Saturday, December 15, 2007

"M" on the 5aA identification

An article "M" sent in word format, that we hope not to have butchered in transformation.

Was the 5A Androstanedial properly identified in the GC-IRMS chromatogram?

Landis was found to have doped with synthetic testosterone. This finding was based on identifying some metabolites of testosterone by gas chromatography (GCMS), in particular the 5A, and then analyzing the carbon content of those metabolites by GC-IRMS to determine whether the testosterone was synthetic.

What I want to establish with my graphs is that the metabolites, in particular the 5A, was accurately identified.


Principals of Identification:

In gas chromatography substances (analytes) pass through a GC column and are recorded as a pattern of peaks, with the size of the peak proportional to the amount of the analyte. Analytes are identified using 2 methods. Each analyte takes a unique time to pass through (elute) the GC column. This is called its retention time (rt), often measured relative to some known substance, called the internal standard. One method is to compare the rt of a sample substance with the rt of a known substance in a “reference material”; if there is a match within a certain degree, here 1%, then one can be sure that the substance is identified. The second method and most reliable method is to analyze the mass spectra of the substance as it exits the GC column with a mass spectrometer. This process is called GC-MS. However, in the GC-IRMS the substance is combusted to carbon atoms after it exits the GC column, so no mass spectra can be analyzed and identification is based on rt. Typically a substance will take longer to pass through the GC-IRMS because of the combustion phase and the exact rt will vary depending on the GC settings used.

“Generally chromatographic data is presented as a graph of detector response (y-axis) against retention time (x-axis). This provides a spectrum of peaks for a sample representing the analytes present in a sample eluting from the column at different times. Retention time can be used to identify analytes if the method conditions are constant. Also, the pattern of peaks will be constant for a sample under constant conditions and can identify complex mixtures of analytes. In most modern applications however the GC is connected to a mass spectrometer or similar detector that is capable of identifying the analytes represented by the peaks.

The area under a peak is proportional to the amount of analyte present.”

Wikipedia, “Gas Chromatography”.

Three metabolites of testosterone are 5A, 5B and P. These occur naturally in urine. Brenna testified that one can compare the pattern of peaks (and corresponding rts) in the GCMS with the pattern of peaks in the GC-IRMS.

With that let us look at the graphs.


I have relabeled the graphs as follows.

I = Internal Standard; B = 5B androstandial; A = 5A androstandial; P = pregnane: a,b,c,d = small unidentified peaks; X = large unidentified peak


Fig 1: click for bigger



The I, B, A, and P in the F3 sample are properly identified in the GCMS because their retention times match those of the Mix Cal Acetate, which is the reference standard here. Further it appears that the mass spectra for each also verify the identity of each analyte. We should also note that the analytes in the blank urine are also properly identified because their retention times match those in the Mix Cal, and that a sequence of peaks has been established based on retention times. Since the analytes in the blank urine match the Mix Cal, the blank urine can also be used as a reference standard.


The question is whether the analytes in the GC-IRMS are properly identified. In particular is the peak identified as A (the 5A) in the GC-IRMS the same peak properly identified as the A (the 5A) in the GCMS.

When we compare the retention times of the IRMS analytes to the those in GCMS Mix Cal Acetate we find that they are off by about 4-6%, but follow the same pattern. Based on the Meier and Brenna testimony we know that the retention times and relative retention times of the analytes in the GC-IRMS cannot match those in the GCMS within a 1% standard because of the delay caused by the combustion time and the general difficulty of matching conditions on different machines. In addition the temperature ramps, and possibly even the columns were different in this case. Meier conceded that one can often only achieve a 2-3% match under ideal matching conditions. Nevertheless the retention times/relative retention times here matched within 4-6% and the pattern of peaks (retention times) also matched.

Since the same substances are in both the GCMS and IRMS F3 sample, we know that all the major (and even minor) peaks that appear in the GCMS must also appear in the GC-IRMS, unless the peak disappeared (contained no carbon).

