Before asking whether collagen supplements work, it is worth following what happens to one.
Collagen is a large protein. Swallow it and your digestive system does to it what it does to any protein: takes it apart. What crosses into your blood is not collagen but amino acids and short peptide fragments. Your body may then use those fragments to build collagen — or to build anything else it needs, because once a protein is dismantled the pieces carry no instruction about where to go. There is no mechanism by which swallowing collagen directs amino acids to your face.
That is not, by itself, an argument that the supplements fail. The real hypothesis is subtler: that certain hydroxyproline-containing fragments survive digestion, reach the skin, and act as signals that prompt cells to make more collagen themselves. It is a reasonable idea. The problem is the state of the evidence for it — animal studies suggest some collagen peptides do reach skin, but the human data are lacking.
So the mechanism is plausible and unproven. Which means everything rests on the clinical trials, and this is where the article you are reading has to be careful, because the collagen literature has a specific and consistent weakness: the thing measured in the study is rarely the thing promised on the tub.
What follows is six claims, in descending order of how much evidence exists, and for each one the same two questions. What was actually measured? And who paid for the measuring?
One thing to establish before the criticism starts: this is an argument about whether collagen works, not about whether it is dangerous. Where trials reported adverse events, they were similar to placebo. Collagen is food protein, and nothing below suggests you are harming yourself. The question throughout is whether you are getting what you paid for.
The question that reorganises everything: what was actually measured?
Almost every disappointed collagen user has been misled by a substitution rather than a lie. A trial measures something narrow, instrument-based and short-term. The marketing translates it into something broad, visible and permanent. Both describe the same study. Only one of them is what the study found.
| The claim on the tub | What the trials actually measured | The gap |
|---|---|---|
| “Younger-looking skin”, “fewer wrinkles” | Skin hydration and elasticity by instrument (cutometer-type devices), over 8–12 weeks | An instrument reading on the forearm or cheek is not the same as looking different to other people |
| “Supports joint health”, “for your knees” | Self-reported pain on a visual analogue scale, in four small trials | Reported pain is a real outcome, but it is not cartilage, and it is wide open to placebo |
| “Stronger nails” | Nail growth rate and counted breakages in 25 women, with no control group | Growth rate is not strength, and without a control group nothing is attributable to the supplement |
| “Builds muscle” | Fat-free mass by body composition scan, in one trial whose numbers were publicly challenged | See the muscle section — the figures were called extraordinary by other researchers |
| “Thicker hair” | Nothing. There is no body of trials to describe | — |
| “Heals your gut” | Nothing in humans | — |
Hold that table in mind. Every section below is an expansion of one of its rows, and the gap column is where the money is made.
[IMAGE 2 — place here, immediately before this heading]
Claim 1: skin — where the evidence collapsed in public
Skin is the claim with the most trials behind it, and in the past eighteen months it has become the claim with the most interesting problem. Three papers tell the story, and they disagree with each other in a way that is unusually instructive.
The one everyone cites. A 2021 systematic review and meta-analysis in the International Journal of Dermatology concluded that ingesting hydrolysed collagen for 90 days reduces skin ageing — improving wrinkles, elasticity and hydration. This is the paper behind most of the reassuring health-site coverage you will find, and it is a real meta-analysis in a real journal.
The one that re-cut the data. In 2025, Seung-Kwon Myung and Yunseo Park published a systematic review and meta-analysis in The American Journal of Medicine covering 23 randomised controlled trials with 1,474 participants, looking at hydration, elasticity and wrinkles. Pooled together, the trials showed gains on all three. Then the authors split them. Among trials with no pharmaceutical or supplement-company funding, there was no benefit. Among higher-quality studies, no significant effect on any measure. The improvements in elasticity came from the low-quality studies. Their conclusion: there is “currently no clinical evidence to support the use of collagen supplements to prevent or treat skin aging.”
