Tuesday, July 08, 2008

Emailer: They still got peak identification wrong

An emailer sends us the following argument about the use of blank urine for metabolite peak identification. It's long and detailed, and it may take a few re-posts to make the formatting reasonable.

This makes reference to the newly released EXHIBITS, particularly EX 26, containing pages LNDD-0001 to LNDD-0519.

Unfortunately, I can't make the "full post on another page" work for this yet, so it's all here on the main page.


Failed Peak Identification in GC/C/IRMS

The basis of the AAF in Floyd’s case is the GC/C/IRMS analysis result.

This AAF is (mainly) based on the isotopic delta-delta values of the testosterone metabolite 5 α-Diol, exceeding the WADA threshold of 3 ‰ as well as LNDD’s threshold of 3,8 ‰ , including LNDD’s measurement uncertainty of 0,8 ‰.

In order to uphold this conclusion the metabolites of interest in this case must be identified in compliance with the ISL. The WADA Code and the ISL are clear on that:

Code Article 3.2 Methods of Establishing Facts and Presumptions

3.2.1 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 presumption 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.
6.4 Standards for Sample Analysis and Reporting.

Laboratories shall analyze Doping Control Samples and report results in conformity with the International Standard for Laboratories analysis.

ISL, page 34

5.4.4.3.1 Uncertainty in identification The appropriate analytical characteristics must be documented for a particular assay.[1] The Laboratory must establish criteria for identification of a compound at least as strict as those stated in any relevant Technical Document.

The relevant Technical Document in this case is TD2003IDCR. It states:
TD2003IDCR

The appropriate analytical characteristics must be documented for a particular assay. The Laboratory must establish criteria for identification of a compound. Examples of acceptable criteria are:

Chromatographic separation

For capillary gas chromatography, the retention time (RT) of the analyte shall not differ by more than one (1) percent or ±0.2 minutes (whichever is smaller) from that of the same substance in a spiked urine sample, Reference Collection sample, or Reference Material analyzed contemporaneously. In those cases where shifts in retention can be explained, for example by sample overload, the retention time criteria may be relaxed.

What the TD2003IDCR states is that one has to compare the retention times of the substances (testosterone metabolites) in the athletes sample with the retention times of the same substance in

  1. a spiked urine sample,
  2. Reference Material or
  3. A Reference Collection sample.

How did LNDD fail to comply with this paragraph?

  1. LNDD did not use a spiked urine sample
  2. The Reference Material (Mix Cal AC) did not contain the substances of interest (5 α-Diol, 5ß-Pdiol and Andro). Therefore a comparison of Retention Times between the athletes sample and the Reference Material is not possible.
  3. LNDD did not use a Reference Collection.

page 677:

1 CYNTHIA MONGONGU - CROSS

17:07:38 2 Q. Do you have an SOP, does
17:07:41 3 LNDD have an SOP on the blank urine,
17:07:45 4 use of the blank urine and the use of
17:07:48 5 the relative retention time technique
17:07:50 6 or method to identify testosterone
17:07:52 7 metabolites?

17:08:03 8 A. No

USADA tried during the CAS Hearing to show that the Blanc Urine is sufficient to serve for peak identification. The use of Blanc Urine in this very case is not possible because of several reasons.

1. Blanc Urine pool 4 is not a Reference Collection by definition of the ISL.

The ISL (page 36) defines Reference Collection:

5.4.6.2 Reference Collections

A collection of samples or isolates may be obtained from a biological matrix following an authentic and verifiable administration of a Prohibited Substance or Method, providing that the analytical data are sufficient to justify the identity of the relevant chromatographic peak or isolate as a Prohibited Substance or Metabolite of a Prohibited Substance or Marker of a Prohibited Substance or Method.

Those criteria aren’t met with Blanc Urine pool 4:

A. “authentic and verifiable administration”

  • We don’t have a proof of origin of Blanc Urine pool 4
  • Blanc Urine pool 4 is a negative control, no administration of the metabolites/substances of interest were conducted to the subjects who gave the urine.
  • The subjects who gave the urine weren’t examined for medical conditions or the use of medications which could exclude them from providing urine.
  • The Blanc Urine pool 4 is a pooled urine of different lab workers and not that of one person
[Note: the whistleblower documents identify at least one AAF that was recanted in mid-2006 by LNDD because of problems with the blank pool. -TBV]

B. “analytical data are sufficient to justify the identity of the relevant chromatographic peak”

  • LNDD/USADA couldn’t prove that the metabolites/substances in Blanc Urine pool 4 are the substances LNDD claimed them to be. LNDD did not verify by mass spectrometry that the metabolites contained in the blanc urine are, indeed, the metabolites. What we have is only LNDD’s assumption that 5 α-Diol in the Blanc Urine is in deed 5 α-Diol.
  • The studies Mongongou claimed she had performed on the Blanc Urine pool 4 weren’t provided. Because of the importance of this studies/verification of the Blanc urine pool 4 one would expect LNDD would have provided them if they really exist.
  • The documents provided- even in the declarations for the CAS Hearing by Buisson- are the pages LNDD0309 and LNDD0310 (SOP E-P-32).
LNDD’s Blank Urine Characteristic (LNDD0309/LNDD0310)

What is LNDD309/LNDD0310?

It is a characteristic of the negative control urine.

Which information is provided on this SOP E-P-32 (LNDD0309/0310)?

