Thursday, November 30, 2006

Duckstrap zeros in on good arguments

Landis is charged with two anti-doping violations, a IRMS CIR positive for exogenous testosterone, and a GC/MS measured violation of a 4.0 limit for T/E ratio.

At DPF, Duckstrap has come up with theories that explain why both conclusions by the LNDD would be incorrect. These theories make sense even to those who are not swayed by the arguments made in the slide set.

[Edited since the original posting for clarification based on comments by Duckstrap.]
[More...]

The IRMS CIR
One of the two charges against Landis is that he had "synthetic" (exogenous) testosterone in his system. This is determined by a test known as the IRMS CIR, which is done on some number of testosterone metabolites. The LNDD found Landis exceeded the allowed ratio on two of four metabolites tested, and declared it a positive test. The WADA standard for the test is vague about how many should be over the limit. Landis has claimed all four should be positive, and others have argued that one or two are sufficient.

At DPF, Duckstrap presents a reasonable case that three or four metabolites should be positive to declare a positive result on the IRMS CIR, based on statistical analysis of published research and other WADA standards.

Duckstrap has reconstructed the "control" data sets from the published studies used to validate the IRMS CIR tests. These are the ones of non-doped samples. He concludes the false positive rate using two metabolites, as done by LNDD, would be around 7%. That is, LNDD would declare 7 of 100 known-clean samples to be doped.

The WADA criteria say the rate should be less than 1%. Duckstrap finds that using three of four would reduce the false positive rate down to 0.68%, and using all four shrinks the error to 0.06%. This means LNDD's use of two metabolites makes no sense, and violates the WADA testing standard of 1%.

By using either three or four metabolites criteria to meet the 1% false positive rate, Landis' results at LNDD would be negative.

Landis says he has scientists and documents showing his CIR test would have been considered negative at other WADA labs that use more stringent criteria than those apparently used by LNDD.

The T/E
The second charge is that Landis had a T/E ratio in excess of the allowed 4.0. This is the ratio of testosterone to epistestosterone, which should be close (around 1.0 +/- natural variance) in absence of doping. LNDD reported a measured ratio of 11.4 in a GC/MS IRMS confirmation test.

Duckstrap has previously argued that the T/E confirmation IRMS tests both seemed to have identified the peaks incorrectly (possibly due to Landis' legal cortisone). He found the TE using the peaks read a different way that should have been reported around 2.0, well within the legal limit.

The complication reading peaks is similar to the following situation. You pour some dry sand through a straw to form one peak. Then you move the straw over a tiny bit and pour even more sand to form another peak. Some of the sand spills over onto the first pile because they are so close together. How much sand is in the first pile? If you don't consider the spillover from the second, you can easily reach the wrong conclusion. The spillover from one pile to the other is called "elution". LNDD doesn't appear to have ever considered the effects of Landis's legal cortisone injections on the results. Duckstrap corrected for the cortisone, and sees no T/E violation.

Together...
Duckstrap makes plausible scientific arguments that LNDD has made the wrong call on both types of test -- better in fact than the slide show. If these are provable, at least to the "balance of probability" standard Landis must meet, these arguments may be good enough to comfortably take to a hearing, even for TBV.

The best detailed explanation of these arguments is at the Wiki.

Landis has hinted there is even more than this, but we have no idea what. Much more may not be needed.

[Darlene, you're the target audience for this explanation. Please let me know if you have trouble following it -- if not, I'll keep trying.]

[End]

12 comments:

marc said...

Great summary of duckstrap's arguments, TbV. I for one am grateful for them.

Marc

Duckstrap said...

dB,
I'd like to clarify a couple things with respect to the confirmatory TE test. First, this test involves gas chromatography/mass spectrometry, not IRMS. It should not be confused with the CIR test (which, confusingly is sometimes referred to as GC/C/IRMS). They are related, but definitely not the same thing.

Second, there is a reasonably complete description of the material related to the confirmatory TE ratio on Thomas A. Fine's Wicki (http://landiscase.wikispaces.com/), complete with links to the WADA documents and chromatograms. It is not a problem of identifying wrong peaks, but the (I think relatively high) likelihood that there are interfering substances that coeluted with T and E in the chromatogram. According to the WADA standard, LNDD needs to show confirmatory evidence that the peaks they base their quantitation of the TE ratio on contains only T and only E. They very clearly collected the right data, but do not show it in the LDP package. As I have previously stated, unless this confirmatory evidence is shown, I believe LNDD's results for this test to be invalid.

Duckstrap said...

Another point of clarification. My analysis examined two plausible scenarios with respect to the control population. One of scenario used assumptions that are relatively generous to LNDD; the other used a case which is also very plausible, but not so generous to LNDD. Unfortunately, given the information we have, we do not know which of these scenarios is the true one (or neither!). The case that is more generous to LNDD suggests a 1.2% false positive rate, while the case TbV cites suggests a false positive rate of 7-8%.

tbv@trustbut.com said...

Duck, thanks, I have done a rewrite to clarify some of the issues. For simplicity, I'm going keep the single 7% argument. Let's have a conversation that gets me to understand the 1.2% case so I can explain it.

We also need cites for the 1% standard.

TBV

Anonymous said...

TBV or Duckstrap:

I'm very impressed with Duck's knowledge about these tests. What is your background?

Anonymous said...

ORG wrote the last comment.

Anonymous said...

The people who ran and researched the tests will probably testify for USADA. Duckstrap's arguments might be easily refuted. Got anything else?

Anonymous said...

ORG here ....

Exactly what will the authors of the study say to refute Duckstrap? Do you know what the problem with Duckstrap's argument is or are you making something up to be provactive?

Anonymous said...

"The people who ran and researched the tests will probably testify for USADA. Duckstrap's arguments might be easily refuted. Got anything else? "

hmm...I would not be so sure of your assumption.

Consider that those who ran the studies did so in a way that sets a specific threshold forwhat they consider "positive". If the LNDD is not adhering to what positivity threshold the study used, look for some very egotistical scientists to argue and debate the particulars of "how" and "why" something is a positive...

Point is you will end up being wrong. Sorry.

Anonymous said...

So Don Catlin is an egoist, not an authority? I hope you are not one of Floyd Landis' advisors. The "science is a house of cards" argument is so reactionary and baseless, and predictable. Like I said, you got anything else?

tbv@trustbut.com said...

Hey, various anons, it'd be nice if for sake of debate clarity you'd at least adopt some handle to distinguish you from each other.

Regarding the immediate debate, I for one don't have any idea what Catlin thinks, or what he'd testify to either under direct or cross-examination. If he does testify, I'll confidently predict it to be the keynote of the TV coverage.

TBV

tbv@trustbut.com said...

ORG asks, "I'm very impressed with Duck's knowledge about these tests. What is your background?"

Duck is a PhD who is a senior scientist at a company in the pharmaceutical field, and an expert in statistical analysis related to drug testing and the decision making of the drug development process. He has been a research fellow at the NIH, and has 17 articles that turn up in a search on Google Scholar.

He prefers to remain anonymous in these discussions.

TBV