The one that audited the arithmetic. In 2026, David Robert Grimes, a biostatistician at Trinity College Dublin, and Anjali Mahto, a UK dermatologist, published a re-examination in the Journal of Cosmetic Dermatology of two earlier meta-analyses — the 2021 paper above and a 2023 one. They reported confused statistical measures, and found that one meta-analysis had reported a non-significant result as significant. After correcting the data, no significant relationship remained between collagen dose and improvement in hydration or elasticity.
| 2021, Int J Dermatol | 2025, Myung & Park, Am J Med | 2026, Grimes & Mahto, J Cosmet Dermatol | |
|---|---|---|---|
| What it did | Meta-analysis of trials | Meta-analysis of 23 RCTs, 1,474 participants | Statistical re-examination of the 2021 and 2023 meta-analyses |
| Headline | 90 days of hydrolysed collagen reduces skin ageing | Pooled benefit disappears in unfunded and higher-quality trials | The underlying arithmetic does not hold |
| Key finding | Improved wrinkles, elasticity, hydration | No benefit without industry funding; no effect in higher-quality studies | A non-significant result reported as significant; no dose–response after correction |
| Conclusion | Effective | “No clinical evidence to support the use of collagen supplements to prevent or treat skin aging” | “Oral collagen supplementation should not be promoted or recommended within dermatological practice” |
Notice the sequence, because it is the whole point. The 2021 meta-analysis is positive. It gets cited by health sites, which is how most people encounter the question. Five years later a statistician re-runs its numbers and the dose–response disappears. Nothing was faked; the paper simply had errors, and nobody checked until somebody did.
Grimes and Mahto also catalogued why these trials are hard to trust even when the arithmetic is right. Most ran only 8 to 12 weeks. Doses ranged from 0.3 to 10 grams a day within a single meta-analysis. Populations differed in age. Some trials gave collagen combined with other bioactive ingredients, so any benefit cannot be attributed to the collagen. And blinding is questionable, because collagen has a taste and a texture that may let participants work out which group they are in. Their summary of the evidence base: “methodologically fragile, inconsistent, and prone to bias,” and its promotion as a proven anti-ageing treatment “premature and not scientifically defensible.”
Two footnotes to the skin claim
Two things readers ask that belong here. First, collagen creams are a separate question with a simpler answer: collagen molecules are too large to cross the outer layer of skin, so a cream cannot deliver collagen into the dermis where it would need to go. Whether a given cream moisturises well is a different matter — many do — but that is the moisturiser working, not the collagen.
Second, where benefits have been reported, they are reported to stop when the supplementation stops. That is worth knowing before you start, because it reframes the purchase: not a course of treatment with a lasting result, but an ongoing monthly cost for as long as you want whatever effect there may be. As the dermatologist Justine Kluk puts it, “not all studies have shown a benefit” — and she notes that many collagen studies are sponsored by supplement companies, which may shape how results are presented.
What the industry said back — and where it has a point
The 2025 meta-analysis did not go unanswered. Three industry organisations responded publicly: BioCell Technology, which sells collagen ingredients; the Collagen Stewardship Alliance, a trade body; and GROW, the Gelatin Manufacturers of the World. None of them is a disinterested party, and you should weigh their arguments knowing that. But several of the arguments are good, and an article that ignored them would be doing exactly what it accuses the industry of doing.
| The objection | Who raised it | Does it have merit? |
|---|---|---|
| “Collagen” is a broad term covering many different ingredients, so pooling them makes comparisons unreliable | BioCell Technology | Yes — and notably, Grimes and Mahto make the same criticism independently. When the industry’s defenders and its statistical critics agree on a flaw, that flaw is real |
| The review did not separate pure collagen peptides from multi-ingredient blends | GROW | Yes. This is the same point, and it is the single biggest weakness in the whole literature — it also undercuts the positive trials, which is why nobody on either side can claim it cleanly |
| The paper’s abstract reports significant benefits while its conclusion says there is no clinical evidence | GROW | A fair complaint about presentation. A reader who sees only the abstract takes away the opposite of the conclusion |
| The subgroup criteria and quality ratings were not published | GROW | Yes, if accurate. Unpublished criteria cannot be checked, and a subgroup split is exactly where undisclosed choices change an answer |
| The funding classifications were wrong — Choi 2014 marked industry-influenced with no direct funding; Sugihara 2015 and Inoue 2016 marked independent despite company ties | Collagen Stewardship Alliance | The most consequential objection, and genuinely unresolved. If the labels are wrong, the headline finding weakens. It is checkable by anyone who reads the three papers; I was unable to open the primary paper to verify it, so this is reported as a live dispute, not settled |
| Most industry-funded studies scored 3–5 out of 5 on the Jadad quality scale | Collagen Stewardship Alliance | Partly beside the point. Myung and Park split trials by their own quality assessment; citing a different scale answers a different question |
| Industry funding does not influence outcomes, because trials are run by outside contract research organisations | BioCell Technology | No. The finding was a correlation in their dataset — benefit appeared in funded trials and not in unfunded ones. Who physically ran the trials does not address that pattern, and the pattern is what needs explaining |
| 17 of the 23 trials were conducted in Asia at lower doses, around 3 g/day | GROW | Worth testing, not yet tested. It is a real hypothesis — that the doses were too low — but it would need a properly powered independent trial at a higher dose to become an argument rather than a possibility |
The honest summary is that both sides are right about the literature and wrong about each other. The industry is correct that the evidence base is too heterogeneous to pool confidently. The critics are correct that what pooling there is depends on funding and on study quality. Those two statements together do not add up to “collagen works.” They add up to: after two decades and dozens of trials, the question has not been answered properly, and the people best placed to fund the answer are the people who benefit from the ambiguity.
