Part 3 (See here for parts 1 and 2)

Dr. Theodosakis then took the podium to present the case for glucosamine and chondroitin. Let me first say that Dr. T. is a young, charismatic man and appears to be a very nice guy. He is certainly an engaging speaker and certainly believes everything he says. I very much doubt that he supports G/CS’s use just because he may have a financial interest; I think he truly believes that these are effective products. That does not mean that I agree with everything that he said.

He went through a lot of information very quickly, but he provided copies of all his slides in his syllabus and encouraged us to review these for details of his argument. I write this with

One of the first things he talked about was our obligation as physicians to ‘first, do no harm’. If you don’t know this already, this is from the Hypocratic Oath that almost all physicians take when they get their MD. I agree that this is a very important principal; his point in bringing it up here is to point out G/CS have excellent side effect profiles. Unlike almost every drug we use, they really have no known side effects. Many of the medications we routinely use certainly do. NSAIDs (non-steoidal anti-inflammatory drugs) for instance (like ibuprofen and naproxen) are well known to have the potential to cause GI bleeding. Acetaminophen (Tylenol) can cause liver failure, and actually is one of the most used medications in attempted suicides. G/CS has never caused a death that we know of.

Therefore, he contends that the benefit/risk ratio of glucosamine and chondroitin are so high (low risk, high benefit, therefore the ratio is very high) that it is almost malpractice NOT to at least try to use them.

The problem with this argument is that it really only addresses one side of the fraction. The risk is indeed very low. But the problem is that the benefits may also be very low. Therefore, the ratio may not be so high as he implies, and theoretically if a product has no benefit, then the benefit/risk ratio is simply undefined (0 in the numerator), no matter how low the risk. So I don’t see this as a very good argument unless we can show that glucosamine actually has some kind of benefit.

He then points out that there are different recommendations on the use of glucosamine and chondroitin depending on what professional society you look at. The ACR doesn’t give the studies of them much credit and does not recommend their use (nor does it discourage their use). EULAR – the European League Against Rheumatism, the ACR’s European counterpart – gives the studies more credence and actually does recommend their use in some circumstances. That’s fine, but it’s not really an argument for G/CS’s efficacy – it’s really just an argument based on authority.

Another argument: these are dietary supplements. In this, these are just like vitamins. If you don’t like G/CS, you also shouldn’t recommend that patients take vitamins or minerals.

Sorry, Dr. Theo, but this just isn’t an argument. Firstly, it’s well demonstrated that deficiency in vitamins is detrimental. If you don’t get enough vitamin C, you get scurvy. If you don’t get enough vit. D you have severe problems with bone growth and hardness (rickets). Deficiencies of folic acid in pregnant women are recognized as being associated with certain birth defects in their children. But deficiencies in G/CS have not been associated with any disease. Vitamins and G/CS simply are not comparable.

Second, there are no studies to my knowledge that show that taking excesses of any vitamin have any kind of beneficial effect. You may think that high doses of vit. C prevents colds but there are no good studies that support this. Same with every other vitamin. It’s the deficiencies that are the problem. So once again, the point fails.

The argument he presented that I liked the least is that traditional drug manufacturers are fighting G/CS tooth and claw, and trying to case doubt on their effectiveness because they are eroding the big drug companies’ profits.

In a word: bollocks!

This is a classic case of setting up a straw man. Now don’t get me wrong; I have my problems with drug companies and their advertising practices, but I’ve never seen evidence that they are trying to undermine G/CS. If anything the objections seem to be coming from the world of academia. The fact is that the drug companies have little interest in OA. Sure, a few of them make drugs like Hyalgan and Synvisc which are supposed to help OA pain, but for the most part they simply aren’t in direct competition. We don’t know why OA happens and there are no medications that stop the progression of the disease. The drug companies really don’t have products that are in competition. Even in the case of Hyalgan and Synvisc, they are available only by prescription and there’s no reason that they cannot be used concurrently with G/CS. I just don’t see the evidence that G/CS is eroding drug companies’ profits.

The trials

This brings us to looking at the trials of glucosamine and chondroitin. This is the meat of all the argument. Do the trials of G/CS show that they are effective in reducing pain when compared to placebo, or compared to other pain medications?

The answer is yes, with reservations. There are a number of studies out there that show that both glucosamine and chondroitin work better than placebo, and work as well as or maybe even better than either acetaminophen or NSAIDs. The problem is that there are also studies out there that show that glucosamine (the more studied agent) is not at all effective compared to placebo. So what studies do we believe?

Dr. Theodosakis argues that we should believe the positive studies. He thinks that the negative trials all have fatal flaws in their design or implementation that make them untrustworthy.

I think he’s picking and choosing his data. During the talk, he spent some time looking at some of the negative trials of glucosamine and showed why he thinks these trials were flawed. An example: a trial by McAlindon et al. from 2004 and conducted over the internet showed no efficacy for glucosamine when compared to placebo. He argues that we should ignore the results of this study because: 1) it was never designed to be ‘real’ study of glucosamine; it was designed to be a validation study to show that trials could be conducted successfully over the internet. 2) The trial wasn’t long enough to show an effect for glucosamine which acts slowly. 3) Many of the subjects reported NSAID use prior to starting the trial.

