Multiple false postives in WADA-sponsored study of isotope test
Thomas A. Fine
I've put an article up on my web page that analyzes the results of
the WADA-sponsored study that allegedly demonstrates that diet
can not cause false positives.
I still have not found complete published results for this study, but I
have found a set of slides from a presentation of the results.
There's nothing in there that's shocking with respect to diet.
However the random variations of isotope levels from one day to the
next in allegedly normal subjects is far larger than anything that
could be considered useful for the purposes of a doping test.
I was able to extract data points from the graphs included in
the slides, and I found among other things, apparent false postives
among the test subjects during the study.
In light of the data in this study, Floyd Landis' results seem
completely consistent with normal natural variations.
Here's the long and gory details:
http://hea-www.harvard.edu/~fine/opinions/testosterone_again.html
tom
Joe King
What are a "C-4" plants?
You wrote: They measured the isotope ratio for Androsterone and
Etiocholanalone, which are testosterine metabolites, and would be the
<<<COMOUNDS>>> targeted for suspicious changes in an athlete."
Joe
"Thomas A. Fine" <fi...@head-cfa.harvard.edu> wrote in message
news:45121d0e$1...@cfanews.cfa.harvard.edu...
Robert Chung
> I still have not found complete published results for this study, but I
> have found a set of slides from a presentation of the results.
Have you e-mailed the lead author and asked for the data for male 4 and
female 2?
pelot...@gmail.com
complexity of a "positive" drug test. It isn't a case that a lab tech
looks into a microscope and sees the synthetic testosterone. It isn't
like a CSI episode where you put liquid into a test tube and it turns
blue indicating guilt or innocence.
The procedure is complex and prone to normal human bias. I'm using
bias in a way that does not imply evil intentions.
I've outlined three studies that show how bias is in everything we do.
http://pelotonjim.wordpress.com/2006/09/19/protocols-are-important/
Jim
Thomas A. Fine
I know, I'm following up on my own article, and changing the subject.
But I think this is important to get out there. I just did a rough
calculation of the false positive rates for the study that I analyzed,
by way of correcting some info on trustbut.blogspot.com.
The official part of the test ran for 28 days. With various data
points missing here and there from various different metabolite tests,
I'll be generous and say there's about 25 viable test points per test
subject, so that's 125 tests, and out of that there were four samples
that exceeded the 3 per mil limit.
That's a 3.2% false positive rate within their study, or an expectation
of a natural "failure" for 1 out of every 31 tests.
"It's foolproof. This analysis tells the difference between endogenous and
exogenous. No error is possible in isotopic readings."
--Jacques De Ceaurriz, head of the Chatenay-Malabry laboratory
tom
In article <45121d0e$1...@cfanews.cfa.harvard.edu>,
Thomas A. Fine <fi...@head-cfa.harvard.edu> wrote:
>Hi,
>
> [ ... ]
>
>http://hea-www.harvard.edu/~fine/opinions/testosterone_again.html
>
> tom
Thomas A. Fine
Photosynthesis is the reason why living things all have less carbon-13
than our surrounding environments - the photosynthesis process doesn't
like carbon-13 as much as carbon-12.
There are two main types of photosynthesis, C3, and C4. (There's long
names for them too.) Most plants are C3 plants, and they end up with
a delta 13C of around -22 to -30. C4 plants don't reject as much
carbon-13, and end up with with delta 13C of around -10 to -14.
So, since we are what we eat, if we eat lots of C4 plants we will
have more and more carbon-13 in our testosterone over time until it
more or less stabilizes.
As far as our diet goes, the main C4 plants are corn, sugar cane,
pineapples, and millet. Almost everything else we eat is a C3
plant.
>You wrote: They measured the isotope ratio for Androsterone and
>Etiocholanalone, which are testosterine metabolites, and would be the
><<<COMOUNDS>>> targeted for suspicious changes in an athlete."