Under point 1 we learn about the codification and the collection time of the Blanc Urine:

  • It was collected between 6.12.2005 and 12.12.2005 and the code is BluP 4.
Under point 2 we learn about pH and density of this BluP4:

  • It has a specific gravity of 1,023 and a pH of 5,58.
Under point 3 we learn that BluP4 underwent a GC/MS screening analysis on the MSD18 instrument.
  • The purpose of this screening analysis by GC/MS was to establish estimates of concentrations for the metabolites.
  • It was a screening analysis and not a confirmation analysis.
  • The sample preparation and the conditions of analysis in screening and confirmation procedures are different and not comparable:

    5.2.4.3 Urine confirmation testing

    All Confirmation Procedures must be documented and meet applicable uncertainty requirements. The objective of a Confirmation Procedure is to ensure the identification and/or quantification and to exclude any technical deficiency in the Screening Procedure. Since the objective of the confirmation assay is to accumulate additional information regarding an adverse finding, a Confirmation Procedure should have greater selectivity/discrimination than a Screening Procedure. (ISL, page 19)
    • This screening was performed on the MSD18 instrument which has a HP1 column.
    • The column used in the MSD18 GC/MS instrument is an Agilent 19091Z-008 column.(USADA0045)
    • The polarity of this column is non polar and therefore different of the DB17 column used in the Isoprime 1 instrument.




    Fig.1: Agilent GC Column Selection Guide, p.81

    • There is no indication or proof that the metabolites of the Blanc Urine Pool 4 were identified by mass spectrometry.
    So the purpose of this analysis was not the identification of the metabolites but the quantification of the supposed metabolites.

    Even if LNDD performed mass spectrometry (which we don’t know, because Landis wasn't provided with the evidence this was done) to identify the very metabolites they did it on a GC/MS instrument with a different column.

    The use of the different columns in the MSD18 GC/MS instrument and the Isoprime 1 instrument would make a comparison of the retention times or pattern between GC/MS and GC/C/IRMS impossible because the elution order may have changed.

    Under point 4 (LNDD0310) we learn about the isotopic values of the supposed six metabolites.

    • The BluP 4 was analysed 3 times on the Isoprime 1 instrument and the mean delta values of those metabolites were calculated as well as the delta-delta values of the testosterone metabolites subtracted by the endogenous reference compound.
    • Those data were obtained under the “optimal” conditions of the Isoprime 1 instrument
    We do not know if the conditions of the Isoprime 1 instrument in the analysis of the BluP 4 were the same as used in the analysis of sample 995474: There is no reference to a SOP on LNDD0309/0310, i.e.to M-AN-41, where the instrumental setup of the Isoprime 1 for IRMS analysis of BluP 4 is described.

    If the argument were made that the metabolites in the BluP 4 were identified only in the GC/C/IRMS analysis, then following TD2003IDCR there must have been a Reference Solution, like Mix Cal Acetate, containing ALL six metabolites otherwise it is the same issue as with sample 995474.

    And this is: there is no identification of 5 α-Diol, 5ß-Pdiol and Andro in GC/C/IRMS analysis.

    Furthermore it is impossible that LNDD had used Reference Standard solutions like Mix Acetate or Mix Cal Acetate for compound identification in Blu P4 because the Reference standard solutions Mix Acetate (used for GC/MS) and Mix Cal Acetate (used for GC/C/IRMS) were first time approved in January [resp. May. (see LNDD0442)]. There were no standard solutions containing the reference metabolites in place before this date.

    Fig.2 LNDD0442

    This argument is supported by the documents which are showing the reference standards used for Mix Cal Acetate. The reference standard documents have a stamp on in it providing the date when those reference standards have been prepared:

    Code

    Substance

    Date

    Reference

    Cal Acetate 1

    5 α-AC

    (Internal Standard)

    17.5.2006

    LNDD0296

    Cal Acetate 2

    5 ß Pdiol AC

    17.5.2006

    LNDD0299

    Cal Acetate 3

    Etio

    17.5.2006

    LNDD0302

    Cal Acetate 4

    11-Ketoetio AC

    21.4.2006

    LNDD0305

    This table shows that the reference standards contained in the Mix Cal Ac were prepared approximately 4-5 month after analysis of Blu P 4.

    Summary of LNDD0309/0310 for BluP 4:

    1. The columns used in GC/MS and GC/C/IRMS analysis of BluP 4 were different. Therefore a comparison of retention times or pattern between those two instruments is obsolete.

    2. There is no proof of identification of the target metabolites in GC/MS analysis for BluP 4, neither by retention time (compared to Reference Standards of that of the same substance or else) nor by mass spectrometry.

    3. There is no proof of identification of the target metabolites in GC/C/IRMS for BluP 4 by comparison of the BluP 4 to reference standards of that of the same substance.

    Bottom-line: LNDD did not provide any proof that the metabolites in the BluP4 were identified according to the TD2003IDCR. Therefore BluP 4 cannot serve as a Reference Collection.

    2. Even USADA did not claim BluP 4 is a Reference Collection

    Furthermore, even USADA and USADA’s witnesses never claimed that BluP 4 is a Reference Collection (In the Hearing Transcripts the term “Reference Collection” is not mentioned once by USADA or witnesses for USADA!).

    3. The use of a Reference Collection is inappropriate in existence/presence of Reference Standards by ISL

    Leave out the above; the ISL states that “Reference Standards should be used for identification, if available.”
    We know that LNDD is in possession of reference standards for the metabolites 5 α-Diol (LNDD0287), 5ß-Pdiol (LNDD0278) and Andro (LNDD0284).

    The ISL paragraph 5.4.4.2.1. continues:

    “If there is no reference standard available, the use of data or sample from a validated Reference Collection is acceptable.”

    By claiming that BluP 4 is a Reference Collection and used for compound identification LNDD also violated this ISL paragraph because the use of a validated Reference Collection (which BluP 4 is not) is only acceptable (emphasis added) if there is no Reference Standard available.

    5.4.4.2 Validation of Methods

    5.4.4.2.1 Confirmation methods for Non-threshold Substances must be validated. Examples of factors relevant to determining if the method is fit for the purpose are: … Standards. Reference standards should be used for identification, if available. If there is no reference standard available, the use of data or sample from a validated Reference Collection is acceptable.

