It's been a while since I mentioned the Autism Omnibus hearings. The Omnibus proceeding is the culmination of all the legal cases brought to the Vaccine Injury Compensation Program by nearly 5,000 families who "feel" that their children's autism was caused by vaccines. Many, but not all, of the plaintiffs blame the mercury in the thimerosal in childhood vaccines, despite there being no good evidence to support such a link. The way that the hearings are being run is that several "test cases" are being chosen by Special Masters, who hear evidence presented by the plaintiffs and the defense presented by government defense lawyers. Unfortunately, thanks to the low bar of evidence required in the Autism Omnibus hearings, the mercury hypothesis may not be a legal loser, even though it's definitely a scientific loser, even though the test cases appear to be nothing more than tugging at the heart strings and the scientists called to testify by the defense routinely demolish the plantiffs' "expert" witnesses.
Indeed, overall things don't appear to be going too well for the first few test cases. Perhaps that's why Representative Dan Burton (R-Indiana), that powerful booster of antivaccination and autism pseudoscience and quackery in the House of Representatives, that man who once had Dr. Rashid Buttar (yes, that Dr. Buttar) testify in front of his committee back when the Republicans controlled Congress, even going so far as to nominate him for an NIH Director's Pioneer Award), has decided to enter the fray. I learn from A-CHAMP that Burton has apparently decided to try to influence the Special Masters by writing them a letter on Congressional letterhead:
A link to a PDF file of the letter can be found here. Some key text:
Further to the deliberations of the Office of Special Masters, I am enclosing a November 2007 article from the Journal of Child Neurology entitled, "Blood Levels of Mercury Are Related to Diagnosis of Autism: A Reanalysis of an Important Data Set." I hope that the Office of Special Masters will take the article's findings into consideration as it carries out its mandate of managing and adjudicating childhood vaccine claims.
Naturally, he can't resist pointedly mentioning his position and power and repeating his demand:
During my tenure as Chairma of the House Committee on Government and Reform, I spent six years researching and hearing testimony from autism advocacy and scientific community about the autism epidemic sweeping our country. Over and over again, the question of causation, namely the use of thimerosal--the mercury-based vaccine preservative--in childhood vaccines was raised...For instance, the enclosed article from the reputable scientific journal...details the most recent findings of M. Catherine DeSoto, PhD, and Robert T. Hitlan, PhD, who reanalyzed an early reported dataset and found a significant relationship between the blood levels of mercury and a diagnosis of autism spectrum disorder.
I hope that ou would agree with me that the Office of Special Masters has a duty and an obligation to review all the sound science relevant to vaccines, mercury, and autism.
This is utterly outrageous. For one thing, the Hitlan paper is a piece of crap. In essence, it's a post hoc statistical fiddling with a dataset in which the authors substituted a one-tail t-test for the two-tail t-test originally used for what appears to be the sole reason of changing a finding that wasn't statistically significant in the original analysis of the dataset to a statistically "significant" finding. But what's even more disturbing is the specter of a powerful and senior Representative throwing his weight around and blatantly trying to influence the outcome of the Autism Omnibus proceeding. This is raw politics, and it strikes me as highly unethical for an elected official to be inserting himself into a high profile court case in such a public manner. You can be bet that if a CDC official had written such a letter asking the Special Masters to take into account, say the Shattuck paper or Fombonne paper from last year showing, respectively, that the "autism epidemic" is mainly due to diagnostic reclassification and that there is no evidence that vaccines, thimerosal-containing or otherwise, are associated with autism, you'd hear the mercury militia crying to high heaven about the attempt to "manipulate" the proceedings. And they'd be right. But apparently Dan Burton can blatantly try to influence the Special Masters, and he gets his letter published on A-CHAMP, separation of powers be damned.
I'm sure any legal eagles out there will correct me if I'm wrong, but if a party who is not a party to a case directly contacts the judge this way and asks him directly to consider a piece of evidence I'm pretty sure that would represent an improper attempt to influence the outcome of the legal proceedings. (In this case, it's also a slap at the plaintiffs' attorneys, who, in Burton's apparent estimation, are doing so poorly that he felt the need to intervene.) If Burton weren't trying to interfere with the legal proceedings and was truly genuinely concerned about what was happening, a far more proper course of action would have been to contact the plaintiffs' attorneys directly and make them aware of Hitlan study. He should not have contacted the Special Masters directly and certainly not publicly. This is especially true because Burton is a powerful Representative, which makes his letter appear even more like an inappropriate attempt to influence the Special Masters' deliberations. Personally, I wouldn't mind it if the plaintiffs' attorneys did enter the Hitlan paper into evidence, because it would be easy to shred, but I do mind credulous antivaccination Representatives interfering in legal proceedings.
Indeed, there's no excuse for such a blatant ploy to influence a judicial proceeding, even if it's cloaked in "letting the Special Masters know about" new research (which really isn't anything new and is poorly done). I can only hope that the Special Masters saw this for the obvious attempt at influencing the proceedings that it was and reported Burton to the Department of Justice and the House Ethics Committee. At the very least, this news should be more widely disseminated, although, sadly, Burton is such a true believer in "mercury in vaccines causes autism" pseudoscience that he is probably beyond being embarrassed.
Oh, wow... I remember the first time I tried to counter the chelationista of the mercury militia several years ago (approximation from memory since that was on another computer):
Chelationista: Dear Group, you should look at the work of Rashid Buttar. He has reported curing several children of autism with his chelation protocol.
Me: Hi, I could not find his papers on www.pubmed.gov. Where did he report this?
Chelationista: He gave congressional testimony in front of Rep. Dan Burton.
Me: I'm sorry but testifying in front of congress is not exactly proof. Do you have anything more "sciency"?
Chelationista: You are horrible, you are trying to keep cures away from our children. You are very dangerous!
Sometime in 2004 someone sent me this:
"We get into this argument everytime you start writing about mercury. I just returned from the DAN! conference where about 1000 people heard many doctors and researchers talk of the problems caused by thimerisol. There is a lot of research showing that it is causing our autism epidemic. You can beleive what you want but please don't stop people from researching vaccines and thimerisol before vaccinating their kids. People like you are dangerous. I just wish someone had educated me on the possible problems of vaccinations before i had my son vaccinated. He is now autistic. The Autism Research Institute is a REAL place and you are showing your ignorance by your statement. It is difficult enough to be a parent of an autistic child but to hear others treat us like we are spewing garbage is very upsetting. We are talking about the truth and soon it will proven."
Still waiting for that "truth", and it looks like Dan Burton is trying to influence what that "truth" will be.
What is much more important than whether the average mercury level in ASD patients vs. non-ASD patients is higher, is how much overlap is there in the distribution of mercury levels in the two populations. Looking at the data sets, there is a gigantic amount of overlap.
I suspect that Burton didn't do this on his own but was lobbied to do it by his constituency. Not all of them are idiots; some of them can see that the omnibus proceedings are going down in flames. They know their only hope at winning the legal lottery is via political pressure and trickery because they don't have the science. That is what this is about, applying political pressure to the court for sure, but maybe also setting the stages for a legislative "solution".
If you look at the Faroe Islands mercury levels
The inter quartile range was 13.1 to 40.4 micrograms/L in cord blood. That is 65 to 201 nM/L. That means that 75% of the children in the Faroe Islands study had cord blood mercury levels above 65 nM/L. In the Hitlan paper, there were only two individuals with blood mercury above 60 nM/L, one ASD and one control.
There was a study of the incidence of autism in the Faroe Islands that included the age cohort tested for mercury at birth and later. In the 1,022 individuals tested for mercury there are at most 5 with autism (out of a total of 1,404 in the age cohort that included the mercury tested individuals).
There are at least 750 with mercury levels higher than the highest seen in the Hitlan paper and at most 5 out of those 750 have autism. If a mercury level greater than 65 nM/L produces an autism incidence of 1 out of 150, by what mechanism do the much lower mercury levels observed in the Hitlan paper cause autism?
if a party who is not a party to a case directly contacts the judge this way and asks him directly to consider a piece of evidence I'm pretty sure that would represent an improper attempt to influence the outcome of the legal proceedings.
I'm no lawyer, but I'm pretty sure this isn't true, at least not if the contact is public. There's a long tradition, at least in Supreme Court proceedings, of "friend of the court" [amicus curiae] briefs, arguments and evidence offered by knowledgable and interested third parties.
by what mechanism do the much lower mercury levels observed in the Hitlan paper cause autism?
I'm tempted to say it's like homeopathy, the less you have the stronger it is ... makes about as much sense as the rest of it.
There's a long tradition, at least in Supreme Court proceedings, of "friend of the court" [amicus curiae] briefs, arguments and evidence offered by knowledgable and interested third parties.
Perhaps, but what Burton did doesn't appear to have been to file an amicus curiae brief. There are specific legal procedures by which such briefs are filed, and they're usually filed by advocacy groups or industries whose interests might be impacted by the ruling of the court. However, he may get away with it under the rule (which I later looked up) that the U.S. or an agent of the U.S. can file such briefs without requiring the permission of both parties.
Even if it's legal, Burton's action appears unseemly.
The more I think about it, the more I think that Burton was put up to this. The lawyers couldn't introduce this at trial because it has no real value to their case. What conclusions could be drawn from a random set of ASD children having a slightly higher mercury level than a random set of non-ASD children? Even if that level reaches the level of "statistical significance". Not a whole lot. The respondent would crush any hint that "statistical significance" means the same as "physiologically significant". That would bring in all the other mercury data in the world, including the Faroe Islands stuff which shows that these levels are completely unimportant as far as autism is concerned.
The plaintiffs are better off trying to influence the court inappropriately with trickery, innuendo and misinformation. They might have a chance with that course of action, their chances are zero via a scientific approach.
The lawyers would have to call a witness, and that witness would have to testify that this is an "important" result. They probably can't find anyone who has any credibility who will do that, and they know that anyone without credibility would be crushed.
Also, Burton is an interested party. He has a family member who he thinks was affected by vaccines (as he appropriately mentions in his letter). There is no way that he can be considered objective or fair.
I suspect that the plaintiff's lawyers have been listening to the cranks pushing the "mercury causes autism" idea, and that those lawyers don't have the technical expertise to understand that their "experts" are delusional and/or are spouting bullshit and nonsense. Until stuff hits the internet and they read the blogs put out by amateur scientists (i.e. not the world class experts that the respondents will call as witnesses). When the likes of me, Orac, Prometheus, Interverbal, DoC and the like can completely blow away the bogus crap being put up there in our spare time as a hobby, what chance do they have against the "real" experts that the respondents will produce? Zero and none. That is why they have gone the political route. I think Burton may try to propose some legislation that "compensates" these plaintiffs. How to do it without breaking the treasury? Maybe this will be a way to get health care for children introduced again. If so, something good could come out of it, provided that quack "alternative" medicine is not funded by it.
While Dan "The Loose Cannon" Burton generally favors quackery, I am quite sure that he endorses the antivaxers because a) he has a child (or grandchild) with autism, and b) he will believe anything.
Too much is being made of the DeSoto paper, where they basically found that a non-statistically signficant difference turned into an *almost* statistically significant difference because of typos that led to errors in Ip et al.
Once again, I call attention to the actual results after correction:
"The mean hair mercury level of children with autistic spectrum disorder was 1.98 Â± 1.05 ppm. The level in the control group was 1.92 Â± 1.58 ppm. There was no statistically significant difference between the 2 groups (P = .79)."
"The mean blood mercury level of the children with autistic
spectrum disorder was 19.53 Â± 15.65 nmol/L as compared
with 14.68 Â± 12.48 nmol/L for the control group (P = .06)"
So the "big deal" found by DeSoto et al. was that the mean blood mercury level of autistics was 19.53 nmol/L vs. 14.68 nmol/L for the controls. Not quite statistically significant, but close. The difference is a factor of 1.33. Whoopty doo! Such a small difference could easily be explained by group differences like diet. Prior studies comparing groups who have different diets have found blood mercury levels to be substantially different.
"However, he may get away with it under the rule (which I later looked up) that the U.S. or an agent of the U.S. can file such briefs without requiring the permission of both parties."
Burton didn't file an amicus brief with the court (which would be sent to the court clerk), he sent a letter to the special masters themselves. I would also guess that "the U.S. or an agent of the U.S." for purposes of amicus brief filing would be the Solicitor General, Justice Department, or some other individual/agency acting *for the United States*, not someone acting as an individual.
