Does chronic polyphenol supplementation modify resistance training-induced strength and muscle morphology adaptations in adults? A systematic review and multilevel meta-analysis of randomized controlled trials
This systematic review with a three-level meta-analysis tested whether taking polyphenols chronically changes how adults adapt to resistance training in strength, in direct muscle morphology and in muscle quantity/lean mass.
Five databases were searched to May 21, 2026 for randomized controlled trials pairing chronic oral polyphenol or polyphenol-rich supplements with structured resistance training, against resistance training alone, placebo or a matched comparator. 11 reports were included, on 11 randomized trials forming 11 independent study clusters; 360 unique participants.
For strength the estimate was small and statistically uncertain (Hedges' g = 0.25, P = 0.060, 95% CI −0.01 to 0.51; low certainty), and likewise for muscle quantity/lean mass (g = 0.13, P = 0.211, 95% CI −0.07 to 0.33; low certainty); for direct morphology it was imprecise and near zero (g = 0.05, P = 0.718, 95% CI −0.21 to 0.30; very low certainty).
The authors conclude that the evidence has not established a reliable added effect, and that with low-to-very-low certainty no firm conclusion is possible about either benefit or lack of benefit.
Participants
11 reports, 11 randomized trials, 11 independent study clusters, 360 unique participants.
Who they were (full text, checked 2026-10-02: section 3.2, Table 1): untrained young men, healthy or sedentary older adults, sarcopenic older men, overweight or obese women and postmenopausal women, in Denmark, Japan, Spain, Iran, the United States and Brazil; analytic samples of 17 to 51 per trial; 154 men, 157 women, sex not reported for 49. No trial enrolled trained strength or endurance athletes.
Protocol
Search: PubMed, Embase, Web of Science, the Cochrane Library and Scopus, from inception to May 21, 2026; PRISMA.
Included: randomized controlled trials in adults pairing structured resistance training with chronic oral supplementation of polyphenols or polyphenol-rich products, versus resistance training alone, placebo or a matched comparator.
Appraisal: RoB 2, PEDro and GRADE.
Model: three-level random-effects models with cluster-robust variance estimation, in R 4.5.1.
Supplements pooled (full text, checked 2026-10-02: section 3.2, Table 2): red/blackcurrant polyphenols, quercetin glycosides, curcumin, matcha, tea catechins, epicatechin, green/black tea polyphenol blends, green-tea powder and soy isoflavones; training two or three times per week for 6 to 39 weeks; supplementation for at least 4 weeks was an inclusion criterion.
Co-exposures in some trials (full text, checked 2026-10-02: section 3.2): caffeine, essential amino acids, a hypocaloric dietary run-in and limited HIIT.
Key Results
Strength adaptations: Hedges' g = 0.25, P = 0.060 (95% CI −0.01 to 0.51; 95% PI −0.36 to 0.86); I² between = 34.5%, I² within = 0.0%; low certainty. Based on 257 unique participants, 8 clusters and 22 effect sizes.
Muscle quantity/lean mass: g = 0.13, P = 0.211 (95% CI −0.07 to 0.33); low certainty. From 280 unique participants in 9 clusters, 13 effect sizes.
Direct morphology: g = 0.05, P = 0.718 (95% CI −0.21 to 0.30); very low certainty. Drawn from three clusters, 115 unique participants, 6 effect sizes.
The primary interpretation held up under publication-bias diagnostics, sensitivity analyses and exploratory subgroup analyses, none of which changed it materially.
The strength estimate was also non-significant under the cluster-robust check (p = 0.105) (full text, checked 2026-10-02: section 3.4.1).
Risk of bias: low in five reports (45.5%), some concerns in six (54.5%), none high; PEDro 6 to 10, median 8 (full text, checked 2026-10-02: section 3.3, Table 3).
One subgroup stood out and did not survive: catechin/flavanol-rich products and lean mass, g = 0.59 (95% CI 0.04 to 1.14; 4 clusters), not supported after the cluster-robust correction (Pdiff = 0.086) (full text, checked 2026-10-02: section 3.4.2, Table 4).
No estimate pointed to a loss: none of the three pooled estimates was negative, and the authors write that the evidence is "likewise insufficient to establish a meaningful interference effect" (full text, checked 2026-10-02: section 4).
