Comparative Effects of Dietary Protein, Creatine, and Omega-3 Supplementation on Muscle Strength, Endurance, and Recovery in Trained Athletes: A Systematic Review and Network Meta-Analysis
This systematic review and network meta-analysis set creatine, dietary protein and omega-3 fatty acid supplements against one another for endurance performance, muscle strength and recovery outcomes in trained athletes.
Scopus, MEDLINE, Embase, Web of Science, Cochrane CENTRAL and SPORTDiscus were searched for randomized controlled trials in people who had done structured training for at least six months. A frequentist random-effects network model pooled the data, and treatments were ranked by Surface Under the Cumulative Ranking curve (SUCRA) probabilities.
In total, 35 trials with 1211 participants were analysed. By the abstract, creatine had the largest effect for muscle strength (SMD = 0.46, SUCRA = 82.4%, 95% CI: 0.29 to 0.63), protein for endurance performance (SMD = 0.28, SUCRA = 85.2%, 95% CI: 0.08 to 0.48) and omega-3 for recovery outcomes (SMD = 0.40, SUCRA = 88.7%, 95% CI: 0.18 to 0.62).
The full text does not support reading these as findings about trained athletes, or as comparisons between the three supplements (see Limitations).
Participants
35 trials enrolling 1211 participants, described in the abstract as individuals engaged in structured training for a minimum of six months.
As the paper describes them (full text, checked 2026-10-02: section 3.2): mean age 24.6 years (range 18–35), 71.4% male; 18 trials in resistance-trained people, 11 in endurance athletes, 6 in mixed-sport athletes.
As the trial list shows them (full text, checked 2026-10-02: Table 1 and reference list, source abstracts read through Europe PMC): at least seven of the 35 trials were run in older adults — vulnerable older women, older men, men 55–75 yr old, women aged 64, sarcopenic women aged ≥ 65, healthy older adults, and physically active adults aged 69 — which contradicts both the stated age range and the stated exclusion of older participants without a training background.
Protocol
Six databases searched from inception through October 2025, English only; two reviewers; Cochrane RoB 2.0; GRADE adapted to network meta-analysis; R packages netmeta and gemtc (full text, checked 2026-10-02: sections 2.1–2.6).
Stated entry criteria (full text, checked 2026-10-02: section 2.2): protein at least 1.2 g/kg body weight per day for a minimum of four weeks; creatine loading (approximately 20 g/day for 5–7 days) or at least 3 g/day for a minimum of four weeks; omega-3 at least 1 g/day combined EPA and DHA for a minimum of two weeks.
Outcome pools (full text, checked 2026-10-02: section 2.2): "muscle strength" = one-repetition maximum, isokinetic peak torque, maximal voluntary contraction; "endurance performance" = time to exhaustion, time trial performance, maximal oxygen consumption, running economy; "recovery" = creatine kinase, muscle soreness ratings, inflammatory markers, functional recovery assessments.
Network geometry (full text, checked 2026-10-02: Figure 3 and Discussion): star-shaped, placebo the only common comparator; the Introduction states that there are no head-to-head trials of the three supplements.
Key Results
| Outcome (abstract) | Ranked first | SMD vs placebo (95% CI) | SUCRA |
|---|---|---|---|
| Muscle strength | Creatine | 0.46 (0.29 to 0.63) | 82.4% |
| Endurance performance | Protein | 0.28 (0.08 to 0.48) | 85.2% |
| Recovery outcomes | Omega-3 | 0.40 (0.18 to 0.62) | 88.7% |
By the abstract, each supplement ranked first for exactly one outcome; every one of these is an effect against placebo, not against another supplement.
The comparisons between supplements were mostly not significant (full text, checked 2026-10-02: sections 3.3–3.5): creatine vs protein for strength SMD = 0.13 (95% CI −0.09 to 0.35); protein vs creatine for endurance SMD = 0.23 (−0.12 to 0.58) and protein vs omega-3 SMD = 0.10 (−0.24 to 0.44); omega-3 vs creatine for recovery SMD = 0.18 (−0.08 to 0.44).Only creatine vs omega-3 for strength (0.30, 0.03 to 0.57) and omega-3 vs protein for recovery (0.23, 0.01 to 0.45) reached significance. So "protein most effective for endurance" is a ranking, not a demonstrated difference from the other two.
