Collagen peptides

Not all collagen is equal. Hydrolyzed peptides differ by molecular weight and peptide profile, and the trials that show benefit used specific preparations at specific doses — details most supplement labels skip entirely.

What each preparation was trialled for

The brand names are not marketing decoration on an interchangeable powder. Each is a distinct peptide fraction with its own set of trials, its own endpoints, and its own strength of evidence. Anyone taking a blend is taking three literatures at once, and they are not equally strong. This is the map.

FORTIGEL — the knee fraction. This is the preparation behind the cartilage claim, and its evidence is the best of the three. Two randomized placebo-controlled trials in physically active people with activity-related knee discomfort — 139 people over 12 weeks, then a deliberate replication in 180 — both found activity-related knee pain falling further on 5 g daily than on placebo, with pain at rest unchanged in both. A third trial extended the result out of the young-athlete population it was built on: 182 adults across a wide age range with functional knee and hip pain reported less pain climbing stairs and kneeling, and physicians rated pain lower at rest and during walking. An earlier 24-week trial of 147 varsity athletes on 10 g of collagen hydrolysate pointed the same way, largest in the knee-pain subgroup. In the replication, the between-group difference was around 6 mm on a 100 mm scale.

TENDOFORTE — the tendon and ligament fraction. Tendon is not cartilage, and this literature stands separately. In chronic Achilles tendinopathy, 2.5 g twice daily alongside calf strengthening produced a 12.6-point VISA-A gain against 5.3 on exercise alone over three months — though by six months the two orders of that crossover had reached the same score, so what the trial shows is a faster arrival rather than a better destination. In 50 athletes with chronic ankle instability, six months of 5 g daily improved perceived ankle stability on two validated questionnaires and reduced re-injuries over the following three months — while instrumented ankle laxity, the objective measure in the same trial, did not differ at all, and the collagen group had begun the trial significantly worse off on every questionnaire it later improved on. On structure the wider tendon literature — not all of it using this preparation — divides by laboratory. Two Freiburg resistance-training trials found tendon getting bigger without getting stiffer: Achilles cross-sectional area up 11.0% against 4.7% on placebo, and patellar tendon area up regionally, with stiffness and strength rising equally in both arms. A 16-week trial without any training programme found the opposite pattern — Achilles and calf stiffness rose on 10 g daily while cross-sectional area did not move. And an independent 15-week MRI trial found neither, with placebo numerically ahead on every mechanical measure.

VERISOL — the nail and skin fraction. This is where the visible effects live, and where the design quality drops. The nail evidence is a single study: 25 women took 2.5 g daily for 24 weeks, nail growth rate rose 12%, broken nails fell 42%, and 64% reached a physician-rated global improvement. It was open-label and single-centre with no placebo group and no control arm of any kind, so every comparison is against the participants' own baseline. The skin evidence is much better built. A double-blind placebo-controlled trial in 114 women aged 45–65 found eye wrinkle volume down about 20% against placebo at eight weeks, with suction-blister biopsies showing 65% more procollagen type I and 18% more elastin, from arms that were unbalanced before treatment and are therefore reported against each group's own baseline. A six-month trial in 105 women lowered cellulite scores against placebo across the whole group, most in normal-weight participants; dermal density separated because the placebo group's fell over the six months and the treated group's did not. The nearest thing to a controlled nail result comes from a different manufacturer's peptide: in 85 East Asian women, nail colour improved faster on 5 g daily than on placebo, though the placebo group reached the same point by twelve weeks.

The same manufacturer has since put out further fractions — one aimed at the skin barrier, which in a placebo-controlled trial left people with moderate atopic dermatitis reaching for 0.94 g of topical corticosteroid against 5.60 g on placebo — so the list is still growing.

Two things follow from the map. The fractions have never been tested as a blend, so the evidence for a mixture is the sum of separate literatures rather than a study of the thing in the tub. And the fraction with the most visible effect has the weakest design behind it.

Dose

The joint trials used 5–10 g daily for at least 12 weeks. The tendon trials spread wider, and where they land turns out to matter — see the tendon entry. Almost nothing in the field has gone higher with a musculoskeletal endpoint, and the exceptions are worth naming precisely, because the gap between what is sold and what is tested widens fast above 15 g.