Thanks, I fixed that.
tom
------------ And now a word from our sponsor ---------------------
For a secure high performance FTP using SSL/TLS encryption
upgrade to SurgeFTP
---- See http://netwinsite.com/sponsor/sponsor_surgeftp.htm ----
aco...@earthlink.net
> Hi,
>
> I know, I'm following up on my own article, and changing the subject.
> But I think this is important to get out there. I just did a rough
> calculation of the false positive rates for the study that I analyzed,
> by way of correcting some info on trustbut.blogspot.com.
>
> The official part of the test ran for 28 days. With various data
> points missing here and there from various different metabolite tests,
> I'll be generous and say there's about 25 viable test points per test
> subject, so that's 125 tests, and out of that there were four samples
> that exceeded the 3 per mil limit.
>
> That's a 3.2% false positive rate within their study, or an expectation
> of a natural "failure" for 1 out of every 31 tests.
>
> "It's foolproof. This analysis tells the difference between endogenous and
> exogenous. No error is possible in isotopic readings."
> --Jacques De Ceaurriz, head of the Chatenay-Malabry laboratory
>
> tom
So are you saying that there's a 96.8% probability that Landis is
guilty?
Andy Coggan
Frank Drackman
<aco...@earthlink.net> wrote in message
news:1158876002.3...@i42g2000cwa.googlegroups.com...
I hope that you are being funny. Of course someone with your credentials
should know that you can't apply group statistics to an individual and
predict the result.
Thomas A. Fine
<aco...@earthlink.net> wrote:
>Thomas A. Fine wrote:
>> That's a 3.2% false positive rate within their study, or an expectation
>> of a natural "failure" for 1 out of every 31 tests.
>
>guilty?
Let's see, either he did, or he didn't. So that's 50/50, right?
(just kidding).
Statistics are a funny thing. My personal favorite is the Monty Hall
problem, but that's neither here nor there...
If we assume that the study of the five individuals would correlate
well to the measurements of world-class athletes in the middle of
competition, then yes, that would be exactly what I would be saying.
However I don't agree with that underlying assumption so I think the
odds that he doped may be more like 75-90%.
Anyway one of the funny things about statistics is the difference
between a population sampling problem, and an individual odds problem.
For example, we can calculate that if 22 tests or more are run, then it
would be more likely than not that you would have a false positive
(again, assuming this 3.2% rate would apply to athletes). But this is
what I really love about statistics: if 105 tests were run, then the
odds of a false positive would be 96.8 percent, while at the same time
the odds that Floyd doped would remain at 96.8 percent. Advice: if you
don't understand how those two things can be true at the same time, DO
NOT take up gambling as a profession. :-)
The presentation of statistics is also funny. If you were considering
some optional surgery to improve a minor problem, and the doctor said
"96.8% of patients end up the same or better", you might react
differently than you would if he said "one of every 31 patients ends up
worse off after the surgery".
Frankly, I don't consider 96.8% good enough for either an optional
surgery that can maim, or for a doping control that destroys careers.
tom
CowPunk
Thomas A. Fine wrote:
> For example, we can calculate that if 22 tests or more are run, then it
> would be more likely than not that you would have a false positive
> (again, assuming this 3.2% rate would apply to athletes). But this is
> what I really love about statistics: if 105 tests were run, then the
> odds of a false positive would be 96.8 percent, while at the same time
> the odds that Floyd doped would remain at 96.8 percent. Advice: if you
> don't understand how those two things can be true at the same time, DO
> tom
And, isn't it amazing how close this is to the actual number of
positives found?
"The UCI did around 12000 tests last year, and about 380 came back
positive. These are just rough numbers off the top of my head.
It worked out to around 3.8% of all tests came back positive. "
Kind of makes you wonder if cycling is as dirty as the sore loosers
would have us believe.
Phil Holman
"Frank Drackman" <Frank...@yahoo.com> wrote in message
news:V5-dnR0oW5Tzio7Y...@comcast.com...
approximate z value of 4 as proof of a positive. That's a very low
probability. Where does this leave the test if the probability of a
false reading is much higher?
