Supplements · Metabolic & Cardiometabolic
green tea decreases blood pressure
In plain terms: Does green tea lower blood pressure?
Strong support Supplements 💰 Industry COI noted🔬 Includes disconfirming
Part of: 🧪 Green tea
RefutedContestedStrong support
consensus score 0.90
A little. Six meta-analyses agree it nudges blood pressure down — but only by about 1–3 points, which is real yet small. A nice habit, not a treatment for high blood pressure.
📅 Last reviewed: 2026-07-15 ⓘ
Evidence ladder
How far up the ladder this claim has climbed. A high consensus on a low rung means "consistent so far," not "proven in people."
Top evidence so far: All trials, pooled (Meta-analysis)
MechanismIn-vitroAnimalObservationalRCTMeta-analysis
How the studies fall
11 support 1 contradict 0 tested null 0 mixed · 12 sources, 12 independent groups
What the evidence shows
Six independent meta-analyses agree green tea lowers blood pressure, but the effect is SMALL and consistent — roughly 1–3 mmHg systolic — with several flagging low-to-moderate evidence quality and high heterogeneity. Real, but not a substitute for antihypertensive treatment.
The evidence (14)
| Source | Grade | Stance | Quality | Finding |
|---|---|---|---|---|
| Taati B et al 2026 · Nutrients | observational | mixed | low | Narrative evidence-based review, explicitly not systematic: 'no predefined PRISMA framework or statistical pooling procedures were applied'. Headline is 31 clinical studies of plant supplements + exercise for BP, but GREEN TEA rests on a 6-study subset (1 acute, 1 short-term 3-week, 4 long-term 6-9 week trials), all of GT combined with exercise. Verdict on that subset is genuinely two-way: acute GT (~636 mg EGCG, ~20 mg caffeine) blunted systolic post-exercise hypotension and induced a diastolic rise; in overweight/obese and T2DM cohorts 'AT alone reduced BP, and GT provided no additional benefit when combined with training'; one trial - by this review's own first author - found GT + low-intensity resistance training better than either alone. Authors concede findings are 'inconsistent and often confounded by the strong BP-lowering effects of exercise itself'. No pooled estimate, no CI, no bias assessment, no independent data: NOT COUNTED (evidence_role: review). |
| Khalesi S, Sun J, Buys N, Jamshidi A, Nikbakht-Nasrabadi E, Khosravi-Boroujeni H 2014 · Eur J Nutr | meta-analysis | supports | moderate | Systematic review and meta-analysis of RCTs (ProQuest/PubMed/Scopus/CENTRAL, articles 1995-2013, random-effects): 'Thirteen studies were included in the meta-analysis.' 'Green tea consumption significantly changed systolic blood pressure, by -2.08 mm Hg (95% CI -3.06, -1.05), and diastolic blood pressure, by -1.71 mm Hg (95% CI -2.86, -0.56), compared to the control.' Green-tea-only exposure - no black-tea arm is pooled in. DENOMINATOR: 13 is the review-wide inclusion count, not the BP pools' count - 'Changes in lipid profile, blood glucose and body mass index were also assessed in the meta-analysis', so the 13 span five outcome families; the per-outcome trial count for the SBP and DBP pools is not stated and the total participant n is never reported in the abstract, so both are cannot-tell from available text. CONDITIONS THE AUTHORS ATTACH: 'Subgroup analysis suggested a greater reduction in both systolic and diastolic blood pressure in studies that included participants with a baseline mean systolic blood pressure of ≥ 130 mm Hg, and studies involving consuming green tea as an extract' - so the headline is a whole-pool average and the larger effect sits in the elevated-BP and extract subgroups. QUALITY: no heterogeneity statistic, publication-bias test or risk-of-bias/GRADE rating appears in the abstract and the full text is subscription-only (Europe PMC: no PMCID, isOpenAccess N), so this pool's reliability is cannot-tell rather than established - moderate, not high. SAME-PAPER FINDINGS NOT HELD HERE: significant reductions in total cholesterol (-0.15 mmol/L [95% CI -0.27, -0.02]) and LDL cholesterol (-0.16 mmol/L [95% CI -0.22, -0.09]); 'Changes in other parameters did not reach statistical significance' (blood glucose, BMI). |