Effects that did not differ from placebo (full text, checked 2026-10-02: Table 2): omega-3 on strength 0.16 (−0.06 to 0.38) and on endurance 0.18 (−0.08 to 0.44); creatine on endurance 0.05 (−0.26 to 0.36), ranked below placebo; protein on recovery 0.17 (−0.04 to 0.38).
Heterogeneity as the authors give it (full text, checked 2026-10-02: sections 3.3–3.5): I² = 38.6% for strength, 34.2% for endurance, 41.3% for recovery.
Risk of bias as the authors give it (full text, checked 2026-10-02: section 3.2): 23 studies low, 9 some concerns, 3 high; no per-study judgments are shown.
GRADE as the authors give it (full text, checked 2026-10-02: Table 4): all indirect comparisons between supplements "Very Low"; omega-3 vs placebo on endurance and protein vs placebo on recovery "Low".
What sat in the "endurance" pool (full text, checked 2026-10-02: sections 3.2 and 3.4): 15 trials, 487 participants; time-to-exhaustion protocols (n = 9), time trials (n = 4), maximal oxygen consumption tests (n = 5). Only 11 of the 35 trials were in endurance athletes.The one included trial whose title is about protein and endurance training reports the opposite of the pooled claim: "Protein Supplementation Does Not Augment Adaptations to Endurance Exercise Training" (Jonvik 2019, 60 recreationally active males).
What sat in the "recovery" pool (full text, checked 2026-10-02: sections 3.2 and 3.5): 18 trials, 612 participants; serum creatine kinase (n = 12), soreness on visual analogue scales (n = 14), repeated performance testing (n = 8).All 7 omega-3 trials (n = 211) are counted in it, although by their own abstracts one measured heart rate and oxygen consumption in 16 cyclists and no recovery outcome, three were resistance-training studies in older adults, and one was a muscular-endurance test in 19 healthy men on 860 mg/day of EPA + DHA.
Practical Application
"Protein, creatine or omega-3 — which should I take for strength, endurance or recovery?" This network meta-analysis (35 trials enrolling 1211 participants, each supplement against placebo only, no head-to-head trials) ranks creatine first for muscle strength (SMD = 0.46, 95% CI: 0.29 to 0.63), protein first for endurance performance (SMD = 0.28, 95% CI: 0.08 to 0.48) and omega-3 first for recovery outcomes (SMD = 0.40, 95% CI: 0.18 to 0.62), but the full text shows the trial list does not match the review's own entry criteria, so none of the three rankings should change what an athlete takes.
What survives: creatine with resistance training raises strength more than placebo — which the base already holds on stronger sources (REF-352, REF-060) — and creatine did nothing measurable for endurance performance here, which is consistent with the base's reading of them.
"Protein is the most effective supplement for endurance performance" is not a usable statement: the pool mixes time to exhaustion, time trials and VO2max, the difference from omega-3 and creatine was not significant, and the best-known included trial found no added endurance adaptation from protein. The base's protein guidance for endurance athletes is a daily total (REF-725, REF-151), not a performance claim, and this paper does not change it.
"Omega-3 is the best supplement for recovery" rests on seven trials of which several were in older adults or measured no recovery outcome, and "recovery" here means blood creatine kinase, soreness scores and inflammatory markers, not performance in the next session.The ISSN stand already says omega-3 may decrease subjective muscle soreness (REF-188) and a review of trial design found the athlete evidence too inconsistent for a dose (REF-189); this paper adds no dose an athlete could rely on.
Not from this paper; the base's own practice: which supplements are worth an endurance athlete's money is in SYNTH-67.
Limitations
A real pooled analysis, but not a real comparison. Every trial compared one supplement with placebo or control; the network has no closed loop (full text, checked 2026-10-02: Figure 3, "star-shaped").In such a network there is no direct evidence between supplements to test against the indirect evidence, yet the abstract reports "no significant disagreement between direct and indirect evidence across all outcomes" and the text gives a consistency test for each outcome. That statement cannot mean what it says.
