October 17, 2006

Floyd Landis: the Evidence

(See the preamble to this post.)

The case assembled by Floyd Landis' support team aims to prove that the athlete's alleged positive test result should be invalidated. In particular, Dr. Baker's PowerPoint presentation and Howard Jacobs' argument for dismissal present a number of points of evidence supporting that claim. I have combed through the laboratory documents to see whether that evidence might lead to Landis' exoneration in this case.

I should say up front that I used the aforementioned two documents (and especially Dr. Baker's slide set) as my guide; I did not go through page by page assembling evidence from scratch. As a result, this is not necessarily a complete argument, as Landis' team claims that there are "dozens of problems" with the tests.

I should also say that I am a physicist, not a biochemist, so I do not have a very deep understanding of the technical data presented in the laboratory documentation. I understand the basic idea of mass spectrometry, but I can't really interpret the raw data. I am not really sure if this is a great hindrance, though; most of the people who review this case (either at the AAA or ultimately at the CAS) will not be medical or scientific experts, either.

In what follows the references of the form "USADA 0000" are to individual pages in the laboratory documentation package. These pages have been provided on archive.com by the blogger at Trust But Verify.

As I noted before, Landis' defence here is based on Article 3.2.1 of the WADA Code. I will repeat that article here to remind you who bears the burden of proof in this matter:

WADA-accredited laboratories are presumed to have conducted sample analysis and custodial procedures in accordance with the International Standard for laboratory analysis. The athlete may rebut this presumption by establishing that a departure from the International Standard occurred.

If the athlete rebuts the preceding assumption by showing that a departure from the International Standard occurred, then the anti-doping organization shall have the burden to establish that such departure did not cause the adverse analytical finding.

The arguments in Dr. Baker's PowerPoint presentation and Howard Jacobs' letter are divided into four rough categories. I will deal with each of these in turn.

1. Bad laboratory documentation: Dr. Baker's presentation shows a number of specific examples of sloppy record keeping by the lab, and I have found at least one more which I will discuss later. The laboratory identification number was incorrect on Landis' T/E results page. The athlete identification number was incorrect — that is, not Landis' — on the same document. The athlete identification number was also incorrect on the specimen transport record, which records the chain of custody of the sample between the collection point and the lab, and on the lab's summary record of the test results.

Another page shows that the sample identification number has been overwritten in a manner inconsistent with the Laboratory Internal Chain of Custody (reference above) — although technically the modification was not made on the Chain of Custody documentation, so the cited reference is not really applicable.

Nevertheless, these mistakes are very embarassing for the lab, and somebody's wrist should definitely be slapped. However, it's probably not going to get Landis off the hook. The lab and sample identification numbers appear dozens (if not hundreds) of times throughout the document; a few errors here and there do not create a significant doubt that the tested samples came from Landis.

To put this in terms of Article 3.2.1, Landis has shown that a departure from the international standard has occurred, but the UCI will not have to work very hard to demonstrate that these errors did not cause his positive tests. If an arbitration panel dismisses Landis' positive over this issue it will be the most technical of technicalities. It's not impossible, but I consider it unlikely.

2. Unexplained variability in the T/E ratio:

In all there were nine vials of Landis' urine subjected to GC/MS (Gas Chromatograph / Mass Spectrometer) analysis at LNDD.

Table 1 — Floyd Landis urine samples subjected to measurement of T, E, and T/E.
Vial Testosterone Concentration
Epitestosterone Concentration
(using Concentration)
References Comment
A Sample Screening Procedure (GC/MS), Aliquot ES04, 2006-Jul-21
11 60.60 13.70 4.9 4.9 USADA 0054, 0055, 0056 T/E exceeds screening threshold (> 4)
A Sample First Confirmation Analysis (GC/MS), Aliquot EC24D, 2006-Jul-22
10 172.23 17.59 10.7 9.8 USADA 0212, 0213 Sample with hydrolysis
11 1.06 0.10 11.2 10.2 USADA 0214, 0215 Sample without hydrolysis
A Sample Second Confirmation Analysis (GC/MS), Aliquot EC24D, 2006-Jul-24
4 61.37 5.20 11.4 11.8 USADA 0092, 0093 Sample with hydrolysis
A Sample Re-Screening Procedure (GC/MS), Aliquot Unknown, 2006-Jul-25
2 49.70 11.10 5.1 5.1 USADA 0057, 0058, 0059 Test not described in aliquot chain of custody table
B Sample Confirmation Analysis (GC/MS), Aliquot EC24D, 2006-Aug-3
4 63.15 5.94 10.9 10.6 USADA 0278, 0277 Sample with hydrolysis
5 61.64 5.75 11.0 10.7 USADA 0279, 0280 Sample with hydrolysis
6 60.18 5.55 11.1 10.8 USADA 0281, 0282 Sample with hydrolysis
7 1.22 0.44 3.6 2.8 USADA 0283, 0284 Sample without hydrolysis

In the table above the "Raw" T/E ratio is calculated using the response amplitudes from the instrument, and the "Concentration" values are calculated after first converting the instrument response to a concentration using a calibration curve. The values don't differ significantly in any case.

