Winter Potpourri
December 10th, 2007
Two articles have been making the rounds on autism “biomed” ‘blogs and lists recently:
and
Curran et al, “Behaviors associated with fever in children with autism spectrum disorders.”
The first of these has been dealt with by other bloggers, but it keeps rearing its ugly head, so, with a degree of reluctance, I will try to deal with it in my own inimitable (if pedantic) manner.
The second article, when I read it, struck me as sort of a “so what?” study. Then I found out that there is a whole sub-culture in the “autism community” that believes some autistic children have been “cured” by a high fever. While this may have originated as a “grass-roots” anecdote, it probably got a big boost from a story by Dan Olmsted, ex-UPI Senior Editor for autism and now stringer for an “biomed”-advocacy ‘blog. My, how the mighty have fallen.
DeSoto and Hitlan:
The origins of this article is a study by Ip et al (June 2004) on hair and blood mercury levels in autistic children compared to non-autistic controls. In this study, the authors found no significant difference between the autistic and non-autistic groups’ hair and blood mercury levels.
Now, this was only “an important data set” to a small group of “researchers” - those who are trying desperately to show that mercury causes autism. More specifically, it is critically important to those who are trying to resuscitate the stillborne hypothesis that autistic children can’t excrete mercury.
This unsupported claim was first made by Holmes et al, in their article of August 2003. Not only did they fail to give a single piece of data to support this outlandish assertion, their entire data set was called into question when, in August 2004, the NHANES study on hair mercury (in 838 children) showed that the Holmes et al control subjects had over sixteen times the hair mercury levels of the children in the NHANES study. The autistic group had hair mercury levels that were over twice the NHANES mean.
Well, for anyone with a brain, that was pretty much “Game Over!” for the Holmes et al study. No matter what caused the huge discrepancy in their analysis, none of their data can be trusted. And even if we allow them to wiggle out of the control discrepancy, the autistic group still had more hair mercury than the NHANES mean, so “poor excretion” doesn’t apply.
In April of 2007, Kern et al tried another go at the “poor excretor” hypothesis. What they came up with a lack of significance for mercury. They tried to put a brave face on it by showing a “statistically significant” correlation between arsenic, cadmium and lead and autism, which they asserted was just as good as showing it for mercury, since they’re all “sulfhydryl-reactive metals”.
This was all well and good, except that their section on statistical analysis doesn’t mention any correction for multiple comparisons (e.g. the Bonferroni correction), so “significant” may be a relative term. At any rate, Kern et al failed to show any statistically significant difference in hair mercury levels between the autistic and control groups, so they didn’t support the assertion (I refuse to call it an hypothesis) of “poor excretion” made in Holmes et al.
Now the husband and wife team of DeSoto and Hitlan are trying to show how an “important data set” that also failed to show a statistically significant difference between the hair (and blood) mercury levels in autistic and non-autistic children was “flawed” and inappropriately analysed.
In a nutshell, DeSoto and Hitlan (D&H) argue that a one-tailed test should have been used because Ip et al knew which direction the difference between autistic and non-autistic groups would be (i.e. that the autistic group would have a higher blood mercury and a lower hair mercury) based (presumably) on the appallingly bad Holmes et al study.
In reality, it appears that D&H are making some unsupported assumptions of their own, by using a test that would show higher “significance” if the difference were in the direction they assume. Given, however, that Ip et al really didn’t know beforehand (or at least, didn’t assume) the direction of the difference, the two-tailed test they used was very appropriate.
In fact, the analysis by D&H appears to be nothing more than post hoc statistical fiddling to get the desired answer.
Some of the other fiddling was to remove outliers - always a tricky proposition and one that need to be very well justified. Their stated assumption is that these values disproportionately skew the results, which is rather a poor reason for removing them. An even better argument could be given for removing the values as the lower end of the scale, since these are at the lower end of detection and cannot be distinguished from zero.
Here’s the original (as corrected by Ip in November 2007) Ip et al data:
As you can see there are two values - one in the autistic group and one in the non-autistic group - that are clearly not grouping with the others. This data gives a one-tailed p-value of 0.028 and two-tailed p-value of 0.057 for the blood values and p-values of 0.40 and 0.80 for the one- and two-tailed t-test on the hair data.