In the GCMS the A followed the B by 21.6 seconds, which is quite close together. Indeed 20 seconds is the outer limit for matching retention times specified by WADA 2003IDCR. In the IRMS the A followed the B by 34 seconds. So there was some further peak separation in the IRMS of the A and B. The B followed the IS by 268 seconds in the GCMS, and by 446 seconds in the IRMS, a significant separation in the IRMS which explains why the retention times and relative retention times are so far off. So it appears that the peaks separated but did not change their sequence.


Fig 2: click for bigger

Could the B or A peak in the GCMS have shifted to some other position in the IRMS rather than in the 4 central peaks? This is highly unlikely given the pattern of peaks and retention time matching. And for the B we know that this is impossible, because it’s retention time matches that of the B in the Mix Cal. TBV and others claim the A could have shifted to the peak at b. But in that case what happened to the b peak, it must be accounted for? The only way for the A to have shifted to b, is for both of those peaks to have switched positions. Moreover, somehow the peak size of the b must increase to look like the A, and the peak size of the A must shrink to look like to b. The probability of this happening is close to zero.


One other corroborative fact is that the retention times of the I, B, A, and P in the IRMS F3 match those in the IRMS Blank Urine, and the I and B match those in the Mix Cal. Recall that the retention times of all of the analytes in the GCMS Blank Urine matched those in the GCMS F3 and Mix Cal. This shows that in this case the retention times of the analytes in the Blank Urine and any reference standard like the Mix Cal should be identical. That is the Blank Urine can be used as a reference standard in the IRMS, and as we have seen all the retention times match with those in the F3, including the 5A. Now it’s true that if the 5A moved in the IRMS Mix Cal then it would move in the Blank Urine also. But we have seen from the Mix Cal retention time that the 5B did not move relative to its position in the 4 central peaks. The internal consistency is just too strong. The 5A did not move, and follows the 5B in the IRMS just as it did in the GCMS. Its retention time matches the Blank Urine.

The fact that the Blank Urine can be used as the Mix Cal in the IRMS suggests that even if the Mix Cal had contained the 5A it would not have provided any more information. If the Mix Cal had contained the I, B, A, and P, and all the retention times had matched would we be any more sure that the A was properly identified than using the Blank Urine. Not really, because if the A had shifted in the Blank Urine, it would have also shifted in the Mix Cal.


The claim that the 5A peak shifted or switched in the GC-IRMS because the chromatographic conditions were different than in the GCMS can be easily tested. Simply duplicate those conditions and run a new GCMS and see whether the peaks shifted. It wouldn’t cost that much. Landis doesn’t do that because he knows it won’t change anything. I note that Shackleton’s published study at figure 3, shows the same sequence of 4 central peaks with the 5B closely followed by the 5A using the same chromatography column as used by LNND. Even if a different column was used for the GCMS as some are claiming the sequence of peaks did not change.

So we have 2 propositions:

1. The A in the IRMS is the A in the GCMS.

2. The A in the IRMS is some other peak in the GCMS, the b.

I challenge those who question the identification to assign a probability to each statement, and back it up with analysis. I assign a probability of 95% to 1, and 5% to 2.



Mix Cal

Bl. Urine


































GCMS F3 retention times: USADA 324
GCMS Mix Cal Acetate retention times: USADA 309
GCMS F3 Blank Urine retention times: USADA 324

GC-IRMS F3 retention times: USADA 351
GC-IRMS F3 Blank Urine retention times: USADA 351
GC-IRMS Mix Cal Acetate: USADA 362


m said...


graphs were truncated.

I'll try sending them again in text format, but the margins may be too wide.

Larry said...

M, I'll try to work my way through this over the weekend. So far, I'm not sure you're saying anything here that you haven't already said before. I've also raised a number of comments under the 5bA anchor thread to the arguments you've made here, and you have not had the opportunity to respond to many of these comments, so I don't know if I should repeat these comments here.

In fact, I don't know whether you really want to argue these points with me or not. If we're going to start, I'd start with your principles of identification, which are incomplete and somewhat misleading.

However, this is primarily a science post, so you may prefer to engage one of the people here whose science acumen is better than mine. You can let me know your preference.

Mike Solberg said...

