Risk falling until 6–8 cups and rising again beyond that is possible biologically, but it is also what you might see if very heavy drinkers differ in other ways. The authors explicitly warned against treating it as a recommended dose.
- Black versus green. In a Chinese population, black-tea and green-tea drinkers may differ by region, income and diet. A difference between them could reflect geography as much as chemistry. The outside expert quoted in the coverage called this finding exploratory.
- Burning-hot tea. This is the most intriguing result, and the most fragile. Temperature was self-reported, and one person’s “burning hot” is another’s “warm”. People who prefer scalding tea may also differ in other habits. The same research tradition gives a useful comparison. A 2018 China Kadoorie Biobank analysis in Annals of Internal Medicine found that burning-hot tea was linked to a much higher risk of oesophageal cancer, but mainly in people who also drank alcohol heavily or smoked. Very hot drinks are classified by the International Agency for Research on Cancer (IARC) as “probably carcinogenic”. Sorting out tea from the habits that travel with it is hard even for cancer, where the mechanism (heat damage to the oesophagus) is direct.
When many comparisons are run, a few striking ones will appear by chance. That does not make them wrong. It does mean they are hypotheses for the next study, not conclusions.
What would change my mind
To be honest about uncertainty, it helps to say what evidence would move the verdict:
- Toward “causal”: a long randomised trial with cardiovascular events as the outcome, or a well-designed Mendelian randomisation study (using genetic variants that affect tea or caffeine intake as a natural experiment) showing a consistent protective effect. Repeated measurement of tea intake over time, rather than a single baseline question, would also strengthen the cohorts.
- Toward “mostly confounding”: cohort studies with very detailed diet and income data in which the association shrinks towards zero, or “negative control” analyses showing similar associations for habits that have no plausible link to the heart.
Verdict
| Prediction | Test strength | Result |
|---|---|---|
| P1: Tea drinkers have fewer events | Weak | ✅ Passed, repeatedly |
| P2: Tea improves risk factors in trials | Moderate | ⚠️ Passed, small effects |
| P3: Tea reduces events in trials | Severe | ❓ Never tested |
| P4: Sub-findings are coherent | Moderate | ⚠️ Mixed; several exploratory |
The fair summary is this: regular tea drinking is consistently associated with slightly lower cardiovascular risk, tea has small, real effects on blood pressure and cholesterol in trials, and the size of the benefit, if it is causal, is modest. The claim that tea protects your heart, in the strong causal sense the headlines imply, has never faced its severe test.
For practical purposes, the evidence supports a few modest conclusions. Unsweetened tea is a reasonable drink that may carry a small benefit. It is almost certainly better than a sugary drink. Letting it cool before drinking is sensible on several grounds. And no amount of tea is a substitute for the things with much stronger evidence behind them: not smoking, staying active, and controlling blood pressure and cholesterol.
Notice that the study’s own best-performing group was the tea drinkers who also did not smoke and exercised. If there is one lesson in the data, it is probably that one.
Sources
- Ye Z, et al. “Habitual tea consumption is associated with lower cardiovascular disease risk.” Journal of Internal Medicine, published online 21 October 2026. DOI: 10.1111/joim.70172. https://onlinelibrary.wiley.com/doi/10.1111/joim.70172
- News-Medical (Wiley release). “Regular tea consumption linked to lower risk of vascular disease.” 21 October 2026. https://news-medical.net/news/20261021/Regular-tea-consumption-linked-to-lower-risk-of-vascular-disease.aspx
- Sci.News. “Tea Drinkers May Have Fewer Heart and Blood Vessel Problems, Large Study Finds.” October 2026 (includes independent comments from Queen Mary University of London and the British Heart Foundation). https://www.sci.news/medicine/tea-consumption-cardiovascular-disease-risk-15145.html
- CNN via CTV News. “How you drink your tea could determine whether you gain or lose heart health benefits.” 21 October 2026. https://www.ctvnews.ca/health/article/how-you-drink-your-tea-could-determine-whether-you-gain-or-lose-heart-health-benefits/
- Greyling A, et al. “The effect of black tea on blood pressure: a systematic review with meta-analysis of randomized controlled trials.” PLoS ONE, 2014. https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0103247
- Hartley L, et al. “Green and black tea for the primary prevention of cardiovascular disease.” Cochrane Database of Systematic Reviews, 2013. https://www.cochranelibrary.com/cdsr/doi/10.1002/14651858.CD009934.pub2/information
- Yu C, et al. “Hot Tea Consumption and Its Interactions With Alcohol and Tobacco Use on the Risk for Esophageal Cancer: A Population-Based Cohort Study.” Annals of Internal Medicine, 2018. https://pubmed.ncbi.nlm.nih.gov/29404576/
- Inoue-Choi M, et al. “Tea Consumption and All-Cause and Cause-Specific Mortality in the UK Biobank.” Annals of Internal Medicine, 2022. https://www.acpjournals.org/doi/10.7326/M22-0041
- National Cancer Institute, Division of Cancer Epidemiology and Genetics. News release on the UK Biobank tea study, 2022. https://dceg.cancer.gov/news-events/news/2022/tea-consumption