So, if you think that Ip et al should have known which way the difference would go, the results are signficant for blood and not hair. If, on the other hand, you think they should have gone into their study with an open mind - since there is no a priori reason they should have known that the difference would go a certain direction - then the results are not significant.
But what about those outliers? Here’s what is looks like when I remove these ”outlying” data points in each group.
This gives a one-tailed p-value of 0.0089 and a two-tailed value of 0.018 for the blood values and 0.14 and 0.29 (one- and two-tailed) for the hair values. That would give a significant difference for blood and not hair, no matter how many tails you use.
So, D&H have deomonstrated that by removing the “outliers”, the blood mercury levels in autistic subjects is higher than that of the control subjects, while their hair mercury is indistinguishable.
One possible explanation of this could be that autistic children excrete less mercury than their non-autistic peers….except that nobody has ever shown - in any mammal (let alone primate) - that mercury is excreted by the hair. In all studies performed to date, the hair mercury is a constant fraction of the blood mercury [note: some animal studies have shown precisely the individual variation in that fraction that the Ip et al data show].
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.
D&H also do not address the numerous other studies that have failed to find a connection between mercury and autism, which is probably why they refer to the Ip at al study as “an important data set”. If they can “turn” Ip et al - so they must think - the wall of data refuting the mercury-causes-autism dead-as-a-doornail hypothesis will crumble.
Except that it won’t. No other study refuting the mercury-causes-autism undead-parrot hypothesis is critically dependant in the Ip et al study. They all stand on their own merits - unlike the myriad “poor excretor” studies that live or die by the finding that hair mercury is an indicator of mercury excretion.
It’s also curious that no matter how D&H twist the data, the hair mercury levels remain essentially the same in the two groups. Curious. Yet D&H insist that their re-analysis supports the Holmes et al conclusions, which were based on autistic children having significantly lower hair mercury than their non-autistic peers.
It’s also curious that the mean hair mercury values in the Ip et al study (and, by extension, the DeSoto and Hitlan article) is about nine times the hair mercury levels found in the NHANES study.
Maybe the Ip et al study says more about living in Hong Kong than it does about autism.
And maybe - just maybe - the Ip et al data set is only “important” if it can be “turned” to support an physiologically un-supportable hypothesis.
Curran et al:
The Curran et al study is an interesting one. As I mentioned above, unless you know the background - the persistent stories of how fever has “cured” autism - it comes across as a “so what” study.
However, this is not how the media - and the “biomed”-advocacy ‘bloggers - have presented it.
To hear it on the radio, television and ‘blogosphere, Curran et al have shown that fever “cures” autism. Unfortunately, this isn’t what their published study says:
“We documented behavior change among children with autism spectrum disorders during fever. The data suggest that these changes might not be solely the byproduct of general effects of sickness on behavior; however, more research is needed to prove conclusively fever- specific effects and elucidate their underlying biological mechanisms (possibly involving immunologic and neurobiological pathways, intracellular signaling, and synaptic plasticity).”
Looking at the data shows exactly how far from reality many of the reports have been.
Curran et al examined the Aberrant Behavior Checklist (ABC - a parent checklist) subscale scores of autistic children during, 1 - 3 days after, and one week after a febrile illness (greater than or equal to 38 degrees C) and compared them to autistic children without fever.
What they found was interesting, but not a “cure” for autism.
The “lethargy” subscale went up during fever and gradually returned to baseline at one week after the fever had passed (no surprise here).
The “irritability” subscale went down slightly but then rose and remained higher than baseline throughout the duration of the testing.
The “hyperactivity”, “stereotypy” and “inappropriate speech” subscales went down during the fever and gradually returned to baseline by one week after the fever.
Bottom line: all subscales were back to baseline - except for “irritability”, which remained slightly elevated - by one week after the fever.
Not exactly a “cure”.
Of course, I expect that the “biomed” advocates will shortly be trotting out recommendations to induce hyperthermia in autistic children in order to effect a “cure”. Doubtless the “IR sauna” manufacturers will be claiming that this study validates their product.
Oh, if only it were that simple.
Prometheus
Filed under: Autism Science, Autism Treatments, Critical Thinking
December 11th, 2007 at 12:32 am
Thank you for this. I was missing a couple of pieces of the DeSoto and Hitlan discussion puzzle. So they’re married? Interesting. I still want to know why the lady with the special interest in hormones and personality disorders suddenly feels qualified to go after mercury levels in children in Hong Kong and speak about it knowledgeably. She and hubby did not write knowledgeably about autism/mercury so it remains a mystery… why did they bother?