Three trials have exceeded it. An independent trial gave 15 g twice daily — 30 g/day, split, for a week of intense resistance training — and measured muscle connective-tissue synthesis with deuterated water and biopsies. Absorption worked: plasma glycine, proline and hydroxyproline all rose above placebo. Synthesis did not follow. Connective-tissue synthesis was 1.97%/day on collagen against 2.00%/day on placebo, and myofibrillar synthesis 1.34% against 1.34%. A second gave 30 g twice daily — 60 g/day, the highest dose ever given in a controlled human trial — to older women for six days, and found muscle protein synthesis rising with whey at every comparison and not rising above baseline with collagen. A third randomized middle-aged active adults to 10 or 20 g daily for nine months with a knee questionnaire: of its five knee subscales only daily-living separated the groups at six months, and no single arm changed significantly within it; pain separated only among participants exercising more than about three hours a week; the 20 g arm improved a general physical-health score in women only; and no endpoint ordered itself by dose.

So the answer about 30 g/day is that no trial has tested it on a joint. The one trial at that dose measured connective tissue in muscle and found nothing beyond placebo; the one trial above it measured muscle and found the same; and the only randomized dose comparison with a knee endpoint did not favour the higher arm. The dose-ordering that does exist comes from horses, where 50 g/day beat 25 g/day on veterinary-rated lameness — but a 450 kg horse on 50 g is taking about 0.1 g/kg, which is what 7 g/day is for a person. Where two doses have been compared head to head in humans, in a skin trial testing 2.5 g against 5 g, both beat placebo and neither beat the other.

There is one consequence of a large daily dose that is measurable rather than speculative. Hydroxyproline, the amino acid that makes collagen collagen, is a precursor of oxalate. In a controlled feeding study, 30 g of gelatin a day against a matched 30 g of whey raised urinary glycolate 5.3-fold and urinary oxalate 43%, with the threshold for a detectable change sitting between 2 g and 5 g. A second, much smaller study at the same 30 g dose found glycolate rising as expected but oxalate and calcium-oxalate supersaturation unchanged, so the finding is not settled. Neither study measured a stone. What they establish is that a gram-scale collagen habit delivers real oxalate substrate, which is worth knowing by anyone who has formed a calcium-oxalate stone and is separately taking gram-level vitamin C.

Timing, and the vitamin C pairing

Timing may matter more than most labels admit: collagen-synthesis markers peak when ingestion precedes loading exercise by about an hour, which is the basis of the pre-training protocol. Splitting the dose has a pharmacokinetic rationale rather than a trial behind it — blood hydroxyproline peptides peak about an hour after a dose and are back to baseline within 24 hours, so nothing accumulates between doses, and the one trial that used a twice-daily schedule used it for convenience rather than to test it.

Vitamin C is a required cofactor for the enzymes that hydroxylate collagen. In the gelatin-timing study, 15 g of vitamin C–enriched gelatin taken one hour before six minutes of rope skipping doubled serum PINP, a marker of collagen synthesis. It was a tiny crossover study using a surrogate marker — promising mechanism, not proof of joint repair. The trials that used vitamin C alongside collagen used about 50 mg of it.

Source and molecular weight matter less than either side of the marketing argument suggests, but not nothing. A crossover study of fish, porcine and bovine hydrolysates at two molecular weights found free hydroxyproline uptake comparable across all of them; total hydroxyproline and Pro-Hyp did separate the products, with porcine ahead of fish and the lower molecular weight ahead of the higher, by factors of 1.2 to 1.4 in six people. Intact gelatin raised plasma Pro-Hyp and Hyp-Gly nearly as well as hydrolysate.

What the whole literature looks like

Individual trials are a bad way to read this field, because there are enough of them now to pool. The best synthesis is a 2024 trial-sequential meta-analysis of 35 randomized trials in 3,165 patients: collagen derivatives improved pain and function with moderate-to-high certainty, and the sequential analysis confirmed enough data has accumulated to call it settled. That is a genuine positive result and this entry's strong tier rests on it.

Read the size, though, not just the sign. The pain effect is a standardized mean difference of −0.35 and the function effect −0.31 — small, and below the 0.5 usually treated as the threshold for clinical meaningfulness. "Real" and "worth noticing" are different findings, and this is the first without being clearly the second.

Who ran these trials

The supporting evidence is often described as "multiple independent randomized controlled trials." Independent is the wrong word: the positive trials cluster tightly around the people selling the ingredient. The 5 g specific-collagen-peptide trials come from the Collagen Research Institute with GELITA support, the 10 g hydrolysate trial used a GELITA product with a Gelita Health co-author, and other positive trials list authors employed by Nitta Gelatin, Suntory, and Lonza.