Claim 2: joints — the best of a thin set
Joints are, slightly surprisingly, the claim where the evidence points most consistently in collagen’s favour. It is also a very small evidence base, and what it measured is not what people hope it measured.
A meta-analysis published in the Journal of Orthopaedic Surgery and Research in 2023 pooled four randomised controlled trials covering 507 patients with knee osteoarthritis, comparing collagen peptide against placebo. Pain scores on a visual analogue scale favoured the collagen group, and adverse events were similar between groups. One included trial’s post-hoc analysis reported that current, average, best and worst pain were all significantly better with collagen peptide.
Two cautions. First, the accessible summary of this meta-analysis is internally inconsistent — its own page summary says clinical outcomes were comparable while also saying VAS scores favoured collagen, and the full report sits behind a paywall. I could not resolve which framing the authors intended, so take the direction of the result but not a magnitude. Second, and more important, look at what four small trials can and cannot establish.
| What joint trials measured | What they found | What they did not measure |
|---|---|---|
| Self-reported pain, visual analogue scale | Favoured collagen peptide across four trials, 507 patients | — |
| Adverse events | Similar to placebo | — |
| Cartilage thickness or joint-space narrowing on imaging | — | Not measured. Nothing here shows a structural effect |
| Disease progression over years | — | Not measured. Trials of this kind run months |
| Need for joint replacement or other hard outcome | — | Not measured |
Pain is a legitimate outcome — if your knee hurts less, that matters, whatever the mechanism. But pain in osteoarthritis is also among the most placebo-responsive outcomes in medicine, which is precisely why four trials are not many, and why “supports joint health” is doing a lot of quiet work. Nothing in this evidence suggests collagen repairs a joint.
Claim 3: nails — one study, and it is worth meeting in person
“Stronger nails” appears on a very large number of tubs. It rests, substantially, on a single paper: Hexsel and colleagues, in the Journal of Cosmetic Dermatology in 2017. The findings are the ones you see quoted everywhere — nail growth increased 12%, broken nails decreased 42%, and almost 90% of participants reported improvement. Here is the study those numbers came from.
| Feature of the study | What it was |
|---|---|
| Participants | 25 healthy women |
| Control group | None |
| Blinding | None — open-label, single-centre |
| What was taken | 2.5 g a day of Verisol, a single branded collagen peptide ingredient |
| Duration | 24 weeks, then 4 weeks without |
| Outcomes measured | Nail growth rate, and counted cracked or chipped nails |
| The authors’ own position | That larger, placebo-controlled studies are needed to confirm the results |
| Institutional links | Conducted with the Collagen Research Institute in Germany, which has worked with the ingredient’s manufacturer for years. The coverage I could access does not state who funded the trial |
Without a control group, nothing observed in those 25 women can be attributed to the supplement rather than to time, season, the attention of being in a study, or the fact that everyone knew what they were taking and was asked whether it was working. That is not a hostile reading — it is the authors’ own reading, which is why they asked for placebo-controlled studies. Nearly a decade later, the claim on the tub still rests on the study that called for better studies.
There is one more detail that puts the whole claim in perspective. The authors noted that before their study there was no scientific evidence for collagen’s benefit on nails at all. So this was the first attempt — and a first attempt, uncontrolled and unblinded in 25 people, is exactly the kind of study that is supposed to justify a proper trial rather than a product line.
Note also the substitution. The outcomes were growth rate and counted breakages. “Stronger nails” is a different proposition, and nobody measured strength.
[IMAGE 3 — place here, immediately before this heading]
Claim 4: muscle — a trial that the field pushed back on
The muscle claim traces to one widely cited trial: Zdzieblik and colleagues gave 15 grams a day of collagen peptides alongside resistance training for 12 weeks to older men with sarcopenia, against placebo, and published the results in the British Journal of Nutrition. The reported changes were large.