December 11th, 2007 at 9:48 am
The fever story reminds me of the (now abandoned) method of “treating” schizophrenia by infecting the patients with malaria.
We still have in our teaching labs microscope slides with blood smears from such patients, to show students Plasmodium malariae. We are instructed to handle these slides with great care because (happily!) the damaged ones can no longer be replaced.
I hope I am not giving anybody an idea for a new biomed treatment of autism.
December 11th, 2007 at 1:30 pm
Maya M,
Unfortunately, there is at least one doctor who thinks we can cure AIDS with malaria:
http://blogs.abcnews.com/theblotter/2007/06/dr_heimlichs_ne.html
Ms. Clark,
Reading the article by D&H, I had to conclude that they truly believed in the mercury-causes-autism hypothesis (after all, they cited Holmes et al as a valid source of information) and so felt a need to show how one of the myriad studies refuting the mercury-autism connection was “flawed”.
I would have liked to see them apply the same degree of scrutiny to the Holmes et al study. Perhaps they could explain the anomalous results.
Prometheus
December 11th, 2007 at 4:41 pm
Dear Prometheus:
Great explanation of the Curran study.
Query: Since a fever makes brain functions less “efficient” (for lack of a better lay term), and since one of the general (or, at least, common) findings re: persons with ASD diagnoses is that they have greater grey matter that neurotypicals, has anyone explored the idea that what is reported is an effect of the fever on someone whose brain contains greater grey matter?
I apologize for the lack of precision in the question.
This would not suggest using fever as a “cure” (although I can see many quackers advocating it and desperate parents falling for it) since a sustained fever would damage the brain and other organs in uncontrolled ways (and for co-morbid neurological conditions, easily risk a “cure” worse than the condition). However, it might suggest answers on how/why the brains of persons with ASD (or some subset) function. If so, then the reason(s) for the reported “beneficial” effects of the fever (assuming they are real), could be isolated, for use in treating ASD symptoms in a way that can be sustained without unacceptable risk of injury.
December 11th, 2007 at 6:05 pm
WFJAG,
The “grey matter” in the brain is the cell bodies of the neurons - the “white matter” is the axons and dendrites of those cells. One hypothesis is that autistic people have more grey matter because there was a failure in one or more steps in neurodevelopment.
Much of neurodevelopment involves “pruning” (a nice way of saying “killing”) neurons - this is the reason that infants have more “brain cells” than adults. If a “pruning” steps fails to occur, you end up with more “grey matter” (you also have more “white matter”, but since the axons are relatively narrow, the difference is less obvious).
This is a rather round about way of saying that it isn’t obvious that there would be a difference in the response to fever.
In fact, it may be that “neurotypical” children (and quite possibly “neurotypical” adults) would show the same pattern as Curran et al showed in autistic children. However, since the ABC subscale scores would (presumably) be low in “neurotypical” children and adults, the change in scores might not be very significant.
I have concerns about artificially elevating the body temperature of any child (or adult, for that matter) because of the potential for harm.
Fever in response to viral or bacterial infection can cause injury on rare occasions, but it is inherently self-limited (in most cases) because the heat is generated by the body and regulated by the hypothalamus.
When body temperature is raised by external heat, there is a very real risk of “overshooting” the “target” temperature. The Curran et al study looked at fevers above 38 deg C; fevers above 42 deg C are often associated with brain injury. Body temperature above 45 deg C leads to brain death in most cases. That’s only 7 deg C (about 12 deg F) above the minimum temperature in the Curran et al study.
This is a very narrow window when you’re looking at externally heating a child (children are more susceptible to overheating than adults) to raise their body temperature.
I know that you’re not suggesting that people try inducing fever to “cure” autism, WFJAG, but I wanted to use the opportunity to emphasize that hyperthermia is not something that parents (or certain practitioners) should “mess with”.
It is entirely possible that the effect seen during fever might lead to some breakthrough in the physiology of autism, which is currently entirely obscure. I hope that is what happens, rather than an outbreak of autistic children dying from hyperthermia.
Prometheus
December 11th, 2007 at 8:19 pm
Thank you.
