The short version: the common objection — "you just digest it into amino acids" — is wrong, and the collagen industry deserves credit for proving it. Specific peptides do survive digestion and circulate.
But that's where the credit ends. Skin hydration and elasticity hold up. Joint pain is a small real-ish signal from four high-risk-of-bias trials. Bone rests on one manufacturer-funded study with a flattering metric and no independent replication in eight years. Hair, gut and muscle claims range from unsupported to publicly dismantled.
This is a field where the manufacturer usually funded the trial, sometimes employed the author, often held the patent, and in one notable case disclosed none of it until a decade later.
The absorption argument, settled
The sceptical objection is that collagen is a protein, protein is digested to amino acids, so swallowing collagen can't deliver collagen anywhere.
That objection is wrong, and the evidence is good. Specific di- and tripeptides — particularly prolyl-hydroxyproline (Pro-Hyp) and hydroxyprolyl-glycine (Hyp-Gly) — survive digestion and appear intact in human plasma after ingestion.
- **Iwai K, Hasegawa T, Taguchi Y, et al. J Agric Food Chem 2005;53(16):6531–6536** — the foundational paper identifying food-derived collagen peptides in human blood, with Pro-Hyp dominant.
- **Ichikawa S, et al. Int J Food Sci Nutr 2010;61(1):52–60** — hydroxyproline-containing di- and tripeptides quantified at high concentration in human blood after gelatin hydrolysate.
- **Shigemura Y, et al. Food Chem 2011** — identification of Hyp-Gly in human peripheral blood.
So the peptides get in. The question is what happens next.
Where the chain breaks
Every step past "it's in plasma" is rodent or in vitro:
- Pro-Hyp stimulates growth of fibroblasts from mouse skin (Shigemura 2009)
- Pro-Hyp stimulates proliferation and hyaluronic acid synthesis in cultured human dermal fibroblasts (Ohara 2010 — in vitro, not in vivo)
- Pro-Hyp and Hyp-Gly enhance growth of primary cultured mouse fibroblasts (Shigemura 2020)
We could find no human study measuring Pro-Hyp in skin or cartilage biopsies after ingestion. The chain "circulating peptide → reaches human dermis → stimulates human fibroblasts → measurably more dermal collagen" has no complete human link.
One more thing worth knowing: plasma Pro-Hyp is generated both from a supplement and from ordinary turnover of your own collagen. Baseline is not zero.
The honest position: absorption is demonstrated, mechanism is not. It's plausible. It isn't shown.
Skin: the one claim that holds up
Two meta-analyses converge here, and crucially they used objective instruments rather than satisfaction surveys.
**Pu S-Y, Huang Y-L, Pu C-M, et al. Nutrients 2023;15(9):2080 — the most fully verifiable one. 26 RCTs, n=1,721**, 2–12 weeks, doses ranging from 372 mg to 12 g/day.
- Hydration: pooled effect 0.63 (95% CI 0.38–0.88), p<0.00001 — measured by corneometer
- Elasticity: pooled effect 0.72 (95% CI 0.40–1.03), p<0.00001 — measured by cutometer
- Sources: 14 fish, 1 bovine, 1 chicken, 2 porcine, 9 unknown
- Risk of bias: missing outcome data in 13 of 26 trials
- Review itself funded by a hospital, not industry
**Miranda RB, Weimer P, Rossi RC. Int J Dermatol 2021;60(12):1449–1461 — 19 double-blind RCTs, n=1,125**, 95% female. Concluded 90 days of hydrolysed collagen reduces wrinkles and improves elasticity and hydration. (Effect sizes and the funding breakdown of included trials are behind a paywall we couldn't get past — we're not quoting numbers we haven't read.)
The problems to keep in mind
- Duration. Median trial length is 8–12 weeks against an outcome — dermal collagen architecture — that changes over years.
- The 32-fold dose range being pooled, from 372 mg to 12 g, in a single meta-analysis.