    ( ISL, page 32)

    Conclusion

    For all those reasons LNDD did not identify the metabolites 5 α-Diol, 5ß-Pdiol and Andro, which are the metabolites on which the AAF in this case is based.

    Without identification nobody can be sure what was measured by LNDD, there is no proof that the peaks in the GC/C/IRMS chromatograms contain the alleged metabolites, therefore there is no AAF.



    [1] (As an aside: There is the question if “analytical characteristics” includes identification of a substance. Then LNDD failed because they have no documentation about peak identification in GC/C/IRMS. LNDD’s technicians testified there is no SOP for peak identification for essay EC31; see transcript quoted above, and this on page 689:


    17:58:06 21 Q. And how do you -- where is
    17:58:10 22 the document that shows in the doc pack
    17:58:13 23 how you originally identified the
    17:58:17 24 target metabolites in the blank urine?
    17:58:21 25 Can you show that to me?

    (800) 325-3376 www.MerrillCorp.com MERRILL LEGAL SOLUTIONS Page 699

    1 CYNTHIA MONGONGU - CROSS

    17:58:25 2 A. It isn't in the doc pack.

    17:58:42 3 Q. Because in order to use the
    17:58:48 4 blank urine method you have to first
    17:58:50 5 identify the target metabolites in the
    17:58:54 6 blank urine, correct?

    17:59:13 7 THE INTERPRETER: I'm 17:59:14 8 repeating the question.

    17:59:28 9 A. Yes.

    17:59:29 10 Q. But that data is nowhere in
    17:59:35 11 the documents we have been provided,
    17:59:37 12 correct?

    17:59:38 13 A. No, there isn't anything
    17:59:47 14 about that in the documentation
    17:59:48 15 package.

    17:59:48 16 Q. And was that document
    17:59:49 17 provided to the COFRAC auditor when you
    17:59:56 18 were accredited?

    18:00:00 19 A. No.

    18:00:01 20 Q. And is the identification
    18:00:08 21 information in the blank urine anywhere
    18:00:10 22 in any other document that we have
    18:00:14 23 received outside of the doc pack?

    18:00:17 24 A. I really don't know.


    Neither do we.





    34 comments:

    Bill Mc said...

    An interesting post of yet another analysis which reaches the conclusion that LNDD's lab work does not provide proof an AAF against Floyd Landis.

    The real question now is, will the Landis case ever be heard in an impartial court (i.e., a "real" court) where reasonable rules of procedure and evidence are enforced?

    If there is any movement in that direction, I for one would be willing to contribute $$$ to the effort.

    Ali said...

    Good luck guys. I hope your further investigations lead you to where you want to get.

    Me, I've gone as far as I dare down that road and have finally "moved on" ... honest. I'm not equipped to deal with cruelty of the magnitude dealt out to Floyd and his family (even the sight of a dead badger on the road is enough to put a downer on my day ... like the one I cycled past today).

    I feel like I'm bailing out and not sticking with the programme, but, in all honesty, I need to and may as well take advantage of a break in this obsession.

    Regards everyone,

    Ali

    Larry said...

    Ali, I'll miss you. Thanks for all your hard work here, for everything you've taught me, and for mostly exempting me from your occasional lawyer-bashing!

    Be good, good luck, and may the path ahead of you be free of road kill.

    Thomas said...

    Hi everybody. I've lurked on Trust But Verify for a long time. You have presented many good arguments but this Emailer is missing something.

    The reference material is seen on USADA 0309. Mix Acetate 001 50 ng injecte.

    The blank urine pool 4 was identified on USADA 311, 315, 0319. Retention times match USADA 0310.

    Then blank urine pool 4 was used to double check IRMS results.

    Thomas A. Fine said...

    It just proves that USADA didn't prove that their AAF is valid.

    But THEY DON'T HAVE TO!

    The lesson learned from all of this is, that if the lab says their work is good enough, then it is. If it is not good enough, you have to PROVE that it is not good enough. It is NOT sufficient to merely prove that the other side failed to document it to your satisfaction. You have to actually show where they did go wrong in the stuff that they did provide to you.

    It's truly a miracle that anyone has every gotten off for anything. Thankfully, after this decision, they'll never have to worry about that happening again.

    tom

    Thomas said...

    Whoops, I meant to write retention times match USADA 309.

    DBrower said...
    This comment has been removed by the author.
    DBrower said...

    Thomas,

    If I understand correctly the argument, the retention times that match in the USADA documents are those for what was in the Mix Acetate 001 50 ng inject, which does not include the 5aA and 5bP in question. (It also doesn't have the andro.)

    So yes, there are some peaks that match between that mix cal and the blank, and different peaks that match between the blank and the sample but are not in the mix cal that was run.

    This says there was never any positive identification of those substances the blank, only a supposition that the peaks over there must be what they are believed to be.

    The concentrations on LNDD309 are given for things assumed to be what is listed (5aA, 5bP, andro), not things proven. We know this because the mix cal that had the substances was not made until after the blank analysis on LNDD0309 was done on Dec 13, 2005.

    That is what I understand the emailer to be arguing, I believe, and it is the bottom part of a discussion we had long ago that wondered if there had been any positive identification in the blanks that were being used to transitively identify the peaks in the sample.

    Some people believe that the single "anchor" from the mix call that was run is good enough, and other people see the absence of positive identification of the 5aA and 5bP peaks in question to be an ISL violation.

    The emailer is showing there was, as I think we supposed, no positive ID in the blank, ever of the 5aA and 5bP. The emailer also points out the column used in the blank analysis in December wasn't the same as that used in the sample runs that also used the blank in July and August, so any ID made in December shouldn't be assumed correct on the other column anyway.

    TBV

    m said...

    What is the relevant education, training and experience of the emailer if known? Is he/she connected to any of the parties or their attorneys?

    m said...

