Wednesday, March 28, 2007

Where are the data files?

Among the things missing from the LDP, and so far not produced by USADA, are the computer data files collected during the confirmatory T/E and GC/MS CIR tests. Evaluation of these is critical for proper interpretation of the results, as we shall see.

At the time of the B sample, Dr. de Boer, Landis' expert, had a problem with the tests as executed and reported:

The identity of the compound(s) belonging to the peaks has not been established according to minimal WADA requirements.

(USADA 368, 369, 370)

This seemed like quibbling at the time to many people. Now, seven months later, some of the implications of this are coming clear in discussion at the Daily Peloton Forum, even to those who have been skeptical of various defense claims.

[MORE]


The CIR is complicated in many ways, from sample preparation, complexity of the machine setup, and interpretation of the resulting measurements.

The sample preparation is finicky because it is necessary to filter out compounds not of interest, because they can interfere with the results. LNDD runs the sample through various steps to produce three different "fractions", each of which is supposed to contain less of the wrong stuff, and more of right stuff than its predecessor. (Less inclined readers may skip the next two paragraphs.)

The subsequent fractions should have more of the particular target substances than the other fractions. In LNDD's IRMS test there are seven target substances: an internal standard, two endogenously produced reference compounds, and four metabolites of testosterone. The internal standard appears in all fractions, since it is added to each by the lab, and the other six substances are each measured in one particular fraction.

The GC-MS instrument is designed to identify compounds in an unknown sample on the basis of two independent properties: 1) the amount of time required for it to emerge from a gas chromatography column (this is the "retention" or "elution" time), and 2) the characteristic size of molecular fragments arising when the molecule is bombarded with electrons (this is a molecule's "Mass Spectrum").

The retention time is not good enough on its own because many different molecules can have the same or very similar retention times--this is called "coelution".

The time measured by GC is similar to dropping things off a tower and seeing how long it takes for them to hit the ground. The speed will be affected by the aerodynamic and mass relationships. While two whiffle balls are likely to hit the ground at the same time, and hardballs much more quickly, table tennis balls might land at the same time as whiffle balls, and meatballs the same time as the hardballs. You need to look at something other than the impact time to distinguish completely, and you need to look elsewhere to determine if you have meatballs and table tennis balls mixed in to your samples.

The mass spectrum is similarly almost unique for each molecule, but not in cases of particular interest here. Epitestosterone and testosterone share the same mass spectrum, but have different GC retention times. Because of the possibilities of interference by unwanted molecules in one measurement, its necessary to look at both properties to be sure that the assay is measuring only the intended substance. The output data need to be carefully examined to ensure that no unintended substances interfered with the measurements.

If, and only if, the compounds in the peak are exclusively of the desired compound will the resulting ratio be correct. If an additional substance is also in the peak, then the result will be wrong, and invalid.

The LDP, as provided, gives only some of the data to justify the claim that the peaks measured were indeed the compounds they are claimed to be. Essentially, LNDD have identified compounds on the basis of the GC retention time alone, and have not used the mass spectra effectively to eliminate the possibility of interference. And some of the data provided in the LDP suggests that the fractions were, in fact, contaminated by other compounds, perhaps even some of the metabolites of Landis' TUEd cortisone.

The answers to these puzzles would be revealed by the data that was collected by the instrument at the time of the test -- the data USADA has so far refused or been unable to produce.

We can think of the following possibilities.
  1. USADA produces the data files, and they show LNDD did the right thing, leaving us with the previous set of arguments.
  2. USADA produces the data files, and they show LNDD didn't do the right thing, invalidating the results.
  3. USADA doesn't produce the files claiming they are not relevant. This is problematic, because WADA technical document TD2003IDCR requires the data to make the positive claim.
  4. USADA doesn't produce the files, claiming they don't exist. This could be because of a "document retention policy" in force at the LNDD.
In the case of data files that show LNDD did the right thing (case 1), we are forced to ask whether they may have been tampered with. Inasmuch as they were not produced before various theories of their relevance were floated, and we have seen LNDD make "corrections" to ostensibly "original" documents already in this case, and in Landaluce have admitted to breaking rules when it seemed expedient, there is now reason to question the veracity of anything that is produced from now on.

In the case of data files that turn up "missing", we are left wondering what to do. They have been identified as quite possibly containing information that proves Landis' case, and their inclusion is supposed to be mandatory for a positive test finding. The arbitrators would be right to throw the case out if they were not produced. We'd then be left with the question whether this would be perceived as a win on a "technicality" or an exoneration.

It certainly seems like the absence of the files would be a violation of the test and reporting protocol that have great relevance on the reliability of the test results, so it is hard to imagine the case being proven without them.

Where are the data files, and why hasn't USADA produced them?




(With assistance from Duckstrap)

5 comments:

swimyouidiot said...

Thanks very much dB (and duckstrap). That really helps me understand this!

One question: Do we know for sure that USADA has not produced the data files? Has this been confirmed by team Landis? We know from Michael Hiltzik that Landis received some documents through discovery that have not been made public. Any chance Landis has these data files, but we are not privy to them?

Anonymous said...

This guy is innocent and a hero. The French and the lab are totally suspect, How to vindicate the American hero?

tbv@trustbut.com said...

SYI, it hasn't been produced according to Baker.

Anon, remember it is a US-funded agency that is doing the prosecution with this data, and it is the one withholding the evidence. This is not a foreign conspiracy, this is domestic CYA in action.

TBV

swimyouidiot said...

Thanks TBV. If USADA presents this data (assuming it still exists) for the first time at the hearing, I am going to be ticked! It would be working in such bad faith!

Anonymous said...

The data should have already been produced. Be prepared to be ticked.