Burton's action strikes me as terribly improper, though I admit I haven't looked up precedential cases dealing with the canons of ethics.
It doesn't appear from a cursory look at the applicable rules of ethics that Burton has violated an explicit prohibition, since he is not "representing" the Autism Omnibus claimants. For me, his actions still don't pass the "smell test," though of course my own opinions on the matter certainly influence that conclusion.
From the House Rules of Ethics:
Unofficial Representational Activities
Several provisions of the Federal Criminal Code and House rules restrict the ability of Members and staff to become involved with outside organizations in ways that require interaction with the Federal Government. Members, officers, and employees are prohibited by 18 U.S.C. sec. 203 from asking for or receiving compensation for "representational services" rendered in relation to a matter or proceeding in which the United States is a party or has a direct and substantial interest. Included are proceedings before any Government department or agency. Additionally, House Rule 43, clause 3, prohibits Members and their staffs from receiving compensation by virtue of influence improperly exerted from a position in Congress.
Even absent compensation, employees are restricted by law and rule from private representation of others before the United States Government or the pursuit of others' Federal claims when not in the proper discharge of official duties. Section 205 of title 18 prohibits employees from acting as agent or attorney for another person or organization in prosecuting a claim against the United States or in connection with "any covered matter." A covered matter includes "any judicial or other proceeding, application, request for a ruling or other determination, contract, claim, controversy, investigation, charge, accusation, arrest or other particular matter." (FOOTNOTE 27)
(FOOTNOTE 27) 18 U.S.C. sec. 205(h).
These provisions have generally been enforced in instances where an official's public duties have conflicted with private interests. Enforcement of the Criminal Code is the responsibility of the Department of Justice.
Another provision that would apply to an employee seeking to represent others in Federal matters is House Rule 41. That rule states:
No person shall be an officer or employee of the House, or continue in its employment, who shall be an agent for the prosecution of any claim against the Government or be interested in such claim otherwise than as an original claimant or than in the proper discharge of official duties.
As with 18 U.S.C. sec. 205, there is no requirement in the rule that the representation involve any compensation.
If a grandstanding dipshit like Burton can do this, surely a Member of the National Academy can be found who can write and explain exactly why, on the science, what Burton says is a steaming pile of it?
In my concept of scientific National Academies (in any country) and their responsibility for guarding the integrity of the scientific process, that is exactly the kind of thing they ought to be doing. Or maybe the Association / College of Pediatrics could do it, or the College of Epidemiologists, or whatever.
Surely there must be some body out there with both the chops and the guts to call this guy on his bullshit?
I must admit I was shocked when I read this letter a couple of days ago. I tend to agree that even if he is well within the law to send this -- I'm guessing he's fine because he's being public about the letter -- it sure smacks of trying to throw weight around by using the letterhead etc.
The plaintiffs could easily have submitted this themselves as there are many studies being submited that don't get directly addressed during the trial, and I'm certainly getting the impression that the special masters are reading them all.
I was even more surprised to see that he was parading this study as being revolutionary new information. It does correct a blatent error in a previous study, and also reveals issues with the methodology used in the original study. It also argues that the correct numbers indicate that there was a correlation in contradiction with the study conclusion.
(The one thing people here tend to forget is that the errors in the original study were pretty serious, and should never have passed peer review, let alone the authors themselves. It's pretty embarrassing really).
At best, this new paper invalidates any strong conclusion the original study may have had. That is valid to enter into the case if the original study had been submitted by the defendants as evidence. But that's it.
As for Mr Burton's intent? This seems to me to be much more like a publicity stunt and possibly an attempt to throw weight around. I honestly don't comprehend the political nuances here. There may be much more going on behind the scenes.
The special masters are in a very difficult position. There is undoubtedly a lot of pressure from the groups that believe that a negative ruling will ruin the vaccination program (many people here have basically said as much), and then there is a lot of pressure from various lobby groups, and now a very powerful politician. There is a lot of money at stake.
However, if I were a special master, and received a letter like this, I would shake my head, read the study, and continue on. I think they are fully qualified to evaluate a simple study like this one.
Burton hasn't got much going for him. Not much to recommend him to anyone as any kind of intellect.
The study he cites is disingenuous at best.
Hopefully the court will see the science and have a "Dover" moment where they realize the mercury causes autism idea is plain bunk.
It does correct a blatent error in a previous study, and also reveals issues with the methodology used in the original study. It also argues that the correct numbers indicate that there was a correlation in contradiction with the study conclusion.
(The one thing people here tend to forget is that the errors in the original study were pretty serious, and should never have passed peer review, let alone the authors themselves. It's pretty embarrassing really).
At best, this new paper invalidates any strong conclusion the original study may have had. That is valid to enter into the case if the original study had been submitted by the defendants as evidence. But that's it.
"Strong conclusion"? There was no "strong conclusion" in the original paper. The conclusion was that there wasn't a correlation, and this "reanalysis" of the data doesn't change that. The DeSoto and Hitlan paper was nothing more than post hoc fiddling with the dataset in order to get the desired result; i.e., in order to convert a result that was not statistically significant to one that was borderline significant. The two-tailed t-test used in the original paper was the appropriate test, not the tortured reasoning to apply a one-tailed t-test used by DeSoto and Hitlan.
If you actually read through the comments on Do'Cs page, you'll see that I was quite engaged in the conversation there.
"Strong conclusion"? There was no "strong conclusion" in the original paper. The conclusion was that there wasn't a correlation, and this "reanalysis" of the data doesn't change that."
That would be a strong conclusion supporting the null hypothesis...
"The DeSoto and Hitlan paper was nothing more than post hoc fiddling with the dataset in order to get the desired result; i.e., in order to convert a result that was not statistically significant to one that was borderline significant. The two-tailed t-test used in the original paper was the appropriate test, not the tortured reasoning to apply a one-tailed t-test used by DeSoto and Hitlan."
Again, I suggest you go back and read the article more carefully, because you are not accurate in your statements at all. They did not re-calculate the data using a one-tailed test. They recalculated the data using the CORRECTED data. They only argued that a one-tailed test SHOULD have been used -- no recalculation was made there. They also argued that the treatment of outliers was not appropriate.
"If the outliers are kept in, and a one tailed test used, does the result end up being significant, or do they even do this?"
They didn't calculate it. There were two outlier data points in this study. The bigger of the two was an unusually high mercury level for someone in the autistics group. The authors felt that by removing this, the two groups were brought closer together, but no stats were run."
You'll also note that the authors of the original study posted a correction, and the journal editor posted an explanation in the same issue that this study was published in. You'll also note that the editor requested the paper from DeSoto et al.
Your accusations of fiddling with the data are certainly not supported by the evidence. Perhaps you only read the original article and not the rebuttal and following discussion?
Since you're drawing some incorrect conclusions from the other discussion, I'll describe it briefly here. I welcome a correction from Interverbal, Prometheus or Do'C if I get anything wrong here.
First, Interverbal did agree that there was some merit in arguing for a difference in the two-tailed vs one-tailed test:
""So much more to say... does it seem a bit pointless to say that usually one picks the type of test (one or two tailed) before the data analysis. Since the data had already been collected...? But, of course, we did realize the test should have been one-tailed before seeing the data set itself. Is DoC suggesting that we should have planned to do a one-tailed test before reading the article?""
"That is a fair rebuttal. If you realized that the test should have been one-tailed before reading the data, then my point doesn't stand."
From my understanding, the issue with the original Authors' paper, is that they chose a hypothesis that should have been matched with a one-tailed test, yet decided to do a two-tailed one. The argument from Interverbal is that in this type of study, a two-tailed hypothesis is more applicable and he has strong arguments on that point.
However, they hypothesis was the choice of the original authors, and DeSoto et al correctly pointed out that based on the choice of hypothesis a one-tailed test should have been used.
The original paper had these two flaws: blatently incorrect data (a transcription error) and a two-tailed test matched to a one-tailed hypothesis. The original Author's conclusion that the null hypothesis is true can be questioned at this point.
Again, if you read the comments by DeSoto, Interverbal, Do'C etc, this much should have been obvious.
The bigger points of contention with DeSoto et al, is that they claimed that when correcting the data errors, that there was a correlation based on data that was almost statistically significant by conventional definition. On Do'Cs board you'll find some good conversation around that topic, with Prometheus pointing out statistical significance is not a line drawn in stone, and that it is mainly convention that people use the 5% mark as the point of significance. However, when using a non-stadard point to determine statistical significance, it would be customary to explain why a non-standard was being used. DeSoto et al. did not provide any explanation.
The other big issue is that after determining that the corrected value (even using the two-tailed test) was effectively statistically significant, that their discussion uses a reference that did not support their statements around mixing up mercury poisoning and Autism symtoms.
There were complaints that the assumption that Autism was increasing were not referenced (and people on the board argue that it is not generally agreed that the rate is actually increasing).
The Authors were contending that the corrected data supported Holmes et al. This was a strong point of contention, with people arguing the lack of merits of that study (Joseph) and lack of references and arguments to support that assertion.
There were also complaints that the Author (DeSoto) was not an expert in Autism and that this was apparent from the discussion portion.
I agree completely that most of these discussion problems were not adequately addressed by DeSoto's rebuttal IMO.
The conspiracy theorists were also out in force, but eventually Do'C shut that down.
However, my main point is that your characterization of the study is pretty inaccurate. The statistical nature of the discussion was not as you indicated, and it was the discussion and conclusions around the meaning of the data that was most in contention.
But I want to be clear, I don't feel it supports Burton's letter because it's value is in pointing out errors in the other study, effectively nullifying their conclusions, not drawing new conclusions.
Schwartz, what is the hypothesis that this study was trying to address? The hypothesis is "Mercury causes autism", therefore people with autism will have mercury levels sufficiently higher than non-autists, so as to be the cause of their autism."
What does the data show? That the mercury levels substantially overlap. There are people with autism with high and low mercury levels, there are people without autism with high and low mercury levels. A person's mercury level does not help predict whether they have autism or not.
The exact and precise position of the average of this tiny data set is unimportant. The important parts of the measurements are the standard deviations, which are very high. That shows that there is a lot of spread in the data that any analysis of data like this is going to be very sensitive to a few "outliers", to people in the tails of the distribution.
There are people with autism, with 1/10 the mercury level of people without autism. How can mercury be causing autism if people with 10x more mercury don't get it? The answer is, not via any mechanism known or recognized, or suspected, or thought plausible, or even thought possible by medical science.
The "mercury causes autism" idea is dead. It is not a close call. The data in this paper doesn't add a bit of credibility to the "mercury causes autism" idea. It adds yet more weight against it because the distributions of the two groups substantially overlap. If mercury were a significant cause, the distributions would have to not overlap. They do overlap. Where the blood mercury levels overlap, mercury isn't "causing" autism; something other than mercury must be causing it. Let's stop wasting time and effort on this dead idea. Let's find the real cause.
The main problem with the DeSoto/Hitlan study is that you can't retroactively apply a statistical analysis to data in order to get the results you want. At least not legitimately.
If you know ahead of time that the difference can only go one direction - for example, if you know that the data from one group will be at or around zero (and that negative numbers are impossible) - then you pick the one-tailed test. If, on the other hand, you have no a priori reason to make that assumption, then the two-tailed test is appropriate.
Ip et al had no reason to assume, prior to looking at their data, that the differences would be in one direction only (i.e. that one group had to have larger values than the other group), so the only appropriate test was the two-tailed test. To quote from my favorite text of statistics:
"Keep in mind that a one-tailed directional test can be applied only if a specific directional hypothesis has been stipulated in advance; otherwise it must be a non-directional two-tailed test."
The DeSoto/Hitlan reanalysis of the Ip et al data should have limited itself to a two-tailed test, since DeSoto/Hitlan only knew which direction the difference would be after they saw the data. Besides, there is nothing about hair or blood mercury that would specify a direction, unless one set of values was zero or near zero (which it wasn't).
We can also argue whether they were justified in removing the "outliers". While it may be emotionally satifying to remove such high numbers, it actually makes little sense. Unless DeSoto/Hitlan can argue that there was a reason (apart from aesthetics) to remove the data, it should remain. Simply being an "outlier" isn't sufficient reason.
It should also be noted that the Ip et al paper isn't exacly a cornerstone of the data refuting the mercury-causes-autism hypothesis. Even if it were to fall, there is still a mountain of data remaining.