    Is it swim?

    Thomas said...
    This comment has been removed by the author.
    Thomas said...

    Sorry I have taken so long to respond.

    See USADA 0309 not LNDD 309. That is confusing you. USADA 0309 identifies retention times for all 7 metabolites of interest in Mix Acetate 001.

    5a Androstanol...........10.70
    Etiocholanolone..........14.37
    Androsterone.............14.63
    5b Androstan.............15.19
    5a Androstan.............15.58
    11 KetioEtiocholanolone..17.09
    5b Pregnane..............19.21

    Retention times for blank pool urine 4 are matched to Mix Acetate 001 reference on USADA 311, 315, 0319. Identification is made by GC/MS.

    For GC/C/IRMS the blank pool urine 4 was used to double check identification.

    For IRMS quantification only 4 metabolites of interest were analysed. See USADA 0354 and USADA 0361 but that is only for quantification not identification.

    I don't understand how an analysis in December affects these identifications. The Mix Acetate 001 reference was analysed 8/4/2006 with the blank pool urine 4 and B sample.

    Larry said...

    M, we are all free here to post comments without revealing identity or affiliation. Even me. Even you.

    This post is not written as expert opinion. The poster may or may not be expert, we don't know. I read emailer's analysis the same way I read your analyses and the same way you read my analyses: they stand or fall based on how well they are argued and reasoned, and on the material we cite in support of our arguments.

    The poster is providing us with an analysis, one I think is pretty good. The reasoning behind the analysis is set forth in detail and the analysis cites its sources. That's all any of us can do here.

    m said...

    Larry,

    If the poster claims some expertise and experience I will give his post more credence, and spend less time checking his claims of fact and what is proper lab practice.

    m said...

    Thomas has pointed out that in the Landis test results the retention times for all of the relevant metabolites in the Blank Urine matched the retention times for the same metabolites in the Mix Acetetate 50, which is the reference collection. This is USADA 309 et seq. in the AAA arbitration exhibits. This is distinct from LNND 309 in the CAS exhibits that refers to the COFRAC validation.

    So there can be little doubt that the Blank Urine contained the relevant metabolites and that they eluted in the proper order and retention times. So use of the Blank Urine retention times to help identify the metabolites in the Landis F3 and other samples was scientifically appropriate. Whether use of the Blank Urine met the legal definition of "reference collection" or more importantly "reference material", or otherwise was legally appropriate is a separate question, which I think can be answered in the affirmative, but am not going to go into detail here.

    The emailer tries to argue that based on the documents in the COFRAC accreditation we can't really tell if the Blank Urine contains the relevant metabolites. This is pretty specious. We know that normal urine always contains the metabolites in question because they are "endogenous". LNND 309 itself says that a GCMS screening was performed, obviously to identify the metabolites so that each metabolite could be analyzed and quantified in the Blank Urine. That is what LNND 309 shows. To argue that the Blank Urine doesn't contain the metabolites because the COFRAC documentation doesn't contain the GC/MS mass spectra is pretty ridiculous. In any case COFRAC accredited the process including its subparts such as the use of the Blank Urine. So I don't think use of the Blank Urine can legally be challenged on that basis.

    That different GCMS columns were used in the COFRAC accreditation document is irrelevant. Whatever column was used in the COFRAC GCMS was appropriate for identifying the metabolites in that GCMS. We are not trying to match retention times in that GCMS with those in the Landis case. Completely different GCMS's were performed in Landis and retention times between the Mix Cal's, Blank Urine, and Landis Samples were matched using those GCMS's.

    Thomas said...

    To clarify what I wrote last night.

    Two different reference materials were used. Mix Acetate 001 for GC/MS identification. Mix Cal Acetate 001A GC/C/IRMS quantification.

    Mix Acetate 001 contained all 7 metabolites of interest. See USADA 0309. Mix Cal Acetate 001A contained 4 metabolites of interest. See USADA 0354. All 7 metabolites were referenced for identification. The Emailer missed that.

    Metabolites of the blank urine pool are identified when metabolites of the sample are identified. See page 842 of LandisCASTranscript.pdf.

    Additionally retention times of the 4 Mix Cal Acetate metabolites are compared. See USADA 0351. tr(s) means retention time as seconds. trr means relative retention time. Relative retention time relates to 5a Androstanol AC. 5a Androstanol AC is added to each blank and sample. It is also present in Mix Cal Acetate 001A.

    In summary

    1.Blank is identified by GC/MS using 7 metabolite reference material Mix Acetate 001.

    2.Blank is used to double check GC/C/IRMS identification.

    3.Mix Cal Acetate affords additional check for 4 metabolites.

    highwheel said...

    This exactly what emailer will tell us. It is: only 4 metabolites in GC/CIRMS are identified by Mix Cal AC. You cannot make any comparison between GC/MS and GC/C/IRMS. What is identified by RT in GC/MS has no value for identification in GC/C/IRMS. This are 2 different instruments. USADA, LNDD and the AAA Panel refuted Meyer Augensteins argumentation about comparing GC/MS and GC/C/IRMS. According to TD2003IDCR you have to identify the substances in every instrument by comparison of the substances in question with "that of the same substance in a spiked urine sample, Reference Collection sample, or Reference Material analyzed contemporaneously."

    Thomas said...

    highwheel,

    GC/MS affords data GC/C/IRMS doesn't. This is why GC/MS is the standard for GC/C/IRMS identification. GC/C/IRMS must be compared with GC/MS. GC/MS identification was made with contemporaneously analysed reference material.

    DBrower said...

    Let me try to summarize what I think are relevant points.

    1. Things in the blank (and other samples) were identified in the GCMS with a mix cal that had all the targets of interest.

    2. We have exhaustively determined that there are no retention time matches between the GCMS and the GC/C/IRMS.

    3. The only identification done in the IRMS portion was against a mix cal that did NOT contain all the targets of interest.

