Treating patients suffering from myalgic encephalopathy/chronic fatigue syndrome (ME/CFS) with sodium dichloroacetate, Comhaire 2018

[Trish]

Thank you. Do you have a double blind trial planned or running yet?

 

[Trish]

Based on this cohort of patients it has been possible to mathematically estimate the probability of positively responding to the dichloroacetate treatment, including the score of the Fatigue Severity Scale (FSS), and the the answer to the question 3, 4, 7 and 9 of this questionnaire, completed before the initiation of treatment.

I have looked up the Fatigue Severity Scale. Here it is:

Read and circle a number.
Strongly Disagree → Strongly Agree

1. My motivation is lower when I am fatigued. 1 2 3 4 5 6 7
2. Exercise brings on my fatigue. 1 2 3 4 5 6 7
3. I am easily fatigued. 1 2 3 4 5 6 7
4. Fatigue interferes with my physical functioning. 1 2 3 4 5 6 7
5. Fatigue causes frequent problems for me. 1 2 3 4 5 6 7
6. My fatigue prevents sustained physical functioning. 1 2 3 4 5 6 7
7. Fatigue interferes with carrying out certain duties and responsibilities. 1 2 3 4 5 6 7
8. Fatigue is among my most disabling symptoms. 1 2 3 4 5 6 7
9. Fatigue interferes with my work, family, or social life. 1 2 3 4 5 6 7

You add up your score.
Less than 36 suggests you may not be suffering from fatigue
36 or more, consult your physician for further diagnosis.
http://nesportandspine.com/sites/default/files/fss.pdf

So Dr Comhaire is saying that the items listed 3, 4, 7 and 9 on this list are good predictors of whether the patient will respond to the treatment.
I find this completely bizarre.

 

[Hutan]

Just to re-iterate, there are risks with this chemical. e.g. from livestrong.com:

Sodium dichloroacetate also is associated with numerous side effects, including nerve damage resulting in weakness and numbness, liver damage and low blood sugar. Additionally, some animal studies have indicated that dichloroacetate causes liver cancer.

​

Dr Comhaire, can you please expand on the personality and education factors that you mention as causal factors?

External factors, including long-lasting severe stress due to emotional, professional or socio-economic reasons, or a traumatic incident or drastic medical intervention, act as provocative factors inducing epigenetic alterations [[9], [10], [11]] among persons with limited stress management skills due to their personality or education.

Is it too much education, or too little, or of the wrong type for the person that is the problem? What evidence do you have for this?

 

[Hutan]

So Dr Comhaire is saying that the items listed 3, 4, 7 and 9 on this list are good predictors of whether the patient will respond to the treatment.

It's more complicated and odd than that. This is the formula from the paper:

The probability of positive response (p) can be calculated using the formula:
logit(p) = 11.87 + 0.330 (duration of disease) − 2.958 (FSS before treatment) + 18.89(item 4) - 18.381(item 7)​

So the following factors increase the likelihood of a response in the small sample of this study:
Longer duration of illness
Lower fatigue score before treatment
Higher score on item 4 (fatigue interferes with my physical functioning)
Lower score on item 7 (fatigue interferes with carrying out certain functions and responsibilities)

 

[Hutan]

However, it is remarkable that there is apparently such formula that is confirmed (without changes) each time when new patients are entered.

The paper reports that after one month of treatment, there were 10 responders out of 22 participants and the formula includes the total fatigue score and items 4 and 7.

Today 33 patients have been followed-up. 13 have been "responders" experiencing significant and persistent improvement of their health condition, and 20 are "non-responders". Based on this cohort of patients it has been possible to mathematically estimate the probability of positively responding to the dichloroacetate treatment, including the score of the Fatigue Severity Scale (FSS), and the the answer to the question 3, 4, 7 and 9 of this questionnaire, completed before the initiation of treatment.

Now, with 33 participants, there are 13 responders. And the formula now includes items 3, 4, 7 and 9 from the FSS questionnaire. So it appears that the formula has changed?

 

[Adrian]

For the scientist among you: the area under the ROC curve is 0.90.

I was a bit confused about this because you seem to have 5 responders and 5 non responders. I assume you did a regression calculation to generate a hyperplane that separates the two groups and this gives the equation you quote and hence the ROC curve.

However, its normal to use different data to train a model (e.g. a regression model or otherwise) from the data used to quote performance. A good ML algorithm will always find arbitrary separations to separate training data. Its the performance on (previously unseen) test data that is interesting. But I don't see how you are doing this with such few data samples.

 

[Indigophoton]

This, indeed is a retrospective calculation, which could be applied prospectively, though there is little question about its validity.