| Peng X, Zhou R, Wang B, Yu X, Yang X, Liu K, Mi M 2014 · Sci Rep | meta-analysis | supports | moderate | Full-text re-read 2026-08-25 (PMC4150247 JATS XML in the local fulltext-cache, 122,393 bytes - no network fetch; the source note said provenance: abstract and tags: abstract-only, both now wrong). Peng X, Zhou R, Wang B, Yu X, Yang X, Liu K, Mi M, Sci Rep 2014;4:6251, doi 10.1038/srep06251. DESIGN AND SCOPE: meta-analysis of 13 RCTs of green tea beverage or extract vs a concurrent control, PubMed/Embase/Cochrane Library to March 2014, PRISMA-based; green tea only, no black-tea arm is pooled in. Blood pressure is this review's ONLY outcome family - unlike siblings gt-khalesi-2014 and gt-onakpoya-2014, whose inclusion counts span lipids, glucose and BMI - so the 13 trials ARE the BP pools' own denominator: 'The mean difference in DBP was reported in 13 trials'. SBP pooled fixed-effects, DBP random-effects; both models agreed. HEADLINE: SBP -1.98 mmHg (95% CI -2.94 to -1.01, P<0.001); DBP -1.92 mmHg (95% CI -3.17 to -0.68, P=0.002). PARTICIPANT DENOMINATOR - CORRECTING THE PRIOR EXTRACT: the stored n=1,367 is the abstract's figure and the paper's own study table does not reconcile it. 'Thirteen trials comprising a total of 1,367 subjects were included in the current meta-analysis' appears once, in the abstract, and nowhere in the Results, Methods, tables or figures; Table 1's 13 per-study subject counts (66, 46, 240, 38, 70, 43, 33, 88, 111, 22, 66, 46, 56) sum to 925 (my arithmetic), consistent with the text's own 'The total number of subjects included in each study ranged from 22 to 240 subjects.' Even double-counting both crossover trials for two periods reaches only 1,061. Which figure is right is cannot-tell from the paper; the trial denominator (13) is sound, the participant one is not, and 1,367 should not be quoted as this pool's n. CONDITIONS THE AUTHORS ATTACH - part of the finding, not a caveat: 'Significant reductions in SBP and DBP were only observed in the low-dose green tea polyphenols, the long-term duration, parallel-design RCTs, high-quality trials subgroups or when the confounding effect of caffeine was removed.' The strata that were NULL for SBP, with their own trial counts (participant n per stratum is not reported - cannot-tell): caffeinated green tea, 4 trials, -1.55 (-5.84, 2.75); high dose >=582.8 mg/d, 7 trials, -1.08 (-2.44, 0.29); <12 wk, 5 trials, -0.44 (-1.97, 1.09); crossover design, 2 trials, +0.31 (-1.54, 2.16); low Jadad (<4), 8 trials, -1.76 (-3.92, 0.40). The whole-pool -1.98 is carried by the decaffeinated, low-dose, >=12-week, parallel, high-Jadad strata. No dose-response: 'The results of meta-regression analysis did not observe a significant dose-responsive effect between green tea and SBP' (P for trend 0.14; DBP 0.16). QUALITY high -> MODERATE, earned both ways. FOR: publication bias was TESTED and null - 'Egger's tests showed no significant publication bias in the current meta-analyses of SBP and DBP (Egger's test: P = 0.62 and 0.81)' - and the pool is stable: 'systematically removing each trial during the sensitivity analyses did not significantly change the overall observed effects of green tea on BP.' AGAINST, from the authors' own limitations: 'First, of the 13 studies, 5 studies were high-quality RCTs, whereas the remaining 8 studies were of low quality.'; 'Second, BP was not the primary outcome in part of the RCTs selected in this meta-analysis and the null findings of secondary outcomes may not be published' - selective outcome reporting, which an Egger test on 13 studies has no power to detect; 'most of the selected studies (11 of 13) did not report the information on sample size calculation'; and the DBP pool needed a random-effects model. CALIBRATION: this was the ONLY one of thirteen counting rows on this claim at high (x1.5). gt-liu-bp-2014 carries null Egger tests (P=0.947/0.653) AND stable sensitivity analyses and is held moderate because its authors report against themselves; gt-ayaz-2023 reports I2=0% with an explicitly tested null for publication bias and is moderate. Peng's self-reported bias signal is stronger than either, so high was out of calibration with its own siblings. STANCE UNCHANGED (supports): both overall pools are significant in the claim's direction; the null strata are conditions on the same object, not a finding pointing the other way. INDEPENDENCE (flag, not fixed): Chengdu Medical College + Research Center for Nutrition and Food Safety, Third Military Medical University, Chongqing; funded by the '12th Five year Plan' National Key Technology Research and Development Program, grant 2012BAI35B02; no competing-interests statement appears in the JATS (cannot-tell, not a declaration of none). Corresponding authors Liu K and Mi M also lead the vault's gt-liu-glu-2013 (PMID 23803878, group_id liu-glu-2013), which this paper cites as 'our previous meta-analysis' - one team holding two group_ids, on different claims, so no double count here. NEAR-TWIN: gt-khalesi-2014 is also a 13-study green-tea BP meta over the same 1995-2014 literature with near-identical estimates (-2.08/-1.71 vs -1.98/-1.92); the two are almost certainly two passes over one set of primaries and should not be read as two independent observations. Twelve of this claim's thirteen counting rows are meta-analyses over that same shared literature. |