They were large enough that six of the most prominent researchers in protein and muscle metabolism — Phillips, Tipton, van Loon, Verdijk, Paddon-Jones and Close — published a response in the same journal in 2016, titled “Exceptional body composition changes attributed to collagen peptide supplementation and resistance training in older sarcopenic men.” Their argument was not about bias or funding. It was about plausibility: the reported gains are far bigger than the entire literature on protein supplementation plus resistance training would predict.
| What the trial reported | The figure | How it compares |
|---|---|---|
| Fat-free mass gained | mean 4.2 kg (SD 3.3) | — |
| Body fat lost | mean −5.5 kg (SD 3.2) | — |
| Between-group difference in fat-free mass | about 1.3 kg | 2.7 times the standardised mean difference in the Cermak meta-analysis and 5.6 times that in the Finger meta-analysis |
| Effect of protein supplementation on fat-mass loss | — | Neither of those meta-analyses found one at all |
A corrigendum has since been published on the original paper.
This is a different kind of problem from the skin literature. There, the issue is weak methods and funding patterns. Here, a single trial produced numbers that specialists in the field considered too good to be credible, said so in print, and the paper required correction. Either way, “builds muscle” is not a claim a shopper should treat as settled — and note that the trial used 15 grams a day alongside a supervised resistance-training programme in men with diagnosed sarcopenia. Even taken at face value, it says nothing about a scoop in coffee without the training.
Claims 5 and 6: hair and gut — where there is nothing to weigh
The last two claims are different in kind from the first four. With skin, joints, nails and muscle there are trials to argue about. With hair and gut, the argument has not started, because the studies do not exist.
For hair, there is no good scientific evidence behind the claims you see — reduced hair loss, faster growth, slower greying. What exists is an anatomical fact: collagen is a component of hair follicles. That is background, not evidence. “Collagen is found in X” is not a finding about what happens when you swallow collagen, any more than “iron is found in blood” means eating iron filings improves your blood count.
For gut health, the reasoning runs that collagen supplies glycine and glutamine, which the intestinal lining uses. The amino acids are real. The inference is not tested in humans. And there is a telling practical signal here: search for evidence on collagen and gut health and what comes back is overwhelmingly supplement retailers’ own blogs rather than trials or reviews. When a claim is widely promoted and the literature search returns marketing, that absence is itself informative.
| Claim | What exists | What does not exist |
|---|---|---|
| Thicker or faster-growing hair | Collagen is a structural component of hair follicles | Trials showing oral collagen changes hair growth |
| Less hair loss | — | Any good evidence at all |
| Slower greying | — | Any good evidence at all |
| Gut barrier repair, “healing leaky gut” | A mechanistic story about glycine and glutamine | Human trials. A literature search returns sellers, not studies |
| Bone density | One trial on a manufacturer’s branded ingredient, promoted by that manufacturer | An independent replication I was able to verify |
These are not cases of “more research needed” in the polite sense. They are cases where a product is sold for a purpose nobody has tested.
The pattern running through all six: who paid
By this point the recurring feature is hard to miss. Myung and Park found benefit in industry-funded trials and none in unfunded ones. Grimes and Mahto reported that an overwhelming number of the studies they examined were industry-funded or had authors with commercial ties to collagen manufacturers. The nail study tested a single manufacturer’s branded ingredient, run with an institute that has worked with that manufacturer for years. The bone claim is promoted most prominently by the company selling the ingredient. Even researching the absorption question turns this up: among the leading results for collagen bioavailability evidence is a press release from Rousselot, a collagen manufacturer announcing a study on its own product line.
Industry funding does not make a result false. Companies fund research into their products because nobody else will, and some of that research is good. What funding does is change the standard of proof you should apply — and more importantly here, it shapes which questions get asked at all. Trials get designed to run 8 to 12 weeks, because that is long enough to move an instrument reading and short enough to be affordable. Outcomes get chosen that are sensitive rather than meaningful. Blends get tested rather than single peptides, because blends are what gets sold.
The result is a literature that is large but not decisive, and the gap persists for a structural reason: the organisation with the strongest interest in a definitive answer is the one that already has customers. Ambiguity is commercially survivable. A clean negative trial is not.
