You have a knee that complains on the stairs, and last night you ended up reading about collagen. Then about glucosamine. Then about a tub costing thirty euros that promises to "nourish the cartilage". An acquaintance swears they fixed it that way. When it came to putting this article on the site, you stopped, and you were right to: on this topic the scientific literature exists, it is abundant, and it is almost entirely against that tub.
Here is the short version first. Of the three best-selling supplements for knees, none has solid evidence of reducing pain. The one dietary factor that really works is not sold in a bottle: body weight. And even there, the effect is real but smaller than you are usually told. Below you will find all the numbers, with the limitations of each study written next to them.
Glucosamine and chondroitin: -0.4 cm on a 10 cm scale
In 2010 the BMJ published a Bayesian network meta-analysis that pooled 10 trials and 3,803 patients with knee or hip osteoarthritis, including only large studies, with more than 200 patients each (PMID 20847017).
The most important thing is what they did before looking at the results: they decided which difference would actually matter. The threshold was -0.9 cm on a pain scale from 0 to 10 cm. Below that value an effect may exist on paper without being enough for you to say "I feel better".
Here is what they found, compared with placebo:
- glucosamine: -0.4 cm (95% credibility interval: from -0.7 to -0.1)
- chondroitin: -0.3 cm (from -0.7 to 0.0)
- the two combined: -0.5 cm (from -0.9 to 0.0)
None of the three intervals reaches the threshold of relevance. On narrowing of the joint space, meaning the gap between the bones visible on an X-ray, the differences are "minute" and the intervals include zero. The authors' conclusion is explicit: glucosamine, chondroitin and their combination do not reduce joint pain and do not affect the narrowing of the joint space. They go as far as writing that health systems and insurers should not reimburse them.
One detail worth knowing: trials independent of industry showed smaller effects than those funded commercially (P = 0.02 for the interaction).
The limitations, stated clearly. Only trials with more than 200 patients were included, and most of the literature on these two supplements consists of small studies, which are left out here. Many comparisons are indirect, reconstructed within a statistical model rather than measured head to head. And the literature search stops in June 2009: this is an older study.
It is not an isolated voice, however. In 2018 the British Journal of Sports Medicine reviewed 20 supplements across 69 studies (PMID 29018060): glucosamine and chondroitin were found to be "ineffective or with small effects of questionable clinical importance". But the most useful finding of that review is another one, and it is the one no advertisement shows you: no supplement, not even those with large effects in the first weeks, maintains a clinically relevant effect on pain in the medium and long term. Not "a small effect": none.
The limitations. The authors themselves qualify everything: "the quality of the evidence was very low". The supplements showing large short-term effects have been studied in a few trials and with few participants. The search stops in April 2017.
And international guidelines do not agree with each other: this is worth saying rather than hiding. A comparison published in Nature Reviews Rheumatology in 2021 (PMID 33116279) sets OARSI and ESCEO side by side: ESCEO recommends pharmaceutical-grade glucosamine and chondroitin sulfate, while OARSI strongly advises against their use. Both start from the same studies. It is a narrative review, not a meta-analysis, and members of both panels appear among the authors: it should be read as a snapshot of the disagreement, not as proof. When two groups of experts read the same research and reach opposite conclusions, the signal is not "it works" or "it does not work": it is that the effect, assuming it exists, is too small to bring those who examine it into agreement.
One final clarification, because it is turned upside down by those who sell these products. The 2019 ACR/Arthritis Foundation guideline (PMID 31908149) mentions chondroitin sulfate once only: as a conditional recommendation and only for hand osteoarthritis. Not for the knee.
Collagen: the meta-analysis that says yes, read to the end
It would be dishonest to write that no positive study exists on collagen. They exist, and they are recent. The point is what happens when you open them.
In 2025 Clinical and Experimental Rheumatology published an updated systematic review of 11 randomised trials and 870 participants (PMID 39212129). Result: a significant improvement in function (mean difference -6.46) and in pain (-13.63), in favour of collagen. The authors conclude that oral supplementation relieves osteoarthritis symptoms in a way that is statistically and clinically relevant.