| Li G, Zhang Y, Thabane L, Mbuagbaw L, Liu A, Levine MA, Holbrook A 2015 · J Hypertens | meta-analysis | supports | low | Systematic review and meta-analysis of parallel and cross-over RCTs of green tea or green tea extract (GTE) versus placebo, random-effects on change-from-baseline, restricted to overweight and obese adults: 'Fourteen RCTs with 971 participants (47% women) were pooled for analysis.' 'Green tea or GTE produced a significant effect on both SBP (mean difference -1.42 mmHg, 95% confidence interval -2.47 to -0.36, P = 0.008; I = 52%, P = 0.01 for heterogeneity) and DBP (mean difference -1.25 mmHg, 95% confidence interval -2.32 to -0.19, P = 0.02; I = 74%, P < 0.001 for heterogeneity), compared with placebo.' POPULATION: overweight/obese only - not a general-adult estimate. DENOMINATOR: 14 RCTs / 971 participants is the review-wide pooled count; the SBP and DBP pools have visibly different heterogeneity (52% vs 74%) yet no per-outcome trial count or participant n is reported, so each pool's own denominator is cannot-tell from available text. EXPOSURE: leaf/beverage and extract are pooled together as one green-tea exposure and no separate estimate for either is given in the abstract - unlike the tea-family metas, no black tea is pooled in. QUALITY: the authors' own verdict is that 'The quality of evidence across studies was low', and DBP heterogeneity is high (I2=74%, P<0.001); against that, 'Similar results were found in subgroup and sensitivity analyses' - stable, unlike the Cochrane green-tea pool. No publication-bias test appears in the abstract and the full text is not in the PMC open-access subset, so bias is cannot-tell rather than absent. THE AUTHORS' OWN FRAMING: 'a small but significant reduction in BP', with 'more high-quality RCTs with large sample sizes' needed before strong recommendations. |
| Hartley L et al. 2013 · Cochrane Database Syst Rev | meta-analysis | supports | moderate | Cochrane review (searches to 12 Oct 2012, RCTs >=3 months, primary prevention) of 11 RCTs and 821 participants in total, of which 7 examined green tea and 4 black tea: 'Trials of green tea were analysed separately from trials of black tea', so this is a green-tea-only estimate. Green tea: SBP MD -3.18 mmHg (95% CI -5.25 to -1.11), DBP MD -3.42 (95% CI -4.54 to -2.30). DENOMINATOR: those figures do NOT come from all 7 green-tea trials - the same sentence says 'only a small number of studies contributed to each analysis', and the review pooled four separate green-tea outcomes; the per-outcome trial count and participant n for the BP pools are not reported in the abstract and the full text is not in the PMC open-access subset, so the BP subset's own denominator is cannot-tell. QUALITY: the authors report their green-tea results were 'not stable to sensitivity analysis' - explicitly unlike the black-tea BP result, which was stable - and conclude 'the results should be treated with some caution'. CONTEXT: 'No studies reported cardiovascular events', so BP here is a risk-factor surrogate in a CVD-prevention review, not an outcome. |
| Xu R, Yang K, Ding J, Chen G 2020 · Medicine (Baltimore) | meta-analysis | supports | moderate | Full-text re-read 2026-08-25 (PMC7015560 JATS XML in the local fulltext-cache, 97,186 bytes; no network fetch - the source note said provenance: abstract and tags: abstract-only, both now wrong). Xu R, Yang K, Ding J, Chen G, Medicine (Baltimore) 2020;99(6):e19047. DESIGN: PRISMA systematic review + random-effects meta-analysis of placebo-controlled RCTs of green tea beverage or extract for >=2 weeks in adults (PubMed/Embase/Cochrane to Aug 2019, English only); multicomponent supplements excluded and green tea required to be the only between-arm difference, so this is a green-tea-only exposure, not a pooled one. DENOMINATOR - and this one is honest: the pooled BP estimate is over '25 comparisons from 24 studies including 859 subjects in the green tea group and 838 subjects in the placebo group', so n=1,697 is the BP pool's own denominator and the whole sample, not a borrowed one. RESULT: SBP MD -1.17 mmHg (95% CI -2.18 