- Proprietary blends. Many products and several trials combine collagen with vitamin C and hyaluronic acid, making the collagen's contribution unattributable. One 2024 trial of a multi-ingredient supplement claimed skin, hydration, elasticity, wrinkle, scalp and hair benefits simultaneously.
- Instrument sensitivity. Cutometer and corneometer readings vary with room humidity and probe placement.
- Sponsorship. The most-cited branded peptides each have their own trial literature largely authored or funded by their manufacturer.
- And note that the 2023 meta-analysis did not systematically tabulate trial funding — in a field this industry-dominated, that omission is itself a finding.
Still: two independent meta-analyses, objective instruments, moderate effect sizes with confidence intervals excluding zero. A sceptic should concede that something is probably happening to skin hydration and elasticity.
Joints: two different products that get conflated
Hydrolysed collagen peptides, around 10 g/day
**Clark KL, Sebastianelli W, Flechsenhar KR, et al. Curr Med Res Opin 2008;24(5):1485–1496. 147 randomised**, 24 weeks, 10 g/day liquid collagen hydrolysate, in athletes with activity-related joint pain.
- Pain while walking: p=0.007, effect size 0.36
- Pain at rest (physician-assessed): p=0.025
- Knee subgroup (n=63): pain at rest p=0.001, effect size 0.45
Funding: GELITA Health GmbH. Co-author Klaus Flechsenhar was in GELITA's Department of Research and Development. Also note 97 of 147 randomised made it into the analysis — a 34% attrition rate — multiple unadjusted outcomes, and subjective pain scales.
The meta-analysis (J Orthop Surg Res 2023;18:782) pooled only four RCTs, n=507 in knee osteoarthritis:
- Pain (VAS): SMD −0.58 (95% CI −0.98 to −0.18), p=0.004, I²=68%
- GRADE: moderate for pain
- All four included studies rated HIGH risk of bias
- Adverse events: OR 1.66 (0.99–2.78), p=0.05 — a borderline signal toward more adverse events
Undenatured type II collagen (UC-II), 40 mg/day
This is a completely different intervention and the logic is opposite.
**Lugo JP, Saiyed ZM, Lane NE. Nutr J 2016;15:14.** 191 randomised, 180 days, three arms: UC-II 40 mg/day versus glucosamine 1,500 mg + chondroitin 1,200 mg versus placebo.
- WOMAC total: UC-II −551 vs placebo −414, p=0.002
- UC-II versus glucosamine/chondroitin: −551 vs −454, p=0.04
- InterHealth Nutraceuticals, which owns UC-II, provided all study materials.
The 250-fold difference matters. Hydrolysed collagen is sold as substrate provision — give the body building blocks. UC-II is sold as oral tolerance — a tiny dose of undenatured collagen trains gut-associated immune tissue not to attack cartilage.
Those mechanisms are not merely different, they're incompatible. UC-II's proposed active principle requires the epitope to stay intact, so heating it, hydrolysing it or stirring it into hot coffee would destroy it. A product cannot be both. Anything marketed as "collagen for joints" without specifying which is marketed by someone who hasn't read either literature.
Bone: one trial, and read the metric carefully
**König D, Oesser S, Scharla S, Zdzieblik D, Gollhofer A. Nutrients 2018;10(1):97. 131 randomised, 102 completed, 5 g/day specific collagen peptides, 12 months**, postmenopausal women.
- Spine T-score: collagen +0.1 vs placebo −0.03; ANCOVA p=0.030
- Femoral neck T-score: collagen +0.09 vs placebo −0.01; ANCOVA p=0.003
- Reported as percentage change: spine +3.0% vs −1.3%; femoral neck +6.7% vs −1.0%
- Bone formation marker P1NP rose in the collagen group only (p=0.007)
- "Part of the study was financially supported by GELITA AG, Germany." The authors declared no conflicts of interest, while the product tested is a GELITA trademark.