Additionally, there remains the "little" problem of physiology. All studies of mercury "excretion" by mammalian hair has shown that it is a passive process, with the mercury in the blood attaching to the sulfhydryl groups of cyteine (an amino acid) in the growing hair strand.
Even if the blood-to-hair ratio of mercury were different in autistic vs non-autistic children, this says nothing about mercury excretion. Hair is an extrememly minor excretory route in humans, and especially so in infant humans (hint: it's because we're not very hairy - at least not compared to other mammals).
So, DeSoto/Hitlan is a "classic" example of how statistical manipulation can be used to "prove" two contradictory hypotheses using the same data.
Final point: the hair mercury levels in the original Ip et al paper and the DeSoto/Hitlan "reanalysis" have a mean value of 1.9 - 2 ppm. The NHANES study - of over 800 US children - showed a mean hair mercury of 0.22 ppm. Might this make a difference?
The thoroughly discredited Holmes et al "study" showed autistic children with only 0.47 ppm (although their controls had a mean of 3.63 ppm - where were they living?) and the Kern et al study found a mean hair mercury of 0.14 ppm (not significantly different from controls).
So, we have two studies that found no difference between hair mercury levels of autistic and non-autistic children (Ip et al and Kern et al) and one that found that autistic children had less hair mercury than controls (although the autistic children had over twice the US mean and the controls had over sixteen times the US mean hair mercury level).
Perhaps it's time to admit that hair mercury is a rough time-integrated measure of blood mercury with a lot of reliability problems and leave it at that.
I think you're misreading my post and changing the topic here. I am not taking the position that this study supports a mercury/Autism link. Quite the opposite as I stated above.
"But I want to be clear, I don't feel it supports Burton's letter because it's value is in pointing out errors in the other study, effectively nullifying their conclusions, not drawing new conclusions."
If indeed the defendants in the Omnibus hearing submited the IP et al study as evidence, this study is perfectly valid in the case to point out the significant issues with that study as originally published.
My point is that this study brings up legitimate questions about the calculations and methodology of the original study and thus nullifies any conclusion they might have drawn. What effect the rest of the discussion has on the overarching body of evidence is for the special masters to figure out.
I'm not going to argue with you the merits of the Mercury/Autism links here. This discussion is about the facts presented in Orac's article: the Burton letter, it's intent, Orac's incorrect characterization of the DeSoto Study, and Burton's mistaken assertions of what the DeSoto et al. study proves.
Schwartz, Ip et al, made an arithmetic error, a transcription error. That is a trivial type of error, one than anyone could make. Even Orac has made those kinds of errors. Correcting that error doesn't "nullify" the study, it corrects the error. Because someone catches the error, doesn't mean they get to apply what ever spin they want to on the study. Scientists who read papers are big enough to understand when conclusions follow from data and if those conclusions are changed by minor changes in that data.
What DeSoto tried to do, was while correcting the error, was to assert that the data now supports the "low mercury excretion" idea of Holmes et al. An assertion that had no plausible physiology behind it in the Holmes paper, and none in the DeSoto paper. DeSoto has no expertise in mercury physiology which would lend the slightest bit of credence to her unsupported assertion.
The DeSoto paper certainly was put out there to influence the "mercury causes autism" debate (in my opinion). I think the lawyers have now realized that it doesn't influence it in the way that they would like. They don't want to put it into evidence because then they would have to answer sticky questions about how it came to happen, such as "did they hire DeSoto to write this"?
If they did hire DeSoto to write this, then how much of the paper is her words and how much is the words of the lawyers? The words of the lawyers are not "evidence", they are "argument". There is a big difference in court between those two things.
What I think happened is the lawyers hired DeSoto to write the paper and they put the words about supporting the Holmes et al idea in there themselves. They won't call DeSoto as a witness because she would be asked how she came up with that idea and she would have to tell them. What I think the lawyers will do is say "look at this peer reviewed paper that says xyz, it must be correct". That isn't how science works.
No one cares about trivial arithmetic errors. Scientists don't spend their time pouring over the literature checking for arithmetic errors they can correct. The only reason there is any attention spent on this paper is because of the mercury causes autism idea. To pretend otherwise is disingenuous.
Orac may make that type of error (among others), but his blog is not published in a journal and he reports on such a large volume and variety of information that it is fully expected that he would make errors.
"No one cares about trivial arithmetic errors. Scientists don't spend their time pouring over the literature checking for arithmetic errors they can correct. The only reason there is any attention spent on this paper is because of the mercury causes autism idea. To pretend otherwise is disingenuous."
You continue to ignore the key point here and you either haven't read the paper, or you're ignoring the majority of it. Even Interverbal stated the following in his article: "The core statistical argument in DeSoto & Hitlan (2007) centers on whether a one-tailed test or two-tailed test is appropriate."
The focus of the paper is not on the transcription (not arithmetic BTW) error but on a re-analysis of the data based on statistical problems they discuss at length in the paper. To trivialize the paper as a correction of an arithmetic error is what is disingenious.
They also discuss and show an analysis that the Blood/Hair mercury correlation has a much higher variance in the Autistic group compared to the control group. Specifically, they found that the relationship between blood/hair levels was further reduced in Autistic children with higher blood levels of mercury -- something not addressed or discussed by Interverbal in the blog anlaysis.
If you just look at the "arithmetic error" it represented a shift in almost 10 points (from p=.15 to p=.056). I hardly consider that trivial especially when it brings the result very close to standard stastical significance. For you to imply that it should of been left alone seems silly given that the original study will continue to be referenced. Additionally, if you read the actual study, DeSoto et al actually explain why they feel that the result should not be cavalierly dismissed as insignificant.
According to DeSoto et al, the value becomes significant (p<0.03) by standard definition when applying a single-tailed test which would match the hypothesis chosen by the original authors.
The incorrect use of a two-tailed analysis despite it not matching the author's original choice of hypothesis is an issue that also changes the results of the study and it certainly points to a study design flaw.
In DeSoto's discussion and they argue that the almost significant result (using 2 tailed test) supports a correlation but as I have already stated, the discussion around the poor excretor hypothesis (especially the one reference) has it's problems that were not addressed adequately in the blog rebuttal.
However, since I have not seen a letter written to the journal, I wouldn't expect to see an official rebuttal either.
Again, I stand by my point that enough issues were raised about Ip et all to question any conclusion they drew. DeSoto et al. certainly discuss some interesting statistical observations on the original data set which could easily be of value in the discussion.
The paper's value to me is primarily questioning the other study's results and this is important if the other study is used as a reference to any material put forth by the defendants in the trial.
As for your conspiracy theories, you can weave all the scenarios you want. The story as stated is that they noticed an error, reported it to the editor of the journal, and were asked by the editor to write the paper. If you have credible evidence to the contrary, I would like to see it in the open.
Sorry, another HTML posting error (I must remember to stop using greater than signs...)
My cut-off sentence in the middle should read:
According to DeSoto et al, the value becomes significant (p less than 0.03) by standard definition when applying a single-tailed test which would match the hypothesis chosen by the original authors.
Statistics are a mathematical tool to help analyze data. Because a result is statically significant does not make it important, or useful, or relevant, or of any value at all in answering real questions that people have.
One tail or two tailed, the minute shift that correcting the typo and shifting the statistics doesn't add even a tiny bit of credence to the idea that mercury causes autism. The distributions substantially overlap. That the averages are slightly different is unimportant.
So the authors Ip et al are bad and should be scolded for making a typo and not catching it. Shame on them. Lets make them wear a dunce cap and sit in the corner and give them a time-out. DeSoto caught the error, lets give her a star and let her wear the Ms Professor cap while Ip et al are in the corner with their time-out.
You want to discount the ideas of Ip et al because they made a typo, and accept the ideas of DeSoto because she found it? Huh? Accept her statement that adds credibility to the completely non-physiologic "poor-mercury excretion" idea simply because she found a typo? Huh? The "poor mercury excretion" idea required toxic blood levels of mercury with hair mercury an order of magnitude lower than what was observed in this study. 100 times higher blood levels with 1/10 the hair mercury? DeSoto's analysis supports that? Huh? A blood to hair ratio 3 orders of magnitude different than what Ip et al measured? Huh?
The trillion dollar question is does mercury in vaccines cause autism?
The Ip et al study adds maybe 0.01% more information on that question. Depending on how you analyze it, how you weight it, maybe it adds 0.0125%. According to the data that I have read, I put the odds that mercury causes autism at way less than 0.00001%, way less than one in a million. Depending how you look at the DeSoto et al study, maybe that changes to less than one in 990,000. It is a tiny change that still makes it extremely unlikely that mercury is in any way causally related to autism. It is no where close to the 50%+ criteria that the Omnibus proceedings will use.
There really is a lot of data that indicates mercury doesn't cause autism. There is so much data that the idea that mercury doesn't cause autism is now the default understanding of the scientific community. It is not an open question any more. The scientific community would be willing to change its consensus opinion, but that would require data, data which demonstrates how mercury causes autism, and which explains why the gigantic amount of data we have so far that shows mercury doesn't cause autism is all completely wrong. The DeSoto reanalysis of the Ip et al study doesn't come close to doing it.
When I read a paper I rarely even think much about the conclusions the authors draw. Sometimes I do, often I don't. I never accept their conclusions if it doesn't follow from the data, or if it contradicts something that is well known. Everyone expects that conclusions in scientific papers may be wrong. People reevaluate ideas in papers all the time. It is the data that is important, not the conclusions.
I appreciate that the mercury causes autism crowd have no science to back up their claims, so they resort to statistical games and disingenuous confabulation of statistical significance with physiological significance.
Maybe their disingenuous confabulations confuse the mercury causes autism "believers", it does not confuse the scientific community. I hope it does not confuse the Special Masters.
You are putting words into my mouth again: "You want to discount the ideas of Ip et al because they made a typo, and accept the ideas of DeSoto because she found it? Huh? "
I will discount the ideas of Ip et al because they based their conclusion on flawed numbers and a flawed methodology. It's quite simple. Both of those flaws change the results of their specific study and certainly can have an effect on their conclusions.
Beyond that, my original position is that if the Ip et al. study is used in the trial as evidence (it doesn't matter how irrelevant according to you), submission of the DeSoto et al. paper is fully justified (not by Mr. Burton mind you).
Again, I'm not arguing about the body of evidence right here, as that would require me to quit my full-time job. That is currently the job of the special masters for the purposes of vaccine damage awards.
The understanding of the science community is based on the evidence available to date, the analysis of that evidence, in combination with human factors such as financing and politics. The scientific understanding also changes over time. Again, I am not debating that here.
I find it amazing that in one breath you complain that people who don't agree with you are "playing statistical games..." etc, and then turn around and imply an incorrect conclusion in a study being put forth as evidence is just fine, because people disregard the conclusions anyways. Just step back and listen to that line of reasoning for a minute.
If you think that the only statistical games being played are on the side of the people pushing the Mercury/Autism link then you are far more naive than I thought.
One last question: Did you actually read the study?
You are mistaken. The only statistical games being played are those being played by those pushing the "mercury causes autism" idea. I know that because I have read their stuff. It is full of crap. They frequently confuse "statistically significant" with "physiologically significant". Often they don't release enough information for anyone else to do an independant statistical analysis on their "data". I know because I requested the raw data, and was denied it. Raw data where the distribution was obviously so skewed that using only the mean would grossly misrepresent the distribution.
Show me a single study about mercury and autism by any other group where "statistical games" have been played. The JAMA article perhaps? Sally Bernard said they played "statistical games", do her assertions hold up?
I don't dispute that the statistics are what they are. I do dispute what those statistics mean. A small difference in the average of two data sets (no matter what the statistical significance) does not by itself add to a demonstration that the small difference is the reason for all of the differences in those two data sets.
I have seen the distributions of the data that Prometheus has put up. It is clear just by looking at the distribution that there is a huge amount of overlap in the mercury levels of the people with autism and those without.
What of the Faroe Islands study, where 3/4 of the children measured at birth had mercury levels higher than all but 2 children in the Ip et al study?
Obviously the updated study with the corrected data should be used. But just because a non-physiological conclusion gets pasted onto the correct data doesn't mean the non-physiological conclusion is correct or should be used.