    4. The IDCR appears to require identification against one of the three candidates on each instrument, because retention times can't be compared across instruments.

    5. USADA appears to have been successful in arguing that some metabolites were identified in the blank in the GCMS, and therefore peaks in the IRMS are the same thing.

    There appears to be a logic gap in 5, where the peaks of the blank in the IRMS were never identified using against one of the three acceptable ISL identified standards.

    We are given to understand by the award that "pattern matching" across the GCMS and IRMS was acceptable to conclude that the peaks identified properly in the GCMS were identified in the IRMS. This is argued to fly in the face of the plain language of the ISL, which does not list "pattern matching" as a peak identification method, requiring one of the three prescribed methods.

    The emailer is arguing that the blank itself cannot be considered one of the three acceptable references in the IRMS because its metabolites were never identified in the IRMS against an accepted standard.

    I think that means: even if the patterns match, it doesn't matter, because the blank hasn't been properly identified in the IRMS either.

    Have I gotten the facts correct and reasonably complete above?

    TBV

    DBrower said...

    Thomas wrote, in part:

    2.Blank is used to double check GC/C/IRMS identification.


    What is the first check of the GC/C/IRMS identification?

    TBV

    DBrower said...

    M wrote:

    Whether use of the Blank Urine met the legal definition of "reference collection" or more importantly "reference material", or otherwise was legally appropriate is a separate question, which I think can be answered in the affirmative, but am not going to go into detail here.

    Please at least outline your theory. I don't believe it was argued, and it seems to be relevant.

    I don't know that anyone is arguing the blanks don't contain the targets of interest. There is question whether specific peaks in the IRMS are the specific things they are asserted to be, since they were never identified against the three specified standards in the IRMS, either contemporaneously or otherwise.

    TBV



    TBV

    Larry said...

    I'm going to take the same question that TBV asked here, in a simple and elegant way, and ask the same question, in my wordy and ponderous way!

    M, Thomas, others -

    IMHO, I think you guys are talking past each other here, and you have not completely framed the question that we need to analyze. In this post, I'm going to spend a lot of time just to frame the question. I'm doing this so that most of the readers of TBV can follow this discussion. The discussion is of critical importance to the Landis case -- I can think of only one other issue that rises to this level of importance.

    What we're talking about here is the suitability of using the blank urine as a reference for identification of peaks in the IRMS chromatograms used by the French lab (LNDD) as the basis for the doping finding (the AAF) against Floyd Landis.

    Let's elaborate. The Landis AAF was based on measurements of the "isotopic values" of substances found in Landis' urine. (It's not essential in this post for me to explain what is an "isotopic value", but I'll be happy to do it separately if anyone asks.) The LNDD accomplished this by using a gas chromatography (GC) machine that (in theory at least) is capable of separating out the substances found in urine so that they can be separately analyzed. Each substance that separately emerges from the GC machine can be referred to as a "peak".

    In order to measure the isotopic values of each peak in the Landis urine, the LNDD used a machine called an isotope ratio mass spectrometer, or IRMS. The GC machine is connected to the IRMS machine, so that the IRMS machine can analyze each peak as it emerges from the GC. The IRMS machine produces a graph called a "chromatogram", that shows us a picture of each peak and indicates the isotopic value of the peak and the time (called the "retention time") when the peak was analyzed by the IRMS machine.

    A key part of the IRMS process is that each peak is incinerated -- the peak is heated at a high temperature so that some of the peak is burnt off, and what remains in the peak is carbon dioxide, or CO2. This will be true of every peak in the sample urine: the GC emits peaks consisting of different substances, (hopefully) one at a time, and the IRMS incinerates the peaks and turns them all into CO2! Only after a peak is incinerated into CO2 can the IRMS machine measure the isotopic value of the peak.

    The incineration required for the IRMS process creates a problem for us. Once the peak is incinerated, we can no longer directly determine what substance was in the peak prior to incineration. But we HAVE to identify the substance in the peak in order to make the test work! We need to know both the isotopic value of each peak AND the substance contained in the peak (or more accurately, the substance that WAS contained in the peak before its incineration) in order to make the test work.

    (To make this discussion read a bit more easily, from this point forward I will simply refer to the substance in an IRMS peak, rather than to the substance formerly contained in the IRMS peak before the peak was incinerated into CO2.)

    So, this is our problem: we need to find a scientifically valid way to identify IRMS peaks. How might we do that?

    Well, one possible way is to hook up the GC machine to a different machine whose purpose is to identify peaks. One such machine is a mass spectrometer, or MS. (When a GC machine is hooked up to an MS machine, we sometimes refer to the combined machines as a GC/MS.) The GC/MS produces its own kind of chromatogram, again graphically showing each peak in the urine sample and the retention time for each peak ... but the GC/MS can also identify the substance contained in each peak.

    So, perhaps, the lab could first run a portion of the urine through the GC/MS to identify peaks, then (quick as a bunny) run a different portion of the urine through the GC/IRMS, and compare the peaks in the GC/MS chromatogram to the peaks in the GC/IRMS chromatogram. Perhaps there's a way for the scientists to effectively combine the information in these two chromatograms, so that we'd know the identity AND the isotopic values for each peak in the sample.

    Is there such a way to compare the two types of chromatograms to get the information we need? Well, THIS issue is a HUGE issue in the Landis case. The Landis team said yes, the information in the two chromatograms can be combined by comparing the retention time for the peaks in the GC/MS chromatogram to the retention time for the peaks in the GC/IRMS chromatograms. If the times match for two peaks, then the two peaks are the same. If you can't match the times (and the Landis team was willing to consider some creative ways to get times to match), then you can't say scientifically that the two peaks are the same. And as it turned out, it was not possible in the Landis case to match GC/MS peaks and GC/IRMS peaks by retention time, which is one of the principal arguments used by the Landis team to argue that no doping case had been proven against Landis.