The sceptical read. That "+6.7% femoral neck" figure is a percentage change in a T-score, not in bone mineral density in g/cm². T-scores are standardised deviations that sit near zero, so a percentage change on them is an unstable and very flattering metric. The absolute shift is +0.09 of a T-score — within DXA measurement noise on a single machine. And a good part of the between-group difference comes from the placebo group declining, not the treatment group gaining.
Replication: the only follow-up is an open-label observational extension of the same cohort, by the same group, on the same product. We found no independent, non-Gelita-funded RCT replicating a bone density effect in eight years.
Muscle: the weakest claim, and the best story in the field
The amino acid problem
Collagen contains no tryptophan and very little leucine. The citable figure comes from a peer-reviewed letter: a 15 g dose of collagen peptides supplies 0.4 g of leucine — about 2.7%. Whey is roughly 10–11%.
Consequence: collagen scores near zero on protein-quality scales as a sole protein source, and grams of collagen should not be counted toward a daily protein target.
**Oikawa SY, Kamal MJ, Webb EK, et al. Am J Clin Nutr 2020;111(3):708–718.** The title is the finding: "Whey protein but not collagen peptides stimulate acute and longer-term muscle protein synthesis with and without resistance exercise in healthy older women."
And when researchers matched the leucine content between whey and collagen, whey still won on muscle thickness over a 10-week resistance training programme — implicating the full essential amino acid profile, not just leucine.
The Zdzieblik affair
This is worth telling in full, because it illustrates the field's funding problem better than any general statement.
**Zdzieblik D, Oesser S, Baumstark MW, Gollhofer A, König D. Br J Nutr 2015;114(8):1237–1245 reported that 15 g/day of collagen peptides plus resistance training in elderly sarcopenic men produced roughly +4.2 kg fat-free mass and −5.5 kg fat mass in 12 weeks** — a 1.3 kg fat-free mass advantage over placebo.
**Phillips SM, Tipton KD, van Loon LJC, Verdijk LB, Paddon-Jones D, Close GL. Br J Nutr 2016;116(3):569–570 — a letter from six of the most senior protein-metabolism researchers alive, titled "Exceptional** body composition changes attributed to collagen peptide supplementation and resistance training in older sarcopenic men." Their argument:
- The between-group fat-free mass difference was 2.7 to 5.6 times larger than meta-analytic estimates for protein supplementation generally
- The dose delivered only 0.4 g of leucine, "insufficient to induce any effect on muscle protein synthesis"
- The 5.5 kg fat loss was about 80% of what older men achieve in six months of hypocaloric dieting
A corrigendum was issued in 2025 — a decade later — adding the omitted disclosures: that Gelita AG provided the study samples and paid for participant compensation and sample analysis, and that co-author Dr Steffen Oesser is a named co-inventor on a patent application filed in 2013 relating to the active substance in the publication.
Oesser's name recurs across the Gelita-adjacent trial literature on joints, bone and skin.
Nails, hair and gut
Nails — one study, and it's worse than "small."
**Hexsel D, Zague V, Schunck M, et al. J Cosmet Dermatol 2017;16(4):520–526. Open-label, single-centre, no placebo, no randomisation, n=24 completers**, 2.5 g/day for 24 weeks. Nail growth +12%, broken nails −42%, 64% global clinical improvement. Only "p<0.05" reported — no exact p-values, no confidence intervals. Gelita acknowledged; no explicit funding or conflict statement.
For a nail-growth outcome with no control arm over a 24-week window, this is close to uninterpretable.
Hair — essentially nothing. The manufacturer announced its own "first peer-reviewed study" showing increased hair thickness in 2025. There's a 2024 mechanism paper in human hair follicle organ culture — ex vivo, not a clinical outcome. Hair claims are unsupported by adequate human RCT evidence.
Gut — the loudest claim with the thinnest file. The only human study we could find is a two-phase mixed-methods study in healthy women, not a placebo-controlled trial. There is no human RCT showing collagen peptides reduce intestinal permeability, treat IBS, or "heal leaky gut." The glutamine and glycine rationale is biochemical speculation. ("Leaky gut syndrome" is also not a recognised clinical diagnosis.)