The first sample noted was subjected to the Screening Procedure for Natural Hormones. Screening procedures for threshold substances like testosterone are meant to flag a sample for further study; in this case, when the T/E ratio exceeds 4. A sample could also be flagged as suspicious if it showed a high concentration of either testosterone or epitestosterone (> 200 ng/mL). There are no numerical accuracy requirements for Screening procedures.

The sample I have called the "A Sample Re-Screening Procedure" is odd for a couple of reasons. First, it was performed after all of the confirmation tests, including the IRMS confirmation. At that point in time, the screening procedure would appear to be more or less irrelevant, so I am not sure why this was done. The handwritten note on the summary results page (USADA 0057) appears to read "vial de conf reinjecte pr screening." Another odd fact is that this test does not appear on the Aliquot chain of custody table (USADA 0011, 0012, repeated USADA 0255, 0256). This is a very clear violation of the Laboratory Internal Chain of Custody (reference above) and therefore another error in the documentation. Add it to the pile in item 1 above.

The Confirmation Analyses are used to definitively assess the T/E ratio in a sample once it has been identified as suspicious by the screening procedure. The samples I noted as "with hydrolysis" are the ones we should focus on, for now. These samples are subjected to hydrolysis to cleave testosterone and epitestosterone from the various compounds that they form in the urine sample (that's this physicist's understanding). The GC/MS then detects the total concentration of T and E in the urine.

The Confirmation Analysis (with hydrolysis) was run on five different vials of Landis' urine sample; two from the A sample and three from the B sample. Of those, four were very self-consistent: the A sample aliquot tested on 24 July, and the B sample aliquots tested on 3 August all showed T concentrations between 60.18 and 63.15 ng/mL, E concentrations between 5.20 and 5.94 ng/mL, and T/E ratios between 10.6 and 11.8. The LNDD claims (e.g. USADA 0101) that the uncertainty for these results is 30% for E, 20% for T, and 30% for T/E. All four of these measurements fall well within those expectations.

The fifth sample, which was actually the first one tested, is way, way off, with T = 172.23 and E = 17.59. It is true that the ratio of T/E is consistent with the rest of the samples, but the variation in steroid concentration remains unexplained.

It is possible that the LNDD will respond with a valid explanation for this variability. At any rate, by itself I do not think it is sufficient to doom the case. It might suggest that the confirmation procedure was run incorrectly in one instance, but it does not directly imply that all of the other results are invalid.

This brings me, however, to Dr. Baker's third claim.

3. Contamination of the sample: Dr. Baker makes a case that Floyd's B Sample was not suitable for testing. The Reporting and Evaluation Guidance for Testosterone (reference above) has this to say about microbial contamination of urine samples (emphasis mine):

The urine Sample is not collected under sterile conditions, and where the circumstances are favourable, the microbes present in the Sample can cause changes to the profile of the urinary steroids. Initially there is cleavage of the glucuronides and sulfates followed by modifications of the steroids’ structure by oxido-reductive reactions. To report an Adverse Analytical Finding of an elevated T/E value, testosterone or epitestosterone concentration or any other endogenous steroid parameters, the concentration of free testosterone and/or epitestosterone in the specimen is not to exceed 5% of the respective glucuroconjugates. Elevated amounts of 5ALPHA- and 5ß-androstan-3,17-dione in the free form also indicate microbial degradation.

The "free" testosterone in a urine sample is provided by running the Confirmation procedure on a sample without subjecting it to hydrolysis. Only T and E that exists "free" in the urine shows up in the GC/MS traces.

So let's go back to Table 1. There were two aliquots subjected to the confirmation procedure without hydrolysis. One of these was the first A sample confirmation on July 22, which showed acceptable levels of free T and free E (about 0.6% in both cases). The second was from the B sample on August 3 — but that sample showed 2% free testosterone and 7.7% free epitestosterone, calculated as a percentage of the mean of the results from the samples with hydrolysis.

Unless I am misreading the documentation or misinterpreting the science (both possibilities) it seems fairly clear to me that Floyd's B Sample should have been ruled too contaminated for testing. The paragraph I quoted above is quite clear on this: with a concentration of free epitestosterone this high, the B sample could not be used to report an adverse analytical finding.