The trials with no commercial stake are mostly the ones that came back null. An independent placebo-controlled RCT of combined undenatured and hydrolysed collagen found no difference from placebo on pain, function, rescue medication or satisfaction — its authors said plainly they ran it because the existing literature was "most being industry-sponsored." A 15-week trial measuring tendon adaptation by MRI found nothing on any outcome. An isotope-tracer study with muscle biopsies found collagen did not raise connective-tissue synthesis above placebo. So did a second tracer trial at twice the dose, and a third comparing collagen against whey.

The pattern has no exception at present. The nearest candidate — nine months, two doses, a knee questionnaire, and a declaration of no competing interests — names the collagen manufacturer in its acknowledgements as the study's funder and the supplier of the product tested. Its findings also sit inside subgroups defined by sex and weekly exercise volume, which is where false positives live. So every positive randomized knee result in this literature has a manufacturer behind it, and the trials without one are the null ones.

None of that makes the meta-analytic result false. It does mean this is a small real effect in a literature with a strong funding signature.

The one imaging trial

Almost no collagen trial has looked at cartilage structure directly, and the one exception is worth stating precisely.

In 2011, thirty people with mild knee osteoarthritis were randomized to collagen hydrolysate or placebo and scanned with dGEMRIC — delayed gadolinium-enhanced MRI, a validated proxy for cartilage proteoglycan content. In the medial and lateral tibial regions, dGEMRIC scores rose on collagen (median +29 and +41 ms) and fell on placebo, reaching significance at 24 weeks.

That is the entire structural evidence base for ingested collagen in human cartilage. It should be read with everything else the same trial found:

• Significant in two of six regions of interest. The other four — femoral and patellar — showed nothing.
• Every T2 measure was null.
• Every clinical outcome was null. Nobody felt better, moved better, or used fewer analgesics.
• The authors call it a pilot and write only that the technique "may be able to detect change."
• It was registered as NCT00536302 with GELITA as lead sponsor, and a Gelita Health co-author.
• It has never been replicated in fifteen years, in a field with substantial commercial motivation to do so.

A follow-up re-analysed serum biomarkers from the same thirty people. The type II procollagen marker PIIANP rose more on collagen (29.9 versus 1.2 percent) — but the paper reports that consistent correlations between the biomarker changes and the dGEMRIC changes "could not be observed." The imaging and the biochemistry did not track each other, which weakens rather than strengthens the mechanistic story.

So: one small industry-sponsored pilot, positive in a third of its imaging regions and null on everything a patient would notice. That is not nothing, and it is a long way from cartilage regrowth.

The veterinary joint trials

Human collagen trials rest almost entirely on questionnaires, and animals cannot fill one in. That makes the two veterinary studies more interesting than their size suggests, because both used measurement instead.

Thirty-one client-owned dogs with naturally occurring osteoarthritis walked over force plates set into a treadmill before and after twelve weeks of bioactive collagen peptides, an approved omega-3 and vitamin E combination, or placebo. Peak vertical force on the affected limb improved on collagen and fell on placebo, and the difference between them was significant; the other two gait measures did not change in any group. Owner-rated quality of life on a validated pain inventory improved more on collagen than on the omega-3 comparator, but not more than on placebo, where it also improved. Accelerometry — total daily activity, the one endpoint that is neither owner-reported nor laboratory-measured — was unaffected by any treatment. The placebo arm held nine dogs, and the study was funded by the ingredient manufacturer with one of its employees as an author.

Thirty-eight horses with mild to moderate osteoarthritis took 25 or 50 g a day for twelve weeks under veterinary orthopaedic examination. The higher dose did more, on six of eight rated parameters. That is the only dose-ordering with a joint endpoint anywhere in this literature, and it comes with the caveat that only one of the two centres had a placebo arm and the dose comparison sat in the other.

The limits of these trials

No collagen trial has demonstrated regrowth of articular cartilage, and no trial has shown a structural change accompanied by a clinical one. The randomized evidence supports less activity-related pain and better tendon-loading tolerance — meaningful, but a different claim. The subjective endpoints are exactly the kind most vulnerable to placebo effects.

What the mechanism work does support is narrower and more interesting than the marketing: gelatin is digested to the dipeptide Pro-Hyp, which reaches articular cartilage in animals and roughly doubles aggrecan expression in chondrocytes — while free proline and hydroxyproline do nothing. So if collagen works, it probably works as a signal rather than as building material. That is a mechanism worth testing properly, and the trial that would settle it — adequately powered, independently funded, with a structural endpoint and a dose above 10 g — is a study anyone could run.