Collagen research supports one finding: tendon cross-sectional area and stiffness rising more with collagen than with placebo alongside hard training, graded strong (GRADE A) by a 2026 systematic review of eight trials, which recommends 15–30 g a day. The same review grades the evidence strong against any effect on muscle strength. And a trial at 15 g a day that it did not include — 39 men, 15 weeks, tendons imaged by MRI — found no difference from placebo.
The table lists the studies the category actually rests on, each with the limit that travels with it. Read the measurement column first. A blood marker, a tendon image and a muscle biopsy are three different kinds of evidence, and most arguments about collagen are arguments between them.
What do collagen studies actually measure?
Mostly proxies. Shaw and colleagues’ 2017 trial, the source of the common 15 g dose, measured PINP — a blood marker of collagen synthesis — in eight men over three days. The tendon trials measure cross-sectional area and stiffness by ultrasound or MRI, and the muscle trials take biopsies. No randomised trial has reported a change in how often people get hurt, which is the question a structural finding is usually stretched to answer.
| Source | Design | Population and n | Dose and duration | What was measured | Finding, with its size | Limit or funding |
|---|---|---|---|---|---|---|
| Shaw et al., Am J Clin Nutr, 2017 | Randomised, double-blind crossover | 8 healthy men | 5 g or 15 g of vitamin-C-enriched gelatin, or placebo, 1 hour before 6 minutes of rope-skipping, 3 times a day for 3 days | Serum PINP; engineered ligaments exposed to the participants’ serum | 15 g roughly doubled PINP; 5 g landed in between | n = 8; a blood marker, not a tendon; gelatin rather than peptides; the ligament data come from a dish |
| Nulty et al., Am J Physiol Endocrinol Metab, 2024 | Double-blind crossover | 8 resistance-trained men, mean age 49 | 0, 15 or 30 g of vitamin-C-enriched hydrolysed collagen before one resistance session | Serum PINP after the session | PINP area under the curve: 30 g above 15 g, both above 0 g (P < 0.05); no PINP change at 0 g | n = 8; one session; a blood marker |
| Buchalski et al., J Funct Morphol Kinesiol, 2026 | Systematic review of randomised trials, narrative synthesis | 8 trials, 257 people (246 men, 11 women), ages 18–52 | About 5–30 g a day, with resistance or plyometric training, for 3–15 weeks | Tendon cross-sectional area and stiffness; strength; performance | GRADE A for larger increases in tendon area and stiffness; GRADE A against an effect on strength; conflicting (GRADE C) for performance | 95.7% men; small, short, varied trials; its tendon trials gave 5 g or 30 g a day, none 15 g; Balshaw’s tendon measurements are not among them |
| Nulty et al., Eur J Sport Sci, 2025 | Double-blind randomised trial | 20 recreationally active men, mean age 47; 9 on collagen, 11 on placebo | 30 g with 50 mg vitamin C after each session; training twice a week for 12 weeks | Tendon area, stiffness and modulus by ultrasound; strength; muscle thickness | Tendon area +6.8 against +1.2 square millimetres; stiffness +661 against +247 N/mm; no group difference on strength or muscle thickness | Nine men took collagen; twice the 15 g dose, taken after training rather than before |
| Balshaw et al., Med Sci Sports Exerc, 2023 | Double-blind randomised trial | 39 young, recreationally active men; 19 on collagen, 20 on placebo | 15 g a day; lower-body resistance training 3 times a week for 15 weeks | Tendon area and aponeurosis area by MRI; tendon stiffness, modulus, elongation and strain | No between-group difference on any tendon measure (0.365 ≤ P ≤ 0.877); both groups changed with training | Young men only; the peptides were a collagen manufacturer’s branded products |
| Bischof et al., Sports Medicine, 2024 | Systematic review with meta-analysis | 19 studies, 768 participants | Collagen peptides with long-term training | Maximal strength; tendon morphology | Maximal strength SMD 0.19, p < 0.01, low certainty; tendon morphology SMD 0.67, very low certainty | Certainty graded low to very low by the authors themselves |
| Aussieker et al., Med Sci Sports Exerc, 2023 | Double-blind randomised trial with tracer infusion and muscle biopsies | 45 recreational athletes (30 men, 15 women), 15 per group | 30 g collagen, 30 g whey or placebo after one resistance session; 5 hours of recovery | Muscle connective and myofibrillar protein synthesis | Connective protein synthesis: no difference between groups (P = 0.09); only whey raised myofibrillar synthesis above placebo | One session, early recovery only |
What is established
One structural finding, conditional on training. Tendon cross-sectional area and stiffness increase more with collagen than with placebo alongside high-intensity resistance or plyometric training: strong evidence, GRADE A, in Buchalski and colleagues’ 2026 review, which recommends 15–30 g a day. Three of the four trials that measured tendon area found a larger increase with collagen — two of them at 5 g a day and one at 30 g — and stiffness separated only in the two trials at 30 g; none of the review’s tendon trials gave 15 g. The clearest single positive trial, Nulty and colleagues’ 2025 study, enrolled men with a mean age of 47, gave 30 g after training, and had nine men on collagen. The finding sits beside a null the review did not include: Balshaw and colleagues’ trial of 15 g a day for 15 weeks, with the tendon imaged by MRI, found no between-group difference on any tendon measure.
And one physiological signal, from small trials. Collagen taken before loading raises PINP in blood in a dose-dependent way: roughly double at 15 g in Shaw and colleagues’ eight men, and higher again at 30 g than at 15 g in Nulty and colleagues’ eight middle-aged men in 2024. That is a synthesis signal in blood. It is not a measurement of a tendon.
What does the tendon research on collagen show?