Phil H
Frank Drackman
"Thomas A. Fine" <fi...@head-cfa.harvard.edu> wrote in message
news:45131959$1...@cfanews.cfa.harvard.edu...
> In article <1158876002.3...@i42g2000cwa.googlegroups.com>,
> <aco...@earthlink.net> wrote:
>>Thomas A. Fine wrote:
>>> That's a 3.2% false positive rate within their study, or an expectation
>>> of a natural "failure" for 1 out of every 31 tests.
>>
>>So are you saying that there's a 96.8% probability that Landis is
>>guilty?
>
> Let's see, either he did, or he didn't. So that's 50/50, right?
> (just kidding).
>
> Statistics are a funny thing. My personal favorite is the Monty Hall
> problem, but that's neither here nor there...
>
I was at an IT company in the 90's when the question ran in Parade Magazine.
There were fierce debates during lunch and much productivity was lost due to
the many people running simulations.
RonSonic
>Hi,
>
>I know, I'm following up on my own article, and changing the subject.
>But I think this is important to get out there. I just did a rough
>calculation of the false positive rates for the study that I analyzed,
>by way of correcting some info on trustbut.blogspot.com.
>
>The official part of the test ran for 28 days. With various data
>points missing here and there from various different metabolite tests,
>I'll be generous and say there's about 25 viable test points per test
>subject, so that's 125 tests, and out of that there were four samples
>that exceeded the 3 per mil limit.
>
>That's a 3.2% false positive rate within their study, or an expectation
>of a natural "failure" for 1 out of every 31 tests.
>
>"It's foolproof. This analysis tells the difference between endogenous and
>exogenous. No error is possible in isotopic readings."
> --Jacques De Ceaurriz, head of the Chatenay-Malabry laboratory
Tom, I know you've got bigger and more interesting fish in the skillet, but for
me, I cannot get past that quote. I take that quote as prima facie evidence that
the LNDD is abjectly and hopelessly incompetent.
Ron
Thomas A. Fine
No, actually I continue to be fascinated by this quote. That's why I put
it here. When he first tested postive, of course I wondered about the
details, that's what I do. But then all of a sudden, this second test
came up, and everyone was claiming it was absolutely foolproof, that
it meant without a doubt there was synethetic testosterone in Floyd.
I know that science never works with such clarity, and rarely even
comes close, so that's when I began my obsession in earnest.
One of the things I did was a survey (informally) of a bunch of
different articles that claimed the test was fullproof. And some
articles would quote one or another expert that would generally
describe the test as highly accurate, or dependable, or those sorts of
words. But the word foolproof, or the claim that you could somehow
"find" or "detect" synthetic testosterone would always only seem to
come up in connecton with Jacques De Ceaurriz, referring back to this
quote.
The point is this: that is the quote from which the crucifixion sprang.
In the world with that quote, thousands of writers around the globe
went directly to "Guilty", and Floyd instantly became the poster child
for hated, lying, cheating dopers. In a world without that quote, the
entire tone of every article that followed would have been different.
Maybe the questions that I'm trying to raise would have been raised
by the mainstream media, were it not for this quote from this "trusted
expert".
So actually, I think this quote is really at the center of the entire
situation. It is in fact emblematic of everything that is wrong in
the anti-doping effort.
tom
aco...@earthlink.net
> <aco...@earthlink.net> wrote in message
> news:1158876002.3...@i42g2000cwa.googlegroups.com...
> > Thomas A. Fine wrote:
> >> Hi,
> >>
> >> I know, I'm following up on my own article, and changing the subject.
> >> But I think this is important to get out there. I just did a rough
> >> calculation of the false positive rates for the study that I analyzed,
> >> by way of correcting some info on trustbut.blogspot.com.
> >>
> >> The official part of the test ran for 28 days. With various data
> >> points missing here and there from various different metabolite tests,
> >> I'll be generous and say there's about 25 viable test points per test
> >> subject, so that's 125 tests, and out of that there were four samples
> >> that exceeded the 3 per mil limit.