I have read a great deal about mercury physiology. hundreds of papers. There is nothing that suggests that the "mercury causes autism" idea is correct, and much which demonstrates it is wrong. No scientific result is looked at in isolation. Looking at this mercury and autism study in isolation would be the wrong thing to do. Each scientific study has to be looked at in the context of the rest of the scientific literature.
Again, I did not say a new conclusion should be reached because the chain of arguments in the study for supporting the poor excreter hypotehsis were not supported by good references. However, the data set still nullifies the Ip et call conclusions, and thus should be submited for the special masters to review if the Ip et al study was referenced by the Plaintiffs. It's that simple.
As for statistical manipulation, I'm sure you know very well that much of it occurs in the selection of the data, exclusion criteria, and choice of calculation. The IP et al is a minor example -- the selection of a non-matching two tailed case to achieve non-statistical correlation despite the design of the hypothesis. The Cochrane group has reviewed hundreds of studies and found many of them lacking due to bad design, some of these studies being epidemiological in nature and thus completely reliant on the statistics. Just peruse their extensive reviews on MMR or Flu vaccine and you'll find plenty of examples.
One other point. I do not doubt your expertise or knowledge in the areas of heavy metal physiology. At this point, I am not convinced at all that Mercury directly causes Autism. I agree with you that study is required in a lot of different areas. I would not propose studying mercury to the exclusion of everything else and I am very doubtful that in isolation mercury poison would cause Autism.
I agree that it is very important to find the mechanisms that result in the condition so we can find the causes but I would not exclude any mechanism at this point.
I do not think that studying genetics in isolation is the right approach as this was looked into exclusively for many years. Medical science has been misled by this problem more than once, that is certain. I think that if the prevalence is as high as people think, then the CDC most certainly dropped the ball as they should have thrown a lot more resources into understanding the problem. I am confident that in a few years, we will certainly know if the prevalence is real or not.
The letter Burton sent reeks of his desire to influence the Special Masters Decision. Burton, a member of the legislative branch, has no business whatsoever interjecting himself into the functions of another branch as he did. I contend that he violated the concept of separation of powers and thus violated the Constitution. He should be removed from office.
I hope you and your fellow bloggers will write separate letters to the Special Masters regarding the study, and also letters to the Justice Dept. regarding Mr. Burton. Maybe something like this will be picked up by TIME?
At best, this new paper invalidates any strong conclusion the original study may have had.
It invalidates the strong wording of the conclusion, but the original conclusion is still essentially correct. There was no statistically significant correlation between either blood mercury levels or hair mercury levels and autism.
Further, DeSoto's suggestion that the results provide support for Holmes et al. is bewildering. Hair mercury levels were slightly higher in autistics, not lower. It doesn't make sense to draw conclusions by comparing a non-significant difference with an almost significant difference IMHO. Furthermore, if you graph blood mercury vs. hair mercury, the trend is very similiar and in the direction you'd expect (unlike what Holmes et al. would predict) in both groups.
"I contend that he violated the concept of separation of powers and thus violated the Constitution. He should be removed from office."
Is this true or just a wish? I am very curious. I would be surprised though, given that he has been around a long time, and probably wouldn't make a mistake that could land him out of office.
I suspect we're pretty close to agreeing, we just might differ on the strength of the conclusion after the adjustment. The poor methodology (choice of hypothesis and lack of matching test) is still an issue that weakens any conclusions of Ip et al. IMO.
I think we agree on the lack of strength of the rest of the arguments in the paper.
One thing though, if you actually read the study, it did contain an interesting analysis about the variability of correlation between hair and blood levels in the Autistic vs the Control group. This was completely ignored by Do'C and Interverbal in their article and I did not see any reference to it until I read the actual study. It certainly has impact to anyone using those methods to judge mercury exposure.
Weakening a conclusion that mercury is not significantly associated with autism, doesn't strengthen a conclusion that mercury causes autism. It is exactly like weakening the case for evolution doesn't strengthen the case for creation.
A conclusion that mercury causes autism would have to have data to support it, so far there isn't any. Not in this study, not in any study. No amount of examination of this study with a microscope is going to find something that isn't there, evidence that mercury causes autism.
The blood to hair ratios observed in this study are not that different than seen in other studies. There is a lot of scatter in the data, which is not unexpected. The major source of mercury is diet. Hair mercury reflects when the hair grew, blood mercury reflects recent mercury consumption. If mercury consumption varies over time (which it does) and mercury excretion varies over time (which is does) so will the correlation between hair and blood.
The time constant for fluctuations in hair mercury is a lot longer than the time constant for fluctuations in blood mercury. Hair mercury is probably a better indication of long term mercury exposure (the exposure most relevant to a mercury causes autism idea).
If the two data sets (hair and blood mercury) don't correlate, one of the data sets can't be discarded and said to be irrelevant. They are both relevant. If one shows a marginal statistically significant correlation and the other doesn't, that marginal correlation is likely due to chance.
No one has ever seen in any organism the blood to hair ratios ASSUMED in the Holmes et al study. Where they ASSUMED toxic blood levels of mercury with hair mercury of less than 0.3 ppm. There is no rational basis for stating that this study supports the ASSUMED blood to hair ratios ASSUMED by Holmes et al.
With no rational basis, there must be an irrational basis for that assertion. Whether that irrational basis comes from being paid off, or from being delusional, or from drinking the anti-mercury kool-ade is irrelevant to me. I don't need to understand the source of someone else's irrationality to know that it is irrational.
Daedalus2u brings up an interesting point.
If we assume that the corrected Ip et al data is supporting the Holmes et al "Just So" story about "poor excretion" of mercury in the hair of autistic children, then the hair:blood ratios of mercury should tell us just how "toxic" these autistic children are.
The Ip et al data (which are published in their entirety) show that the hair:blood ratio is 0.12 in the autistic subjects (range: 0.04 - 0.38, SD: 0.05) and 0.15 in the non-autistic subjects (range: 0.08 - 0.33, SD: 0.05).
For the Holmes et al data, this would correspond to blood mercury levels of 3.9 mcg/L (range: 1.2 to 11.8) in the autistic subjects and 24.2 mcg/L (range: 11.0 to 45.4) in the "control" subjects.
For reference, the "normal range" for blood mercury is up to 10 mcg/L.
So, rather than the Ip et al data supporting the Holmes et al "Just So" story about mercury and autism, it seems to indicate just how screwed up the Holmes et al study really was.
If we use the hair:blood ratios from the Ip et al study, it suggests that the autistic children in the Holmes et al study were pretty much below the "toxic" blood mercury range and the "control" children were well into the "toxic" range.
So, it would seem that the real message from the Holmes et al study, as illuminated by the Ip et al data (even after redaction by DeSoto and Hitlan), is that elevated blood (and therefore brain) mercury levels are negatively correlated with autism.
In other words, higher mercury levels - again, according to the Holmes et al data - seem to prevent autism.
Now, of course, that is simply silly. And so is the Holmes et al study. For some reason, the poor authors got a "bad batch" (or is it a "hot lot") of data and made a silly and unsupportable claim: that autistic children are "poor excretors" of mercury.
That DeSoto and Hitlan bought this "Just So" story from Holmes et al is merely indicative of their willingness to overlook data in the pursuit of a really good story.
"Weakening a conclusion that mercury is not significantly associated with autism, doesn't strengthen a conclusion that mercury causes autism."
You are correct. However, weakening the conclusion of the Ip et al study removes it's credibility from the pile of evidence used by the defendants in the trial. If the defence present the Ip et al study as evidence supporting their position, then the DeSoto study can be used by the Platintiffs to reduce or eliminate it's conclusion from the body of evidence.
Eric Fombonne, one of the witnesses for the defence has cited the Ip et al study in two of his papers. This makes the DeSoto study more than fair game in this trial.
"A conclusion that mercury causes autism would have to have data to support it, so far there isn't any. Not in this study, not in any study. No amount of examination of this study with a microscope is going to find something that isn't there, evidence that mercury causes autism."
If you look to the root of the argument, these hair/mercury level studies are really asking a couple of questions:
1) Do autistic children show hair mercury levels in proportion to their overall mercury exposure measured against controls?
2) If not, does this represent an inability of autistic children to excrete mercury.
The first question on its own is very important if hair mercury levels are used in any mercury studies involving Autistic children to determine mercury exposure.
If it is shown that Autistic children have significantly different hair mercury level correlation with exposure, then it is valid to seek to understand why the hair mercury levels are different and if it does indeed indicate a problem with excretion.
I agree that one can't conclude Mercury causes Autism from any of these studies. All they provide is a piece of the puzzle of differences between Autistic children and controls. I don't think any of the papers claim that this proves Autism is caused by mercury. However, the invalidated Ip et Al conclusion certainly states that Mercury does not play a neurotoxic role in causing Autism.
"The blood to hair ratios observed in this study are not that different than seen in other studies. There is a lot of scatter in the data, which is not unexpected. The major source of mercury is diet."
If you read the DeSoto study there is a measurable difference in the scatter between the Autistic children and the Control group.
"Another way of looking at it, the relationship between blood level and hair excretion may be different for persons with autism than those without autism. Levine's test of equality of variance indicated the variance in hair mercury was not evenly distributed between the autism and control groups (F = 5.98, P = .017). We calculated the correlation for persons whose circulating levels of mercury were in the top quartile separately for the autism and control groups. The correlation between blood and hair levels of mercury was r = .91 for the control group (accounting for 84% of the variance). For the autistic group, the correlation was r = .73, meaning only about 55% of the variance in the hair mercury levels was attributable to the blood mercury level differences."
"To check the hypothesis that hair excretion was overall lower than would otherwise be predicted based on a certain blood level in the autistic group, a best fit regression line was calculated (y = 10.3, x = â2.48) indicting that for each unit increase in hair level, blood level increased by 10.3 units. A t test on the residuals showed that autistic participants were significantly more likely to have lower hair mercury levels than would be predicted as a function of their blood levels, t(133) = â2.92, P < .005;
see Figure 1). It should also be noted that the presence of unequal variances or nonrandom residuals (in this case, autistic persons are both more likely to have greater variability at high levels of circulating mercury and a lower hair value for a given blood level) are both violations of important assumptions of the t test; a t test of hair mercury is therefore probably not a valid means to predict autism diagnosis as a function of mercury exposure."
"The time constant for fluctuations in hair mercury is a not longer than the time constant for fluctuations in blood mercury. Hair mercury is probably a better indication of long term mercury exposure (the exposure most relevant to a mercury causes autism idea)."
This statement is only valid after you can prove that hair mercury levels in Autistic children correlate equally with controls and overall mercury exposure. The data in this study shows that they do not and that the correlation decreases as mercury exposure increases. Indeed, the study seems to show very clearly that the variability in Autistic children is higher than controls.
"If the two data sets (hair and blood mercury) don't correlate, one of the data sets can't be discarded and said to be irrelevant. They are both relevant. If one shows a marginal statistically significant correlation and the other doesn't, that marginal correlation is likely due to chance."
If one can show that the hair mercury level does not correlate with time correlated blood samples, than it seems obvious that the time correlated blood/mercury sample would be a better measure of exposure than the hair mercury level. It would also be important to determine why in Autistic children the correlation did not follow as closely as controls, as it would indicate a difference in biological mechanisms involving mercury.
"There is no rational basis for stating that this study supports the ASSUMED blood to hair ratios ASSUMED by Holmes et al."
Good thing they never stated that at all (kind of making Prometheus' post interesting but not too relevant to the DeSoto paper. Here is exactly what they stated (since I'm still assuming you didn't read the study):
"Holmes et al. performed the most direct test of the hypothesis that autistic children may be deficient in terms of ability to remove mercury from circulation."
Here they are discussing the Hypothesis that Autistic children have difficulting removing mercury from circulation, and they note that the Holmes study attempts to study it directly compared to other studies they mention.
They then proceed to summarize the conclusions around this HYPOTHESIS and how it will relate to the results of the Ip et al data. Again the focus is on a discussion of the hypothesis, not much more.
"In the autistic group, severity of autism was inversely related to hair mercury levels. This means that the more severe autistic cases actually had less excretion of mercury. Furthermore, among the normal children, hair levels of mercury were correlated to the mother's mercury exposure (as would of course be expected). But among the autistic children, there was no linear relation between the mother's mercury exposure and excretion of mercury in the hair. As the authors state, this pattern of results is easily understood if one considers "detoxification capacity of a subset of infants,"6 (p 6) such that the bodies of those diagnosed with autism appeared to be less able to excrete and/or metabolize the mercury they were exposed to."