Types, sources, and whether marine is worth it
Type I is skin, bone and tendon — about 90% of body collagen. Type II is cartilage. Type III sits alongside type I in skin and vessels. Bovine hide gives types I and III, marine (fish skin and scale) predominantly type I, chicken sternal cartilage type II.
Does source matter? For hydrolysed peptides the logic says largely no — hydrolysis destroys the triple helix and the type distinction, leaving di- and tripeptides whose composition differs only modestly. The 2023 meta-analysis pooled fish, bovine, porcine and chicken trials and didn't find source to be the discriminating variable, though with 9 of 26 trials not reporting source at all, that comparison was never properly powered.
We found no head-to-head RCT of marine versus bovine hydrolysate on any skin or joint endpoint. The marine premium rests on absence of evidence.
The defensible non-marketing arguments for marine: somewhat lower average molecular weight in some preparations, and avoiding bovine material for dietary or religious reasons. "Better absorbed" has no comparative outcome trial behind it.
Topical collagen doesn't penetrate. Native collagen is roughly a 300 kDa triple-helical molecule. The widely cited 500 Dalton rule (Bos & Meinardi, Exp Dermatol 2000) holds that compounds above about 500 Da don't meaningfully cross intact stratum corneum — collagen exceeds that by around 600-fold. Topical collagen works as a surface humectant and film-former. It holds water on the skin, which genuinely improves short-term hydration readings and feel. It does not enter your dermis or become your collagen.
Dose, and what Australians are actually buying
| Indication | Dose in the positive trials |
|---|---|
| Skin | 2.5–10 g/day, 8–12 weeks |
| Joint pain (hydrolysate) | 10 g/day, 24 weeks |
| Joint osteoarthritis (UC-II) | 40 mg/day, 180 days |
| Bone | 5 g/day, 12 months |
| Nails | 2.5 g/day, 24 weeks (uncontrolled) |
| Muscle | no effective dose established — 15 g failed to raise muscle protein synthesis |
Australian retail, 1 October 2026:
- Vital Proteins Collagen Peptides, 284 g — A$40.99. 10 g per scoop, ~28 serves, bovine types I & III, single ingredient. A$1.46/day, about A$44/month at a trial-equivalent dose.
- Dose & Co Marine Collagen, 200 g — A$44.99. 10 g of marine peptides per serve, 18 serves, with added vitamin C. A$2.50/day, about A$76/month. The marine premium works out at roughly 71% per effective dose.
Small-format "beauty collagen" tubs can't deliver 10 g/day for long — a 120 g tub is twelve days at trial dose.
Collagen coffee and collagen water are the clearest consumer problem. We tried to verify the per-serve collagen in Australian collagen-coffee products and could not: at least one popular product names marine collagen in the ingredients but does not disclose grams per serve. That non-disclosure is itself the finding. Products in this format typically carry a fraction of the trial dose, at a premium price.
(Hot coffee is irrelevant for hydrolysed peptides — they're already denatured. It would be fatal to a UC-II product's proposed mechanism.)
The vitamin C question
The biochemistry is uncontroversial: prolyl-4-hydroxylase and lysyl hydroxylase are ascorbate-dependent, and severe deficiency causes collagen failure. The question is whether adding vitamin C helps someone who isn't deficient.
**Shaw G, Lee-Barthel A, Ross ML, Wang B, Baar K. Am J Clin Nutr 2017;105(1):136–143. Randomised double-blind crossover, n=8 healthy men**, vitamin C–enriched gelatin before rope-skipping. 15 g roughly doubled a collagen synthesis marker versus placebo.
But the design cannot answer the question it's cited for: there was no gelatin-without-vitamin-C arm. Eight men, an acute biomarker, no clinical outcome.
We found no trial isolating the effect of adding vitamin C to a collagen supplement in people with adequate vitamin C status. Australians with any fruit or vegetable intake are replete. An added 40–80 mg of ascorbic acid has no plausible marginal effect — and its presence makes the collagen's independent contribution unattributable in any trial of that blend.
