Tea in modern science II

一From a promising molecule to a human outcome#

A laboratory mechanism, a short attention test and a study following tea drinkers for years answer different questions. This second part asks what human research can support, where results disagree, and how safety changes with the preparation and the person.

Read Tea in modern science I for the chemistry of cultivar, processing and brewing, including why an infusion cannot be equated with a capsule. Here, L-theanine is considered separately from caffeine combinations; self-reported sleep is distinguished from measured sleep, and changes in risk markers from prevention of disease.

The aim is to make research useful to someone who enjoys Chinese tea without turning a tea style, mountain, vintage or chemical constituent into a treatment claim.

二Reading the evidence: five questions before believing a headline#

Ask what was studied, in whom, against what comparison, for how long, and with which outcome. A randomized controlled trial assigns an intervention and can support a causal conclusion about that intervention under its test conditions. A trial of capsules in adults with raised cholesterol does not establish the effect of a gaiwan of white tea in a healthy teenager.

A prospective cohort follows people's existing habits. Statistical adjustment can reduce measured differences in smoking, alcohol use, diet, income or activity, but cannot guarantee that the groups otherwise match. People may change tea habits because of illness; self-reported cups and strength are imperfect exposure measurements. An association remains an association.

A systematic review uses explicit search and selection methods; a meta-analysis statistically pools studies. Pooling does not remove their bias, turn a surrogate marker into a clinical outcome, or make unlike preparations equivalent. A wide confidence interval means the estimate is imprecise. A small statistically detectable change may be practically minor.

Cell cultures, isolated enzymes and animal models test mechanisms. Chemical assays measure molecules or reactions. These are valuable steps in research, but cannot by themselves show what happens to a person drinking tea. Evidence labels below describe study design; they are not an invented numerical health score.

三Alertness and attention: plausible, measurable, short term#

Human randomized trials. Two 2011 crossover studies compared brewed black tea with a matched placebo drink. Participants performed better on an attention-switching task and reported greater alertness after tea. The study addresses acute attention under test conditions, not intelligence, prevention of dementia, or long-term work performance. The authors were affiliated with a tea manufacturer.

The 2025 systematic review tabulates these trials as 26 and 32 participants receiving two servings totaling about 100 mg caffeine and 46 mg theanine. This is useful exposure information, but it is not a guaranteed composition for Chinese hongcha brewed at home.

Caffeine is a credible contributor to feeling more awake. Comparing caffeinated tea with a caffeine-free placebo cannot, by itself, determine how much of an effect comes from theanine, caffeine, other tea compounds, or their interaction. Prior caffeine abstinence and withdrawal can also complicate interpretation. Enjoying focused time over tea is compatible with the evidence; describing a cultivar as a proven cognitive enhancer goes further than these trials allow.

四Theanine, calmness and cognition: keep the formulation in view#

Systematic review of randomized trials. A 2025 review included 50 trials, with 15 contributing to meta-analyses. Theanine plus caffeine showed small improvements in some attention measures over one to two hours; for attention-switching accuracy in the second hour the standardized mean difference was 0.33 (95% CI 0.13–0.54). Several other estimates were uncertain. The literature search ended in August 2023, and tea-industry funding and author employment were disclosed.

An SMD is measured in study standard deviations: 0.33 does not mean 33% better concentration. Capsule doses and isolated-ingredient combinations cannot be translated into a number of ordinary infusions without composition data. The review itself calls for more beverage and tea-equivalent-dose research.

Theanine findings do not demonstrate that tea cancels caffeine's sleep effects or treats anxiety. Nor can short cognitive tests establish protection against neurodegenerative disease. These are separate questions requiring different populations, follow-up and outcomes.

五L-theanine, stress and sleep: promising signals with important limits#

Human randomized trial. In a 2019 double-blind crossover study, 30 adults without major psychiatric illness took 200 mg L-theanine daily or placebo for four weeks each, with a washout between periods. Some sleep questionnaire subscales improved more with L-theanine. Improvements in depression and trait-anxiety scores compared with each person's baseline did not establish superiority over placebo; the overall sleep score also did not clearly differ between treatments. The manufacturer funded the study and two authors were its employees. This small trial does not establish treatment of an anxiety disorder or insomnia.

Systematic review and meta-analysis. A 2025 sleep review identified 19 articles involving 897 participants, with 18 contributing to meta-analysis; its search extended to September 2024. It found improvements in some subjective outcomes. For overall subjective sleep quality, the standardized mean difference was 0.43 (95% CI 0.04–0.83; 12 studies). That is neither 43% better sleep nor a number of extra minutes. The authors emphasized the shortage of studies isolating pure L-theanine and the unresolved dose and duration.