to -0.16, P=.02); DBP MD -1.24 mmHg (95% CI -2.07 to -0.40, P=.004) - the SBP interval stops 0.16 mmHg short of null. HETEROGENEITY: I2=43% (P=.01) SBP, I2=57% (P=.0002) DBP, and 'substantial heterogeneity persisted in most subgroup analyses'. QUALITY, both directions: the authors tested for and found no publication bias (funnel plots symmetrical; Egger P=.674 SBP, P=.270 DBP), but their own Jadad stratification splits the result - 'a significant reduction in SBP was found in low Jadad scores and a significant reduction in DBP was found in high Jadad scores', so NEITHER outcome survives in both quality strata - and they report 'Of the 24 studies, 11 were of high quality and the remaining 13 were of low quality' while their Results section says twelve and twelve. Trials ran 3-16 weeks only. NO DOSE GRADIENT, running backwards: 'the BP reduction effect was greater in the low catechin dosage group than in the high catechin dosage group' (median split at 615 mg/d catechins; per-stratum trial counts are in Table 3, which is a caption-only stub in this XML), and 'our meta-analysis did not identify an optimal dosage'. Caffeine did not modify the effect (9 decaffeinated vs 12 caffeinated comparisons). Larger effect in the 15 high-normal/hypertensive comparisons than in the 10 normotensive ones, with no interaction test reported. INDEPENDENCE FLAG: a ninth pass over largely the same primary RCTs as the other meta-analysis rows on this claim; Xu cites Peng 2014 and Liu 2014 as priors and reports being 'generally consistent' with them. |
| Ghoflchi S et al. 2025 · Clin Ther | meta-analysis | contradicts | low | Meta-analysis of RCTs in metabolic syndrome: no effect on SBP (SMD -0.42, P=.36) or DBP (SMD -0.24, P=.53) - a null on a directional claim. The only BP benefit is a women-only subgroup (SBP SMD -1.74) whose trial count and n the abstract does not report; that magnitude is ~1.7 pooled SDs against a whole-literature estimate of ~1-3 mmHg. Authors state the results carry 'overall high heterogeneity and publication bias'; k, total n and every confidence interval are absent from the abstract. |
| Liu G, Mi XN, Zheng XX, Xu YL, Lu J, Huang XH 2014 · Br J Nutr | meta-analysis | supports | moderate | Meta-analysis of 25 RCTs (1,476 subjects; MEDLINE/EMBASE/CENTRAL to May 2014) of BLACK AND GREEN tea, split 6 acute trials and 21 long-term trials (1,323 subjects). Green tea is reported as its own stratum of the long-term pool, so this is a green-tea estimate and not a pooled-exposure one: 'when stratified by type of tea, green tea exhibited a significant reduction in systolic BP of 2.1 (95 % CI -2.9, -1.2) mmHg (I2 = 21.8 %) and a decrease in diastolic BP of 1.7 (95 % CI -2.9, -0.5) mmHg (I2 = 59.9 %)' - Fig. 3 subtotals -2.05 (-2.94, -1.15) and -1.69 (-2.88, -0.49), which the abstract rounds. DENOMINATOR: those figures are NOT from 25 trials. The green-tea stratum is 14 of the 25 trials; participants in those 14 sum to 1,031 from Table 1 (my arithmetic - the paper gives no stratum n; it reconciles, 1,031 green + 292 black = the 1,323 the authors report for the long-term pool). The black-tea stratum (7 trials) is a separate subtotal and the 6 acute trials are excluded from this estimate entirely. DURATION - CORRECTING THE PRIOR EXTRACT: the green-tea numbers are NOT the >=12-week result. They come from all 14 long-term green-tea trials, which ran 3-24 weeks. The >=12-week figure is a different subgroup - 9 trials, black AND green tea pooled, SBP -2.57 (-3.48, -1.65), DBP -2.15 (-2.98, -1.32) - versus <12 weeks, 12 trials, SBP -0.81 (-1.79, 0.16), non-significant; duration was the only subgroup reaching a between-group difference (P=0.010 SBP). There is no green-tea-specific duration estimate. EXPOSURE: 9 of the 14 green-tea trials used green tea EXTRACT, 5 a beverage; in 4 the control arm was a low-polyphenol green tea matched for caffeine (high- vs low-polyphenol tea, not tea vs none) despite the text describing a 'tea-free control', and 2 were open-label with no-intervention controls. ALSO FOUND: 'The acute intake of tea had no effects on systolic and diastolic BP.' QUALITY - moderate, earned both ways. FOR: 'the funnel plots were symmetrical and Egger's tests indicated no significant publication bias (P=0.947 for systolic BP and P=0.653 for diastolic BP)'; 'The sensitivity analyses showed that the significance in the pooled changes in BP were not altered after the removal of the six trials with a cross-over design or the five trials with low quality.'; 18/25 trials Jadad >=3; no COI, public