And if the B Sample cannot confirm the A Sample results, the International Standard for Laboratories (reference above) states clearly that the test is negative: The B Sample result must confirm the A Sample identification for the Adverse Analytical Finding to be valid. The mean value for the B Sample finding for Threshold Substances is required to exceed that threshold including consideration of uncertainty. (International Standard for Laboratories)

If the “B” Sample confirmation does not provide analytical findings that confirm the “A” Sample result, the Sample shall be considered negative and the Testing Authority notified of the new analytical finding.

It is also interesting to consider the hypothesis of contamination with respect to the sample-to-sample variability I discussed in point 3 above. No measurement of free T and E was made in the A sample second confirmation analysis, so there is no evidence of contamination in that sample. However, the concentration results are consistent with the B sample results, which do show evidence of contamination. It might be that the only uncontaminated sample was the first confirmation of the A sample. Note that the aliquot used in the second confirmation of the A sample was prepared for confirmation on the afternoon of July 23 and not tested until almost 24 hours later (ref. USADA 0256).

This speculative explanation does not cover the B sample contamination; however, any microbes present in the sample had an additional 10 days to act on the refrigerated B sample while it was stored at LNDD. Perhaps this was sufficient time to degrade the sample.

4. CIR results did not support a "positive" finding: I've already stated that I think that Landis is ultimately going to win his appeal due to the evidence that his B sample was contaminated when it was tested.

That outcome, however, might still leave the public with the feeling that Landis was a cheater who "got lucky" on the test. After all, his clearly uncontaminated A Sample showed a T/E ratio of approximately 11/1, and the IRMS results on that same sample showed evidence of exogenous testosterone. The IRMS results have been treated by most (including me) as the "smoking gun" in the case.

Landis has an answer for that, too: the IRMS test results do not show evidence of exogenous testosterone, by WADA's own standards.

The argument hinges on the fact that the IRMS test was performed on four different hormone ratios, and more than half of them came up negative when the laboratory standard is applied.

Table 2 — Floyd Landis urine samples subjected to IRMS testing
Metabolite-Reference Blank Urine Δ‰ Sample Δ‰
A Sample IRMS, Aliquot EC31, 2006-July-23 (reference USADA 0186)
Etio - 11 Kétoétio -0.87 -2.58
Andro - 11 Kétoétio -0.48 -3.99
5ß Adiol - 5ß Pdiol -0.55 -2.15
5α Adiol - 5ß Pdiol -1.59 -6.14
B Sample IRMS, Aliquot EC31, 2006-Aug-3 (reference USADA 0352)
Etio - 11 Kétoétio -1.08 -2.02
Andro - 11 Kétoétio -0.08 -3.51
5ß Adiol - 5ß Pdiol -0.67 -2.65
5α Adiol - 5ß Pdiol -1.60 -6.39

The threshold for a positive IRMS finding for exogenous testosterone is Δ‰ = -3.0. The IRMS uncertainty, according to LNDD, is 0.8. That means, according to the International Standard for Laboratories (reference above), that a positive must show Δ‰ of -3.8 or less.

By this standard, the A sample shows two metabolites as positive, and the B sample shows only one.

Landis' team asserts that this is not enough; they argue that all four ratios must exceed the -3.8 threshold to make a positive test. The Reporting and Evaluation Guidance (reference above) is ambiguous on this point; it neither specifies which metabolites should be measured, which reference steroids should be used, nor does it specify how many or which pairs must exceed the threshold.

Depending upon the nature of the endogenous steroid suspected to have been administered, the metabolites analysed could be testosterone, epitestosterone, androsterone, etiocholanolone, the androstanediols, DHEA, or other relevant metabolites while the urinary reference steroid usually analysed by the Laboratories is one of, pregnanediol, pregnanetriol, cholesterol, 11-hydroxyandrosterone or 11-ketoetiocholanolone. …

The results will be reported as consistent with the administration of a steroid when the 13C/12C value measured for the metabolite(s) differs significantly i.e. by 3 delta units or more from that of the urinary reference steroid chosen.

It seems pretty clear, under the Reporting and Evaluation Guidance, that the LNDD could legitimately have tested for a single ratio; and if they had chosen to use the 5α Adiol - 5ß Pdiol ratio as their single measure, it would have returned an unambiguously positive result.

But they didn't do that, and I can't really predict how things are going to go here. It will come down to a question of precedent, I think; whether there is an undocumented standard for the choice of metabolites and the application of the threshold, and whether the LNDD can prove it. This will become an important question if my prediction is incorrect and the B sample results are ruled to be valid.

No comments: