Both of MJ’s objections are, therefore, completely spurious.

Third, this was the third post on the HBOT study and a third attempt to discredit it, using a third angle – which is what the comment was referring to.

That critique was written by Prometheus, BTW. You’ll note that the critique did raise a number of valid points, e.g. there was no controlling for repeated measures. This is not a hypothetical.

Meanwhile, your critiques raise hypothetical and unlikely points, e.g. maybe some relevant variable did not match after randomization.

Yet, you are able to conclude that the new RCT is not a real RCT, but the Rossignol RCT is a good RCT; a target of mean attempts to discredit it.

There is no evidence that they matched on ANY variable other than age and hours of ABA.

You do realize that the normal expectation after randomization is that variables match, right? There’s little chance that an arbitrary variable will not match.

To say that there’s no evidence that some variable matched is misleading. Chances are that it did.

For that matter, there’s no evidence of matching in Rossignol et al. of a number of variables, such as:

– Prevalence of intellectual disability in each of the groups.
– Hours of TEACCH, RDI, etc.
– Performance in the embedded figures test.
– CARS score.
– ADOS score.

Let’s imagine that one or more of these variables did not match. Wouldn’t that completely throw off the numbers? Yet, there’s no evidence that they did match. Therefore, by MJ’s reasoning, Rossignol et al. can be dismissed in its entirely simply because of this. My imagination tells me so.

1. I am not a troll.

2. I have read the text of both studies, both the current one (CARD) and Rossignol, several times in fact. Have you?

3. You seem to be confused about which study you are talking about.

a) The CARD study is matched on age and hours of ABA but did not publish any data about the makeups of the HBOT and placebo groups. This is commonly published as Table 1 in RCTs but was not published with this study.

b) Rossignol did match/randomize on other criteria. This list included age, gender, severity of symptoms, and other treatments. They did publish the makeups of their groups as a Table 1. I linked to the Rossignol study and pointed out Table 1.

As a result, while the Rossignol study is on the weak side, the CARD study is much weaker because they either did not publish the data on the matching/randomizing on other criteria or they did not perform said matching/randomizing. Either way is a problem.

Ok, first of all, that was directed to Sullivan, not Prometheus. Second, that is hardly “quite upset” unless you and I have very different meanings for the word upset. Third, this was the third post on the HBOT study and a third attempt to discredit it, using a third angle – which is what the comment was referring to.

“What I think you’re trying to say is that variables should match across groups at the beginning of the trial.”

They should match at beginning and at the end. Some characteristics are going to be constant (ie gender) while others should be held constant (other treatments). For example, what would happen to the results of the study if the participants started another new therapy in the middle of the trial, one that was not controlled for? That would invalidate the results.

“Is there any evidence that some variable was unmatched?”

There is no evidence that they matched on ANY variable other than age and hours of ABA. If the groups were matched on any other criteria is wasn’t published with the results.

“You seem to be suggesting that variables were unmatched, but the authors just didn’t report this very important finding.”

You are correct, it is possible that they did match and just failed to report this. I just don’t find it very likely that they did and somehow just forgot to publish it. My best guess is that the dropouts basically destroyed the matching between the groups so the matching was omitted rather than call attention to the fact.

And where exactly did I do all of this?

@MJ: In this comment.

The goal of any RCT is to measure the change from one variable while all others are held constant.

There’s no requirement that any group variables remain constant, as far as I’m aware. What I think you’re trying to say is that variables should match across groups at the beginning of the trial.

Is there any evidence that some variable was unmatched?

If there were it would have been reported in the text of the, like almost every other RCT out there.

You seem to be suggesting that variables were unmatched, but the authors just didn’t report this very important finding.

After a randomization, roughly only 5% of any variables you can come up with should be unmatched statistically.

And you don’t think the things you mentioned would greatly weaken the results?

They are both very speculative. They attempt to argue for hypothetical and unlikely scenarios (e.g. non-matching of characteristics after randomization) which the researchers were either too lazy to check, or withheld information on.

“You’re not directly promoting HBOT, MJ, but when Prometheus posted his critique of Rossignol et al. you appeared quite upset and accused him of manipulating the data from the paper.”

And where exactly did I do all of this?