funding. AGAINST: the authors report against themselves - 'the lack of blinding of participants or investigators to the intervention in a number of studies increased the risk of expectation bias', only 8 of 25 trials reported allocation concealment, and green-tea DBP heterogeneity is I2=59.9%. INDEPENDENCE (flag, not fixed): Fuwai Hospital, Beijing - Liu G, Mi XN, Zheng XX, Xu YL, Lu J, Huang XH. Co-authors Zheng XX and Xu YL lead gt-zheng-2011 and gt-zheng-glu-2013, so one team holds three group_ids (different claims, so no double count here). This Liu G is NOT the Liu A of gt-li-bp-2015 (McMaster). The 14 green-tea primaries are the same literature pooled by gt-hartley-2013, gt-khalesi-2014, gt-onakpoya-2014, gt-peng-2014 and gt-li-bp-2015. |
| Mahdavi-Roshan M, Salari A, Ghorbani Z, Ashouri A 2020 · Complement Ther Med | meta-analysis | supports | moderate | Systematic review and meta-analysis of RCTs of green OR black tea AS A BEVERAGE (PubMed/Scopus/Web of Science/ProQuest, Feb 1995 to 20 Jul 2019), restricted to adults with elevated BP or hypertension: 'The search strategy generated 1119 trials, of which finally five trials fulfilled the criteria for being included in the current study.' THE HEADLINE NUMBER IS A POOLED-EXPOSURE ESTIMATE, NOT A GREEN-TEA ONE: 'According to nine measurements derived from 5 trials on 408 individuals, it was found that regular tea intake resulted in the reduction in SBP (weighted mean difference (WMD): -4.81 mmHg, 95 %CI: -8.40 to -1.58, P = .004) and DBP (WMD:-1.98 mmHg, 95 %CI: -3.77 to -0.20, P = .029)' - 'regular tea intake' there is green AND black tea combined, so -4.81/-1.98 is not this claim's exposure and must never be quoted as a green-tea effect size. It is also the largest estimate on this claim by 2-4x, against a literature of ~1-3 mmHg. WHY THIS ROW STILL COUNTS RATHER THAN GOING OFF-SCOPE: the paper does stratify by tea type - 'Categorized studies, according to the tea type, revealed that the hypotensive effects of green tea were more pronounced compared to black tea' - so an identical row would NOT vote the same way on a black-tea claim. But that subgroup is reported as a DIRECTION ONLY: no green-tea WMD, no green-tea trial count and no green-tea participant n appear in the abstract, and the full text is subscription-only (Europe PMC: no PMCID, isOpenAccess N, hasPDF N), so all three are cannot-tell from available text. This row supports the claim on that stratified sentence, not on the headline. DENOMINATOR: the pooled figures come from nine measurements pooled from only five trials totalling 408 participants - trials contribute more than one measurement each, and whether that within-trial correlation was handled is cannot-tell from the abstract. NOT STABLE: 'however, excluding the most heterogeneous trials showed that regular tea intake might reduce SBP and DBP by about -3.53 and -0.99 mmHg, respectively' - roughly 27% of the SBP effect and 50% of the DBP effect disappears on the authors' own sensitivity analysis. QUALITY: 'Three out of 5 articles showed a low risk of bias', i.e. two were not; k=5 is too few to test publication bias and none is reported, so bias is cannot-tell rather than absent. Held at moderate, matching gt-hartley-2013, which carries the same instability signal. ALSO FOUND (not held as claims here): meta-regression - 'the longer the duration of tea intake (>=3months), the higher the decrease in both SBP and DBP', concordant with the duration subgroup in gt-liu-bp-2014; and 'None of the studies reported any side effects', which is absence of reporting in five small trials, not demonstrated safety. INDEPENDENCE (flag, not fixed): Guilan University of Medical Sciences, Rasht (Mahdavi-Roshan M, Salari A, Ghorbani Z, Ashouri A); its five primaries are drawn from the same literature pooled by gt-liu-bp-2014, gt-hartley-2013, gt-khalesi-2014, gt-onakpoya-2014, gt-peng-2014, gt-xu-bp-2020, gt-yarmolinsky-2015 and gt-rezaei-2025 - twelve of this claim's thirteen counting rows are meta-analyses over one shared primary literature. |