In the same abstract, however, there are two numbers that change everything: I² = 88% for pain and I² = 75% for function. I² measures how much the studies disagree with each other. A value of 88% means that almost all of the variability between results is not due to chance: the studies are measuring different things, in different people, with different tools. Averaging them produces a number that describes none of them well. You can see it in the confidence interval for pain too: from -20.67 to -6.58, the distance between "almost imperceptible" and "enormous". And 870 people divided among 11 studies means fewer than 80 participants per study.
Then there is the earlier meta-analysis, from the same research group, published in International Orthopaedics in 2019 (PMID 30368550). The total WOMAC score improves (-8.00; from -13.04 to -2.95; p = 0.002). But when they break it down into its parts:
- pain: -0.22, interval from -1.58 to +1.13, p = 0.75. Not significant. The interval includes zero and also includes a worsening.
- functional limitation: -0.62, p = 0.81. Not significant.
- stiffness: -0.41 (from -0.74 to -0.08, p = 0.01). Significant, but of tiny size.
- VAS: -16.57 (from -26.24 to -6.89). Significant.
Put simply: the overall score improves, but the pain component, exactly the one you would be buying the tub for, does not reach significance.
The limitations, in both directions. The 2019 review does not declare in its abstract how many studies and how many patients it included, and does not report the numerical heterogeneity, the risk of bias or the funding. The 2025 one has no dedicated section on limitations, does not say who funded the included trials, and does not even specify which pain scale was used. Anyone selling you collagen will quote the 2025 conclusion and not the 2019 subscores. Now you have seen both, and you can decide for yourself.
None of these studies, it should be said, shows rebuilt cartilage. They report symptom scores.
Why a supplement does not reach where you think
The mechanism is more mundane than it is made out to be. Collagen is a protein: when you swallow it, digestion breaks it down into amino acids, which the body then uses wherever it needs them. Glucosamine is a modified sugar and meets a similar fate. None of these molecules carries a postal address reading "right knee". Eating cartilage is no more direct than eating nails to make your nails grow.
There is then the second half of the problem, which concerns how we judge whether something has worked. In the 2013 JAMA trial on vitamin D (PMID 23299607), knee pain fell in both groups: -2.31 in those taking vitamin D and -1.46 in those taking placebo. Hundreds of people took an inert pill for two years, and at the end their knee hurt them less.
That is why "my cousin took it and feels better" proves nothing. Osteoarthritis pain comes and goes on its own, and the very act of doing something has a measurable effect. To separate the product from expectation you need a comparison with placebo, and it needs to be blinded. The same reasoning applies to much of what is read about natural remedies for osteoarthritis.
Vitamin D: only if you are truly deficient
This is the chapter where the answer changes depending on your blood test, so it needs explaining in order.
The most solid trial is the VIDEO study, published in JAMA in 2016 (PMID 26954409): 413 participants with knee osteoarthritis, already selected for having low vitamin D (25-hydroxyvitamin D between 12.5 and 60 nmol/L), randomised in double blind to 50,000 IU per month or placebo for two years. The blood level rose markedly in the treated group, as expected. On the rest: no significant difference in the annual change in tibial cartilage volume, no difference in the WOMAC pain score, no difference in cartilage defects or bone marrow lesions.
The safety data should also be reported, because supplementing is not free: adverse events in 56 participants in the vitamin D group versus 37 in the placebo group (P = 0.04).
The 2013 trial mentioned above says the same thing in 146 people: no reduction in pain, and cartilage lost to the same extent in both groups (-4.30% versus -4.25%, P = 0.96).
The "but", and it must be told in full. In 2025 a post hoc analysis of the same VIDEO study, in Rheumatology Advances in Practice (PMID 40677617), looked for subgroups: below a threshold of 43 nmol/L the total WOMAC score improved more than with placebo (-256.41 versus -72.10, P = 0.0060), while above that threshold there seemed to be an effect on cartilage and not on symptoms.