    Predictably, USADA did not agree with the Landis team on this issue. By the end of the first arbitration (the AAA arbitration), USADA had argued that no one could ever expect to match retention times between a GC/MS chromatogram and a GC/IRMS chromatogram -- USADA argued that the "plumbing" of these two machine combinations is too different to expect urine peaks to emerge and be measured at comparable times.

    But USADA DID claim at the AAA arbitration that GC/IRMS peaks could be identified by reference to GC/MS peaks. The way you would perform the identification is by "pattern matching". You'd look at the GC/MS chromatogram for a pattern of peaks -- first a big peak, then a medium sized peak, then a small peak, then another big peak, and so forth -- then see if you could find the same pattern on the GC/IRMS chromatograms (big, medium, small, big, etc.). If you could match patterns, then you could match peaks: the first big peak on the GC/MS is the same as the first big peak on the GC/IRMS, then the medium sized peaks would be the same, and so on. The Landis team claimed that "pattern matching" was subjective and not truly a scientific criteria for peak identification, but the majority of the AAA arbitration panel disagreed and convicted Landis on the pattern matching.

    Flash forward to the recently concluded CAS arbitration. By the time this case had been argued to the CAS arbitration panel, the position of USADA had shifted somewhat. Yes, USADA still argued that LNDD had used pattern matching to help it identify IRMS peaks. But USADA ALSO claimed that LNDD used a SECOND technique to identify IRMS peaks -- and that, in fact, the scientific and legal validity of LNDD's IRMS peak identification PRIMARILY rested on this second technique.

    What was this second technique? This second technique is the testing of the "blank urine" that we've been discussing in connection with emailer's post on "They Still Got It Wrong".

    USADA now argues that LNDD definitively identified the IRMS peaks in the Landis urine sample by comparing (1) the retention times for the IRMS peaks in the Landis sample, to (2) the retention times for IRMS peaks in the blank urine. Under this method of identification, we're no longer comparing IRMS peaks to GC/MS peaks -- instead, we're comparing IRMS peaks to other IRMS peaks, so comparison of retention times should work (according to USADA). Pursuant to this method, if the retention time for an IRMS peak in the blank urine chromatogram matches the retention time for an IRMS peak in the Landis chromatogram, then we can conclude that the substance in the athlete's IRMS peak is the same as the substance in the blank urine IRMS peak.

    But there's a problem with this method, a hole in the method that TBV has pointed out. It's fine to say that we can identify IRMS peaks in an athlete's sample by reference to IRMS peaks in the blank urine. But how do we FIRST identify the IRMS peaks in the blank urine?

    Of course, we cannot identify the IRMS peaks in the blank urine by comparing them to the IRMS peaks in the athlete's sample. That would be circular.

    We cannot identify the IRMS peaks in the blank urine by comparing them to the GC/MS peaks for the blank urine. That comparison is just as problematic as the comparison we discussed above, where both sides tried to compare GC/MS peaks from the athlete's sample to IRMS peaks for the athlete's sample. Either we'd have to make this comparison using retention times as argued by the Landis team, which USADA says doesn't work, or we have to use pattern matching, which no one seems to think is good enough standing alone.

    (Besides, if there was a valid method for identifying blank urine IRMS peaks by reference to blank urine GC/MS peaks, then we WOULDN'T NEED to use the blank urine AT ALL for identification purposes -- we could use this method directly to identify the athlete's IRMS peaks by reference to the athlete's GC/MS peaks.)

    The blank urine might be helpful in IRMS peak identification if it were easier to analyze than an athlete's sample. But this is not the case. Blank urine is "live" urine, from human beings, and is supposed to resemble the urine from a non-doping athlete. It is every bit as complicated as an athlete's urine, must undergo "sample preparation" just like an athlete's urine, and is just as difficult to analyze as an athlete's urine.

    So the question on my mind is: how does using blank urine help us identify IRMS peaks in an athlete's sample?

    I think this is the question TBV is asking, not in a nutshell!

    Unknown said...

    Larry said, " I can think of only one other issue that rises to this level of importance."

    Larry,
    I look forward to your thoughts on this other important issue. Can you remind me what this issue might be, so I can eagerly anticipate your more thorough analysis?

    Thanks,
    Cal

    Larry said...

    Cal, IMHO, the other critical issue is testosterone metabolism: if (as LNDD and USADA claimed) Landis used exogenous testosterone, how is it that (as LNDD and USADA claimed) this exogenous testosterone is evident in only one of the four measured testosterone metabolites?

    I understand that testosterone metabolism is a VERY complicated matter and that the following analogy is simplified to an extent that might be seen by some as irresponsible, and by others as close to criminally irresponsible, but consider a green empty beer bottle. Now, in your mind, take an imaginary hammer and smash the bottle. You'll end up with a lot of broken pieces of glass, and they'll all be green.

    Now, imagine exogenous (artificial) testosterone molecules, which according to USADA and LNDD will contain on average fewer C13 atoms than normal. Sort of like the molecules are green like the bottle. Now, with your imaginary hammer, smash (metabolize) those molecules. LNDD and USADA would have you believe that it's normal for some of those smashed pieces NOT to be green! (i.e., not to be short of C13 atoms)

    How do you do that? How do you smash a green bottle and end up with broken pieces of brown glass?

    OK, testosterone metabolism is not nearly as simple as smashing glass ... I'm still trying to figure out what the story is here. From my reading, the BEST case that USADA can make is that Landis has an extremely unusual metabolism. Which would be a coincidence in a case full of coincidences. But I have a lot more reading to do.

    Thomas said...