Nice argument, m. That's the clearest and most complete statement of the argument you've made. Currently, I am thinking it is right, but still pondering a few things.

If it is right, what do you think it says about specificity? It shows that they did measure some 5aA and 5bA and P in those peaks (in both GCMS and IRMS, but does it show that they measured ONLY those substances in each? I'm still cogitating on that.


m said...


This doesn't say much about specificity, or whether some other substance could be hidden under the 5A peak in addition to the 5A.

I've previously said, following OMJ, that it's up to Landis to show the likelihood that there is some substance in his urine which elutes at exactly the same time as the 5A.


This is a restatement of my position intended to make it more understandable, since some have appeared to have misunderstood it.

I didn't respond to your arguments, because I think most of them go to your idea of general legal/scientific principals regarding identification, and don't deal with the specific F3 retention times and graphs, and because I didn't have the time. I do disagree that any of those arguments discredits the identification that I believe I have shown.

Larry said...

M -


I think that many of my arguments ARE specific and DO address the F3 retention times and graphs. In particular, I was interested to hear your reaction on my posts showing from the data that there's no necessary relation between MS and IRMS peak heights. (Anyone interested in confirming this can look at the S17 "B" sample F2 MS and IRMS peaks.)

In any event, since you've characterized this post as a restatement of your earlier posts, and since I've commented extensively on your earlier posts, I'm sure you understand my position on what you've written here. We're in fundamental disagreement on the issue of peak identification. Let's let it go at that. If you want to take me up on any of my previous comments, you know where to find me!

Russ said...

Your understanding of the y axis from wikipidia, for MS can't be correct.

Here is a quote from item (link later):
MS identifies substances by electrically charging the specimen molecules, accelerating them through a magnetic field, breaking the molecules into charged fragments and detecting the different charges. A spectral plot displays the mass of each fragment. A technician can use a compound's mass spectrum for qualitative identification. The technician uses these fragment masses as puzzle pieces to piece together the mass of the original molecule, the "parent mass."

Here is the link:

The y axis, in ms appears not to detect quantity as you think but rather identity by spectrum analysis.


Larry said...

Russ, you're talking about a mass spectrum graph, and you've described it correctly, but M is describing a different kind of graph.

You're describing a mass spectrum graph. The x-axis on the mass spectrum graph shows the type of ion (by mass-charge). The graph itself contains a number of vertical lines, corresponding to the ions produced in the process you've described. The graph looks kind of like a bar code.

M is describing a total ion chromatograph. The x-axis for the total ion chromatogram shows the retention time for each ion. So this graph looks like a line graph, with a number of jagged peaks, each peak representing a substance. The area described by each peak represents the amount (relative to the other peaks on the graph) of the substance described by the peak.

I think that M's description of chromatography is less than ideal, but the information he quoted from wikipedia is correct to the best of my knowledge.

the Dragon said...


You and by reference OMJ continue to say FL MUST prove x,y,or z.

How is this done practically speaking?

The ONLY time that the sample was in FL's possession was immediately before it's collection. From that point forward it is/was in the control and possession of WADA World.

That sample was conveniently used up (was this the controlling intent of the several 4-5 hours of tests with no documentation?). Records on the hard drive were supposedly erased, WADA World withheld testing results to the last possible moment to cripple their analysis by FL side. I'm sure you see all those circumstances innocent, my take is quite different.

Again, how can FL PROVE anything? Your answer seems obvious...GOTCHA!!!


m said...


The chances of any other any other substance eluting at exactly the same time as the 5A and being completely hidden under the 5A peak is minuscule.

At a minimum, Landis should identify such a substance. There are reference libraries of substances with their retention times out there. Secondly he must show how such a substance might occur in human urine. Thirdly he must show how it might occur in his own urine, much less in the F3 urine sample after it had been fractionated. He has not done any of that. He has not raised this argument at all.

He faces a huge hurdle, because it is pretty evident from the blank urine GCMS chromatograms that there is likely no substance hiding under the 5A in normal human urine. So he will have to show that it is something unique to himself. This is why Mike has made such a big to do about Landis taking Cortisol.

