Collagen for tendon (not cartilage)

This entry exists to separate two claims the marketing merges. Tendon and cartilage are different tissues with different collagen, different vasculature and different turnover, and most of the evidence people cite for "collagen builds connective tissue" was collected in tendon and ligament.

That evidence is worth understanding on its own terms — partly because it is better designed than the cartilage literature, and partly because that is exactly why it has produced clearer negative results.

The founding result

Eight healthy men, a randomized double-blind crossover, and 5 or 15 g of vitamin C–enriched gelatin taken an hour before six minutes of rope skipping, three times a day for three days.

The 15 g dose doubled serum PINP, a marker of type I collagen synthesis. Serum taken from the supplemented subjects, applied to engineered ligaments in a dish, increased their collagen content and improved their mechanics.

It is an elegant study and it generated the whole pre-training timing protocol. It also has an n of eight, uses a systemic circulating marker that reports whole-body collagen synthesis rather than anything tissue-specific, and its tissue arm was in vitro.

The lab's own replication

Two years later the same group ran it again in ten recreationally active men, comparing placebo against 15 g of vitamin C–enriched gelatin, hydrolysed collagen, or a gummy containing both.

Their finding, in their own words: PINP "tended to increase ∼20% from baseline in the gelatin and HC interventions but not the placebo or gummy. However, large variability was observed, which precluded significance for any treatment."

The gummy formulation — the convenient consumer format — was inert.

That is not a refutation, and the direction was preserved. But when the originating lab cannot reach significance on its own endpoint at the same dose, the effect is smaller or noisier than the first paper suggested.

When the endpoint moved to tissue

Two independent trials asked whether any of this shows up where it would have to.

Fifteen weeks, MRI, resistance training. Thirty-nine young men took 15 g of specific bioactive collagen peptides or placebo daily through a standardized three-times-weekly lower-body programme, with patellar tendon cross-sectional area, aponeurosis area and mechanical properties measured before and after.

The result was nothing. Verbatim: "No between-group differences were detected for any of the tendinous tissue adaptations." The p-values ranged from 0.365 to 0.877. Both groups adapted — and placebo was numerically higher on every mechanical measure, with tendon stiffness up 20.9 percent against 17.3 percent on collagen.

Isotope tracing with biopsies. Forty-five recreational athletes received infused labelled phenylalanine, did resistance exercise, then took 30 g whey, 30 g collagen, or nothing, with muscle biopsies over five hours. This is the gold-standard method — it measures the fractional synthesis rate of connective protein directly rather than inferring it from a blood marker.

Muscle connective protein synthesis: 0.072 %/h on whey, 0.068 on collagen, 0.058 on placebo, p = 0.09. No significant difference between groups. Whey did raise myofibrillar synthesis; collagen raised nothing above placebo.

What survives

The positive tendon trials are real but come with a pattern. A 14-week trial found Achilles tendon cross-sectional area up 11.0 percent on 5 g collagen peptides against 4.7 percent on placebo — its stated primary endpoint — but tendon stiffness and muscle strength were null between groups, and stiffness in fact rose about twice as far on placebo. The collagen group also began with the smaller tendon and ended at roughly the size the placebo group started at, which is part of why its percentage gain is the larger one. Thirteen of the 40 men who started did not finish. The study comes from the manufacturer-affiliated group in Freiburg, was part-funded by GELITA, and has a Collagen Research Institute co-author who holds collagen-peptide patents.

The same group found the same shape at the knee. Fourteen weeks of high-load knee-extensor training with 5 g of collagen peptides increased patellar tendon cross-sectional area more than placebo at two measurement sites — while stiffness, muscle cross-sectional area and maximal strength all rose equally in both arms. Size again, without the mechanical consequence.

One trial reports the mirror image, and it is worth the attention for being the rare tendon study with no training co-intervention. Sedentary young men taking 10 g daily for sixteen weeks gained Achilles tendon stiffness (d = 0.38) and gastrocnemius stiffness (d = 0.59), with a matching rise in rate of torque development, while cross-sectional area moved in neither group. Stiffness without size, from a manufacturer's research group, and reported as within-group changes rather than a clean between-arm comparison.

