Collagen

The collagen sold as a powder or in capsules is animal protein, obtained mostly from the skin and cartilage of cattle, pigs, chickens or fish. In its original form it is a long molecule, coiled into a triple helix. To turn it into a supplement, that structure is almost always broken up and cut into small fragments.

Two products with little in common are sold under the same name. The more common one is hydrolyzed collagen, also called collagen peptides, which is taken by the gram, by the spoonful. The other is undenatured type II collagen: it is extracted from the cartilage of the chicken sternum without breaking its structure, and it is taken by the milligram, in a capsule. Each has its own trials, and this guide counts them separately.

In the European Union, collagen and the conventional hydrolyzate of bovine, porcine or avian origin are marketed as a food supplement. European Union food supplement legislation does not require prior authorization of the product by the competent authority before it is placed on the market. In Spain, whoever places it on the market must notify the competent authority (AESAN or the autonomous community) by sending a copy of the label, either beforehand or at the same time. In the United States it is sold as a dietary supplement, also without prior approval. New forms, such as collagen peptides from eggshell membrane, do need authorization as a novel food.

A different mechanism has been proposed for each product, and none is demonstrated in people.

Of the hydrolyzate it is said that some small fragments containing hydroxyproline are absorbed intact and act as a signal on the cells that build the matrix of the skin and the cartilage (fibroblasts and chondrocytes). The idea rests on laboratory and animal studies.

For undenatured type II collagen an immunological mechanism is proposed. A small dose of collagen with its structure intact, on passing through the lymphoid tissue of the gut, would induce oral tolerance: the immune system would learn not to attack the collagen of the body’s own cartilage. The largest trial of this product presents it expressly as a hypothesis and calls for studies to clarify the mechanism.

Skin. In women, trials with collagen peptides find small improvements in hydration and elasticity after 8 to 12 weeks, measured with instruments. When one meta-analysis kept only the high-quality trials that also had no industry funding, the improvement stopped being significant.

Knee osteoarthritis. Undenatured type II collagen, at 40 mg per day, reduced symptoms versus placebo in several trials, all sponsored by the manufacturer. In this guide, hydrolyzed collagen has no primary trial in osteoarthritis.

Bone. A single trial, in postmenopausal women, found higher bone mineral density after a year. It did not measure fractures.

Sport. Combined with strength training, in the trials that measured tendon and muscle in most detail, these grew somewhat more than with placebo, but they did not gain more strength, nor the tendon more stiffness. Recovery after exertion and muscle soreness do not improve across the body of studies.

Sleep. There is a single small trial, and its two primary objectives were not met.

Every primary trial this guide rests on has a company that sells the ingredient behind it. And since none compares collagen against an equivalent protein, it remains unanswered whether what is observed is an effect of collagen or of taking more protein.

Skin

The two trials that opened the field, from 2014, used the same branded ingredient for eight weeks. In the first, 69 women aged 35 to 55 took 2.5 or 5 grams per day, and skin elasticity improved versus placebo at both doses. In the second, with 114 women aged 45 to 65 and 2.5 grams, the volume of the wrinkles around the eyes fell by 20% versus placebo, and the difference was still there four weeks after stopping. In 2024, a trial in 85 Chinese women measured by ultrasound 6.3% more dermal density than with placebo, although its elasticity was only better at four weeks and all of them used the same moisturizer.

Knee osteoarthritis · hydrolyzed collagen

The best candidate from our search, a six-month Spanish trial with 120 patients, had vitamin C in the formula, so it does not allow the effect to be attributed to collagen. There are two Korean placebo-controlled trials, from the same manufacturer, that find improvements versus placebo on the WOMAC questionnaire and no change in joint space. We do not treat them as primary: one is declared exploratory by its own authors, and the funding of the other is unconfirmed.

Knee osteoarthritis · undenatured type II collagen

The largest trial has its card in § 08 (Lugo 2016). Another, in five hospitals in India with 101 patients, found at twelve weeks that total WOMAC fell 32.47 points with collagen versus 13.84 with placebo, on a scale of 0 to 96. The improvement was similar to that of glucosamine with chondroitin, although the trial was not designed to demonstrate that they were equivalent. Both were funded by the company that makes each ingredient.

The only trial without a manufacturer’s money, carried out in Thailand with 68 patients, found no differences from placebo on any outcome. It did not test this product alone but a combination of two types of collagen, and the null result of a mixture says nothing about each component separately.