I assume that Landis had blood samples and probably urine samples taken during training and possibly even during or after the race. He might start with those. The reason he hasn't is because his experts have told him it won't yield anything.

Ali said...


You said:

"So we have 2 propositions:
1. The A in the IRMS is the A in the GCMS.
2. The A in the IRMS is some other peak in the GCMS, the b.

I challenge those who question the identification to assign a probability to each statement, and back it up with analysis. I assign a probability of 95% to 1, and 5% to 2."

I'd be interestd to learn how you derived these probabilities. You must have accountered for all factors potentially effecting peak position (column characteristics, temperature ramps, combustion, etc).

If you can describe your methods, it would certainly add weight to your arguement.

How did you derive these figures ?


floyd said...

"I assume that Landis had blood samples and probably urine samples taken during training and possibly even during or after the race. He might start with those. The reason he hasn't is because his experts have told him it won't yield anything."


You start with assumptions and finish the paragraph with statements that you claim to know about those assumptions. You are better than that "M", maybe you should work on your brief for a while, after all it's due on Jan. 31.

But remember, in a real court, assumptions like that have little value, as do scientific tests with (in your own words) a 95% chance of being correct. So take comfort in knowing that this is an arbitration, but don't get too comfortable because Mr Mclaren appears to be working on other things and couldn't be on the panel.

Also, sorry to hear about your first loss last week, hang in there, you'll get used to it.

the Dragon said...


You're good.

Can you please provide the detailed chain of custody documents to prove that the sample was not tampered with? NO!!!

Can you provide documentation that neither tampering or cherry picking did not occur during those several 4-5 hours of test runs? NO!!!

Can you provide the Mass Spectra Data which LNDD chose to erase? NO!!!

I am sure you believe FL has those records as well. I'd hang him until he provides them...


ps: should FL have such other sample, and they proved his point, you and OMJ would be the first to say it means nothing. Wasn't there a book & movie about this 30 odds ago? Catch-22

Russ said...

Larry, m,
I am not disputing the running investigation you are doing, my concern is simply to clear up
what I think is a bit of confusion. I think the wiki is like a middle school intro to mas-spec
and confuses things. The MS is a mass identification thing. Duck calls the quantification a
"TIC" or toatal ion current plot. The fine points are useful in understanding proper scientific
identification versus quantification.
I suggest reading the entire article linked from scientific,org (not that long) but I have
selected a number of quotes from it herein that I think informative to various points being

Quote from Duckstrap on DPF

The plots you are referring to are "TIC", or total ion current, plots. Essentially this is the chromatogram of all ions in the sample. They are not specific for any particular ion. LNDD have labeled the location of several of the peaks of interest, but these plots, of themselves, do not provide any definitive identification of those peaks.

More selected quotes from the prior link:
Note this paper is in three sections
1. GC 2. MS 3. GC/MS

from "Gas Chromatography"

" Proper scientific practice requires that the GC technician compare the spectral output with a known standard sample of the suspected substance. The standard sample must be analyzed with the same instrument, under the same conditions, immediately before and immediately after analyzing the unknown specimen. If the resulting three spectral outputs do not agree, the technician can not make a reliable identification of the specimen based on the GC analysis."

" Less than ideal spectral peaks may indicate less than ideal analytical procedures or equipment. The technician can readily observe whether the output exhibits unsatisfactory results. Ideally, the spectral peaks should be symmetrical, narrow, separate (not overlapping), and made with smooth lines. GC evidence may be suspect if the peaks are broad, overlapping, or unevenly formed. If a poorly shaped peak contains a steep front and a long, drawn-out tail, this may indicate traces of water in the specimen."

"Crucial Factors

GC analysis is highly reliable if the instrument is properly maintained, the technician follows proper procedures, and the interpretation of the results is competent. While some factors rarely affect GC analysis, some factors are absolutely essential for the use of reliable GC evidence. In all cases a technician must process a standard sample containing a verified composition identical to the presumed contents of the collected specimen. This standard sample must be processed before and after the collected specimen under identical conditions. Any output from the collected specimen that does not match the standard sample is inconclusive. If tabulated reference data exists for the relevant conditions, the specimen data must match the reference data.