| Yarmolinsky J, Gon G, Edwards P 2015 · Nutr Rev | meta-analysis | supports | moderate | Abstract re-read 2026-08-25 (NCBI efetch, 3,251 bytes; full text NOT obtained - Europe PMC reports no PMCID, isOpenAccess N, and the Unpaywall-listed OUP free PDF returned a Cloudflare interstitial). Yarmolinsky J, Gon G, Edwards P, Nutr Rev 2015;73(4):236-46, doi 10.1093/nutrit/nuv001. OFF-SCOPE ON THE SUBJECT (#539 pooled exposure). The eligibility criterion is a pooled tea exposure: 'The selection criteria included randomized controlled trials of adults whose blood pressure was within hypertensive or prehypertensive ranges and in which the applied intervention was green or black tea'. The result is reported for that pool and nothing else: 'Meta-analyses of 10 trials (834 participants) showed statistically significant reductions in systolic blood pressure (MD -2.36 mmHg, 95%CI -4.20 to -0.52) and diastolic blood pressure (MD -1.77 mmHg, 95%CI -3.03 to -0.52) with tea consumption.' So does the conclusion: 'Consumption of green or black tea can reduce blood pressure in individuals within prehypertensive and hypertensive ranges'. THE TEST FIRES: this row, unchanged, would vote identically on a black-tea-lowers-blood-pressure claim, so it measured neither member of the pool. HOW THIS DIFFERS FROM ev-gt-mahdavi-2020, the other pooled row on this claim: mahdavi survived the same test at reread #657 because its abstract stratifies by tea type ('the hypotensive effects of green tea were more pronounced compared to black tea'), direction-only but green-specific. This abstract contains no counterpart - no tea-type subgroup, no green-tea estimate, no green-tea trial count, no green-tea direction. Followup 168 recorded these two as the same shape pending this read; they are not, and the convention question it raises (does a direction-only tea-type subgroup earn full meta weight?) now applies to mahdavi alone. CANNOT-TELL: whether the paywalled full text carries a tea-type subgroup is unknown from available text - absence in an abstract is not a finding about the paper, and if the full text is obtained and does stratify, this ruling should be revisited. DENOMINATOR: 10 trials / 834 participants is the green-AND-black denominator; the green-tea trial count inside it is not reported, so -2.36/-1.77 mmHg must never be quoted as a green-tea effect size. SEARCH DATE: 'The CENTRAL, PubMed, Embase, and Web of Science databases were searched for all relevant studies published from 1946 to September 27, 2013' - the earliest search on this claim, so its trials sit inside the literature later re-pooled by gt-onakpoya-2014, gt-liu-bp-2014, gt-khalesi-2014, gt-peng-2014, gt-li-bp-2015, gt-xu-bp-2020, gt-mahdavi-2020 and gt-rezaei-2025. QUALITY held at moderate (unchanged): the authors qualify their own result - 'further investigation with studies of longer duration and stronger methodological quality is warranted to confirm these findings' - and no publication-bias test is reported in the abstract, so bias is cannot-tell rather than absent; this matches how gt-hartley-2013 and gt-mahdavi-2020 are held. INDEPENDENCE (flag, not fixed): London School of Hygiene and Tropical Medicine, a genuinely distinct author team from every sibling, but data-independent from none of them - followup 167's shape, twelve meta-analysis rows over one shared primary literature carrying ~98% of this claim's weight. ALSO NOTED: despite the title's 'secondary prevention of cardiovascular disease' framing, the pooled endpoint is blood pressure only; no cardiovascular event outcome appears in the abstract. |
| Onakpoya I et al. 2014 · Nutr Metab Cardiovasc Dis | meta-analysis | supports | moderate | Systematic review and meta-analysis of RCTs of green tea (Camellia sinensis) on blood pressure AND lipid parameters (Medline, Embase, Amed, Cinahl, Cochrane Library plus hand searches of bibliographies; 'The reporting quality of included studies was assessed using a checklist adapted from the CONSORT Statement.'): 'As many as 474 citations were identified and 20 RCTs comprising 1536 participants were included.' 'A meta-analysis revealed a significant reduction in systolic blood pressure favouring green tea (MD: -1.94 mmHg; 95% CI: -2.95 to -0.93; I(2) = 8%; p = 0.0002).' EXPOSURE: green tea only - no black-tea arm is pooled in, unlike gt-mahdavi-2020, gt-liu-bp-2014 and gt-yarmolinsky-2015; whether beverage and extract trials are pooled together is not stated in the abstract. DENOMINATOR: 20 RCTs / 1536 participants is the review-wide inclusion count, not the SBP pool's. The same 20 trials span blood pressure and the lipid outcomes ('Similar results were also observed for total cholesterol ... and LDL cholesterol'), and no per-outcome trial count or participant n is reported, so the SBP pool's own denominator is cannot-tell from available text. CORRECTING THE PRIOR EXTRACT: 'no significant DBP' overstates what this abstract says. No diastolic estimate, confidence interval or p-value appears anywhere in it; DBP is merely