December 13th, 2007 at 11:15 am
I am working on an analysis of the Curran study and hope to have it up in a few days. I think it is mediated through NO from the iNOS expressed as part of the immune response. I don’t think it has anything to do with temperature per se. I will address the seemingly paradoxical observation by NTs that fever makes their brains function less well, and how could it make an ASD brain function better. That relates to percolation in neural networks and which side of the percolation threshold the network is on.
The use of “fever therapy” was mostly used on neurosyphilis, where it was actually the “standard of care” for several decades. At the time, a large fraction of institutionalized mental health patients were suffering from GPI, generalized paralysis of the insane. The malarial agent of choice wasn’t P. falciparum but rather P. malariae which had a milder course but with more frequent bouts of fever (and less chance of a fatal outcome). People actually were “cured” of neurosyphilis, they recovered their abilities to think, work and to live in society. Without treatment, a diagnosis of neurosyphilis was a death sentence in a few years. Many thousands of people were treated this way and recovered. People who would have died without treatment.
http://www.pubmedcentral.nih.gov/articlerender.fcgi?tool=pubmed&pubmedid=4557082
The originator of fever therapy won the Nobel Prize for his work. He did it carefully, with much thought and planning, and with consideration for his patients before, during and after treatment. The “trials” he did were (at the time), state of the art. Not up to modern standards, but much of his work was over a century ago. They are a sharp contrast to what the anti-mercury crowd is doing.
December 22nd, 2007 at 9:29 pm
Prometheus,
You have clearly shown your bias in this writeup:
Do you still maintain that the Holmes study values are comparable with the NHANES study results?
1 a)The ages for NHANES are 12 - 60 months, with no mean given
b) The ages for Holmes were 12-24 months with a mean of 17 months
2 a) The NHANES study only used 1 months worth of mercury accumulation in the hair
b) The Holmes study used the full sample (average 17 months)
Looking at the two studies I would absolutely expect the Holmes values to be higher.
You are also being misleading in that you discuss the DeSoto paper as if it were an independent analysis. Instead, it is a paper designed (and commissioned) to correct the problems with the Ip et al study by the editor of the journal.
You say:
“In a nutshell, DeSoto and Hitlan (D&H) argue that a one-tailed test should have been used because Ip et al knew which direction the difference between autistic and non-autistic groups would be (i.e. that the autistic group would have a higher blood mercury and a lower hair mercury) based (presumably) on the appallingly bad Holmes et al study.”
You are cherry picking the discussion section and making it look very biased when in fact, they discuss the both options in detail.
“Given that this is the first direct test of this hypothesis and considering the potential importance of finding a relation between mercury blood levels and autism, it is just as important to avoid a false negative as a false positive. As the original authors have now currently calculated, the obtained difference suggests that there is probably a real difference (specifically that the chance that a real effect exists is about 94%, or, conversely, that the chance the null effect is true is less than 6%, which misses the conventional p=.05 or 5%mark of statistical significance). Given the close value to conventional significance, most researchers would not call this a firm rejection of the hypothesis, but might say it was marginally significant. Most researchers facing a P value of .056 would not want to categorically state that results “indicate that there is no casual relation between mercury level . . . and autism.”1 It concerns us that the original authors would want to let this conclusion stand in light of
the new P value (which differs markedly from the .15 previously reported in 2004).”
If you look at the ip et al study:
“We aim to see if there is increased mercury exposure in children with autistic spectrum disorder.”
That sounds pretty directional to me and it is exactly what DeSoto et al point out.
“Whether to use a one-tailed test or a two-tailed test can be decided based on considering what would happen if the results ended up in the opposite direction of what one suspects. In this case, it would mean that the blood mercury levels were lower in the autistic group. Would this support the original hypothesis? (No!) However, if this were to happen, that is, if the autistic group were significantly lower in their blood mercury levels than the normal group, the researchers would find themselves in the incongruous position of having to accept their hypothesis that autism is related to elevated levels of mercury in the blood! The key point here is that their hypothesis was directional, and a one-tailed test should have been used. In this case, the just missed significance of their new analysis using a two-tailed t-test (P = .056) would have reached a conventional level of statistical significance (with P less than .03).”
That discussion is not biased as you allege in your article.
You say:
“Some of the other fiddling was to remove outliers - always a tricky proposition and one that need to be very well justified. Their stated assumption is that these values disproportionately skew the results, which is rather a poor reason for removing them. An even better argument could be given for removing the values as the lower end of the scale, since these are at the lower end of detection and cannot be distinguished from zero.”
I think you made a mistake here. They described exactly the criteria and justification for removing 2 outliers that ip et al left in:
“Outliers were removed prior to statistical analysis. An outlier is defined as a score that is “substantially greater or less than the values obtained from any other individual.” 10(p521) Outliers have an unduly large influence on the outcome of a statistical test. What actually qualifies as an outlier differs depending on the research question and the statistician analyzing the results; however, values greater than 3 standard deviations either above or below the mean generally qualify as extreme cases.11 Within the Ip et al. data, there were 2 such values that were not removed prior to our reanalysis. These 2 values were more than 3 standard deviations above the mean, and both of these values were far from any other score. (Other scores were within 3 points of the next individual; these 2 scores were each 15 or more points away from any other score in the distribution.) To avoid the appearance that these 2 outliers were reemoved to influence the statistical outcome as opposed to objective criteria for cleaning a data set, it should be noted that the biggest outlier of the 2 was an unusually high blood mercury level of 98, which was in the autistic group. To be clear if anything, removal of the outliers resulted in a more conservative test as it actually decreased the mean difference between the 2 groups.”
You are alleging bias, but the reaoning is well explained, and in fact discusses the very topic of bias and the fact that their manipulation
lowered the average mercury exposure of the Autistic group.
You say:
“Except that it won’t. No other study refuting the mercury-causes-autism undead-parrot hypothesis is critically dependant in the Ip et al study. They all stand on their own merits - unlike the myriad “poor excretor” studies that live or die by the finding that hair mercury is an indicator of mercury excretion.”
The ip et al study is referenced by a number of prominent researchers claiming that Autism in not related to Mercury exposure including Fombonne who was a witness in the Omnibus hearing.
You say:
“Yet D&H insist that their re-analysis supports the Holmes et al conclusions, which were based on autistic children having significantly lower hair mercury than their non-autistic peers.”
Please point out the section where this is stated in the paper, as I can not find it. This is all I can find:
“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 not congruant with what you stated in your writeup.
“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.”
That is exactly what DeSoto et al did. They showed that the data supports the Ip et al hypothesis. From Ip et al: “We aim to see if there is increased mercury exposure in children with autistic spectrum disorder.”
To quote DeStoto et al. “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.”
“It’s also curious that the mean hair mercury values in the Ip et al study (and, by extension, the DeSoto and Hitlan article) is about nine times the hair mercury levels found in the NHANES study.”
You again harp on the absolute value comparison to the NHANES study, but unless I’m missing something, you’re still comparing apples to oranges which is a pretty flawed argument.
December 24th, 2007 at 11:33 am
Schwartz,
If you can’t see the significance of the difference between hair mercury values reported in the Holmes et al study and that reported in the NHANES study, then I’m not sure that I can explain it to you.
Let me try, though.
[1] Holmes et al obtained hair saved from the first haircut of autistic and non-autistic children. They got samples from 94 autistic children (ages ~3 - 17 years at the time of the study) and from 45 non-autistic “control” subjects (ages ~3 - 12 years at the time of the study).
The children were from 11 to 24 months old when the hair was cut, which means that the hair samples were at least one year and at most 16 years old.
[2] The NHANES study obtained hair samples from 838 children ages 1 - 5 years (12 - 60 months) and analyzed at once. The subjects were chosen at random as part of a nation-wide health and nutrition survey. The “mean” age at the time of the sample is the same as the “mean” age of 1 - 5 years olds in the US census - about 3 years.
You’ll note that this includes almost all of the 11 - 24 month time period when the Holmes et al hair samples were collected.
[3] The Holmes et al study had the hair analyzed for mercury, which was reported as parts per million - this was equivalent to the NHANES study, which reported the values as micrograms of mercury per gram of hair.
The Holmes et al study analyzed the entire hair length, which gives an unknown time period, since there was no record of how short that hair was cut at the time of “collection”. Based on societal norms, it is likely that the hair was not cut to the base, and so only reflected the earliest hair growth.
Since hair growth in the first year of life is highly variable, it is expected that the amount of mercury in this hair sample - which numerous studies have shown is dependent on the rate of hair growth - would also be highly variable.