The limitations here are decisive. It is a post hoc analysis: the subgroups were sought after seeing that the main trial was negative. The 43 nmol/L threshold was not decided in advance; the data themselves found it with a classification algorithm, and this exposes the analysis to a concrete risk of overfitting. The authors use the conditional: "could alleviate", "might improve". And the result is internally odd, because the benefit on symptoms appears below 43 nmol/L and the one on cartilage above it, a divergence with no obvious biological explanation.
The honest reading is this: if you have a deficiency documented by a blood test, correcting it makes sense for many reasons and perhaps for the knee too. Taking it "because it is good for the joints" without knowing your level, no. And the level is set by a doctor, not by an article.
Body weight: the only lever with serious evidence
Here everything changes, but less than you might expect.
The figure you will probably already have heard, "1 kg less is worth 4 kg less load", comes from a study published in Arthritis and Rheumatism in 2005 (PMID 15986358): 142 sedentary adults, overweight or obese, with documented knee osteoarthritis, analysed with three-dimensional gait analysis. A 1 kg reduction in weight corresponded to reductions of 40.6 N and 38.7 N in the compressive and resultant forces at the knee: about 4 units less load for every unit of weight, at every step. Multiplied by the thousands of steps in a day, that is not small.
The limitations, stated clearly. It is an association estimated with regression models in 142 people, not a clinical outcome. It measures the load during walking, not pain: from it you cannot deduce how much less it will hurt. The sample is highly selected and does not extend to people of normal weight or to the young. And the authors themselves use the conditional about clinical relevance.
The next step was taken by the same researchers in the IDEA trial, published in JAMA in 2013 (PMID 24065013): 454 adults over 55, 18 months, three groups (diet plus exercise, diet alone, exercise alone). Those doing diet plus exercise lost an average of 10.6 kg. Knee loads were lower in those on diet (2,487 N versus 2,687 N for exercise alone, a difference of 200 N, P = 0.007), and interleukin-6, a marker of inflammation, was lower in both diet groups. On pain, measured on a scale from 0 to 20: 3.6 in the diet-plus-exercise group, against 4.8 for diet alone and 4.7 for exercise alone.
The limitations. The main outcomes of the study were mechanical and inflammatory: pain was a secondary outcome, and the difference of 1.02 points out of 20 is modest. The study was single blind, which is unavoidable when the intervention is following a diet and going to the gym. Above all, diet and training were supervised for 18 months inside a university centre. It is also worth noting that diet alone lowered the load, but pain improved only where movement was added to the diet.
And outside the university centre? They measured that too, in JAMA in 2022 (PMID 36511925): 823 people, the same kind of programme but delivered in urban and rural counties, with a blinded assessor. Weight fell: -7.7 kg against -1.7 kg in the control group, a difference of 6 kg. Pain, the primary outcome of the study: WOMAC 5.0 versus 5.5, a difference of -0.6 points on a 0 to 20 scale, where the authors had pre-set the minimal clinically important difference at 1.6. Their conclusion, verbatim, is that the difference is statistically significant but small and of "uncertain clinical importance".
This is the number that keeps all the rest of this article honest. Weight remains the dietary lever with the best evidence: the ACR/Arthritis Foundation places weight loss in overweight or obese people with knee or hip osteoarthritis among the strong recommendations, alongside exercise, and OARSI in 2019 (PMID 31278997) lists among the basic treatments for the knee education and structured land-based exercise "with or without dietary weight management": weight alongside movement, not instead of it, and no supplement on the list. But "the best lever" does not mean "the knee stops hurting". If you want the full picture on body weight and the knee, that is where we covered it.
One finding from the 2022 study that is usually not reported: the study-related adverse events were almost all in the active arm (9 injuries out of 10, 6 strains out of 7, all 6 falls). Moving more is not zero risk, and that is exactly why the load should be increased by a criterion and not by feel, as we try to explain when talking about safe exercise when you are starting from overweight.
"Anti-inflammatory diet": what lies beneath that phrase
In 2025 the European Journal of Clinical Nutrition published a systematic review of 9 randomised trials and 898 participants on dietary interventions in osteoarthritis (PMID 40295866). Overall, diets improved pain (SMD -0.67), function (-0.62) and weight (-3.18 kg).