The paper will then later discuss how the Ip et al data supports that hypothesis. This study never states that Ip et al data supports any data measured in the Holmes study at all. All of the discussion revolves around the hypothesis against which they measure the new data.
Again you allege conspiracy without evidence. The data and calculations published by DeSoto et al, and the journal are wide open for analysis. I have not yet seen any criticism about the statistical analysis they performed around the difference in scatter or any mention of it at all. This certainly appears to be an observation of some value in addition to the lessening correlation as blood/mercury levels become quite high.
To quote from the editor of the journal where the correction, data, letter, and DeSoto paper were published:
"Dr. Brumback states, " One of the philosophical myths associated with science is that invalid information published in scientific journals will be exposed by scientists who find that a particular study cannot be replicated. Unfortunately, this is rarely the case."
"...interest in replicating the published results is lacking because there is no glory in being the second person to describe a particular phenomenon, and research funding agencies (such as the US National Institutes of Health) do not provide grants to investigators wanting to confirm the results of other investigations. Thus, it is actually a fluke when problematic publications are identified."
Unless you're alleging the editor of the journal is a co-conspirator (without evidence), I think that the evidence backs up the statements that DeSoto et al found an error, and were asked by the Editor to publish a paper on it and a re-analysis of the data.
If anything, the original Ip et al authors may have had other interests as they state in their conclusion regarding the use of Chelation: "Some have proclaimed that chelating therapy for suspected mercury poisoning cures those autistic children with a higher mercury level. Our pilot study demonstrates that this is not based on hard-core evidence."
So this study that you so vehemently argued earlier only adds "0.01%" to the question of whether mercury causes autism can be touted by the defence to have a lot more importance than you infer. Since the mercury excretion hypothesis is one avenue in explaining how mercury could be a contributing factor, and this study is one of a small handful addressing this particular hypothesis, I think your weight attribution context needs to be re-evaluated. Additionally, the ip et al study states that it alone invalidates Chelation therapy, so nullifying this conclusion certainly has merit for the plaintiffs in the case. For the Plaintiffs, the Ip et al study conclusions were damaging until shown to be flawed by DeSoto et al.
The DeSoto study also shows that Autistic children appear to react to mercury different than the control group and that could also support some hypotheses put forth by the Defence.
If there are indeed other reasons or biological mechanisms for you to discredit the overall Autism/Mercury link, than that is good and should be discussed in the appropriate forum.
However, since the Defence and their witnesses use the Ip et al study in their body of evidence, it is incumbant on the Plaintiffs to use DeSoto et al.
"I contend that he violated the concept of separation of powers and thus violated the Constitution. He should be removed from office." (Quoting me)
Is this true or just a wish? I am very curious. I would be surprised though, given that he has been around a long time, and probably wouldn't make a mistake that could land him out of office.
I worked for the federal government, and we had rules regarding contact with Congresscritters. I often had to review these contacts, and I never saw one of this nature. For the most part, they were passing on constituent inquiries.
Burton is not merely passing on a constituent inquiry. He is trying to influence a decision.
As for older being wiser, it does not apply always to concresscritters. What happens is that the ones in secure districts tend to get sloppy from their security.
I ran this by a friend who is a member of the NYS Legislature. He was mighty uncomfortable with it, to say the least. He assured me that he would not do such a thing.
I truly think he broke the rules. He will probably get away with it.
"Holmes et al. performed the most direct test of the hypothesis that autistic children may be deficient in terms of ability to remove mercury from circulation."
That's one of the strangest things that DeSoto and Hitlan say in their article, since Holmes et al didn't actually test their hypothesis.
If Schwartz had actually read Holmes et al, it would have been clear that Holmes et al didn't measure blood mercury or even mercury exposure.
They collected some historical data - fish consumption, dental "fillings", RhoGam, etc. - and then adjusted the weighting until they came up with a "formula" that matched the historical data to the control hair mercury.
Then they used that "formula" to calculate what the hair mercury should have been in the autistic group.
Since the hair mercury in the autistic group was lower than the "formula" value, they asserted (without ANY data to support their claim) that autistic children didn't "excrete" mercury as well as their typical peers.
Of course, since the control group's hair mercury levels were over sixteen times the US mean (as shown in the NHANES study published a year later), there is a lot of doubt about the validity of the "formula" used by Holmes et al and - as a result - their "poor excretor" hypothesis remains unsupported.
While I can agree to disagree with the statistical manipulations exercised by DeSoto and Hitlan, it is their unquestioning acceptance of the Holmes et al "Just So" story about autistic "poor excretors" that I find most troubling.
Schwartz, you really don't know what you are talking about. Your "analysis" is filled with assumptions that not only are not warranted, they are in fact wrong.
In none of these studies was mercury exposure actually measured. None, zero, nil, nothing. To measure mercury exposure, one needs to measure how much mercury has been ingested. Has mercury in the diets of these children been measured? No, it hasn't. So nothing is actually known about actual mercury exposure, unless one assumes that blood levels, or hair levels, or something else that has been measured is actually also a measure of mercury exposure. But then if one assumes the level reflects exposure, then one cannot use the level to determine a difference in hair mercury vs. exposure because the exposure is assumed.
Hair mercury can only correlate to blood mercury when the hair is actually growing. This correlation is pretty good. That is not something that can be told from this data set because the blood mercury was not measured while the hair was growing.
While hair can be considered an "excretion" pathway for mercury, it is at most a very minor pathway. The reason it is minor is because hair growth is slow. In a year, a person may grow a few 10's of grams of hair, lets say these children grew 30 grams of hair in a year. The average hair mercury was about 2 ppm. That would be about 60 micrograms of mercury in 30 grams of hair. 4 ounces of tuna at 0.5 ppm has as much mercury as a child would put in their hair in a year. What fraction of ingested mercury ends up in someone's hair? A few percent at most. What about the other 95%?
The major excretion pathway is via the stool. Was stool mercury measured in any of these children? No, it wasn't. Not in the Ip et al study, not in the Holmes et al study, not in the DeSoto reexamination. Trying to infer excretion kinetics by only observing a very minor excretion pathway and without knowing total mercury ingested or total mercury excreted is a useless exercise. A reliable estimate cannot be made.
You say that my statement "The time constant for fluctuations in hair mercury is a not longer than the time constant for fluctuations in blood mercury. Hair mercury is probably a better indication of long term mercury exposure (the exposure most relevant to a mercury causes autism idea)." is only valid if I prove that ASDs have the same correlation between hair and blood mercury as non-ASDs. To which I say "huh?" Hair grows only because it is supplied with blood. If that blood contains mercury, hair growing at a time of a certain blood mercury level will reflect that mercury level. Since hair accumulates over time, where blood doesn't, mercury measurment of hair will reflect the average of mercury in that hair over the length of time that the hair grew. Blood mercury is an instantaneous point measaurement. It reflects the amount of mercury in the blood at the time of measurment.
If there is some variability in the ratio of hair to blood mercury, that variability can only add to the variability reflected in the actual variability in blood vs the averaging that occurs in hair.
Was the gender of the ASD children and the controls matched? In this paper
on page 312, they mention that females have faster mercury excretion kinetics that males, both for methylmercury and inorganic mercury, and that males have higher hair mercury than females. Males have a 3-4 times higher incidence of ASDs than females. If the sample was not sex-balanced, that alone could explain the (very small) difference. On page 311 they list hair mercury levels in 35 different countries. They are all over the map, literally. From averages of 7.9 in Kenya (n=71) to 5.0 in Japan and the UK (n=1916 and n=1223) to 0.3 in Nepal (n=45). Do Nepalese suffer from "low mercury excretion disorder"? If we apply the "logic" of Holmes et. al, and DeSoto, they do.
The Ip et al data and the DeSoto reanalysis shows that none of the ASD children had a toxic level of mercury in their blood, or in their hair. There were some differences in the two populations, and those differences could easily be explained by things that were not measured. There is no need to invoke some completely non-physiological fantastic made up ad-hoc fairy tail of "low mercury excretion disorder".
You are correct, I have not yet purchased the Holmes et al study. However, the DeSoto study goes on to describe exactly what you did in terms of the approach to estimating mercury exposure so it is not new information for me.
I'm curious if you feel that the use of surveys in estimating exposure is an invalid method?
I have read at least two other studies that used surveys to estimate exposures. At a minimum, the methodology outlines that the two clusters of children (autistic vs control) have different functions of hair excretion compared to estimated mercury exposure.
Upon first glance at the study you referred (I assume I found the right one) you'll note that they collected hair from children 1-5 years old. That is a very different timeframe from babies first haircuts and I would not expect the mercury levels to be the same at all since these samples would definately be time sensitive. As DeSoto stated about the Holmes paper, that the baby first haircut was used as an indicator of in-utero and very early life exposure, very different from haircuts of 1-5 years old (a very large range as well). It is certainly not proof of error since they're measuring very different exposure timeframes and ages of children.
Why do you think an absolute number comparison is valid?
Going back to Holmes, the absolute measurement isn't the most important point in testing that kind of hypothesis -- using a survey to estimate exposure is obviously not going to give you accurate absolute numbers. It's all about the correlating function of estimated mercury exposure vs hair excretion in the two clusters of children. That is a statistical exercise. It is possible that Holmes et al, had problems with their statistical work -- if that's the case, then I'm interested in reading about it.
Back to the real point of DeSoto et al. Even if they weren't entirely accurate in using the word "test", their intent of the analysis is to show that the Ip et al numbers support the Holmes Hypothesis. I did not find any mention of Holmes et al's absolute values in the DeSoto paper, nor does it try to correlate any absolute numbers to it.
Additionally, aside from the parts of the DeSoto study I quoted above, I could not find any other reference to the Holmes et al study at all. I might have missed other references to Holmes (I searched but found no others), and I don't see any passage in the study where they state an "unquestioning acceptance" of the study at all. In fact, the whole exercise was to use the Ip et al data to test the hypothesis which is exactly what they proceeded to do.
I can certainly agree that their introduction discussion and use of the study reference claiming that Mercury poisoning mimicked Autism is a stretch at best (totally inaccurate at worst), but the methodology of the calculations and application against the Holmes hypothesis does not appear to have any fundamental flaws.
You consistently miss the specific point of this analysis and go off on tangents.
Are you trying to support the Ip et al conclusion?
Your post implies that the Ip et al and holmes et al methodology is completely flawed because without continuously measuring the mercury level in blood, one can never test the correlation with mercury hair levels. Since the half life of Methyl Mercury is estimated to be 21 days and Ethyl Mercury 2-8 days, I would expect a very spikey graph blood/mercury graph and of course those hourly fluctuations could never be reflected in hair growth right?
Of course we can also ignore the data that shows a correlation between estimated blood mercury and hair mercury levels in the controls because it must be coincidence right?
So your assertion (not that I agree with it -- and I am curious who else supports you there) that all of these studies are fundamentally flawed effectively makes all of the experts for the defence that reference the Ip et al study (or similar ones) as credible evidence are guilty of being duped by a significantly flawed methodology. That would also apply to anyone using hair to estimate mercury exposure in any study because we've never been able to show that autistic children excrete mercury in hair at the same rate as controls.
So now if all those defence experts were so easily duped by such a flawed methodology, that begs the question as to what other studies they were duped by.
So if we go back to my original point, you're basically saying that the platiniffs don't need the DeSoto study to invalidate the conclusion of Ip et al, since you've shown that this methodology is fundamentally flawed and therefore was already nullified.
OK, I'll accept the final result, even if I disagree on how you got there.
Very interesting information. I'm assuming that the political capital involved in actually making an issue of this would not be worth expending, or proving intent to influence is too difficult in court.
That is very unfortunate that this type of behaviour is just accepted as par for the course.
Thanks for the insight.
Schwartz, The specific points of your analysis are flawed. They take one aspect of one paper and extrapolate non-physiological assumptions where they make no sense.
The "experts" do know the limitations of the techniques they use and the data that they collect. It is non-experts who take that data and make speculative assumptions and then wildly extrapolate to regions very far from where the data or any understanding of physiology suggests is reasonable.
Everyone who understands mercury physiology knows that mercury blood levels do fluctuate and that hair levels are an integration of those fluctuations (transduced through the blood-hair transfer function).