The NHANES study cut the hair to the scalp and analyzed only the most proximal centimeter, which corresponded to the last month of growth. This allowed them to compare the results with the blood mercury results they obtained (reported here: http://jama.ama-assn.org/cgi/content/full/289/13/1667 ).
Yes, there are differences in the study designs. However, these differences fail to explain why Holmes et al got hair mercury levels in their control subjects that were sixteen times higher than the much larger and much more carefully done NHANES study.
Given that hair mercury is a passive reporter of blood mercury, the Holmes et al results would suggest (in fact, would insist!) that the “poor excretors” are the non-autistic children.
Schwartz, it’s all right if you want to ignore the fact that the Holmes et al study found hair mercury levels in their “controls” that had a mean value over five times higher than the 95th percentile of the NHANES study.
It’s your choice to ignore the warning signs and continue blindly following the people telling you that there is convincing evidence that mercury and autism are “connected”.
Just don’t be too disappointed when I don’t follow you on your path. I follow the data, and the data isn’t going that way.
Prometheus
December 24th, 2007 at 1:36 pm
I liked your old “How Science Works” post:
http://photoninthedarkness.blogspot.com/2007/07/how-science-works.html
But there, you only talked about how science does NOT work. You’re a great writer, so I’d love to hear you write about how science DOES work. Specifically, what differentiates science from pseudoscience or bad science or non-science?
I’ve listed some of the problems with common attempts to separate science from pseudoscience and non-science:
http://www.iidb.org/vbb/showthread.php?p=5052902#post5052902
December 24th, 2007 at 2:08 pm
Luke,
Good point! I’ll add that to my short list of “Things to do soon.”
It is often hard to concisely define what science is, while it is often much easier to say what it isn’t. As far as I know, the debate about a simple set of rules to differentiate between science and pseudoscience is still raging.
Prometheus
December 25th, 2007 at 12:54 am
prometheus,
I look forward to reading it.
Yes, the debate is still raging. To my knowledge, demarcation has failed. Rather, many philosophers see science as a “programme” wherein some theories and experiments are more successful - or more scientifically rigorous - than others. Sometimes pseudosciences grow to become accepted as science (as with osteopathy & continental drift). It’s a comlicated world.
December 26th, 2007 at 11:10 pm
Prometheus,
You have pointed to the fact that the measurements are very different from each other in that the baby first hair cuts could have some indeterminate variability due to unknown lengths and variability of hair growth in early years, in addition to the age of the samples.
In the NHANES study, the hair sample is certainly biased toward the older age (scalp measurement), while Holmes is most certainly biased toward the younger part of the age (end of the hair). The NHANES study mentions that boys measure higher, but that the difference was insignificant, however, they don’t provide a number, nor do they publish the specifics which is very frustrating — they published the insignificant difference between hair treatment but not by boy/girl. You’ll note that Holmes used a 4:1 male/female ratio in their study. Given the age differences, it’s almost certain that NHANES would only be capturing very limited mercury exposure due to vaccines let alone no chance of pre-natal exposure. NHANES also discusses that their data is lower than numerous other studies in US and international populations.
The Holmes et al study appears to have used a slightly different process for measuring the mercury (although I might be wrong there, but that’s the way it read to me). Without being specific about a flaw in one of the methods, or a flaw in the study and control group selection you have a limited basis with which to determine the more accurate number. I do not know the measurement process Ip et al used, but one would think that you should at least state an assumption there as well since you’re questionning the measurements.
The NHANES study also notes that the variability was high and they used log transforms on the hair levels to normalize the distributions. I would be interested to know what the Ip et al ratio of blood/mercury levels were compared to NHANES to see if it was consistent. If it was consistent with NHANES, then it would probably indicate a sample difference. You’ll note that NHANES does discuss their sample limitations. It is also interesting that in NHANES they had some extremely high outliers (rightfully discarded for the analysis) that they confirmed through followup tests as being accurate.
You claim you’re following the data, but at best, you’re reading and commenting on data points in isolation of context without stating any key assumptions or any valid discussion (unlike NHANES which identifieds and discusses the differences).
If you want to claim with confidence that you can throw out two whole datasets as unreliable, then it still appears that you’re basing that statement on a lot of assumptions, not concrete evidence. There is not enough information when comparing these data points to determine which ones are incorrect as you are doing and there are almost no other datasets that I’ve found that measure hair mercury levels from below 12 months of age. There is certainly enough difference to ask the question though.