So the following factors increase the likelihood of a response in the small sample of this study:
Longer duration of illness
Lower fatigue score before treatment
Higher score on item 4 (fatigue interferes with my physical functioning)
Lower score on item 7 (fatigue interferes with carrying out certain functions and responsibilities)

A good ML algorithm will always find arbitrary separations to separate training data. Its the performance on (previously unseen) test data that is interesting

So, for what it's worth, here's my n=1 prediction and result, based on the published formula and method.

First, just to note that, according to my reading around, sodium dichloroacetate is considered safe at a dose of 400mg a day if taken with B1 in particular, and alpha lipoic acid, to prevent neuropathic damage (which is reversible anyway, when it's stopped). (People using it for cancer take much higher doses).

That being the case, and since it's plausible that it might influence energy production, I put my numbers into the equation, and got a P of 0.99, ie, a high probability of success was predicted.

I have been ill more than 3 times longer than the median duration of patients in the paper, and had high scores on all the fatigue measures relevant to the formula.

I decided to give it a try. I thought about enlisting the aid of my carers to blind test it, but decided against, so I knew I was taking it. It's quite bitter tasting when dissolved in distilled water, which was how I took it, since that was cheaper than buying the capsules, which were anyway the wrong dose.

I took 400mg per day, the dose used in this study, obtained from the same company as in the study, for 30 days, the same length of time as in the trial. I also took, or was already taking, all the other elements of the nutriceutical.

I had mild nausea and gas a few times during the first three days, and no other side effects. I don't know which of the elements of the formula caused this. I don't usually get nausea, so it's likely it was the DCA and/or associated supplements that caused it.

Result: it made no discernible difference. DCA had no effect on my ME.

 

[Indigophoton]

I've had a message from someone trying to work out this formula in their own case, and was reminded that there is seemingly an error in the paper. So to save anyone else puzzlement, the number you calculate is not p, but logit (p) where,

logit (p) = ln {p/(1-p)}

So, e^logit (p) = p/(1 - p), and solve for p.

In Comhaire's Fig.2, logit (p) runs from 0 to 1. I asked him about this, but he didn't clarify. Presumably a typo; presumably it should be p, not logit (p) on the axis.

For anyone with brain fog or math-aversion, m.wolframalpha.com will do the calculations for you. There are apps for mobile devices.

 

[ME/CFS]

I agree, I really don't understand why it wasn't done with this trial. It would have been easy to use the same patients, and run it for 2 months, with each patient getting the treatment for 1 month and placebo for another, with neither patient nor researcher knowing which they were getting first. Then unblinding the results afterwards would give a double blind with all the patients acting as active treatment and control for a month, so direct comparison could be made. And surely use at least SF36 physical activity scale, not just a fatigue scale.

Edit to add: It's possible that this treatment may in fact be effective. I really hope it is, as a simple nutriceutical treatment with small danger of side effects is much more appealing to me than a scarily powerful cancer drug with serious side effects. But we can only judge it on the information provided here - and this information tells us nothing. Such a wasted opportunity.

This is a proof of principle, pilot, pragmatic trial. It is what it is, not more or less.

 

[ME/CFS]

At best all one can say about this trial is that it supports the need to have a proper double-blinded trial. In the responders it really does look like a decent response in self-report score (though with the emphasis on 'self-report'!) - but they were given 'extensive explanation on the details of the trial' which is almost guaranteed to increase placebo effect.

One thing that isn't clear is, whilst further investigation found other potential causes for fatigue in the non-responder group (which is to be expected, especially using Fukuda, due to the high rate of misdiagnosis of ME/CFS), whether the same investigation was performed in the responder group. Or is it only those who didn't respond who were subject to this investigation for other causes of their symptoms? If that latter I would think that's a gigantic red flag.


I understand that you recommend exactly that what I have written in my paper.

As for the claim 'Remarkably, it seems to be predictable which patients will or will not respond favourably to the nutriceutical treatment' I would need more than a small pinch of salt. Too few patients to do a proper statistical analysis and so the predictive formula included in the paper is most likely worthless.

Overall, it would be good to see this tested in a more rigorous trial; prior to that, I wouldn't be rushing out to source Sodium Dichloroacetate any time soon.

 

[ME/CFS]

Thank you for this further information, Dr C.

I am having difficulty understanding your claim of being able to predict which patients will respond.

I understand you have calculated which of the fatigue scale questions correlate most closely with treatment response in your patient cohort to date.

What I am not clear about is whether this has successfully predicted response in a further cohort.