Verdict: formulation talking point.
The Australian regulatory position
Collagen products sit in two regimes depending on marketing:
As a food — collagen coffee, collagen water, collagen in yoghurt — governed by the FSANZ Food Standards Code, where therapeutic claims are prohibited and only permitted health claims may be made. "Collagen" has no permitted health claim, so these products can legally say almost nothing. Hence the reliance on imagery, the word "beauty," and implication.
As a listed complementary medicine (AUST L) — the sponsor may select from the Permitted Indications List and must hold evidence. But the TGA does not assess that evidence before the product goes on sale. AUST L listing is self-certification plus an automated electronic process, with post-market audits of a sample.
An AUST L number is not a statement that the product works. That's the single most important thing for an Australian buyer to understand.
We also found no Australian independent product testing of collagen powders — no label-claim verification, no heavy metals testing on marine collagen.
On sustainability: marine collagen is made from fish skins and scales, which are by-products of existing fishery and aquaculture processing — a valorised waste stream rather than new fishing pressure, which is a genuine positive. The caveats are that species and fishery of origin are almost never disclosed, and there's no collagen-specific chain-of-custody standard.
What I'd tell someone spending $60 a month
- Your most likely real benefit is modest, measurable skin hydration and elasticity after 8–12 weeks. Not wrinkle reversal. Not your gut, hair, bones or muscles.
- Buy plain bovine hydrolysate at 10 g/day, about A$44/month. Skip the marine premium — you're paying ~71% more for no demonstrated outcome difference — and skip the added vitamin C, which is doing nothing your diet isn't.
- Never buy collagen coffee or collagen water for the collagen. The dose is a fraction of trial dose and frequently isn't even disclosed.
- If it's your joints: decide which product you want. Hydrolysate at 10 g, or UC-II at 40 mg. They are not the same intervention and they differ by 250-fold.
- If it's your protein intake: buy whey or any complete protein. This isn't a close call, and collagen grams shouldn't count toward your target.
- And read every trial with the funding line in view. In this field the manufacturer usually paid for it.
Frequently asked
Does collagen actually work? For skin hydration and elasticity, probably yes — two meta-analyses using objective instruments found moderate pooled effects with confidence intervals excluding zero. For joint pain, a small signal from four high-risk-of-bias trials. For bone, one manufacturer-funded trial with no independent replication. For muscle, hair and gut, the evidence ranges from absent to actively negative.
Isn't collagen just digested into amino acids? No, and this is the one place the industry's argument is sound. Specific di- and tripeptides — Pro-Hyp and Hyp-Gly — survive digestion and appear intact in human plasma. What hasn't been shown is the next step: that those circulating peptides reach human skin or cartilage and do anything there. That evidence is rodent and in-vitro only.
Is marine collagen better than bovine? There is no head-to-head trial of marine versus bovine hydrolysate on any skin or joint endpoint. Hydrolysis destroys the structural differences between sources, leaving peptides of broadly similar composition. In Australia the marine premium runs about 71% per effective dose.
How much collagen should I take? The trials that found something used 2.5–10 g/day for skin, 10 g/day for joint pain, and 5 g/day for bone — all over 12 weeks to 12 months. Undenatured type II collagen is a separate product used at 40 mg/day. Most "collagen coffee" and collagen water products carry a fraction of any of these.
Does collagen help build muscle? No. Collagen contains no tryptophan and about 2.7% leucine against whey's 10–11%. A trial in older women found whey stimulated muscle protein synthesis and collagen did not, and even when leucine was matched, whey still produced more muscle thickness. Collagen grams should not count toward your protein target.
Does topical collagen do anything? It works as a surface humectant — it holds water on the skin and improves how skin feels and measures short-term. It cannot penetrate the skin barrier: native collagen is around 300 kDa, roughly 600 times the ~500 Dalton threshold for meaningful absorption.
Last reviewed 1 October 2026. Industry funding is noted wherever the published papers disclose it. Prices are Australian retail at the date of review. This page is general information, not medical advice.