Everyone who has read enough about mercury physiology knows this. There is still a reasonable correlation between blood mercury and hair mercury, provided that the blood mercury stays reasonably constant over that time period that the hair grew. When people have a stable diet and stable mercury exposure, their blood mercury levels do stay pretty constant and so the blood-hair transfer function will stay pretty constant too.
When there is a mismatch between blood mercury (which is an instantaneous point value) and hair mercury (which is an integrated value over months of time), the simplest explanation is that the blood mercury was not constant over that time. The only medical reason for measuring mercury is to find toxic levels. Fluctuations well below toxic levels have no known medical relevance. A toxic level will show up in hair, or in blood with pretty high reliability. If either one showed a toxic level, you would treat. If neither one showed a toxic level you wouldn't.
With the right techniques you can measure the mercury content of a single hair as a function of length, and see the changes that occur over time. People have done this, and they have seen some of these fluctuations. The tests don't have great fidelity because there isn't very much sample in a single strand of hair.
These studies are not "flawed", rather the data they collected has its limits and one cannot take small differences in the data and wildly speculate that those differences mean what you want to assume that they mean. That is what DeSoto did. The absolute differences in the two populations is small, small compared to differences observed in other populations. What does a small difference mean, even if it is "statistically significant"? Not a whole lot. When children in the Faroe Islands have mercury levels an order of magnitude higher with no apparent excess of autism, what do these small fluctuations mean? Maybe all they mean is that in this group, the people with autism ate a little more fish per month than the people without autism. Maybe there is a disparity in income between the families of children with and without autism, and that reflects in the type and amount of fish that they eat. A change in fish consumption by 1/4 might be enough to account for the entire difference. Eating fish 3 times a week instead of 4. Eating big portions instead of small, eating high mercury fish instead of low mercury fish.
Scientific papers are not meant to be read in isolation. They are scientific communications from one expert in the field to another expert in the field. The experts in the field know the background and the literature, and so they understand the limits of the data and the studies.
I think the Holmes et al paper is available on the SafeMinds website. You could probably find it with Google scholar.
Burton's Balogna about a Desoto driven by a Hitman:
1. DeSoto and Hitlan/Holmes/IP (or is that LP) did not prove the mercury exposure came from vaccinations.
2. While the Causation was Questioned it was Correlation addressed in the 20 hearings held
3. Drinking the Epidemic Koolaid (he uses the E word more than once)
While spending MY TAX DOLLARS FRIVOLOUSLY (Yours too.) entertaining
4. (in)credible national and international guests (scientists)
To discuss (testifying is not properly documenting with a Study) (Ah swear ah Saw a UFO Sir)
5. Thimerosal/Mercury in Vaccines
Whose source wasn't even proven under item 1.
6. He claims there is a growing body of evidence and DOES NOT CITE IT. (or link to a good reference point)
Just like most of the CrankSupporters I see trying to forward their cause as weblog respondents.
7. He States that it is the DUTY of the Office of Special Masters to review all sound science relevant to the matter.
Definitely trying to exert some influence.
Almost forgivable, but it looks like hasty typing or dictation, there is a grammatical problem where the letter uses 'aware this' instead of 'aware of this'.
Finally he makes a 'forgone conclusion'
8. "... millions of children, adults and families afflicted with autism linked to a childhood vaccine."
He just fed the conspiracy nutz everything they needed in number 8, even if it was inadvertent. Lets hope they don't realize it.
Can somebody find me a larger paper bag to put over my head?
I'm bookmarking this stuff for the primary elections when Burton (who is an insurance salesman, not even a lawyer) is facing an ER physician for the Republican nomination.
McGoff for Congress.
fusilier - embarassed in Indianapolis
There are a couple of flaws in your logic. You state: "The only medical reason for measuring mercury is to find toxic levels."
I would argue that measuring mercury is important for DETERMINING toxic levels. That has not been done for all types of mercury as the toxicity of Ethyl Mercury is not well studied.
Who's definition of toxicity are you using? The published CDC level? How is it that the published toxicity level for lead is continually lowered over a period of decades, and now they are saying that toxicity occurs at levels below the current definition? You have no proof of toxicity, just like you have no idea what effect low levels of continuous mercury exposure will have over a long period of time, because it has not been effectively studied.
As for methodology, you either agree or disagree with the methodology used by Ip et al. If you agree with the methodology, then you have to agree with the conclusion of DeSoto et al, because they used the same data, and tested it against the same hypothesis.
If you don't agree with Ip et al, then their conclusion is already nullified and your defence experts clearly don't "understand the limits of the data and the studies". That calls into question the professional judgement of those defence experts then.
Take your pick. Either way, the Ip et al conclusion is nullified.
If you look on Safeminds (which I did long before your post) you will not find the Holmes study. But again it doesn't matter. As I stated earlier, Prometheus has not pointed out anywhere in the DeSoto study where they even reference Holmes et al outside the quotes I already pointed out, let alone the data from the study. Your statements (along with his) that they wholesale supported that study is an exaggeration. Please point out the places if I missed something here.
I'm also still waiting for a logical reason for comparing the absolute values of the two studies mentioned. Your own descriptions of mercury measurement above point out how flawed that logic is.
What you say about measuring the Mercury in hair makes sense. Presumably when measuring blood and hair at the same time, there is probably a decent chance of correlation since it takes time to clear mercury from the blood the a current blood measurement is likely indicative of past exposure unless a recent spike had been encountered.
So how do you explain a good correlation within the control group, and a much different correlation in the Autistic group? Coincidence?
If you search via Google Scholar on the title, it shows up in 17 versions. "Reduced Levels of Mercury in First Baby Haircuts of Autistic Children"
It is avialable on the SafeMinds website.
The correlations between hair and blood mercury are not that different in the autism group in the Ip study. What did Prometheus calculate? 0.12 vs. 0.15 in the control group. What is the "difference" that needs to be explained?
Let me try and understand your "logic". Because many well done studies show absolute mercury levels and absolute mercury differences between different healthy groups in mercury levels that are much larger than the tiny difference measured by Ip et al, we should ignore those studies and ignore the absolute mercury level and focus on the tiny difference found in the data of Ip et al? Huh?
You are the one who wants to ignore reliable data, it is you who needs to explain why ignoring reliable data will lead us to a more acurate understanding of reality. My experience is that ignoring reliable data never leads to a more accurate understanding of reality.
The level where lead causes adverse effects is being dropped because of data that shows there to be actual adverse effects at the levels under consideration. There isn't any corresponding data on mercury.
The best and largest data sets of mercury show no or only tiny effects at levels an order of magnitude larger than what Ip et al found. The effects that are found do not mimic the symptoms of autism. Not at any level of mercury.
If the "mercury causes autism" idea was correct, then the incidence of autism should track mercury levels in the population. Places with very high mercury levels such as the Faroe Islands should have very high autism levels. They don't. If mercury causes autism, then there should be a very high incidence of autism in the Faroe Islands. There isn't a high incidence there.
That data, that there isn't a high incidence of autism in the Faroe Islands proves that mercury at the levels measured in the Faroe Islands doesn't cause autism. The levels measured in the Faroe Islands in non-autistic children were much higher than what Ip et al measured in autistic children.
I appreciate that you don't want to face that data because you want to cling to your belief that mercury causes autism. It doesn't. There is no data that suggests it does. It is time to stop beating the dead horse that is the "mercury causes autism" idea and divert those resources to finding the real cause and developing real treatments.
If you go to page 279 of the Holmes et al paper, you'll find that the "first baby haircut" occurred between 11 and 24 months of age (mean 17.7 months). The NHANES data looked at ages 1 - 5 years (12 to 60 months). While the overlap is not exact, the NHANES data does overlap the Holmes et al sampling age.
Now, another big difference between the Holmes et al data and the NHANES data is that NHANES analyzed their hair samples immediately, while Holmes et al analyzed the hair about 6 - 7 years after it was cut. As you might imagine, this time delay between cutting the hair and analyzing it is not insignificant.
As for the idea of collecting mercury exposure data by survey, you should look at the questions they asked. It is also significant that they are asking questions about dental work and diet 8 years (on average - the range is 3 - 17 years) after the child in question was born.
Asking dietary questions - even about relatively recent meals - is known to yield inaccurate information. How much more so after several years.
Actually, the biggest complaint I have about the way they "analyzed" the Holmes et al data is that they concocted a "formula" that took the survey information and adjusted the wighting until they came up with a good fit to the "control" hair mercury levels.
They then applied that same formula to the survey information from the "autistic" group and report that it predicted much higher hair mercury levels - ergo, they develop a fantasy about "poor excretors".
However, since we now know that the Holmes et al "control" hair mercury was completely off the scale (over 5 times the 95th percentile of the NHANES data), the whole analysis collapses, including their "poor excretor" hypothesis.
Is there anything unclear about that?
That is still a very big difference in ages. If one considers that most Thimerosal exposure occurs between 2 - 24 months in the form of vaccines, it is a very notable difference. Childhood vaccine applications (and thus ethyl mercury exposure) drop off significantly after 2 years of age. Also, I couldn't find the mean age listed in the NHANES study, but the Holmes study mean age of 17 months would capture most Vaccine exposure.
The NHANES study also didn't include vaccinations in their survey which means they certainly would have missed any vaccine correlation. Given that Ethyl Mercury half life is much shorter than Methyl, the outliers (3 of them) could have been caused by recent vaccinations. (I guess in 1999, they hadn't realized the level of mercury exposure from Thimerosal in vaccines). We don't know how many children under 2 that they tested unless I missed the data somewhere (that is possible). That alone still casts doubt on the validity of your comparison.
It is interesting to note that 2 of the 3 outliers excluded due to unusually high levels of mercury were ages 1 and 3.
" As you might imagine, this time delay between cutting the hair and analyzing it is not insignificant."
You have to be more specific than that. I was not aware that heavy metal components within hair would degrade over time? Are you suggesting that the mercury level might increase over time? If anything it might undermeasure the real value.
It also looks to me like the Holmes et al study included the full hair sample in their analysis, while the NHANES study only used a small amount to correlate with the small period of time of their survey -- vs Holmes much longer period of time.
If the average age of the Holmes study is 17 months, than you are probably looking at a 17 month cumulative exposure vs the very limited 30 day exposure (from the small hair sample).
Given the sample differences, I would expect the hair/mercury level to be far higher from the Holmes study. You are certainly not comparing apples to apples here.
Did you consider that?
My eldest daughter is 6 years old now. I know exactly what our diet was like (particularly the fish part -- we ate a lot of fish) in her first two years of life since new parents often pay a lot of attention to things like that. I know lots of parents that write down a whole lot of information about their kids from early ages. I also know exactly how many, and what type of dental work I have. Dental work is usually quite memorable to some people (given the unpleasant nature of the experience I would venture). I agree that surveys will not yield exact absolute numbers and I'll admit, I may not qualify as a typical example, but both arguments are anecdotal.
As for their statistical formula, I am interested in what your specific objection was. Measuring exposures to Thimerosal seems reasonable, so I assume your objection is to the Amalgam and diet components. The diet component is clearly the least reliable of the bunch. Is there a specific issue with using the Square of the Amalgam exposure? If so, you still have a fairly reliable number in the Thimerosal exposure.
Even if you don't like the correlation calculations (and the authors note the weaknesses of the study) between hair and estimated mercury exposure, you still have a very distinct difference in hair mercury levels between the Austistic children and the matched controls. This alone mertis further investigation.
So to summarize your comparison:
In one study they measure the total cumulative hair mercury level to an average age of 17 months.
The NHANES study measures the cumulative hair mecury of the previous 30 days for children aged 1-5 with no average given.
Looking at that, I would fully expect the first study's measurements to be much higher especially given that Thimerosal exposure is highest during the first 24 months.
Am I missing something here?
I'm glad you linked the study because I was about to buy it. I did search the research page on the Safeminds site and it is not linked there (so I'm not going crazy), but obviously they have it on the server and google found it.