“Now you’re putting the current “RCT” in scare quotes, suggesting it’s not a real RCT or something to that effect.”

That is exactly what I am saying, minus the “scare”. The goal of any RCT is to measure the change from one variable while all others are held constant. If you have a large enough effect, a large enough group, or a set of well balanced groups you can get close to this. This particular study fails on all accounts.

“I guess it’s possible you’re just a stickler for good science”

No, I just dislike nonsense.

“There might not have been matching on some variables after randomization. We just can’t be sure there was”

If there were it would have been reported in the text of the, like almost every other RCT out there.

“Even though the drop-outs were split evenly”

It was not actually even – 30% of the placebo group dropped out while 21% of the HBOT group did. But even if the numbers were dead even, losing that many of the partipants would distort the makeup of the groups. And even if it didn’t, I would expect the authors to deal with the problem directly and talk about it, which they did not in this case.

“That’s pretty much it, isn’t it?”

And you don’t think the things you mentioned would greatly weaken the results?

Sullivan, you said –
“I note that your response was missing actual substance.”

Sullivan, talking to you sometimes is like talking to a child. I make a comment, you reply to something else. I point out that your response is lacking substance – so you reply that my response was lacking substance. It is like child repeating “I know you are but what I am” over and over again.

“If you have something from that paper you would like to quote that I missed, please do.”

You said –
“Although where he got the idea that gender was considered in that study is not clear to me. They only mention “male/female” once, and never mention gender in the text.”

Which is one of the reasons I pointed out Table 1 in the Rossignol study – when you look at the table you can see that gender was taken into account.

“The fact that the two groups had similar characteristics is very different from being able to say that the researchers randomized based on those factors”

Do yourself a favor and go read up on how a RCT should work. I would recommend this book – http://www.amazon.com/gp/product/052170958X/

Still not going to rise to the baiting. Why do you continue.

I note that your response was missing actual substance. You defend yourself by pointing me to the paper that I just quoted from.

If you have something from that paper you would like to quote that I missed, please do.

Here’s what they say about table 1:

At the onset of the study, the use of nutritional supplements, medications, and applied behavioral analysis (ABA) therapy was similar in both the treatment and control groups (p = ns), see Table 1.

The fact that the two groups had similar characteristics is very different from being able to say that the researchers randomized based on those factors—which is what you said:

The randomization was just done on 2 factors – age and hours of ABA. They did take into consideration other factors like gender, initial severity of autism, or other styles of therapy currently being used (the Rossignol group did).

Just below Table 1 is the section “Randomization and Allocation”

From the 62 children who were enrolled in the study, 33 were randomly allocated to the treatment group and 29 were randomly allocated to the control group. To achieve this allocation, a random allocation sequence (1:1) was generated and stratified on both the participant’s group (treatment or control) and center, and this sequence was equilibrated every eight patients. The allocation sequence remained concealed to all investigators, study participants, parents, nursing staff, and all other clinic staff.

This may seem familiar, as I quoted it above.

If you think that I am promoting HBOT please provide a quote of exactly where I said that.

You’re not directly promoting HBOT, MJ, but when Prometheus posted his critique of Rossignol et al. you appeared quite upset and accused him of manipulating the data from the paper. Now you’re putting the current “RCT” in scare quotes, suggesting it’s not a real RCT or something to that effect.

I guess it’s possible you’re just a stickler for good science, who knows :)

So basically your objections to the study can be summarized as follows:

– There might not have been matching on some variables after randomization. We just can’t be sure there was.
– Even though the drop-outs were split evenly, it’s possible the authors left out some details from their intent-to-treat analysis that could potentially be relevant in showing that the drop-outs were significantly of a different nature between groups.

That’s pretty much it, isn’t it?

But really, who knows? This DAN! cabal are constantly on the search for bizarre/ off-label uses of otherwise effective therapies (chelation, HBOT, GFCF)used for other conditions I really believe that we’ll be seeing these guys advertising the efficacy of leeching and cupping in the near future.

But for all who decry the existance of questionable autism therapies, know this: They flourish because the medical community has NOTHING to offer.

There are suspect medical differences between autistic children and the rest of the population. But the medical community is concentrating on genetics. It is safer. Genetics won’t impugn the practice of vaccinating our children for everything.