“It’s your choice to ignore the warning signs and continue blindly following the people telling you that there is convincing evidence that mercury and autism are “connected”.”
It’s interesting that when debating specific points of discussion with you (among others), that you insist on employing broad brush statements like this. What do my very specific issues with your article have to do with my or other people’s position on Mercury and Autism? What would Interverbal call that? Hermatic fallacy, or appealing to your reading audience perhaps? I’m trying to validate details of your analysis because I am interested in differing opinions on the same data and I’m trying to understand all of the data as accurately as possible. That is hardly blind acceptance of a hypothesis or theory. I read your articles because you are clearly familiar with the scientific method and are able to perform statistical analysis yourself. Ignoring the biases in the commentary, your strongest factual argument against the studies is the question of data accuracy. That obviously deserves to be analyzed carefully so why shouldn’t one question the non-described details of your analysis of it?
When I first read Do’C and Interverbal’s analysis of DeSoto et al, I formed an opinion on the data much closer to his review — you can confirm that by my early posts on the topic. After reading the rebuttal I remained unconvinced to change my fundamental opinion. However, after reading the actual study and finding discrepencies in the tone of the review vs what was actually in the study, I find myself taking a much less polarized stance on the topic. Your review here contains even more bias and grossly exaggerates several aspects of the actual study writeup. You then appear to take any questionning of your broad brush statements (or confidently dismissive statements) as indications of fanaticism despite a lack of explanation for any discrepencies you point out or even limitations in your assumptions.
If your objective is to provide objective information and discussion for those trying to sort through the information, then that is not a good way of going about it. If your objective is to entertain an audience that you feel is reading your blog, then perhaps you’re meeting your objective. That would be unfortunate.
December 31st, 2007 at 6:25 pm
Schwartz,
You can continue to rearrange the deck chairs on the SS Holmes et al for as long as you like - the study still sank. There are no explanations - not even their “poor excretor” story - that can explain away the extraordinarily high hair mercury levels in their control group.
In fact, their “poor excretor” story doesn’t even explain why the autistic kids would have lower hair mercury than the controls - it would be the opposite, since hair (as numerous studies have shown) simply reflects the blood mercury.
Again, you are free to rationalize the results as you see fit.
Prometheus
January 1st, 2008 at 8:02 pm
I have now posted my analysis of the fever paper. Glad that is done.
January 8th, 2008 at 2:28 pm
I follow the Austism Street blog on the Desoto/Hitlan’s autism article (http://www.autismstreet.org/weblog/?p=158) and thought you might be interested to know that the authors posted this FAQs website. I found it useful.
http://www.uni.edu/desoto/desoto_hitlan_autism.html
January 8th, 2008 at 4:37 pm
Connor,
I “surfed” on over to the apologia (FAQ) section you mentioned. Some of the “answers” show a rather dissapointing lack of understanding. I’ll be addressing one of them later.
On the “one-tail” vs “two-tail” issue - DeSoto argues, essentially, that there is nothing in the “rules” to prevent her from using a one-tailed test. I’d argue it from the other direction, that there is nothing to ALLOW her to use a one-tailed test.
Unfortunately, there is no Supreme Court of statistics to which we can appeal for a ruling. As a result, we’re left with the “I say toe-may-toe, you say toe-mah-toe…” argument.
In short, I found her “FAQ” section less useful, except as an insight into her lack of understanding about physiology.
Prometheus
March 5th, 2008 at 4:07 pm
Look, I don’t know, but my daughter has Autism and seizures. We have also noticed that she seems much more “with us” when she has a fever. What does that mean? Heck if I know, but this blog sort of poop-poos on it for some reason.
FJH
March 5th, 2008 at 4:56 pm
FJ Hurst,
I think you summed up my assessment of the study quite nicely.
Kids with autism show transient improvement when they have a fever, but it does not result in any long-term improvement. There is no indication of what this might mean, nor is there any suggestion that this might lead to a useful therapy.
How does that “poo-poo” the study?
I was simply pointing out that the folks who have been trying to find more significance in this study are doing the equivalent of reading tea leaves or sheep entrails. There is nothing more here than a confirmation of what parents have reported.
It would have been interesting to see the results of a similar study of “typical” children with fever, or children with a different developmental disorder, such as Down’s Syndrome. As it is, it brings nothing more to the table than a simple confirmation of parental reports.
Prometheus