absent from the conclusion's list - 'Green tea intake results in significant reductions in systolic blood pressure, total cholesterol, and LDL cholesterol.' Whether DBP was pooled and null, or not pooled at all, is cannot-tell; the abstract asserts no null. The row therefore supports this claim on the systolic result alone. THE AUTHORS' OWN FRAMING: 'The effect size on systolic blood pressure is small, but the effects on total and LDL cholesterol appear moderate', and 'Longer-term independent clinical trials evaluating the effects of green tea are warranted.' ADVERSE EVENTS - the authors' own counter-signal: 'Adverse events included rash, elevated blood pressure, and abdominal discomfort', i.e. raised blood pressure appears among the harms of the same review whose headline is a blood-pressure reduction; no rates are given. QUALITY - moderate, not high. FOR: SBP heterogeneity is low (I(2) = 8%), the interval excludes null, and eligibility, quality assessment and extraction were done in duplicate ('Two reviewers independently determined eligibility, assessed the reporting quality of the included studies, and extracted the data.'). AGAINST: 'There were variations in the designs of the RCTs'; the CONSORT-adapted reporting-quality assessment is described but its RESULT is never reported in the abstract; and no publication-bias test, funnel plot or GRADE rating is mentioned, with the full text outside the PMC open-access subset (BioC pmcoa returns no result), so bias is cannot-tell rather than absent. The prior 'high' read a missing bias test as a clean one - the absence-settles-nothing error. Held level with gt-ayaz-2023, which reports I(2)=0% and an explicitly tested null for publication bias and sits at moderate. SAME-PAPER FINDINGS NOT HELD HERE: total cholesterol MD -0.13 mmol/l (95% CI -0.2 to -0.07; I(2) = 8%; p < 0.0001) and LDL cholesterol MD -0.19 mmol/l (95% CI -0.3 to -0.09; I(2) = 70%; p = 0.0004) - the vault holds claim-green-tea-lowers-ldl-cholesterol and carries no row from this source. INDEPENDENCE (flag, not fixed): Oxford CEBM primary-care group - Onakpoya I, Spencer E, Heneghan C, Thompson M. The vault holds two further Onakpoya sources (s22844674, stv-onakpoya-2015) under a different group_id, 'onakpoya', but on other claims, so no double count here. Its 20 primaries are the same 1995-2014 literature pooled by gt-hartley-2013, gt-khalesi-2014, gt-peng-2014, gt-liu-bp-2014, gt-li-bp-2015, gt-xu-bp-2020, gt-yarmolinsky-2015 and gt-rezaei-2025 - twelve of this claim's thirteen counting rows are meta-analyses over one shared primary literature. |
| Hoseini E, Razavi BM, Alavi MS, Hosseinzadeh H, Roohbakhsh A 2025 · Avicenna J Phytomed | animal | supports | low | Full-text re-read 2026-08-25 (PMC12441192 JATS XML in the local fulltext-cache, 104,503 bytes; no network fetch — the source note said provenance: abstract). Controlled rat experiment, 14 groups of n=5 (70 male Wistar rats, 180 g), isolated purified green tea catechins ECG or EGCG given INTRAPERITONEALLY once daily for 11 days; SBP by non-invasive tail-cuff at endpoint. Olanzapine 5 mg/kg raised SBP: 'Mean SBP was significantly higher in rats receiving 5 mg of olanzapine than in the control group'. Against the olanzapine group, EGCG worked at every dose ('all doses of EGCG reduced SBP increased by olanzapine') but ECG at only two of three — 'A dose of 40 mg/kg of ECG decreased SBP. There was a significant decrease in mean SBP in rats treated with ECG (20 and 40 mg/kg) and olanzapine' — so the previously stored '10-40 mg/kg' range overstated ECG, which did not work at 10 mg/kg. The catechin-ALONE arms give a direct effect against normal controls at the top dose only: 'Moreover, administering 10 or 20 mg/kg of EGCG did not change the mean SBP. However, 40 mg/kg of EGCG reduced mean SBP compared to control animals'. Quality low on reliability, not size: n=5 per group, no blinding described, one-way ANOVA/Tukey across ~10 outcomes with no multiplicity handling, and the paper's own concluding paragraph credits 'EGC (20 mg)' — epigallocatechin, a compound not administered in any of the 14 groups. Authors' own limit: 'The justification, however, requires well-controlled clinical studies.' |
| Yildirim Ayaz E, Dincer B, Mesci B 2023 · Altern Ther Health Med | meta-analysis | supports | moderate | Meta 9 RCTs, n=680 healthy adults (345 vs 335, mean age ~36): SBP -2.99 mmHg (95% CI -3.77 to -2.22, P<.00001) / DBP -0.95 mmHg (95% CI -1.62 to -0.27, P=.006); I2=0% on both, authors report no publication bias. |