Bone

The trial has its card in § 08 (König 2018). Besides density, it measured markers of bone turnover: these are all indirect indicators, and so far without an independent replication. A later four-year follow-up had no placebo group and kept fewer than a quarter of the participants.

Nails, hair and cellulite

Nails. The only placebo-controlled trial that studied them measured only their color: at four weeks they were somewhat less yellow. It did not measure growth, hardness or breakage. The one that did measure growth and breakage had no placebo group. Hair. None of the placebo-controlled trials we have reviewed measures its growth; what circulates is indirect evidence or evidence from combined formulas. Cellulite. A six-month trial with 105 women, carried out with the ingredient of the company that makes it, found less cellulite than with placebo, especially in women of normal weight.

Tendon and muscle

In sport, almost all trials combine collagen with the same exercise program in both groups. In the best-measured trial, using magnetic resonance imaging, 15 grams per day over fifteen weeks of strength training increased total muscle volume by 15.7% versus 11.4% with placebo, and maximal strength did not differ between groups. The same happens in the tendon (Jerger 2022 card in § 08).

Recovery and muscle soreness

After an effort designed to damage muscle, the largest trial saw strength and jump evolve better with collagen, but at no measurement point did the groups come to differ, and muscle pain was the same. The review that gathers these studies finds no effect either on recovery or on muscle soreness.

Joints in active people · undenatured type II collagen

In 96 healthy people who play sport and notice knee discomfort, type II collagen improved active knee flexion by 3.23 degrees versus 0.21 with placebo at 24 weeks. A 2026 trial, from a Spanish manufacturer, did not meet its primary objective, the time it takes for discomfort to appear during exercise; its favorable results come from secondary measures and from subgroups.

Sleep and cognition

Thirteen male athletes with sleep problems took 15 grams of collagen or placebo one hour before going to bed, one week of each. The two primary outcomes were how long they took to fall asleep and what share of the night they spent asleep, and neither changed. They woke up fewer times, and on an attention test of 14 responses they got them all right with collagen, versus 97% with placebo: less than half an answer of difference. It is also the only trial of collagen on cognition that we know of.

What the regulator said

The meta-analysis and its dispute. Myung and Park 2025 gathered 23 trials and 1,474 participants up to June 2024, with two quality tools (RoB 2 and the Jadad scale). Its literal conclusion is that there is no clinical evidence supporting the use of collagen supplements to prevent or treat skin aging. The authors declare no conflicts and no funding. A second letter in the same journal supports the finding: the benefits were not evident in high-quality or independent trials.

The umbrella review. Ravindran 2026 gathers 16 systematic reviews with meta-analysis, 113 trials and 7,983 participants, and its results by outcome are in Figure 1. It is the weighty piece with no declared industry link, and the one that declares the quality of its own reviews according to AMSTAR-2: one high, four low and eleven critically low. Its general conclusion speaks of consistent and clinically significant benefits, and coexists with non-significant results in skin roughness, mechanical properties of the tendon, strength recovery and muscle soreness.

Muscle in sarcopenia. The Zdzieblik 2015 trial analyzed only those who completed it (53 of 60), the collagen group started with more fat mass and less fat-free mass, and the placebo was calorie-free silica, so part of the effect could be due to the extra energy. The article states that the analysis was carried out by the investigators independently and, in the same paragraph, attributes the statistical analysis to a co-author from the institute linked to the manufacturer.

Recovery. In Bischof 2023, with 75 men randomized and 55 analyzed, the statistical interaction favored collagen in isometric strength, force production and jump, but the article states that the groups did not differ at any time point. It also found no changes in body composition, with the same dose of 15 grams, the same manufacturer and the same research environment as Zdzieblik 2015.

Type II, range of motion. In the trial of 96 active people, knee extension improved within the collagen group but not versus placebo (p = 0.3115); in the largest osteoarthritis trial, flexion did not differ between groups. Any claim that this product improves knee extension versus placebo rests on a within-group comparison. In the Indian osteoarthritis trial, a seven-day placebo run-in removed those who responded to placebo before the groups were drawn, which widens the expected difference. And a third trial, in Indonesia, reports WOMAC reductions far larger than any other, which we do not reproduce because they have not been verified against its results table.

The 2026 trial. Möller 2026 randomized 75 people, below the 100 planned, did not correct for multiple comparisons, and its abstract does not make clear that the primary objective failed.