In fairness, the industry’s sharpest reply to the 2025 meta-analysis is precisely about funding labels — the Collagen Stewardship Alliance says some trials were misclassified in both directions. That dispute is unresolved and worth watching. The broader point survives it: nobody independent has yet run the trial that would settle this. On reading funding and other quality signals on a supplement, see our guide to how to tell if a supplement is actually good quality.
Six verdicts, and what would change each one
A verdict with no stated conditions is just an opinion. So each one below comes with the evidence that would move it — which is also the standard any future study or advert should be held to.
| Claim | Verdict | What would change it |
|---|---|---|
| Skin | Unproven, and weaker than it appeared two years ago. Benefit vanishes in unfunded and higher-quality trials; a leading positive meta-analysis failed a statistical re-examination | An independent, adequately powered trial: six months or longer, one defined peptide at one dose, no other active ingredients, and blinded-assessor or photographic outcomes rather than instrument readings alone |
| Joints (pain) | Possible, thinly supported. Four trials, 507 patients, pain scores favouring collagen, no structural outcomes | Larger independent trials, and at least one with imaging outcomes so the question of cartilage is actually addressed |
| Nails | Untested. One open-label study of 25 women with no control group | The placebo-controlled trial the original authors themselves asked for in 2017 |
| Muscle | Contested. One trial, with results specialists called implausible in print, and a corrigendum on the paper | An independent replication at 15 g a day alongside supervised resistance training |
| Hair | No evidence. Anatomical background is not a finding | Any trial |
| Gut | No evidence in humans. A mechanistic story, and a literature search that returns sellers | Any human trial |
The question that replaces “does collagen work?”
Notice that “does collagen work?” never got answered here, and could not have been. The question hides three variables, and until you fill them in it has no answer at all: work on which outcome, measured how, in a trial funded by whom? Fill those in and every row above becomes answerable. Leave them out and the same body of evidence supports a confident yes and a confident no — which is exactly the condition the marketing depends on.
So the useful rewrite is not “is collagen good?” but “what, specifically, am I expecting — and has anyone independent measured that?” For five of the six claims the answer today is no.
One practical note to end on. Collagen is protein, and an incomplete one. If your protein intake is low, food and ordinary protein sources fix that more cheaply and with better evidence. If your intake is already adequate, then the thing you are buying is the specific hypothesis in the opening section — that particular peptide fragments survive digestion and signal to skin cells. That hypothesis is plausible and currently unproven in humans. Buying it is not foolish; it is funding an experiment. The honest way to do that is to treat it as one: pick a single claim you care about, give it a defined period, decide in advance what a real change would look like, and stop if it does not appear. That is more rigour than most of the trials managed.
References
- Myung S-K, Park Y. Effects of collagen supplements on skin aging: a systematic review and meta-analysis of randomized controlled trials. The American Journal of Medicine, 2025. Reported with the industry responses in Industry reacts to meta-analysis concluding collagen supplements show no proven benefit for skin aging, NutraIngredients
- Grimes DR, Mahto A. Hydrolyzed collagen for skin aging meta-analyses are unreliable. Journal of Cosmetic Dermatology, 2026 (doi: 10.1111/jocd.71208). Reported in Scientists find major flaws in the research behind collagen supplements, ScienceAlert, 7 October 2026
- Analgesic efficacy of collagen peptide in knee osteoarthritis: a meta-analysis of randomized controlled trials. Journal of Orthopaedic Surgery and Research 2023;18(1):694. Summary via OrthoEvidence ACE Report
- Hexsel D, et al. Oral supplementation with specific bioactive collagen peptides improves nail growth and reduces symptoms of brittle nails. Journal of Cosmetic Dermatology, 2017 (doi: 10.1111/jocd.12393). Reported in Collagen supplements may boost nail growth and appearance: Human data, NutraIngredients
- Phillips SM, Tipton KD, van Loon LJC, Verdijk LB, Paddon-Jones D, Close GL. Exceptional body composition changes attributed to collagen peptide supplementation and resistance training in older sarcopenic men. British Journal of Nutrition 2016;116(3):569–570. Record
- Zdzieblik D, et al. Collagen peptide supplementation in combination with resistance training improves body composition and increases muscle strength in elderly sarcopenic men: a randomised controlled trial — CORRIGENDUM. British Journal of Nutrition
- Collagen products for skin, hair and nails: what the evidence shows — ZOE, including dermatologist commentary and the limitations of the nail study
- Rousselot. New fundamental study on bioavailability adds to the literature on the mechanisms behind Peptan’s health benefits — cited as an example of manufacturer-generated absorption research, not as independent evidence