Put together, the positive trials do not agree with each other about what collagen changes. One set finds a tendon that is bigger but no stiffer; the other finds one that is stiffer but no bigger; the independent MRI trial finds neither. That is a literature with a signal in it, not yet a mechanism.

Two independent trials at a higher dose

Then the picture moved. Two trials from the same independent group, both declaring no conflicts of interest, tested 30 g a day — double the highest dose in any of the null trials — taken with vitamin C alongside training.

Twenty middle-aged men, twelve weeks: patellar tendon cross-sectional area rose 6.8 mm² against 1.2 on placebo (p = 0.027), stiffness 661 N/mm against 247 (p = 0.009), and Young's modulus 0.21 GPa against 0.09 (p = 0.018). All three structural measures moved together, which none of the earlier trials achieved. Strength and muscle thickness gains did not differ.

Eleven professional female footballers, ten weeks: stiffness rose 15.4 percent against 4.6 (p = 0.002) and Young's modulus 14.2 percent against 3.4 (p = 0.004), while cross-sectional area showed a training effect with no group difference at all. Stiffer, not larger.

These are nine, eleven, six and five people per arm, and effect sizes from samples that small carry very wide uncertainty whatever the p-values say. But the funding story that organised the earlier sections does not explain them. Dose comes closer: every trial that moved tendon stiffness alongside training used 30 g, and none of the 5 to 15 g training trials moved it. The pattern is not clean, though, and the exception is two paragraphs above — the one trial run with no training at all raised stiffness on 10 g. Nor were the 5 g trials null throughout; they made tendons measurably bigger. What no trial has done is randomise dose against a tendon mechanical endpoint, which is the experiment this question now needs.

Ligament shows a related split. Fifty athletes with chronic ankle instability took 5 g daily for six months — the longest randomized collagen exposure in this corpus. Perceived stability improved on both the Cumberland Ankle Instability Tool and the Foot and Ankle Ability Measure (both p < 0.001), and re-injuries were fewer over the following three months. Instrumented ankle laxity did not differ at all.

That trial needs one more fact beside it, and the paper puts it first among its own limitations: randomisation left the collagen group markedly worse off at the start on all three questionnaires, and by the end it had arrived at roughly where the placebo group began. The placebo group, already close to the top of the scales, had little room to move. The comparison was never adjusted for that.

The most clinically interesting result is in injured tendon rather than healthy tendon: a crossover pilot in twenty people with chronic mid-portion Achilles tendinopathy found collagen peptides plus calf strengthening improved VISA-A by 12.6 points against 5.3 for exercise alone over the first three months. By six months both orders had arrived at the same score, which is why the authors describe the peptides as accelerating the exercise programme rather than adding to it, and the trial's other measure — tendon microvascularity — fell equally in both. Twenty people, and a pilot funded by the manufacturer. But it is the one design where a substrate effect might plausibly matter, because the tissue is actually trying to repair.

What this means for the cartilage question

Two things, and they cut opposite ways.

The tendon literature is the most rigorous test the collagen hypothesis has faced, because tendon is biopsiable and imageable in ways articular cartilage is not. What it has produced is a rough dose gradient: tendon size responding at the doses the joint trials use, tendon mechanics responding at several times them, in samples small enough that the next trial could go either way. Nobody has run that test for cartilage at any dose.

But tendon is not cartilage. It is type I collagen, better vascularised, and adapts to mechanical load through different signalling. Pro-Hyp's measured effects — aggrecan expression, glycosaminoglycan deposition — are chondrocyte behaviours with no tendon equivalent. A null result in tendon does not mechanically transfer to cartilage any more than a positive one would.

The only tissue-level evidence of a collagen effect in humans comes from tendon, at a dose few people take, in trials of eleven and twenty people — and that a dose-ranging trial with a tendon mechanical endpoint would settle more than another symptom study ever will.