Then you look inside, and you understand who is doing the work:
- reduced-calorie diets: pain SMD -0.85, function -0.95, weight -3.13 kg. All with p below 0.0001.
- Mediterranean diet: pain SMD -0.27, interval from -1.14 to +0.60, p = 0.54. Function -0.28, p = 0.29. Not significant on either.
What gets measured, when a diet benefits the knee, is the reduction in calories. Not a list of foods that "switch off inflammation".
The limitations, and here they are large. There are very few trials for each type of diet: 2 on the Mediterranean diet, 1 on an anti-inflammatory diet alone, 1 low-carb, 1 plant-based. An interval running from -1.14 to +0.60 does not say "the Mediterranean diet is not useful": it says "on this outcome it is not demonstrated", which is a different thing. The authors themselves conclude that individual eating patterns "are promising but require further investigation".
What to do, in order of importance
1. If you are overweight, weight. It is the only dietary factor with strong recommendations in the guidelines. Expect a realistic order of magnitude: in the programme closest to real life, 18 months produced a difference of about 6 kg and a small effect on pain. The load per step, however, really does come down, and the rest of the body is grateful.
2. Movement, which is not diet but is what in the trials moves the pain. In the IDEA trial, pain improved only in the group that added exercise to the diet, not in diet alone. That is why no serious programme for an osteoarthritic knee consists only of what you put on your plate: here is how to set up movement when you have knee osteoarthritis.
3. Calories, not individual foods. If you choose an eating pattern, choose it because you can keep it up for months, not because a list of foods is defined as anti-inflammatory. In the data, the pattern matters less than the deficit.
4. Vitamin D only with a blood test in hand and the advice of a doctor. Not "by eye", not "it cannot hurt": in the largest trial, adverse events were more frequent in those taking it.
What is not worth it. Glucosamine, chondroitin and collagen, if you are buying them for pain. Of the three, collagen has at least some positive meta-analysis, but with 88% heterogeneity and with the pain subscore not significant when isolated. If you decide to try them anyway, do so knowing you are buying a small, unproven possibility, and set a time limit and a spending limit on the trial instead of renewing the order for years. And if you take other medicines, talk to your doctor or pharmacist first: "natural" does not mean "free of interactions".
The same goes for the other question that always comes up, namely whether all of this can cure osteoarthritis. The short answer is no, and anyone promising it to you in a jar is selling you something.
When you need a doctor
Diet has nothing to do with these situations, and none of them should be managed by reading an article:
- a knee that is warm, red and swollen together with fever or general malaise: this picture can indicate a joint infection and needs to be seen the same day.
- sudden swelling that appears within hours for no apparent reason.
- a knee that really locks, that stays stuck and will not straighten, or that gives way suddenly, making you lose your footing.
- inability to bear weight after an injury, or visible deformity.
- night pain that does not change with position, especially if accompanied by weight loss you did not seek, night sweats or unusual tiredness. Unintended weight loss is never good news: it needs investigating, not celebrating.
- swelling of the knee together with a painful, tight and warm calf: a venous circulation problem needs to be ruled out.
If your knee is warm and swollen, we have written a whole article on how to tell the different pictures apart, but the rule does not change: faced with these signs, you go to a doctor.
What we can do, and what we cannot
Let us start with what we cannot do, because on this topic that is the most important part.
A physiotherapist does not prescribe diets. It is outside the professional scope: the eating plan is set by a doctor, a dietitian or a nutrition biologist. We can explain what studies say about weight and joint load, as we have done here, and we can build the movement programme that sits alongside that path. If someone offers you rehabilitation and an eating plan in the same package, they are stepping outside their remit.
To be fair, something has moved in research: in the POWER trial, published in the British Journal of Sports Medicine in 2024 (PMID 38637135), 88 Australian participants with knee osteoarthritis received, by video conference, from physiotherapists trained for the purpose and with meal replacement products, a very low-calorie diet: the active group lost 8.1% of their weight against 1.0% in the control, and 71.1% reported a general improvement in the knee against 47.6%. But in the same study "there were no differences between the groups in any other secondary outcome": pain and function measured with questionnaires did not differ, despite eight percentage points less weight. That protocol, in a research setting with dedicated training, does not authorise any physiotherapist in Italy to prescribe you a diet.