| Rezaei M et al. 2025 · Blood Press | meta-analysis | supports | moderate | Systematic review and dose-response meta-analysis of RCTs of green tea supplementation vs control or placebo in adults (PubMed, Scopus, Web of Science through January 2024; random-effects WMD; PICOS-defined eligibility): 'A total of 36 RCTs were included. Green tea reduced SBP (WMD: -1.08 mmHg; 95% CI: -1.98, -0.18; I2 = 85.0%) and DBP (WMD: -1.09 mmHg; 95% CI: -1.67, -0.50; I2 = 74.0%).' CORRECTING THE PRIOR EXTRACT: it carried neither confidence interval and attached I2=85% to both outcomes. 85.0% is the SBP figure; DBP is 74.0%. Both intervals exclude null, but this is the SMALLEST estimate on the claim and its SBP upper bound reaches -0.18, nearer null than any sibling's. DENOMINATOR: 36 is the review-wide inclusion count. No total participant n appears anywhere in the abstract, and no per-outcome trial count is reported for the SBP and DBP pools - the differing I2 values leave open that the two pools are not the same 36 - so both are cannot-tell from available text (Europe PMC: no PMCID, isOpenAccess N; publisher full text returns Cloudflare 403; BioC pmcoa: no result). EXPOSURE: green tea only - no black tea is pooled in, unlike gt-mahdavi-2020, gt-liu-bp-2014 and gt-yarmolinsky-2015; whether beverage and extract trials are pooled together is not stated (cannot-tell). CONDITIONS THE AUTHORS ATTACH - part of the finding, not a caveat: 'Subgroup analyses showed stronger effects in participants with elevated baseline BP (SBP >=120 mmHg), those consuming <500 mg/day green tea, studies <=8 weeks, Asian populations, and women.' The headline -1.08 is a general-adult average and the effect concentrates in those strata; not one of the five subgroups is given its own trial count, participant n, estimate or interval, and the complementary strata are not reported at all, so whether they go null is cannot-tell. The pattern also runs the wrong way for a dose-response - larger at LOWER dose and SHORTER duration - and the paper's distinguishing analysis returned null: 'No clear association was found between BP changes and tea dose or intervention duration.' QUALITY moderate, HELD. The title advertises a GRADE assessment but the abstract never reports the certainty rating, and no publication-bias test or funnel plot is mentioned, so both are cannot-tell rather than clean (absence settles nothing). Against that: this is the largest pool on the claim, prespecified by PICOS, random-effects, and the authors report against themselves - 'Given the heterogeneity and lack of clear dose-duration effects, green tea is better considered a complementary approach for managing blood pressure.' NOT lowered to low: gt-li-bp-2015 sits at low on its authors' own verdict that 'the quality of evidence across studies was low'; I2=85% alone is heterogeneity, not a reliability verdict, and this row is held level with gt-peng-2014, gt-liu-bp-2014 and gt-ayaz-2023 at moderate. STANCE UNCHANGED (supports): both pooled intervals exclude null in the claim's direction. INDEPENDENCE (flag, not fixed): Rezaei M, Akhavan N, Fathi F, Alavi SM, Fadaii M, Dehzad MJ, Askarpour M (Shahid Beheshti / UNLV / Mazandaran / Tehran / Shiraz / Kerman). Dehzad MJ and Askarpour M are the same pair behind the vault's s36804260 (Cytokine 2023 curcumin meta, same 'GRADE-assessed systematic review and dose-response meta-analysis' title template), stored truncated as '[Dehzad]' under group_id curcumin-inflam - a different claim, so no double count here, but an overlap the vault's stored author field cannot see. Its 36 primaries are the same 1995-2024 literature pooled by gt-hartley-2013, gt-khalesi-2014, gt-onakpoya-2014, gt-peng-2014, gt-liu-bp-2014, gt-li-bp-2015, gt-xu-bp-2020, gt-yarmolinsky-2015 and gt-mahdavi-2020 - twelve of this claim's thirteen counting rows are meta-analyses over one shared primary literature. |
Disagree, or know a study we missed?
We grade by evidence, not opinions. The way to weigh in is to point us to a study we haven't cited (check the evidence table above first), or to flag a problem with one we have. Every submission is reviewed; if it holds up, the grade updates and shows in Science Changes Its Mind.
Opens a short form. You'll sign in with Google so submissions are tied to a real account — we don't display your identity, and we only accept a link we can verify (PubMed, DOI, ClinicalTrials.gov).
Educational only, not medical advice. Grades and scores reflect published evidence weighted by study design and quality; see the methodology.