If not, it would seem that all you are doing is calculating a prediction formula on one cohort, then testing it on that same cohort, in which case, of course it will work.

What I found is that the calculated formula alows for rather clear differentiation between reponders and non-responders. The (p) in the "logit(p)" means the probability that a particular person does belong to the group of responders (probability between 0 and 1). It is commonly called the "predicted probability".

 

[ME/CFS]

I've had a message from someone trying to work out this formula in their own case, and was reminded that there is seemingly an error in the paper. So to save anyone else puzzlement, the number you calculate is not p, but logit (p) where,

logit (p) = ln {p/(1-p)}

So, e^logit (p) = p/(1 - p), and solve for p.

In Comhaire's Fig.2, logit (p) runs from 0 to 1. I asked him about this, but he didn't clarify. Presumably a typo; presumably it should be p, not logit (p) on the axis.

For anyone with brain fog or math-aversion, m.wolframalpha.com will do the calculations for you. There are apps for mobile devices.

THis comment is perfectly correct. The vertical axis shows (p) not logit (p). Thank you!

 

[wdb]

This is a proof of principle, pilot, pragmatic trial. It is what it is, not more or less.

I just struggle to understand the point of such a trial it doesn't tell us anything, people have good days and bad days, it is impossible to know how many participants would have reported feeling better at the end point if they had had no treatment at all and we know from studies of homeopathy and such like that open label subjective trials are all but guaranteed to give positive outcomes from placebo effect alone. I'm sorry but it just seems like a massive waste of everyone's time especially when a placebo group could have been included so easily for such a treatment.

As for as for subdividing the responders after the trial, I might do a trial where I ask participants to try to predict the result of a coin toss, after the results are in I will divide them into two sub-groups, those with psychic powers (guessed correct), and those without psychic powers (guessed wrong), I probably won't bother with any controls that would be excessive for a pilot study.

 

[ME/CFS]

This isn't research - it's a sale pitch.


He has multiple such patents, and is completely resistant to using appropriate methodology. He writes papers to support his business, not to learn anything about ME/CFS or what works to treat it.

 

[ME/CFS]

I refuse to answer absurd personal attacks.

 

[ME/CFS]

I have looked up the Fatigue Severity Scale. Here it is:

You add up your score.
Less than 36 suggests you may not be suffering from fatigue
36 or more, consult your physician for further diagnosis.
http://nesportandspine.com/sites/default/files/fss.pdf

So Dr Comhaire is saying that the items listed 3, 4, 7 and 9 on this list are good predictors of whether the patient will respond to the treatment.
I find this completely bizarre.

It is strange indeed. The formula generated by the logistic regression analysis with stepwise elimination also includes the total score of the FSS. It is an observation, not an explanation. I try to inform those among you who are interested in the evolution of my research. I hope you appreciate that, and that you also understand the work to be in progress. Some findings intrigue me as much as they do intrigue you.

 

[ME/CFS]

Thank you. Do you have a double blind trial planned or running yet?

The trial will be initiated as soon as, and provided that the DCA is officially accepted and registered as a food supplement.

 

[Alvin]

The trial will be initiated as soon as, and provided that the DCA is officially accepted and registered as a food supplement.

Why do you need it accepted as a food supplement, afaik this is not the case for pharmaceuticals

 

[Diwi9]

@ME/CFS - Thank you for contributing to this thread. I'm interested in learning about your follow up work with sodium dichloroacetate. I see that it is also being investigated as a treatment for Leigh's Syndrome to treat lactic acidosis.

ETA: Looks like sometimes referred to as "Leigh's Disease" otherwise Leigh Syndrome.

 

[ME/CFS]

Just to re-iterate, there are risks with this chemical. e.g. from livestrong.com:

Sodium dichloroacetate also is associated with numerous side effects, including nerve damage resulting in weakness and numbness, liver damage and low blood sugar. Additionally, some animal studies have indicated that dichloroacetate causes liver cancer.

​

Dr Comhaire, can you please expand on the personality and education factors that you mention as causal factors?

​

Is it too much education, or too little, or of the wrong type for the person that is the problem? What evidence do you have for this?

Several papers, and my own experience from history taking, suggest that "things in the past" and lack of emotional binding during infancy and youth are reported by some patients. This, evidently, is their own subjective feeling. Also patients mention that they feel vulnerable when exposed to external pressure. Other patients have been educated in a spirit of authoritarian parenthood (you MUST do this, MUST do that, you may not spend time on things that are not useful, etc.). Once again this is the subjective feeling of (commonly female) patients. I, as an observer, can not objectively assess these claims. The latter is probably not so important, since it is the personal emotional feeling of the patients themselves that counts.