That it is less settled than its headline. A 2026 systematic review grades larger increases in tendon size and stiffness with loading as strong evidence, from trials that gave 5 g or 30 g a day, and recommends 15–30 g. Balshaw and colleagues gave 39 men 15 g a day through 15 weeks of resistance training, in a trial the review did not include, and imaged the tendon by MRI: no between-group difference on any tendon measure. Both findings belong in any honest summary, and neither cancels the other.
The MRI trial is the easy one to leave out of a summary, and the 2026 review left it out: it lists the same trial’s companion paper on muscle, not its tendon measurements. At 15 weeks it ran longer than any of the four tendon trials the review did assess, which lasted 10 to 14 weeks, and at 15 g a day it studied a dose none of them used. The long account of it is on does collagen actually work, and the dose arithmetic on how much collagen for tendons.
Why do collagen studies disagree?
Because they differ on almost every axis that matters. Doses run from about 5 g to 30 g; timing runs from an hour before training to straight after it; tendons are imaged by ultrasound in some trials and MRI in others; training trials last 3 to 15 weeks; and samples are small — the eight trials in the 2026 review total 257 people, 246 of them men. With that spread, a null trial and a positive trial can both be correct.
How reliable is the evidence on collagen?
It depends on the outcome, and the graders use different scales. The 2026 review in J Funct Morphol Kinesiol rates the tendon findings GRADE A — its own letter grade of recommendation, not a GRADE certainty level — though a 15 g, 15-week MRI trial it did not include found no difference, and reports “strong evidence (GRADE A) against an effect on muscle strength”. Bischof and colleagues’ 2024 meta-analysis of 19 studies found maximal strength SMD 0.19, p < 0.01, on low-certainty evidence, and tendon morphology SMD 0.67 on very-low-certainty evidence. Small trials, mostly men, mostly short.
Does a rise in PINP mean collagen is working?
No. PINP is a blood marker of type I collagen synthesis, and it cannot say which tissue the new collagen went into. Shaw and colleagues’ 15 g arm roughly doubled it in eight men, and Nulty and colleagues’ 30 g arm raised it further than 15 g in eight others. Neither trial measured a tendon. A marker that moves is a reason to run a structural trial, not a substitute for one.
What the research does not show
- Strength. The 2026 review grades the evidence strong, GRADE A, against an effect on muscle strength. The 2024 meta-analysis that did find one measured maximal strength at SMD 0.19, p < 0.01, on low-certainty evidence. Those two numbers travel together, and neither supports a strength claim.
- How often people get hurt. No randomised trial in this literature has reported a change in it, at any dose. Tendon images and blood markers are not counts.
- Anything for muscle protein. Thirty grams after resistance exercise did not raise muscle connective protein synthesis above placebo, and only whey raised myofibrillar synthesis in the same trial (Aussieker et al., 2023). Collagen is not a protein supplement and is not sold as one here; collagen vs whey protein explains why.
- Women, or anyone over 52. Eleven of the 257 people in the 2026 review were women, and its oldest participants were 52.
- The dose question. The structural trials here used 15 g (Balshaw, no difference) and 30 g (Nulty, 2025, a difference), in different men, with different timing and instruments, so between them they cannot settle 15 g against 30 g. Thirty grams raised the blood marker further than 15 g, which is a marker, not a tendon.
- A strength finding in the article printed next to Bischof’s. Dolan and colleagues’ 2024 meta-analysis, the next article in the same issue of Sports Medicine and easily confused with it, examined the bone-turnover markers P1NP and CTX-1 after a single exercise bout under different nutrition interventions. Strength is not among its outcomes.
The cosmetic claims the category also sells, and why TARE makes none of them, are covered on does collagen actually work.
Where Under Load sits
Under Load is collagen peptides, Type I and III, bovine (grass-fed, Brazil), one ingredient, 350 g. The declared serving is 10 g, two 5 g scoops, 35 servings. The protocol on this site is a directions statement, not a serving: 15 g, three scoops, 30–60 minutes before training, with 50–100 mg of vitamin C — 23 whole doses per tub. We use 15 g because it is the bottom of the 15–30 g daily range the 2026 tendon review recommends with training, and the table shows how contested that ground is. Its certificate was run by ViaCore, the manufacturer’s own laboratory, on lot 121125450: 9.8 g measured per 10 g serving; 1.5 servings is 14.7 g by arithmetic. The documents are on Testing & Certification and the protocol on the protocol page.
The long-form pages: how much collagen for tendons, collagen peptides vs gelatin, bioactive collagen peptides vs collagen peptides and how to read supplement research. The same evidence map for the other two ingredients: creatine research summary and magnesium research summary. Every form compared side by side: supplement forms compared.
Sources
- Shaw G, Lee-Barthel A, Ross ML, Wang B, Baar K. Am J Clin Nutr 2017;105(1):136-143. PubMed 27852613
- Nulty CD, Tang JCY, Dutton JJ, et al. Am J Physiol Endocrinol Metab 2024;327(5):E668-E677. PubMed 39259166
- Buchalski A, Jeanfavre M, Altorelli C, Leff G. J Funct Morphol Kinesiol 2026;11(1):130. PubMed 41900537
- Nulty CD, Phelan K, Erskine RM. Eur J Sport Sci 2025;25(4):e12281. PubMed 40100255
- Balshaw TG, Funnell MP, McDermott EJ, et al. Med Sci Sports Exerc 2023;55(11):2083-2095. PubMed 37436929
- Bischof K, Moitzi AM, Stafilidis S, et al. Sports Med 2024;54(11):2865-2888. PubMed 39060741
- Aussieker T, Hilkens L, Holwerda AM, et al. Med Sci Sports Exerc 2023;55(10):1792-1802. PubMed 37202878
- Dolan E, Dumas A, Esteves GP, et al. Sports Med 2024;54(11):2889-2906. PubMed 39136851