These findings concern tested formulations and reported outcomes. They do not establish that an evening gaiwan helps sleep, that L-theanine neutralizes the caffeine in tea, or that increasing tea intake reproduces a supplement trial. The caffeine evidence below remains relevant even when a tea contains L-theanine.

六Sleep: an effect worth noticing before seeking benefits#

Human intervention evidence. A 2023 caffeine meta-analysis estimated an average reduction of about 45 minutes in total sleep time across its included experiments. Doses, timing and participants varied, so this is not the predicted loss from one cup of tea. It does show that feeling able to fall asleep is not a complete test of sleep effects.

EFSA notes that even a single 100 mg dose near bedtime can affect sleep in some adults. A long evening gongfu session can extend the period of caffeine intake even when each pour is small. There is no universal cutoff that suits every person or every brewing method.

If evening tea coincides with difficulty sleeping, an earlier and smaller session provides a more practical experiment than choosing by liquor colour. Count coffee, energy drinks, chocolate and supplements as well. 'Aged', 'ripe', 'roasted' and 'smooth' do not mean caffeine-free, and theanine is not a reliable antidote.

七Cardiovascular outcomes: what the large Chinese cohort found#

Observational association. The China-PAR project followed 100,902 Chinese adults from 15 provinces for a median 7.3 years. Habitual drinkers had lower adjusted hazards of atherosclerotic cardiovascular disease incidence, cardiovascular death and all-cause death than never or non-habitual drinkers. The respective hazard ratios were 0.80 (95% CI 0.75–0.87), 0.78 (0.69–0.88) and 0.85 (0.79–0.90).

A hazard ratio of 0.80 is an estimated 20% lower relative hazard, not a 20-percentage-point reduction in an individual's chance of disease. Absolute implications depend on baseline risk and follow-up. The study's modeled life-expectancy comparison is not a promise that starting tea adds years to life.

Tea intake was self-reported rather than assigned. Smoking, drinking, socioeconomic circumstances, diet and other behaviours can differ between groups despite statistical adjustment. This important Chinese cohort supports further investigation of habitual tea drinking; it cannot establish a dose to prevent cardiovascular disease or prove that every tea category has the same association.

八Blood pressure: small pooled changes, considerable disagreement#

Systematic reviews of randomized trials. A 2025 synthesis of 36 green-tea trials estimated changes of −1.08 mmHg systolic (95% CI −1.98 to −0.18) and −1.09 mmHg diastolic (−1.67 to −0.50). Between-study heterogeneity was high, and there was no clear relation between response and dose or duration. These are modest average changes across interventions, not a predicted response to a specific tea.

A separate 2020 review of green and black tea found no statistically clear blood-pressure or lipid effect in 14 trials with 798 participants lasting 4–24 weeks. Its authors rated the evidence low to moderate and emphasized limited statistical power. A null estimate is not proof of precisely zero effect, just as a small positive estimate is not proof of disease prevention.

Differences in participants, initial blood pressure, formulations and study selection help explain conflicting reviews. A trial that measures blood pressure over weeks does not establish fewer strokes over years. Tea should not be presented as a replacement for prescribed blood-pressure treatment.

九Cholesterol: put the size and preparation beside the result#

Systematic review of randomized trials. A 2020 analysis pooling 31 trials and 3,321 participants estimated a reduction in LDL cholesterol of 4.55 mg/dL (95% CI 2.80–6.31 mg/dL lower), with total cholesterol also modestly lower. Preparations included green-tea interventions rather than one standardized Chinese loose-leaf recipe. This does not establish the same response for oolong, white tea or pu'er.

A 2023 synthesis of 55 trials also found modest changes in several cardiovascular risk markers, including lipids. Its supplement-inclusive evidence cannot be used to promote concentrated products simply because the pooled average is favourable. Safety, magnitude, consistency and relevance of the tested preparation remain separate questions.

The reasonable reading is that some green-tea preparations may modestly affect lipids in some populations. There is no validated way to choose a cake, mountain or roast level to achieve a particular LDL change. Neither the biomarker evidence nor cohort associations justify stopping established treatment.

十Glucose and diabetes: a small average is not a treatment#

Systematic review of randomized trials. In a 2020 synthesis of 27 trials involving 2,194 people, green-tea interventions reduced fasting glucose by an average 1.44 mg/dL (95% CI 0.62–2.26 lower). Fasting insulin and HbA1c did not show clear changes. The fasting-glucose difference is small, and a short trial cannot establish prevention of diabetes or its complications.

A later 2023 synthesis reported a modest HbA1c change as well, illustrating how inclusion criteria and additional studies can shift pooled estimates. These reviews overlap in their underlying literature; counting them as independent confirmations would exaggerate the evidence.