Here are the problems: 1) Though it is absolutely true that this trial was designed to test the feasibility of doing a study over the internet, that doesn’t invalidate the fact that this was a well designed, well reported, truly randomized and truly double-blind placebo controlled trial. Just because it had another purpose does not invalidate the whole study. 2) He fails to point out that many of the early studies of glucosamine that were strongly positive were even shorter than this one. Blaming it on the length of the trial given this fact does not make sense. 3) Many of the subjects reported NSAID use prior to starting the trial. So what? I have never heard that prior NSAID use somehow makes people less sensitive to pain forever, or even for anything longer that a day or two. I’m just not sure why this matters.

Many of his arguments are like this. He brings out what he perceives to be flaws in the negative studies while ignoring many of the problems in the positive studies. He fails to point out that some of the positive studies appeared to have insufficient allocation concealment – that is, they did not make sure that either the subject or the investigator had no way of guessing what treatment (placebo or active therapy) the next subject would receive. This is well known to be a major factor in the design of a trial which can bias the results. He also puts more emphasis on the length of the trial period that I think it deserves. The GAIT trial was 24 weeks long which is as long as many of the previous positive glucosamine trials. It was negative.

He’s especially guilty of this with the GAIT trial. This was a very well designed trial with a very well defined protocol and endpoints, but he poo-poos it by pointing out potential faults whether they’re really significant or not. Examples:

    Perceived flaw: There was a high placebo response, therefore it was not possible to see the true effects of glucosamine.
    Answer: the placebo response is on the order of magnitude of other OA trials. This was not exceptional. Other trials of G/CS have shown them to be clearly superior to placebo. Why aren’t they here?
    Perceived flaw: Glucosamine dosing was three times a day instead of once a day, which is the preferred method. It will not be as effective if dosed this way.
    Answer: Most previous trials used three times daily dosing. Only one recent trial used once daily dosing after the manufacturer started claiming that this was more effective. So why is this suddenly such a big problem? Because one of the manufacturers of glucosamine says it is?
    Perceived flaw: GC/S is widely available over the counter and maybe some of the subjects were using it without telling the investigators.
    Answer: NSAIDs are also widely available over the counter. Why should we expect subjects to use one ‘illegal’ therapy preferentially to the other?
    Perceived flaw: Because the investigators used a total of six pills per day (3 glucosamine or placebo, 3 chondoitin or placebo) the patients probably missed some doses and were getting a reduced amount of drug and therefore weren’t getting the full effect.
    Answer: This applies to the placebo group as much as the G/CS groups. Why should we expect a differential effect?

The point here is that the perceived flaws just aren’t. The whole point of a placebo controlled trial is to eliminate differences between the arms so that only what the patient actually receives makes a difference. This is the standard which all drug trials must meet. The GAIT trial met it. Before we throw it away as irrelevant Dr. T. needs to make a much stronger argument that it deserves to be ignored, and he simply has not done it.

He also argues that the single sub-group analysis that shows that the combination of glucosamine and chondroitin are effective should not be ignored. I’ll come back to this argument in part 4.

Finally, he makes an argument that really just stunned me at the time, and still does.

As background you must know that when there are many arms in a trial and many potential outcomes – as there are in the GAIT trial – there is a very well understood problem called the problem of ‘multiple comparisons’. What this states is that the more comparisons you do, the more likely it is that one of them will be positive just by chance.

For instance, if you flip 2 nickles once, 25% of the time 2 heads will come up together. If you flip 2 pairs of nickles, the chance that one of the two pairs will show 2 heads is higher than 25%. That makes sense, right? The more pairs you flip, the more likely it is that at least one of those pairs is going to show two heads. If you flip 20 pairs of nickels, it’s pretty likely that at least one of those pairs will show two heads.

The same thing happens in a trial. The more comparisons you make between placebo and active drug – the more endpoints you assess – the more likely it is that you’re going to find a significant difference between the two at least once just by random chance.

In the GAIT trial, there were multiple endpoints, so the investigators made an appropriate statistical adjustment to make sure that they did call one of the comparisons positive when it was simply due to chance. This is a well-established and necessary procedure that has a very long track record in clinical trials.

And yet, Theodosakis suggests that if we ignore this adjustment, we’ll find that many of the results of the trial are actually positive. In other words, he’s suggesting that we ignore a well-established statistical procedure that is absolutely necessary, and that when we do so we’ll find that the results make much better sense.

What do you think? Does that make sense? Should we really ignore this necessary procedure because in doing so we’ll find that non-significant results become significant results? Isn’t this exactly what we’re trying to prevent? Isn’t it much more likely that he’s seeing results that appear to be positive just by random chance? The fact is that Dr. T. should know better.

Anyway, I think I’ve made my case. The point is that Dr. T. engages in forms of arguments that are routinely used by advocates of questionable beliefs; from Intelligent Design supporters to HIV deniers, from those who believe in UFOs to those who believe in ESP. He raises straw men, picks and chooses his data, and ignores data that contradicts his views.

In the next part, I’ll summarize Dr. Felson’s arguments against G/CS and further explain why I think his point of view is correct.