I have read the paper on glucosamine and chondroitin by Juni et al (here, but you may need to be a BMA member to see it).
They find no significant effects from these supplements in arthritis in the papers they examined, but no side effects either.
This runs counter to my clinical experience, of patients happy to have lost their pain, and taking less erosive pain killers. Yes, I know. Anecdotal. But happy patients mean a lot to me.
I could find no obvious flaws in the paper, but then to do that you would have to be a statistician. Incidentally, a statistician reviewed the statistics of a load of papers published in the Journal of the Royal College of Psychiatrists (of which I am a member) and found that the fast majority were statistically flawed.
However, the fact that they looked at joint space deterioration is significant, even if the results were not statistically significant. NSAID painkillers make no claim to increase joint space (in fact they may decrease it).
One of the difficulties with non-patentable drugs and treatment modalities is that there is no incentive for manufacturers to pour squillions into research, as investment towards receiving 20 years of patents royalties. So the work does not get done adequately.
Pace Juni et al, if I get a patient with year old with early osteoarthritis coming in, besides all the usual stuff - painkillers, weight loss, exercise - I will still advise him to try a 3 month course of glucosamine, drink more water, and try the effect of a magnet. I'm sorry if this will infuriate EBM zealots, but if I see him skipping in in 3 months with a smile on his face, pain free, and not taking analgesics, I cannot see that as a bad thing. Meta analyses are useful, they have their place, but they are not Holy Scripture.