    TBV and Larry,

    The pattern matches are so evident between GC/MS and GC/C/IRMS I believe this matter is a nonstarter. Sample patterns match. Blank urine pool patterns match. The 4 metabolites in the Mix Cal Acetate 001A match the sample and blank urine pool IRMS analysations. Add in Mix Acetate 001 and there is too much evidence to overcome.

    Larry,

    I as you do about testosterone metabolism. See other IRMS samples the LNDD lab judged positive. See LNDD 0436 at usada-ex 26-lndd-0001-0519.pdf. One sample is similar to that of Landis. Sample no. 15. But the T/E of that sample is 33.6.

    I want to believe in Floyd. I am not naive enough to rule out cheating but I believe he would not have put himself through what he has if he cheated to win the Tour de France.

    DBrower said...

    Thomas,

    We have here been over pattern matching, and not reached any particular conclusion.

    USADA certainly argued that pattern matching was acceptable, yet we haven't identified a particular place in the ISL where that is identified as a method that should suffice; we have only comparision of retention times with a tolerance to one of the three standard types, none of which LNDD has done for the 5aA and 5bP in the IRMS tests.

    One of the reasons this is important is that it could (should?) have caused a formal "burden flip", where USADA was obliged to prove that the failure to meet the ISL did not cause the AAF. Neither panel chose to declare a burden flip on the IRMS identification, forcing USADA to prove the assertions made.

    It may be that USADA could prove that the identification issue did not cause the AAF -- but doing it under a burden flip is harder for them than without it, where a presumption of correctness is held. Thus, throughout, when The Panel says "X is not an ISL violation, and if it were it would not have caused the AAF, it is making the latter determination as a hypothetical, under the presumption it didn't cause it. Had a violation been declared, there would have had to be more evidence presented to make those cases, and USADA didn't do much of that. Instead, USADA wanted to make sure that no violations were declared.

    What we have here with identification is an example of where, it seems to me, a reasonable person could have declared an ISL violation, and given USADA the opportunity to prove it didn't cause the AAF. By declining to declare the violation, there was really no reason for The Panel to have discussed whether it would have caused the AAF, because that point was never really argued. In fact, it is probably improper for the panel to consider the second half when considering the first, because of the changing burdens of proof.

    As to the metabolite issue, there is much to be learned from LNDD 0436 -- how many AAFs did LNDD declare, and how many of them were single metabolites with T/Es like Landis? It would seem that is a very atypical positive. That raises questions about whether it is a positive, leading us to validation and positivity criteria -- which we should REALLY take to another post.

    I'd be delighted if someone finds the validation material in EX 26 and compares it to the positives LNDD has declared, and then compares it to the Landis sample(s). If emailed, I'll put it up as a post to start that discussion.

    thanks,
    TBV

    DBrower said...

    Thomas,

    We have here been over pattern matching, and not reached any particular conclusion.

    USADA certainly argued that pattern matching was acceptable, yet we haven't identified a particular place in the ISL where that is identified as a method that should suffice; we have only comparision of retention times with a tolerance to one of the three standard types, none of which LNDD has done for the 5aA and 5bP in the IRMS tests.

    One of the reasons this is important is that it could (should?) have caused a formal "burden flip", where USADA was obliged to prove that the failure to meet the ISL did not cause the AAF. Neither panel chose to declare a burden flip on the IRMS identification, forcing USADA to prove the assertions made.

    It may be that USADA could prove that the identification issue did not cause the AAF -- but doing it under a burden flip is harder for them than without it, where a presumption of correctness is held. Thus, throughout, when The Panel says "X is not an ISL violation, and if it were it would not have caused the AAF, it is making the latter determination as a hypothetical, under the presumption it didn't cause it. Had a violation been declared, there would have had to be more evidence presented to make those cases, and USADA didn't do much of that. Instead, USADA wanted to make sure that no violations were declared.

    What we have here with identification is an example of where, it seems to me, a reasonable person could have declared an ISL violation, and given USADA the opportunity to prove it didn't cause the AAF. By declining to declare the violation, there was really no reason for The Panel to have discussed whether it would have caused the AAF, because that point was never really argued. In fact, it is probably improper for the panel to consider the second half when considering the first, because of the changing burdens of proof.

    As to the metabolite issue, there is much to be learned from LNDD 0436 -- how many AAFs did LNDD declare, and how many of them were single metabolites with T/Es like Landis? It would seem that is a very atypical positive. That raises questions about whether it is a positive, leading us to validation and positivity criteria -- which we should REALLY take to another post.

    I'd be delighted if someone finds the validation material in EX 26 and compares it to the positives LNDD has declared, and then compares it to the Landis sample(s). If emailed, I'll put it up as a post to start that discussion.

    thanks,
    TBV

    Larry said...

    Thomas -

    If I'm following you correctly, you're not solely relying on the blank urine as an exclusive or even a sufficient method of IRMS peak analysis. You're kind of combining evidence from a number of sources, including RTs from the mix cal and pattern matching?

    OK, fair enough. I think that your argument falls into the "what else could these test results mean?" mode of analysis. I think that this is the crux of the analysis I've read from OMJ at DPF, whose opinion I respect even though I don't agree with it. The "what else could it mean?" approach is, I think, the essence of the reasoning behind the AAA opinion, I think it's the essence of the conclusion reached by Caitlin at UCLA, and based on the conflicting testimony of the witnesses, I think that a number of reputable scientists would agree with your assessment. But given the Landis witnesses, we can guess that there are at least some scientists that hold a contrary opinion.

    I'm not a scientist, so I can't really say who's right as a matter of science. I AM a lawyer, so my natural tendency is to focus on the rules. And as emailer correctly points out, the WADA lab rules require the labs to have criteria in place for the identification of substances. We can argue whether the criteria has to be in writing and part of the lab's SOP (I think so), but at minimum the lab must establish criteria.

    So, let's put to one side the question of whether we can assemble enough evidence from various sources to identify the peaks, and ask instead, do we see evidence here that LNDD had established CRITERIA for peak identification?