The Spanish trial with vitamin C. Carrillo-Norte 2024 reports zero dropouts among 120 patients aged up to 81 over six months, which is unusual in osteoarthritis, and at baseline 71.7% of the placebo group were taking analgesics versus 28.3% of the collagen group, an imbalance it does not adjust for and that bears directly on its primary outcome, pain.

A pattern of reading. In ten articles across the three fields, the abstract is more favorable than its results table. That is why every figure in this guide comes from the tables.

The trials have been done mostly in adult women (skin), in people aged 40 to 75 with knee osteoarthritis, in postmenopausal women (bone) and in men who train for strength, whether young or older. Outside those groups, what this guide describes has not been tested. Pregnancy, breastfeeding, childhood and kidney or liver disease are in § 11.

In sport, the International Olympic Committee consensus on supplements treats it as low risk and with few data, without backing it as a performance aid. The International Society of Sports Nutrition has no position of its own on collagen: in its protein documents it mentions it in passing, as a lower-quality protein for building muscle and poor in leucine.

In the trials, hydrolyzed collagen has been taken at between 2.5 and 15 grams per day, depending on what was being studied. Undenatured type II collagen has been taken at 40 milligrams per day, in one capsule. In skin, the changes appeared at 8–12 weeks. These are the amounts used in the studies, not a recommendation: we have not found any dose or upper limit set by an official body.

Doses used in the trials in this guide. These are protocol doses, not recommended regimens. The table does not select by result: it includes the trials that did not meet their primary objective, flagged as such.

The timings in the table (at night, fasting, alongside training) are conventions of each protocol, not the result of comparing timings against each other.

In the 2014 elasticity trial, the two doses tested, 2.5 and 5 grams, improved versus placebo. In the umbrella review, higher daily doses were associated with better mechanical properties of the tendon, but longer interventions with a smaller effect, a paradox its own authors call for investigating.

Many products combine collagen with vitamin C. The trials that do so do not allow us to know what part of the effect belongs to collagen, and that is why none appears as primary in this guide. Whether the combination adds anything is a question this material does not allow us to answer.

In the Indian osteoarthritis trial, the 40 mg of type II collagen were spread across six daily capsules totaling 2,700 mg, made up with glucose; the placebo group took the same amount of glucose.

Part of the hydrolyzed collagen reaches the blood as fragments of two or three amino acids; the rest, as single amino acids, just like any protein. Whether those fragments act on the skin or the cartilage is, for now, a hypothesis.

The study that demonstrated it, in 2005, gave healthy volunteers between 9.4 and 23 grams of hydrolyzed gelatin on an empty stomach. The most abundant fragment in blood was prolyl-hydroxyproline (Pro-Hyp). The fragments containing hydroxyproline reached their peak between one and two hours later, at 20 to 60 nanomoles per milliliter of plasma, and by four hours they had fallen by half.

Along with Pro-Hyp, six other fragments containing hydroxyproline appeared, among them Ala-Hyp and Pro-Hyp-Gly. A 2023 review indicates that Pro-Hyp and Hyp-Gly are absorbed quickly, remain in plasma for up to 24 hours and are eliminated in the urine.

For undenatured type II collagen there are no absorption data in people among the studies in this guide. Its proposed mechanism acts in the gut and does not depend on its passing into the blood.

No. Gelatin, hydrolyzate and branded peptides are the same material, cut into fragments of different sizes. Undenatured type II collagen is another product, with a different dose and a different proposed mechanism. A good share of the trials has been carried out with branded ingredients, and the authors of one of them warn that their results cannot be extrapolated to other ingredients.

Forms of hydrolyzed collagen and its relatives

“Hydrolyzate” and “collagen peptides” are near-synonyms in commercial use. When the label names a branded ingredient, its composition is standardized by the manufacturer. When it does not name one, the label does not allow you to know which supplier or which specification lies behind it.

Undenatured type II collagen

Its order of magnitude is tens of milligrams, not grams.

A substantial share of the clinical evidence has been generated with patented ingredients from manufacturers that sell to other companies, not with generic collagen.

The content of the capsule depends on how it is measured. The same 40 mg of material appear in the trials as 10 mg, 10.4 mg or 1.2 mg of native type II collagen, depending on the analytical method. The manufacturer attributes this to a change in ELISA sample preparation introduced in 2014, in a letter signed by an employee of its own, which is not an independent verification.

The name misleads in one specific case. There is also a hydrolyzed type II collagen, sold mixed with chondroitin and hyaluronic acid. Despite the name, it is a different product from the one in the type II tables in this guide.

