Collagen has drifted a long way from the supplement shelf. It now turns up in coffee creamers, gummies, protein bars and “beauty” blends, sold on two promises that are usually printed side by side: firmer skin and quieter joints. This year’s literature makes those two promises easier to pull apart than the packaging does. The skin evidence has grown into something you can summarise — small effects, measured in weeks, with a lot of caveats attached. The joint evidence was recently tested in the strictest format available, and the result was a null.

What 25 trials of hydrolysed collagen add up to

The largest recent attempt to summarise the skin literature is a systematic review in the European Journal of Clinical Nutrition, covering 25 randomised controlled trials with sample sizes from 13 to 236 participants. Most enrolled healthy volunteers; seven included people with a chronic condition.

Counted trial by trial rather than pooled into one number, skin hydration improved over placebo in 10 of the 15 trials that measured it, elasticity in 10 of 13, and wrinkle depth or volume in 9 of 10. Water loss through the skin — a barrier measurement rather than a cosmetic one — behaved less consistently. Adverse events were uncommon, and the authors describe the supplement as generally well tolerated.

Their own verdict is deliberately modest: “preliminary evidence” that collagen “may offer dermatological benefits”, with high-quality, standardised, adequately powered, longer-term trials still needed. The reason for the caution is stated in the paper. The majority of the trials it included were rated as having a high risk of bias.

Two features of this review are worth knowing before the counts are read as a ruling. Its literature search closed on 1 April 2023, so everything published since sits outside it. And two of the three authors list a sports-nutrition consultancy as their affiliation; the disclosure statement records one author’s employment at a healthcare consultancy and no conflicts specific to the review.

A second synthesis, and a much smaller number

A separate meta-analysis in Frontiers in Medicine, published in March, pooled 19 trials with 1,341 participants and looked at peptides broadly — oral and topical, collagen-derived and synthetic. Hydration and brightness both improved significantly.

Wrinkles did too, but the pooled effect was small: a mean difference of 0.27, with a p-value of 0.04. A subgroup analysis put the effect for oral polypeptides at 1.5 (p = 0.01), which is where most of the signal came from. Elasticity and skin density were not significantly changed.

The authors reported heterogeneity of roughly 100% — statistical shorthand for the trials disagreeing with each other far more than chance would explain. They attribute it to the obvious culprits: different peptide types and molecular weights, doses running from 600 mg to 10 g a day, interventions lasting 4 to 12 weeks, and measurement methods that “were not consistently described”. A device that reads skin hydration is not the same instrument as a dermatologist’s wrinkle scale, and pooling them produces a number that points in a direction without settling a dose. The review was registered in advance (CRD420250652779).

The joint claim, tested against placebo

The knee-pain question got a cleaner test in JB JS Open Access in September. Researchers in Thailand randomised 216 adults with mild-to-moderate knee osteoarthritis to one of four arms: undenatured collagen type II at 40 mg daily, prescription-grade crystalline glucosamine sulfate, diacerein, or placebo, for 24 weeks with follow-up to 36. Two hundred and eight participants (96.3%) completed.

The primary outcome was the total WOMAC score, a 220-point index of pain, stiffness and function where higher is worse. At week 24, none of the three supplements separated from placebo. The differences were −4.7 points for collagen (95% CI −18.6 to 9.1), −8.1 for glucosamine (−21.8 to 5.7) and −12.3 for diacerein (−25.9 to 1.3). Every confidence interval included zero, and every point estimate sat at or below the threshold usually treated as the minimum clinically important difference. Results at 36 weeks were the same story, as were secondary outcomes including pain scores, joint space width and functional performance. Side effects were mild and evenly spread.

The authors’ conclusion is blunt: the findings “do not support their routine use as monotherapy for knee OA”. A few limits are worth stating. The trial was single-country, 90.9% of participants were women, disease was mild to moderate, and 24 weeks is short for anything structural. It also tested a specific product — 40 mg of undenatured type II collagen — which is not the same thing as the hydrolysed peptides in the skin trials, and that distinction gets lost in the aisle.

The bone meta-analysis that reads better in the abstract

A third synthesis, in Frontiers in Nutrition, looked at collagen peptides and musculoskeletal outcomes and found bone mineral density at the femoral neck and spine rising with supplementation. It also found substantial inconsistency between trials on that measure (I² = 80.1%), muscle performance improving with a confidence interval that only just clears zero (standardised mean difference 0.60, 95% CI 0.05 to 1.15), and larger effects when collagen came with calcium and vitamin D.

The authors suggest collagen peptides “could be used as an adjunct therapy for managing osteoporosis”. That is their reading of pooled trial data, not a recommendation from any guideline body, and the paper itself notes that bone mineral density is a surrogate for fracture risk rather than the outcome patients care about.

Where the consumer references still stand

Harvard’s Nutrition Source entry on collagen, last revised in November 2024, is a useful check on the marketing. It explains the biology plainly: collagen is a major structural component of bone, skin, muscle, tendon and cartilage, and with age the collagen network in the deeper skin layers shifts from an organised weave into a disorganised one. Applied to the surface, collagen fibres are too large to pass the skin’s outer layers, and the page notes that research has not shown shorter chains — peptides — do better.

Its line that “human studies are lacking” now trails the trial literature, which is the ordinary lag of a consumer reference page. What still holds is the practical caution: supplements vary in what they actually contain, and a collagen habit can become a reason to skip the behaviours that protect collagen in the first place, such as getting enough sleep and not smoking. The same page is unimpressed by bone broth, which it describes as carrying only small amounts of bone minerals, with studies having found lead and cadmium in samples.

What the label does and does not guarantee

In the United States, dietary supplements do not go through pre-market approval for safety and effectiveness. The FDA’s position is that manufacturers and distributors are responsible for evaluating the safety and labelling of their products before selling them. That is a different arrangement from drugs, and it is why independent testing of supplement contents keeps finding mismatches between the tub and the certificate of analysis.

Reading the aisle

None of this is a scandal. Collagen is a protein, it is generally well tolerated in the trials that studied it, and a small cosmetic effect on skin measurements is a real finding rather than a marketing invention — provided it is described as small, measured over weeks, and inconsistent between trials.

What the record does not support is the stronger claim that travels with the category: that a scoop rebuilds cartilage or resolves an aching knee. The one Level I trial to test that head-to-head against placebo found nothing, and the skin evidence, while larger, is still dominated by trials its own reviewers rated at high risk of bias. Anyone weighing a supplement against a diagnosed joint problem — or thinking of it in place of treatment — is better served by a conversation with a healthcare professional than by a label, and our medical disclaimer sets out the limits of what this site can tell you. For context on how similar checks have gone elsewhere in the supplement aisle, see our reads on probiotics and on the greens-powder trial that tested safety rather than benefit.