    I think the clear answer to this question is "no". Peak identification by "pattern matching" is not criteria in the absence of some kind of written direction as to how to do it. Peak identification by using the mix cal is a great idea, but if someone from LNDD had intended this as a "criteria", they would have specified use of a mix cal containing all 4 testosterone metabolites (especially since LNDD had such a mix cal and used it for other purposes). I don't know how LNDD could have come up with "criteria" for use of the blank urine, for the reasons I've previously specified.

    Switching over to the question of testosterone metabolism, oh boy, this is a complicated question. I've spent part of the evening reading the testimony of Dr. Amory at the CAS. There's a lot to learn and absorb. I suppose it's possible for a person to use exogenous testosterone and show the CIR results that LNDD measured for Landis, but the odds seem to be heavily against it.

    Thanks for your reference to LNDD 0436, this is terrific evidence for how unusual LNDD's measurments for Landis really were.

    m said...

    Thomas states

    "The pattern matches are so evident between GC/MS and GC/C/IRMS I believe this matter is a nonstarter. Sample patterns match. Blank urine pool patterns match. The 4 metabolites in the Mix Cal Acetate 001A match the sample and blank urine pool IRMS analysations. Add in Mix Acetate 001 and there is too much evidence to overcome."

    Exactly! Not only pattern matches but retention time matches. I have made extensive posts in the past pointing this out.

    Unfortunately the audience on this site is very resistant to being persuaded by those facts and arguments. Although at one time I thought I had made a dent.

    Vic e-Teacher said...

    Grat post re GC/IRMS GC/MS larry

    Larry said...

    rdk, thanks!

    M, be nice. This "audience" has given you a long, polite and arguably fair hearing, and we have debated you on the substance. I have acknowledged to Thomas that the evidence assembled by USADA (an incomplete mix cal, plus a pinch of pattern matching, plus whatever the blank urine is worth scientifically) appears to add up to a "what else could it be?" kind of argument that persuaded a lot of people with scientific bona fides. I may personally be resistant to buying into this sort of patchwork of evidence, in some part because it feels like it was assembled after the fact by USADA rather than representing the "criteria" that LNDD was required to have in place under TD2003IDCR, in part because it's obvious that LNDD could have done a much better job of this by using the mix cal with all of the relevant metabolites. But I have acknowledged that the patchwork IS evidence of a sort.

    I think TBV has a strong point that a "what else could it be" kind of argument might have been most fairly put forward explicitly by USADA, in the context of an ISL burden shift.

    M, I'm determined to be an honest reporter here, and not to be argumentative. It's fair to judge how well I succeed at this, but this is all I'm trying to achieve these days. Together, you and I serve the audience here when we explore issues cooperatively and find common ground. If you want an argument, I'll go along to an extent, but my heart's not in it. If you want to help us explore the issues and make sure we understand the argument on the other side, then I'll be a lot more enthusiastic.

    The scientific arguments regarding peak identification seem to resemble the scientific arguments all over the Landis case: Landis had his experts, USADA had their experts, it was a classic battle of the experts. As you know, it's impossible for judges, juries and arbitration panels to listen to competing experts and make a decision based on which expert is "right". By definition, the experts are more learned in their area of expertise than we can ever be -- we can't be smarter than they are. So as you know, judges/juries/panels do one of two things with a battle of experts: either they throw out all of the expert evidence and make their decision on other grounds, or else they choose to believe one side's experts based on what they perceive to be superior credentials. I think the CAS went in the second direction, as USADA had all of the ADA experts on its side. You can understand how frustrating this is for some of us, knowing as we do that ADA experts cannot testify in favor of an accused athlete.

    So ... as a lawyer I turn back to the rules, where things are clearer (I think) and where emailer focused his/her/their attention. I AM curious to hear your reaction to emailer's arguments (and while I might wish for a response that seeks common ground, that may be harder to do when we cross from the scientific into the legal arguments).

    Thomas A. Fine said...

    So we're working on the transitive property here where we've shown that A=B, but we just hope that B=C, because even though we can't prove it, they look kinda similar. And that's the basis for saying that A=C. Does that about sum it up?

    Dang larry, you opened the door for me, what do you expect me to do. So I'll just review it one more time. I know of only one thing that has been established in studies to boost production of 5-alpha metabolite compared to other T metabolites. This same substance has been shown to short term boost general testosterone production. And this same substance has been shown to rapidly incorporate it's carbon atoms into cholesterol (from which all testosterone is made).

    Alcohol.

    And if the alcohol is barley-based, it should have a very negative d13C, which should skew all testosterone metabolites towards the negative (because of the production boost), and should skew the 5-alpha particularly negative (because of the bigger boost).

    Of course if Floyd had actually been drinking beer the night before, that would almost be a slam dunk, but what are the odds of that?

    Now, to be fair, rumor has it that another substance can also cause a similar profile. Allegedly T-gel can do this, and allegedly these results are peer reviewed (by an informal non-standard definition of "peer reviewed" which seems to mean "I showed it to some people"). But so far any such studies, if they exist, have not been published anywhere that I can find.

    tom

    daniel m (a/k/a Rant) said...

    Tom,

    Interesting you would make that point. Though the name escapes me right now, wasn't there an Irish athlete who was cleared of a testosterone positive after having been out for a bender the night before he got tested?

    I'd love to see the documentation behind that decision, especially to see if the IRMS results were similar to Floyd's.

    Thomas A. Fine said...

    Gareth Turnbull

    He was prosecuted based on T/E results, because while the CIR test was done, the results were inconclusive. I'd love to see those results too.
    I think they tested his CIR in one lab, then
    retested again elsewhere and on different metabolites.

    He was cleared because of the effect of alcohol on T/E, and because of the unlikelihood of doping benefit while NOT racing.

    tom