For an individual tea drinker, formulation, baseline glucose, medication use and trial duration matter more than a headline saying 'tea lowers sugar'. Unsweetened tea and a sweetened tea beverage are different interventions. Research on a plain infusion should never be transferred automatically to bottled sweet tea or milk tea.

十一A Taiwan oolong trial: interesting, but not enough to prescribe#

Human randomized trial. In Miaoli, Taiwan, 20 adults with type 2 diabetes who continued prescribed medicines consumed 1,500 mL of oolong tea daily or water for 30 days each in a crossover study published in 2003. Glucose fell during the tea phase; its reported mean changed from 229 to 162.2 mg/dL, while the water phase did not show a comparable improvement.

Those are within-phase values, not a clean estimate of the difference caused by tea. The trial was small, short and conducted in medicated patients, with a substantial beverage volume. It does not show that 1,500 mL is an appropriate dose for everyone, establish long-term HbA1c or complication outcomes, or justify changing medication.

The study deserves a place in Chinese-tea science because it tested an actual oolong intervention in Taiwan. It also deserves restraint: a striking early result needs larger, well-controlled replication, chemical characterization and careful assessment of crossover effects before becoming practical medical advice.

十二Weight and metabolism: do not sell tea as a slimming intervention#

Systematic review of randomized trials. Cochrane's review of green-tea preparations included 18 studies with 1,945 completers, generally lasting 12–13 weeks. The weight changes were small, statistically unclear overall and unlikely to be clinically important; weight maintenance was not convincingly improved. The reviewed preparations often contained concentrated ingredients rather than ordinary brewed tea.

A transient change in energy expenditure, fat oxidation or an enzyme assay does not establish meaningful sustained weight loss. Even when a later meta-analysis finds a small mean difference, it must be assessed against formulation, adherence, baseline weight and study bias. Results for green-tea products cannot be used to claim that ripe pu'er 'burns fat' after a meal.

Choosing plain tea instead of a sugary drink changes what someone consumes, but that substitution is a different question from claiming a pharmacological slimming effect. A tea's value can be its taste and place in a routine without requiring a weight-loss promise.

十三Cancer and longevity: the strongest claims need the strongest evidence#

Systematic review of mixed human designs. A 2020 Cochrane review covered 142 studies and more than 1.1 million participants. Findings differed by cancer site and study design, and did not establish green tea as a cancer-prevention intervention. A large cumulative sample does not overcome inconsistent exposure measurement, confounding or small randomized trials.

Laboratory inhibition of cancer cells is not evidence that drinking tea treats an existing cancer. Nor do animal lifespan findings or human mortality associations establish an anti-ageing prescription. The relevant clinical outcomes require long follow-up, suitable comparators and careful attention to competing explanations.

Terms such as 'detox', 'anti-cancer' and 'rejuvenating' should therefore not appear as established properties of a tea entry. A guide may describe the research question and its uncertainty; a product card should not turn it into a therapeutic promise.

十四Oral health: a mouthwash trial is not a tea-drinking trial#

Systematic review of randomized trials. A 2021 review included 18 studies with 870 participants. Some plaque and periodontal measures improved with green-tea formulations, but several pooled estimates were imprecise and the authors found insufficient evidence to recommend them as first-choice treatment for gingivitis, periodontitis or caries.

Delivery matters here. A rinse, gel or dentifrice can place a particular concentration directly against oral tissue for a set time. That is not equivalent to swallowing an infusion. Reduced bacterial counts or plaque scores are also not identical to fewer cavities or teeth retained over years.

This is an active research area rather than a reason to substitute tea for established oral hygiene or dental care. The guide should distinguish topical preparations, ordinary unsweetened tea and sweetened beverages, and should avoid presenting antibacterial activity in a dish as a demonstrated dental benefit from drinking.

十五The microbiome and fermented teas: a mechanism still being tested#

Mixed evidence. A 2019 review found six human trials alongside 18 mechanistic studies of tea and the gut microbiome. Such an evidence base can identify hypotheses, but is too heterogeneous to establish that pu'er or Fuzhuan produces a defined clinical benefit through a particular bacterial change.

A 2023 study of 1,179 men and 1,078 women in Shanghai examined tea habits and stool microbiota. It adds valuable Chinese population evidence, but remained observational. Differences in microbial composition do not themselves prove improvement in health, and diet, medicines and other exposures can also affect the microbiome.

Fermentation during tea manufacture and a beneficial effect on the drinker's intestinal ecosystem are separate claims. 'Fermented' does not automatically mean a beverage is a validated probiotic. Naming organisms in stored tea is not enough: a probiotic claim would need evidence about the defined organism, viability, dose and benefit under the actual conditions of use. See Ageing and storage for what ageing changes in the tea itself.