Origin matters for another reason: fish collagen is an allergen (§ 11), and jellyfish collagen is generally considered a novel food in the European Union, with no specific authorization on record as being in force.

A 2025 analysis, paid for by the Italian Ministry of Health, set out to measure metals in fish and jellyfish collagen supplements. Of 65 brands examined, only 4 declared on the label which species the collagen came from, and the study had to limit itself to those. In its 120 samples, none exceeded the European limits for lead, cadmium or mercury, the estimated daily intake stayed below the tolerable one, and neither the jellyfish nor a mackerel skin extract prepared by the authors themselves had detectable metals. Differences between brands were significant only for arsenic, higher in the cod one. The study describes four products, not the market.

For arsenic, the regulation sets no limit at all in food supplements, so no collagen supplement can be declared compliant or non-compliant on arsenic. The study measured total arsenic without separating the inorganic form, which is the toxic one, and calculated the risk as if all of it were, the most cautious assumption. Even so, the estimated intake came to around 16% of the tolerable one.

In the European Pharmacopoeia, the gelatin monograph sets, among other limits, a maximum of 10 ppm of chromium. There is no official monograph for collagen peptides.

Whether it is the collagen or the protein. The placebos in these trials are maltodextrin, silica or other substances without protein. We have found no trial comparing collagen against an equivalent protein in skin, joint or bone, so it is not known whether what is observed belongs to collagen or whether any protein intake would produce it.

A trial without sponsorship. What would change this guide most is an independent trial, in skin or in osteoarthritis, with measurements taken using instruments.

Bone for real. Fractures, and a replication of the only bone density trial.

Time. The skin, joint and bone trials last between two months and a year.

The liver. This guide has not yet consulted the hepatotoxicity databases. Until it does, it neither asserts nor rules out a signal.

Why the tissue grows more than its function improves. The authors of the tendon trial propose three explanations: deposition of material that does not bear load, changes in the material properties of the tendon, or methodological factors. None is proven.

A meta-regression that contradicts itself. In the umbrella review, more dose improves the mechanical properties of the tendon and more duration worsens them, and associations appear that are hard to believe, such as fasting glucose being related to the study’s year of publication. It suggests confounding between variables rather than a biological effect.

Arsenic without a surname. Without distinguishing inorganic arsenic, a low total value can reassure or mislead depending on the real inorganic share, and in marine collagen organic arsenic, which is not toxic, tends to dominate.

Fish allergy. Fish collagen is a recognized allergen, and cases of anaphylaxis have been published after taking hydrolyzed collagen or fish gelatin. In the European Union, fish is an allergen that must be declared on the label, and that includes the marine collagen sold as a supplement.

Pregnancy, breastfeeding, childhood, kidney or liver disease. There are no trials in these groups, and the absence of data is not evidence of safety.

Kidney stones. Collagen is rich in hydroxyproline, which the body converts in part into oxalate, and oxalate excretion in the urine increases with dose. It has not been demonstrated that this causes stones, nor do we know of trials in people who already form them, but it is a reason for anyone who has had them to discuss it with their doctor.

Side effects. In healthy adults and over the short term, the trials show no serious attributable adverse effects; what is most often described is mild digestive discomfort. No trial was designed to study safety.

This guide describes what the published studies have measured. It is not medical advice and does not replace consulting a healthcare professional.

In a metabolic study, 30 grams of gelatin over three days raised urinary oxalate by 43% versus the same amount of whey protein, and glycolate by a factor of 5.3. With loads of 5 and 10 grams the change was significant; with 1 or 2 grams, it was not. Another study, with pure hydroxyproline, found no changes at 500 or 1,000 mg per day; at 2,000 mg oxalate rose without reaching the level of those who form stones, and returned to its starting value four weeks after stopping.

A meta-analysis of 35 trials in osteoarthritis, with 3,165 patients, concludes that collagen derivatives did not increase the risk of dropout or of adverse effects versus control.

There are no drug interactions documented with sufficient quality on record, which describes the lack of data, not the absence of risk.

In competitive sport, the strict liability of the World Anti-Doping Agency applies: the athlete is responsible for any substance detected, wherever it came from. There are no specific data on contamination of collagen with prohibited substances. In general, a 2026 review estimates that around 9–15% of the supplements analyzed in the studies reviewed contained prohibited substances or undeclared drugs, above all in pre-workout, weight loss and muscle gain products; the figure refers to those analyses, not to the market as a whole, and does not include collagen data.