The population is not what the title says. The paper states ages 18–35 and excludes older people without a training background; its Table 1 and reference list include Gualano 2014 ("vulnerable older women"), Chrusch 2001 ("older men"), Eijnde 2003 ("men 55–75 yr old"), Rodacki 2012 ("elderly women"), da Cruz Alves 2022 ("sarcopenic older women"), Lee 2022 ("older adults") and ten Haaf 2019 ("physically active older adults") (full text, checked 2026-10-02: Table 1, references 15, 32, 35, 39, 40, 42, 44).Three of the seven omega-3 trials are in this group.
Trials that do not meet the stated criteria. Short studies with no four-week supplementation (Areta 2013, a 12 h protein-synthesis study with no strength, endurance or recovery outcome; Rindom 2016, training periods of 5 days; Eddens 2017, supplements for a total of 4 days); trials without a placebo arm although 62 studies were excluded for exactly that (soy vs whey, protein-source and protein-distribution comparisons; omega-3 trials whose control was training alone); an omega-3 trial below the 1 g/day minimum (full text, checked 2026-10-02: Table 1 against section 2.2 and section 3.1; doses and designs from the source abstracts).
Counts in text and table disagree. The text says all trials were "published from 2008 to 2025"; Table 1 lists trials from 2000, 2001, 2002, 2003 and 2006. The text says sample sizes "ranged from 16 to 72 participants, with a median of 32" and, one paragraph earlier, "ranging from 16 to 116 participants per study"; the table's median is 30.The strength analysis reports 942 participants "receiving active interventions" plus a placebo node of n = 496, which is more than the 1211 in the whole review. Figure 2B gives 882 males of 1211, which is 72.8%, not the stated 71.4% (full text, checked 2026-10-02: sections 3.1–3.3, Table 1, Figures 2B and 3).
Table 1 does add up, but not to the source papers. Its 35 rows sum to exactly 1211 (protein 583, creatine 417, omega-3 211).Against the source abstracts, the sample sizes or durations differ for several: Jonvik 2019 is listed as 44 participants and 10 weeks (abstract: 60, 12 wk); Okut 2025 as 40 (abstract: thirty, no placebo); Morishima 2020 as 24 (abstract: nineteen); Hansen 2020 as 26 and 4 weeks (abstract: twenty-four, six-week); da Cruz Alves 2022 as 28 and 16 weeks (abstract: thirty-four, 14 weeks); Rindom 2016 as 24 (abstract: 12, crossover); Eddens 2017 as 30 (abstract: twenty-four); Kritikos 2021 as 20 (abstract: ten, crossover); Kim 2023 as 21 (abstract: thirty-two).Nine of the 20 source abstracts read disagree with the table on sample size. Completers or counting crossover arms twice may explain some, not all (full text, checked 2026-10-02: Table 1; source abstracts via Europe PMC).
One "included trial" is not a trial. Table 1 row "Wang et al., 2024", n = 25, "4-week creatine supplementation with complex training", cites reference [5], which in the reference list and in Crossref (10.3390/nu16213665) is a systematic review and meta-analysis of 23 studies (full text, checked 2026-10-02: Table 1 and reference 5). All 20 references spot-checked by DOI resolve to real papers; the problem is what the review says they are.
GRADE text and table disagree, and the grades are generous. The text rates certainty "moderate" for creatine vs placebo on strength and omega-3 vs placebo on recovery; Table 4 rates omega-3 vs placebo on recovery "High" and creatine vs placebo on strength "Moderate". Indirectness is marked "Not serious" for every placebo comparison despite the older-adult and non-athlete trials (full text, checked 2026-10-02: section 3.6 and Table 4).
No per-study effect sizes, risk-of-bias table or data supplement (the only supplement is the PRISMA checklist); the pooled numbers cannot be recomputed. No protocol registration. Funding: none; conflicts of interest: none declared (full text, checked 2026-10-02: section 2.1, back matter).
Outcomes are pooled as standardized mean differences across different tests; no result is expressed in watts, seconds or kilograms.
How this entry was checked and changed
Why this level
lowered 2026-10-02 from the nominal A- of a network meta-analysis after reading the full text: the included-trial table breaks the review's own entry criteria (older and sarcopenic adults, acute studies, non-placebo comparators, one cited "trial" that is a meta-analysis), sample sizes differ from the source papers, and the pooled numbers cannot be traced to trial-level data