十六Iron absorption: timing matters, especially when iron is low#

Human controlled meal study. Tea can reduce absorption of nonheme iron, the form found in many plant foods and fortified foods. A 2017 nonrandomized stable-isotope study in 12 iron-replete, nonanemic UK women tested 200 mL tea with an iron-containing porridge meal or one hour later. Simultaneous tea inhibited absorption; the one-hour delay attenuated that effect.

This is evidence about absorption from a test meal, not proof that ordinary tea drinking inevitably causes anemia. It also does not establish that exactly one hour solves the problem for every meal or every person. Overall iron intake, iron status and dietary context matter.

For someone with low iron or prescribed iron supplements, keeping tea away from the dose and discussing timing with a clinician or pharmacist is sensible. The point is not to compensate with a supposedly 'gentler' tea category: a light liquor is not a validated measure of its effect on iron absorption. Pregnancy is another context in which individual nutrition advice matters.

十七Pregnancy and caffeine: count all sources, not just cups#

Official guidance. EFSA identifies up to 200 mg of caffeine per day from all sources as an intake that does not raise safety concerns for the fetus. This is guidance for total exposure, not a target to reach or a guarantee for every pregnancy. Hong Kong's Centre for Food Safety likewise advises pregnant and lactating women to avoid excessive caffeine and remember that milk tea contributes.

Loose-leaf tea does not come with a reliable cup-to-milligram conversion. A session with several infusions, coffee later in the day and a supplement can contribute to the same total. Follow individualized pregnancy advice, especially when caffeine content is uncertain.

For healthy nonpregnant adults, EFSA's reference values are up to 200 mg in a single dose and 400 mg across a day. They are not recommended consumption amounts, and a person can experience sleep disturbance or other unwanted effects below them. Adult figures should not be repurposed as advice for children.

十八Brewed tea and concentrated extracts have different safety profiles#

Official safety assessment. EFSA's 2018 review considered traditional green-tea infusions generally safe at customary exposure, while recognizing rare, unpredictable reports of liver injury. For supplements, daily doses at or above 800 mg EGCG were associated with elevated liver enzymes in intervention studies. The panel could not identify a safe supplemental dose from the available evidence.

The 800 mg figure is therefore not a boundary below which every extract is safe. Nor is it a recommended dose or a reason to convert capsules into 'equivalent cups'. Extracts differ in concentration, composition and dosing pattern; a single bolus can differ from drinking tea across a day.

Health marketing often treats extracts as a more efficient way to obtain tea's benefits. That skips the need to demonstrate benefit and safety for that exact preparation. Ordinary enjoyment of tea supplies no reason to add an EGCG capsule. If liver-related symptoms or other concerning reactions develop while using a supplement, seek medical advice rather than experimenting with a different dose.

十九Medicines: one demonstrated interaction is enough to avoid blanket claims#

Human randomized pharmacokinetic trial. In a 2020 crossover study, 11 healthy volunteers took nadolol with water, with 150 mL of brewed green tea, or one hour after tea. Simultaneous tea markedly lowered drug exposure; the geometric-mean AUC ratio was 0.371 (90% CI 0.303–0.439). A one-hour interval did not remove the effect. The tested tea contained about 166 mg EGCG, so this is a characterized preparation rather than a universal estimate for any cup.

The finding is drug-specific and does not mean tea interacts with every medicine in the same way. It does show that brewed tea, not only capsules, can matter. NCCIH also highlights reduced atorvastatin exposure with green-tea extract.

A person taking medicines should mention substantial tea consumption and any extracts to a pharmacist or prescriber. Do not change the medicine or invent a timing rule from an unrelated study: the one-hour interval examined for meal iron cannot be transferred to nadolol.

二十What remains uncertain—and how to recognize a better study#

For Chinese loose-leaf tea, major gaps remain in long-term randomized research on ordinary drinking, especially for white tea, yellow tea, distinct oolong styles, raw versus ripe pu'er, and other heicha. The absence of robust trials is not evidence that these teas are harmful or inert; it limits claims about specific health outcomes.

A useful new study would report the actual leaf, preparation, caffeine and relevant polyphenol content; distinguish beverage from extract; specify participants and medicines; preregister outcomes; use an appropriate comparison; measure adherence and harms; and last long enough for the outcome in question. Independent replication matters, particularly where commercial funding is involved.

For observational research, repeated exposure measurements and careful handling of smoking, alcohol, diet, socioeconomic differences and illness-related habit changes would strengthen interpretation. For laboratory work, matching doses to achievable human exposure and testing metabolites would improve relevance.

No current evidence summarized here supports assigning a health score to a tea, prescribing a universal number of gongfu cups, or ranking categories as treatments. A more defensible practice is to choose tea for taste, record preparation when dose matters, protect sleep, attend to iron and medication context, and keep the distinction between an interesting mechanism and a demonstrated human outcome.