Also in the "no" column: at a distance we cannot put our hands on you. No palpation, no ligament laxity tests, no measuring of swelling with a tape. We do not prescribe tests or medicines, and we do not replace the reading of an X-ray by the person who must read it.
What we can do is everything else, and that is not little: reconstruct the history of your knee, watch you move and understand how you go up and down stairs, how you get up from a chair, where the load is poorly distributed. From there a progressive programme is built that you can do at home, adjusted along the way, and we tell you clearly when the path is instead a doctor.
The first consultation is free and lasts twenty minutes by video call, with physiotherapists registered with the TSRM PSTRP register. It serves to understand whether your case can be handled this way and to tell you even when the answer is no. Subsequent sessions cost 29,90 euro (up to ten years old they last 50 minutes and cost 59,90 euro), and you pay only if you decide to continue. If you want to start from there, you can begin with the questionnaire.
Sources
- Wandel S, et al. Effects of glucosamine, chondroitin, or placebo in patients with osteoarthritis of hip or knee: network meta-analysis. BMJ, 2010. PMID 20847017
- Liu X, et al. Dietary supplements for treating osteoarthritis: a systematic review and meta-analysis. British Journal of Sports Medicine, 2018. PMID 29018060
- Simental-Mendia M, et al. Effect of collagen supplementation on knee osteoarthritis: an updated systematic review and meta-analysis of randomised controlled trials. Clinical and Experimental Rheumatology, 2025. PMID 39212129
- Garcia-Coronado JM, et al. Effect of collagen supplementation on osteoarthritis symptoms: a meta-analysis of randomized placebo-controlled trials. International Orthopaedics, 2019. PMID 30368550
- Arden NK, et al. Non-surgical management of knee osteoarthritis: comparison of ESCEO and OARSI 2019 guidelines. Nature Reviews Rheumatology, 2021. PMID 33116279
- Messier SP, et al. Weight loss reduces knee-joint loads in overweight and obese older adults with knee osteoarthritis. Arthritis and Rheumatism, 2005. PMID 15986358
- Messier SP, et al. Effects of intensive diet and exercise on knee joint loads, inflammation, and clinical outcomes among overweight and obese adults with knee osteoarthritis: the IDEA randomized clinical trial. JAMA, 2013. PMID 24065013
- Messier SP, et al. Effect of diet and exercise on knee pain in patients with osteoarthritis and overweight or obesity: a randomized clinical trial. JAMA, 2022. PMID 36511925
- Dietary interventions in osteoarthritis: a systematic review and meta-analysis of randomised controlled trials. European Journal of Clinical Nutrition, 2025. PMID 40295866
- McAlindon TE, et al. Effect of vitamin D supplementation on progression of knee pain and cartilage volume loss in patients with symptomatic osteoarthritis: a randomized controlled trial. JAMA, 2013. PMID 23299607
- Jin X, et al. Effect of vitamin D supplementation on tibial cartilage volume and knee pain among patients with symptomatic knee osteoarthritis: a randomized clinical trial. JAMA, 2016. PMID 26954409
- Effect of vitamin D supplementation on knee symptoms and cartilage loss in patients with knee osteoarthritis: a post hoc analysis of the VIDEO study. Rheumatology Advances in Practice, 2025. PMID 40677617
- Bannuru RR, et al. OARSI guidelines for the non-surgical management of knee, hip, and polyarticular osteoarthritis. Osteoarthritis and Cartilage, 2019. PMID 31278997
- Kolasinski SL, et al. 2019 American College of Rheumatology/Arthritis Foundation guideline for the management of osteoarthritis of the hand, hip, and knee. Arthritis Care and Research, 2020. PMID 31908149
- Robson EK, et al. Effectiveness of a very low energy diet delivered by physiotherapists via telehealth for knee osteoarthritis: the POWER randomised controlled trial. British Journal of Sports Medicine, 2024. PMID 38637135