Practical Application
"Will a daily polyphenol supplement — green tea extract, quercetin, curcumin, berry extract — help or hurt what I gain from strength training?" Across 11 small randomized trials (360 participants, none of them trained athletes) pooled in a three-level meta-analysis, neither was shown: strength g = 0.25 (95% CI −0.01 to 0.51, P = 0.060; low certainty), lean mass g = 0.13 (95% CI −0.07 to 0.33; low certainty), direct muscle size g = 0.05 (very low certainty). The authors' conclusion is that a reliable additional effect is not established and that firm conclusions on benefit or lack of benefit are not possible.
Read it as "no reason to buy it for gains, and no evidence in these trials that it blunts them" — not as proof that polyphenols are harmless to adaptation. The intervals are wide enough to hold a small benefit and a small loss.
Not from this paper: the base's antioxidant caution rests on a different supplement and a different adaptation — 1000 mg vitamin C plus 235 mg vitamin E a day blunted mitochondrial markers during endurance training (REF-908). This meta-analysis neither extends that caution to polyphenols in strength training nor lifts it.
Not from this paper: short courses for recovery are a separate question — tart cherry juice around one marathon (REF-009). The paper's own introduction separates acute recovery from long-term adaptation and analysed only the latter.
Inference, not tested here: for an endurance athlete adding gym work the result transfers only loosely — the participants were untrained or older adults, and no trial combined the supplement with endurance training.
Not from this paper; the base's own practice: where polyphenol supplements sit among supplements to skip is SYNTH-67.
Limitations
It is a real meta-analysis, properly modelled, of a thin evidence base. Effect sizes were pooled with three-level models that handle several outcomes per trial, with cluster-robust checks, leave-one-out and pre-specified sensitivity analyses; the review was PROSPERO-registered and used RoB 2 (appropriate for randomized trials) plus GRADE. The weakness is the input, not the method (full text, checked 2026-10-02: sections 2.6-2.8).
Certainty as the authors give it: low for strength and lean mass, very low for direct morphology. Each outcome rests on fewer than ten independent trial clusters (8, 9, 3), so subgroup, dose and publication-bias analyses are exploratory by the authors' own statement.
"Polyphenols" is nine different products, differing in compound, dose, formulation and absorption; a class-level average may hide a compound that works and one that does not. No dose-response analysis was possible.
Not athletes. The title says "adults"; the trials are untrained young men, older, sarcopenic, overweight or postmenopausal adults (full text, checked 2026-10-02: section 3.2). Nothing here was measured in trained lifters or endurance athletes.
Attribution is blurred in several trials (full text, checked 2026-10-02: sections 3.2 and 4): caffeine in the product, essential amino acids alongside, a hypocaloric run-in, HIIT mixed into the training; one trial had no placebo; training loads included 40% 1RM and isometric-only programmes (Table 2).
Data handling: some change-score SDs were reconstructed with an assumed correlation of 0.5, some values were read from figures, and seven reports sought could not be retrieved (full text, checked 2026-10-02: sections 2.5, 2.8, 3.1).
Small count disagreements inside the paper (full text, checked 2026-10-02: Table 1 vs section 3.2): the group sizes in Table 1 add up to 361 participants and 158 women, where the text gives 360 and 157; one trial listed among the strength sources shows only muscle thickness and mass as its outcomes in Table 2. None of this moves the conclusion.
Lean mass was mostly estimated from body-composition devices, not measured as muscle size; the direct measures come from three trials.
Safety could not be evaluated: adverse-event data were available in five reports and systematically assessed in one (full text, checked 2026-10-02: section 3.2).
Funding: Keep-Capital University of Physical Education and Sports Collaborative Innovation Laboratory of Sports Science (Keep is a commercial fitness company); the authors declare no conflict of interest (full text, checked 2026-10-02: Funding, Conflict of interest).
How this entry was checked and changed
Why this level
lowered from A-: the authors' own GRADE is low (strength, lean mass) and very low (direct morphology); 8, 9 and 3 independent trial clusters per outcome, each trial small; nine chemically different supplements pooled as one class; no trained athletes among the participants