This is taken from the Ip et al data:
"To check the hypothesis that hair excretion was overall lower than would otherwise be predicted based on a certain blood level in the autistic group, a best fit regression line was calculated (y = 10.3, x = â2.48) indicting that for each unit increase in hair level, blood level increased by 10.3 units. A t test on the residuals showed that autistic participants were significantly more likely to have lower hair mercury levels than would be predicted as a function of their blood levels, t(133) = â2.92, P less than .005; see Figure 1). It should also be noted that the presence of unequal variances or nonrandom residuals (in this case, autistic persons are both more likely to have greater variability at high levels of circulating mercury and a lower hair value for a given blood level) are both violations of important assumptions of the t test; a t test of hair mercury is therefore probably not a valid means to predict autism diagnosis as a function of mercury exposure. We performed an analysis of covariance (ANCOVA) with autism diagnosis as the independent variable and hair mercury level as the dependent predictor using blood levels as a covariate. Results indicate that hair level may be related to diagnosis of autism, not as a predictor in terms of absolute value, but such that for equivalent circulating levels of mercury in the body, those with ASD excreted less than normal such that F(1,134) = 3.9 and P = .05. To sum, the relationship between blood levels of mercury and mercury excreted in the hair is reduced for those with autism compared with nonautistic persons; furthermore, the difference between autistic and nonautistic persons is most pronounced at high levels of mercury."
That certainly seems to indicate a statistically significant difference in the ratios and also shows how the Ip et al conclusion is flawed. This is based on real data in the Ip et al study.
"Let me try and understand your "logic". Because many well done studies show absolute mercury levels and absolute mercury differences between different healthy groups in mercury levels that are much larger than the tiny difference measured by Ip et al, we should ignore those studies and ignore the absolute mercury level and focus on the tiny difference found in the data of Ip et al? Huh?"
You forget, these differences were compared with controls and the result was statistically significant. Why do you continue to ignore that?
You are continually trying to move the discussion away from the details of the study in question. If you claim that the measurements from this study and the statistically significant differences are not applicable, than you are again saying that the whole study is incapable of answering the hypothesis proposed by the study itself.
"The level where lead causes adverse effects is being dropped because of data that shows there to be actual adverse effects at the levels under consideration. There isn't any corresponding data on mercury."
You're right and until a few years ago there wasn't any data on low levels of lead exposure to determine the appropriate level of toxicity. Here is the kicker. They still set a level of lead toxicity incorrectly based on no credible data.
Since very few studies have been done on long term low level mercury exposure, I can safely say that your "defined" level of toxicity is as much a guess as the previous and current levels of lead toxicity. It is a meaningless number until you have data (which you don't).
"I appreciate that you don't want to face that data because you want to cling to your belief that mercury causes autism. It doesn't. There is no data that suggests it does. It is time to stop beating the dead horse that is the "mercury causes autism" idea and divert those resources to finding the real cause and developing real treatments."
I really dislike the way you try to boil a problem down to a simple flawed analogy and then put words in my mouth (again).
If I thought that Mercury was the sole factor in causing Autism (which I don't) then maybe, your analogy of Faroe islands would be valid (which it's not). Of course we would also have to assume that all kinds of mercury are the same and have the same effect in the human body (which they don't). We would also have to assume that the Faroe islands residents were comparable in all other respects to the general worldwide population (which is very unlikely), and we would have to assume that only a single element is the cause of all autism cases. Yes, in that case you're right, we would see something. I guess one of those assumptions must be false.
You'll also note that the DeSoto study does not say high levels of Mercury correlate with Autism diagnosis. They actually say that the mismatch between hair and blood levels is a predictor of Autism diagnosis, so that pretty much rules out the analogy to high levels in the Faroe islands.
Please remember, this specific discussion is about the DeSoto et al study and it's pertinance to the Omnibus trial. Not my personal views of Autism and the role that mercury exposure might play or how we might allocate funds -- I don't feel all or even the majority of funding should investigate mercury regards to Autism either so stop implying that I do.
Both the Holmes study, and the Ip et al study found statistically significant differences between the mercury hair content of Autistic children vs controls. (Holmes found lower absolute values, Ip et al shows differences in relationship to blood level). You continue to choose to ignore that data.
Schwartz, a "statistically significant" difference does not mean a "meaningful difference" or a "causal difference" or an "important difference". Prometheus pointed out the excellent statistical correlation between shoe size and reading ability in grammer school. Does that mean that reading ability resides in the feet?
Holmes et al didn't measure mercury consumption. They asked a few questions about a few sources of mercury, and ignored everything else.
Mercury in hair may not be completely stable. There are differences in measured mercury levels due to hair color. Similiarly there are differences based on latittude of residence. Some of that is thought to be from UV reduction of bound mercury and liberation of mercury as vapor. If so, then there could be signficiant migration of mercury in a sample of hair, liberated at sites where it is exposed to light, migrating to darker regions. Depending on how the hair specimens were sampled for analysis, (the part exposed to the light, or not), there could be large differences even in the same specimen depending on storage and sampling.
The Ip et al data were from samples taken at the same time, that means the hair mercury would reflect blood mercury levels months earlier when the hair was growing, not the instantaneous blood level on the day of sampling.
The difference is 0.12 vs. 0.15. Where is the large difference that needs explaining? There was large scatter in the ratio for both groups of children. The distributions largely overlap.
If the absolute level of mercury isn't important, then what you are saying is that mercury is not a causal factor in the development of autism. That is what the Autism Omnibus trial is about. Did the mercury in vaccines cause autism in the 5,000 plaintiffs?
Ethyl mercury might be different than methyl mercury which might be different than inorganic mercury, which might interact with every other element in the periodic table. The idea that mercury causes autism is an idea that has no data to support it. When an absence of data is found to support a hypothesis, a scientist will abandon it as a failed hypothesis. A scientist doesn't change the hypothesis making it ever more complicated in the absence of data. The "low mercury excretion" idea is an ad hoc made up idea simply to explain why Holmes et al didn't measure higher levels of mercury in the ASD children's hair. There is no data to support it, no phsyiology to support it, nothing to support it except the belief on the part of Holmes et al that ASD children must have toxic levels of mercury even if that mercury cannot be measured. Science is not about belief, it is about data.
There may be things that are unexplained in the Holmes and the Ip studies. That does not give one free license to assume what ever one wants to provide an explanation. There are multiple explanations that are much more plausible than the "low mercury excretion" idea. None of those more plausible explanations can be excluded on the basis of the data that was measured in those studies.
They didn't measure shoe size. By the Prometheus correlation, that will reflect reading ability. Actually, there is also a correlation between shoe size and a diagnosis of autism. Children with a shoe size below a certain level don't have a diagnosis of autism.
You can cling to the non-physiological notions expressed by Holmes and DeSoto. I prefer to accept the understanding of physiology that is laid out in thousands of other papers rather than discard them (which I must do) to accept the non-physiological notions of Holmes.
It is your perogative to continue to focus on the aspects of the data (and sentences) you like and ignore the rest.
"Prometheus pointed out the excellent statistical correlation between shoe size and reading ability in grammer school. Does that mean that reading ability resides in the feet?"
Not when you have another variable that is correlated with feet size that explains the difference. You don't have one here. Please stop with the invalid analogies.
"If the absolute level of mercury isn't important, then what you are saying is that mercury is not a causal factor in the development of autism. That is what the Autism Omnibus trial is about. Did the mercury in vaccines cause autism in the 5,000 plaintiffs?"
I never stated absolute levels of mercury weren't important. I stated that for the purposes of studying hair/blood mercury correlations, the absolute levels weren't important. Same goes when comparing levels to controls as long as you are measuring the controls the same way as your study group.
Even still, Prometheus' complaints about absolute levels still don't stand up to scrutiny without further data since the study he's comparing to is measuring different amounts at different ages. A rough calculation actually shows the amount to be underestimated since using simple math the value should be 17 times higher on average (one month vs 17). But, of course that calculation is far too simplistic a forumula and I wouldn't expect to see results like that.
Additionally, mercury doesn't have to cause Autism to be a problem. All it has to do is be a contributing factor -- you continue to put words in my mouth and simplify the problem to a ridiculous level.
"Mercury in hair may not be completely stable. There are differences in measured mercury levels due to hair color. Similiarly there are differences based on latittude of residence. Some of that is thought to be from UV reduction of bound mercury and liberation of mercury as vapor. If so, then there could be signficiant migration of mercury in a sample of hair, liberated at sites where it is exposed to light, migrating to darker regions. Depending on how the hair specimens were sampled for analysis, (the part exposed to the light, or not), there could be large differences even in the same specimen depending on storage and sampling."
You'll note the NHANES study referenced by Prometheus did test for hair treatments and found no difference at all. You'll also note that most of the controls in the Holmes study were matched for location.
However, if it is such a big problem, then you will have a problem with every study using hair to measure mercury, meaning ip et al as well. Then we end up in the same place again, the study methodology is flawed and Ip et al, and we have to question their judgement along with anyone who references them? Which is it?
I find it interesting that you are arguing that statistical significance doesn't allow you to prove or disprove a hypothesis. Either the methodology of this type of study is flawed or not. If it is not flawed, then a statistically significant result should support the hypothesis. If not, we go back to the previous argument.
"The difference is 0.12 vs. 0.15. Where is the large difference that needs explaining? There was large scatter in the ratio for both groups of children. The distributions largely overlap."
First, DeSoto et al specifically looked at the scatter which showed that the study group's scatter was different from the scatter of the control group and it was statistically significant.
I am sure you realize that the reason we perform statistical analysis on data instead of looking at absolute values is to determine correlations and patterns that may not be obvious when looking directly at the absolute values. If you find a statistically significant result, then one can state support for the hypothesis with a greater than 95% percent certainty. If you want to start debating the statistics, then you should start being a lot more specific in your issues.
"The idea that mercury causes autism is an idea that has no data to support it."
You keep going back to this. We're talking about DeSoto et al, and Ip et al. Until a couple years ago, the idea that levels of lead below 10 micrograms per deciliter caused serious problems had no data to support it. I guess lead just suddenly became a problem in the last couple of years.
"There may be things that are unexplained in the Holmes and the Ip studies. That does not give one free license to assume what ever one wants to provide an explanation. There are multiple explanations that are much more plausible than the "low mercury excretion" idea. None of those more plausible explanations can be excluded on the basis of the data that was measured in those studies."
To quote from the study summary:
"Moreover, the hair sample analysis results offer some support for the idea that persons with autism may be less efficient and more variable at eliminating mercury from the blood."
That is hardly stating an absolute conclusion as you allege -- again you exaggerate things you don't like into false allegations. It says that the data offers some support to the idea.
I'll ignore the shoe size analogy for obvious reasons.
"You can cling to the non-physiological notions expressed by Holmes and DeSoto. I prefer to accept the understanding of physiology that is laid out in thousands of other papers rather than discard them (which I must do) to accept the non-physiological notions of Holmes."
I am not clinging to anything, you are again attributing a position to me. I am looking at the data being presented in both cases: Ip et al, and DeSoto et al.
To quote Prometheus: "So, unless D&H (or Holmes et al, Kern et al, or Whoever et al) have shown that hair mercury level says anything about mercury excretion (by the organs that really excrete mercury - the kidneys and liver), then all they've shown is that - in this group of children, the autistic group had a higher blood mercury."
Yes, that's exactly what they did.
This is from Ip et al: "We aim to see if there is increased mercury exposure in children with autistic spectrum disorder."
I maintain that DeSoto et al is a valid submission to nullify the Ip et al conclusion in the Omnibus trial.
I think that the physiological aspect should absolutely be investigated. You have data from two studies that indicate that there is a problem with the current physiological understanding around autistic children, blood mercury levels, and hair mercury levels. That certainly merits investigation.
In the case of Prometheus's correlation of reading ability and shoe size, yes, we do know that there is another factor that actually correlates better and that is in fact more directly related to reading ability and that is age. We don't know if there is a confounding factor related to ASDs that modifies mercury absorption or excretion. That is because we don't know what causes ASDs.
That confounding factor might be oxidative stress. Mercury physiology is highly related to thiol physiology and thiol physiology is highly related to oxidative stress. People with ASDs are in a state of oxidative stress. It could be diet, people with ASDs do tend to have different eating habits than people without ASDs. Neither of these were measured, considered or accounted for in either the Ip study or the Holmes study. You want to assume there is no confounding factor, and that we can put blinders on and ignore every other bit of research that has ever been done and only focus on this one study. It could be ethnicity. The Ip study was done in Hong Kong. Were the two groups racially balanced? Hong Kong has a lot of visitors from mainland China.
You want to assume there is no confounding factor, and then use the mercury data to show that there is no confounding factor. You can't do that. If you assume the only important factor is mercury, any analysis you do will only confirm that.