> >>
> >> That's a 3.2% false positive rate within their study, or an expectation
> >> of a natural "failure" for 1 out of every 31 tests.
> >>
> >> "It's foolproof. This analysis tells the difference between endogenous
> >> and
> >> exogenous. No error is possible in isotopic readings."
> >> --Jacques De Ceaurriz, head of the Chatenay-Malabry laboratory
> >>
> >> tom
> >
> > So are you saying that there's a 96.8% probability that Landis is
> > guilty?
> >
>
>
> I hope that you are being funny.
I would call it more of a snide remark, wouldn't you?
Andy Coggan
RonSonic
But as quotes go, it is among the more nonsensical. Perhaps reporters are more
ignorant than even I expect (a shocking low standard) but I'd expect more than a
few to have their ears perk up at "foolproof" and "no error is possible."
Nothing in human experience meets that standard, certainly not any medical
procedure or test.
I'd like to see a poll of actual working lab techs to see if there is a single
one working anywhere who also feels that ANY test is foolproof, that no error is
possible.
>Maybe the questions that I'm trying to raise would have been raised
>by the mainstream media, were it not for this quote from this "trusted
>expert".
The guy on the science desk should have started laughing the moment he heard it.
>So actually, I think this quote is really at the center of the entire
>situation. It is in fact emblematic of everything that is wrong in
>the anti-doping effort.
Can't argue with that.
Ron
dude.a...@gmail.com
The odds of at least 1 false positive if someone were tested 15 times
would be what? Would the test subject be only 51.6% guilty? Of course,
if there were at least 2 false positives in 15 tests, would our test
subject be 77.4% guilty?
If there is a significant chance of false positives, then doesn't
effective testing require a large enough number of samples/tests to
create to opportunity to observe a total number of positives
significantly larger then number of expected false positives? So my
final question would be, at what rate of false positives would testing
require more than 1 positive to have enough significant to indicate
'guilt beyond a reasonable doubt'?
Frank Drackman
<aco...@earthlink.net> wrote in message
news:1158927587.3...@d34g2000cwd.googlegroups.com...
Thomas A. Fine
> <aco...@earthlink.net> wrote:
>>Thomas A. Fine wrote:
>>> That's a 3.2% false positive rate within their study, or an expectation
>>> of a natural "failure" for 1 out of every 31 tests.
>>
>>So are you saying that there's a 96.8% probability that Landis is
>>guilty?
Ugh. You may have been joking, but I blew right by it. A combination
of lack of sleep, and the 15 years since my last statistics class.
Really, it must be more the sleep, since it's much more a thought
problem than something I'd need to look up.
To actually figure out the probability that he's guilty we need to know
what percentage of found positives are true versus false. This is not
knowable, without a firmer number for false positives, a firm number for
true positives, and a statistically significant number of positives
from the cycling community that would allow us to calculate the
actual doping rate from the first two numbers. Which is all pretty
much impossible in a practical sense.
Basically, you got me. Sucks to be me.
tom
Howard Kveck
RonSonic <rons...@tampabay.rr.com> wrote:
> But as quotes go, it is among the more nonsensical. Perhaps reporters are
> more ignorant than even I expect (a shocking low standard) but I'd expect
> more than a few to have their ears perk up at "foolproof" and "no error is
> possible."
I don't know that reporters are more ignorant now, but they (or at least many of
them) sure seem to have gotten lazier than they were a decade or two ago. Especially
in allegedly "trivial" areas like this. Just getting the quote seems to be good
enough - no real need to give it any critical thought. i wonder if De Ceaurriz (of
the Chatenay-Malabry lab) is an actual lab guy or just a manager. If he's a real lab
tech, then that statement is profoundly stupid (or arrogant). If he's just a
manager, then his ignorance is almost excusable.
--
tanx,
Howard
Never take a tenant with a monkey.
remove YOUR SHOES to reply, ok?