The Ip et al study did not find toxic levels of mercury in ASD children. They did not find that ASD children had a ratio of hair mercury to blood mercury such that toxic levels in blood could occur with sub-ppm levels in hair (as Holmes et al assumed).
Every piece of data that can be measured can be measured wrong, and can have artifactual errors in it.
Usually when hair mercury is measured, it is measured soon after sampling, not many years later as the samples in Holmes et al were. Do we know what happens to the mercury level in hair samples when they have been stored for many years? Under variable conditions? If I were to store samples, I would want them in a sealed container that I knew was impermeable to what ever it was that I wanted to measure. How were these samples stored? Is it even known? If we don't know how storage changes samples, then we have to assume something. You want to assume that nothing happens. Maybe that is correct, depending on how the sample was stored, maybe not. They tested for some hair treatments. They did not test for storing the samples for many years in envelopes in a dresser or a locket with other infant memorabilia. Do such storage conditions change mercury levels? Do they cause migration of mercury in the sample? Without testing, the answers to those questions are unknown. I would want to know the storage conditions, the hair color, and perhaps some other things. I might want to analyze the containers the hair was stored in to see if the container picked up or lost mercury.
I agree that the physiology should be investigated. Why haven't the plaintiffs done so? They have had many years, they have spent a great deal of money on legal fees, many of them have spent a great deal on chelation treatments, yet there are no studies showing mercury levels in these 5,000 plaintiffs. There are no studies that measure mercury levels in ASD children. Why is that? I suspect it is because the data would not show what they want it to show. There are many scientific questions that could have been answered regarding mercury physiology in ASD children. I think the reason those studies have never been done (or have never been published if they were done) is because the plaintiffs don't want to answer a scientific question; they only want to show that mercury causes autism, whether it does or not.
The plaintiffs could certainly have actually measured mercury ingestion and mercury excretion, even if just with a normal diet. That would have shown if ASD children had "mercury efflux disorder". The plaintiffs either didn't do the study, or are not showing the data.
It may very well be that people with autism do have a slight perturbation to their mercury physiology. There are reports that females excrete mercury more rapidly. Females do tend to have a lower level of oxidative stress, due to their increased nitric oxide levels from estrogen. People with autism do have lower NO levels and higher levels of oxidative stress. Metallothionein is expressed in response to oxidative stress. Metallothionein is the major storage site of mercury. Oxidative stress could well cause a perturbation in mercury physiology and in mercury excretion. But so what. That is like shoe size and reading ability.
When there is a plausible confounding factor, that confounding factor can't simply be ignored the way that you want to do.
All of the problems you discuss can apply to any study including the ones reference by the CDC and IOM with the exception of direct physiological evidence (little of which exists).
The irony is that evidence based medicine puts a very high weighting on epidemiology which is rife with the type of issues you discuss. In fact, much of it is even worse because it relies on the subjective definition of Autism which has changed over time and it also relies on data collection methods that are rife with inaccuracy. This is also the bulk of the evidence used to disprove any association between Autism and mercury. If you go back to the big IOM conference in 2004(?) you'll note that most if not all of the evidence presented to disprove a Mercury/Autism association was epidemiological in nature. Most if not all of the evidence presented by the Autistism lobby groups was physiological in nature and deemed not as credible.
You ask why the plaintiffs don't do physiological studies. Only recently have lobby groups obtained the financial wherewithal to fund such things. You seem to ignore that most of the research funding is dictated either by government or big pharma neither of which had any interest in funding physiological studies about mercury -- witness the lack of any safety data for thimerosal, a compound grandfathered because it was so old.
And they are now funding these studies. Unforunately, there are very restrictive statutes of limitation in which they can bring a case forward.
If you follow the stories of the parents in the Omnibus trials, they have spent a lot of money trying to diagnose and help their children. How could they ever fund a large enough study to satisfy the rest of the industry? It is impossible. Performing a credible study on the 5000 applicants would be very difficult and expensive as it would have to involve a lot of experts, be properly designed, and must involve a large enough control group.
Do I think this study is conclusive proof that Mercury causes Autism? No. I am quite aware that there can be a lot of confounding factors and this study only shows a correlation. Alone, it is not proof of causation. However, the data collected by the study supports the conclusions by the Authors (DeSoto et al) and as such should be considered in the body of evidence. It can also point to areas for further research.
I am not satisfied by the explanation given by Holmes et al
on why the hair mercury level does not match the blood mercury level. But that does not change the data in these two studies which indicate that there is some difference in the physiological treatment of mercury in Autistic children. It could be causative, or it could be an effect of some other Autism feature. Only further study will determine the answer.
I also think that a lot more effort should go into classifying different types of Autism and the related physiological problems far more acurately. I find it completely astounding that the CDC hasn't been able to even accomplish that much given that they've been tracking this problem for a long time now.
I've noticed that you seem to use Mercury and Thimerosal interchangeably. If we are discussing thimerosal in vaccines as a possible cause of autism, the subject of "the big IOM conference in 2004", the strongest evidence against an association is the fact that autism rates haven't wavered in response to removal or reduction of thimerosal in vaccines.
You said: "Alone, it is not proof of causation. However, the data collected by the study supports the conclusions by the Authors (DeSoto et al) and as such should be considered in the body of evidence. It can also point to areas for further research."
How does it support the author's conclusions and how on earth does it point to areas for further research? I can only see it pointing to further research on mercury as a possible cause of autism. How much time and how much research do we really need to answer this question?
The Ip et al study was not designed to examine small differences in mercury physiology between people with ASDs and people without them. It was a simple study designed to see if there were large differences. They did not find large differences, they only found small differences. Differences that are smaller that those that have been observed in many other populations with seemingly no adverse effects from the different mercury levels, and with seemingly no excess of autism.
Ip et al didn't need to look at myriad confounding factors to answer the question they were looking for, are there large differences in mercury levels between ASD and non-ASD children. The small differences that show up in their data might be real, they might be artifacts. In either case they are irrelevant to the question Ip et al were trying to answer, are there large differences in mercury levels in ASD vs. non-ASD individuals. The answer to that question is no, there are not large differences in mercury levels.
If you want to answer the much more complex question of is mercury physiology different in ASD vs. non-ASD individuals, it takes a much more complex study. It would require actually looking at mercury ingestion and excretion over time. The only way to do that is to measure it. Measure mercury intake by various routes and measure mercury excretion by various routes. If we had that data on a number of different ASD and non-ASD individuals, we might see a difference, we might not. No doubt there are some idiosyncratic differences between individuals independent of their ASD status. There are known differences related to gender, age, diet, perhaps ethnicity, hair color, sun exposure, and probably a lot of other things too. Once you control or at least account for other known differences (and perhaps as yet unknown differences too), then you could figure out if there are ASD differences.
Once you understand if there are differences in mercury physiology in ASD vs. non-ASD individuals, you might try to determine how those differences are related to ASDs. Is the mercury level causal? Is there some pre-disposing factor where mercury will cause an ASD in one individual but 1000 times more mercury won't cause it in another individual? That is an extremely complex question.
You can't do the experiment of exposing people to different mercury levels and seeing if they get an ASD. Each individual is genetically unique, so unless you had myriad clones the experiments wouldn't tell you anything except for the individuals you were testing. Even identical twins are not identical genetically. They sometimes have subtle gene copy number variations (which some suggest might be what causes ASDs, I think they are wrong, but they think they are right).
A question you might be able to answer is "is there some mercury level below which ASDs do not happen"? Ip et al found multiple people with ASDs with blood mercury levels of 5nM/L. There are a number at the same level, so I presume that is the detection limit. If there is a mercury level below which ASDs don't happen, it is lower than 5 nM/L. How much lower? If we believe the Holmes et al measurements on hair levels, their data implies blood mercury levels lower than what Ip et al measured. Ip et al found a sizable fraction (5-10%) of their ASD group had very low mercury levels, less than 10 nM/L. In some of the other papers on mercury and autism, there have been individuals with autism that had no detectable mercury on chelation challenge. If there is no detectable mercury present, how it mercury causing their autism? If there is no mercury level below which ASDs do not happen, how can mercury be considered to be a cause of autism?
One way of doing hypothesis driven science is to make a hypothesis and attempt to falsify the hypothesis that is to find data that shows the hypothesis to be false. Ip et al had the hypothesis that there would be large differences in mercury levels in ASD and non-ASD individuals. They found small differences not large, so they falsified the hypothesis that there are large differences.
My question to the proponents of the "mercury causes autism" idea is what data would convince you that the hypothesis was false? If there is no such data, then the "mercury causes autism" idea isn't a hypothesis about reality, it is a belief independent of reality, in other words, it is a supernatural belief. People are entitled to have supernatural beliefs, it is called freedom of religion. However freedom of religion also prevents people from imposing their supernatural beliefs on others.
Evidence based medicine does have difficulties with confounding factors in studies. Anecdotal based treatments have those same difficulties but to a much much higher degree. Simply because you try to account for confounding factors does not make your study worse than a study that completely ignores them.
The Ip et al study hypothesis was: "We aim to see if there is increased mercury exposure in children with autistic spectrum disorder."
The DeSoto et al paper showed that the data in the Ip et al study does support the hypothesis.
"We have reanalyzed the data set originally reported by Ip et al. in 2004 and have found that the original p value was in error and that a significant relation does exist between the blood levels of mercury and diagnosis of an autism spectrum disorder."
The areas for further research are contained in Results section where they perform exploratory analysis on the data set and identify several distinct characteristics between the study and control groups as discussed in my numerous posts above.
As for the discussion of the IOM conference on Thimerosal, that only came up as part of the discussion with Daedalus2u surrounding the value of statistical analysis vs physiological evidence. My comment there was to respond to his comments about why the plaintiffs don't do any physiological studies.
This discussion is centered around the value of the DeSoto paper in the Omnibus trial and I have repeated numerous times why it is valid to be submitted (not by a politician mind you).
There are plenty of other articles on Blogs that talk about the Thimerosal elimination evidence and the quality of data that is used to justify both an epidemic and current climbing rates.
Most of your discussion may very well be accurate, however, the reality is that Ip et al designed a study with the following hypothesis:
"We aim to see if there is increased mercury exposure in children with autistic spectrum disorder."
They are not talking about large or small differences as you state. They are looking for an association that is shown to be statistically significant. An analysis of the data shows that there is a statistically signicant result meaning that there is at least a 95% chance that the data supports the hypothesis. That is the core of what the DeSoto paper says. It also points out other interesting patterns in the data that they think merits further investigation.
You can easily choose to disagree with their reasoning, you are also welcome to disagree with the way these types of studies are done (a valid argument separate from this one) but the data is what it is and this is the way this type of science is currently conducted (and statistics calculated), so it is valid to add to the body of evidence.
I agree that they found a "statistically significant" association of elevated mercury with a diagnosis of autism in the group they tested. But so what? There was a lot of scatter, the distributions completely overlap.
They didn't actually measure "exposure", what they measured was mercury in blood and hair. To estimate "exposure" from blood and hair levels, one needs to assume kinetics of absorption and excretion. To do that, one needs to assume that people with autism are "the same" as people without autism, so that the kinetics are "the same". If the kinetics are not "the same", then the "exposure" (the mercury consumed in the diet) might bear no relation to what was measured. I think the kinetics are quite similar, perhaps not identical. The variation between individuals seems to be much greater than the difference between the two groups. Whether that reflects greater differences in kinetics or in diet remains unknown.
The difference they found was small in terms of what is known about what mercury does in people, where the known toxic limits are and what is the normal variation seen in larger populations.
They did not find anything that supports the idea that mercury causes autism.
If all the Autism Omnibus plaintiffs have is a study that says maybe people with autism on average process mercury a little bit differently than people without autism, and maybe they should do more research, they have nothing. They need to show that it is more likely than not, that the mercury in vaccines caused autism in the 5,000 cases.
"If all the Autism Omnibus plaintiffs have is a study that says maybe people with autism on average process mercury a little bit differently than people without autism, and maybe they should do more research, they have nothing. They need to show that it is more likely than not, that the mercury in vaccines caused autism in the 5,000 cases."
Yes, that is true. However, DeSoto et al can easily serve the purpose of nullifying the opinion of those that used Ip et al to support their own position. From that perspective, it's purely a legal strategy at this point. Their own paper suggests that more research is necessary at any rate.