If advance notice of GC testing is available, an adverse party should observe the procedure. If a retained consultant or the knowledgeable attorney observes the technician's use of the GC instrument, important information can be recorded. The technician's preparation of the specimen and the subsequent injection can be observed for errors or malfunctioning equipment. The observer should record the instrument's make, model, serial number, injection temperature, column temperature, carrier gas flow rates and pressure, identify the type of detector used, and observe any manipulation of the data by use of a computer. Ensure that the technician properly starts measuring the time at injection and records the time of elution. Any discrepancy in the time will produce an erroneous retention time. If the procedure can not be observed, the adverse party should seek all pertinent information (experimental conditions, measurements, instrument identification) and hard copy output. "

"Providing that the interpretation of the output correctly determines the parent mass, MS identification is conclusive. "

from "Mass Spectrometry (MS)"
from "Description of process"

" In the acceleration chamber the charged particle's velocity increases due to the influence of an accelerating voltage. For one value of voltage only one mass accelerates sufficiently to reach the detector. The accelerating voltage varies to cover a range of masses so that all fragments reach the detector.

The charged particles travel in a curved path towards the detector. When an individual charged particle collides with the detector surface, several electrons (also charged particles) emit from the detector surface. Next, these electrons accelerate towards a second surface, generating more electrons, which bombard another surface. Each electron carries a charge. Eventually, multiple collisions with multiple surfaces generate thousands of electrons which emit from the last surface. The result is an amplification of the original charge through a cascade of electrons arriving at the collector. At this point the instrument measures the charge and records the fragment mass as the mass is proportional to the detected charge.

The MS instrument produces the output by drawing a array of peaks on a chart, the "mass spectrum." Each peak represents a value for a fragment mass. A peak's height increases with the number of fragments detected with one particular mass. As in the case of the GC detectors, a peak may differ in height with the sensitivity of the detector used."

"Analysis of Output

Each substance has a characteristic mass spectrum under particular controlled conditions. A technician can identify a specimen by comparing the specimen's mass spectrum with known compounds. Quantitative analysis is possible by measuring the relative intensities of the mass spectra.

Usually a mass spectrum will display a peak for the unfragmented molecule of the specimen. This is commonly the greatest mass detected, called the "parent mass." Like the picture on a puzzle box, the parent mass is used to fit the pieces together from the other peaks in the mass spectrum. The parent mass reveals the mass of the molecule while the other peaks indicate the molecule's structure.

Determining the parent peak and consequently the molecular mass of the specimen is the most difficult part of MS analysis. Identifying the parent mass is outside the scope of this article. Assuming that a technician can correctly determine the molecular mass, the technician makes an educated guess of the specimen's identity and compares the mass spectrum to reference spectra for confirmation. The mass spectra for larger molecules containing carbon are complicated and require tedious calculations that are subject to error. Computers are commonly used for spectral analysis. "

from "Limitations"


If the interior pressure in an MS instrument is too high, erroneous results may occur. As the specimen molecule breaks up, the fragments accelerate. If a fragment collides with another fragment, then these two fragments may combine to make a new particle. In this event, the detector will register the mass of this new particle on the mass spectrum. The reference spectra for comparison are produced under low pressure conditions which minimize collisions between fragments. A technician would find a spectral peak where one is not expected. In the puzzle analogy, this is similar to finding pieces from a different puzzle in your box and trying to make these extraneous pieces fit. As this is impossible, any MS analysis under high pressure conditions would depend greatly on guesswork by the technician. "

"Technician's Skills

As in the puzzle analogy, knowing the shape of a piece of the molecule helps to join the pieces together. To determine the specimen's molecular structure before fragmentation, the technician needs to employ skill and art to determine the molecular structure from mass spectra patterns. Computers and databases can assist, but a human expert is necessary to distinguish between likely and unlikely answers. Alone, a computer can not determine molecular structures as well as a competent human. This causes the weight of MS evidence to depend greatly on the technician's qualifications and proficiency with MS spectrum analysis. "

"Crucial Factors

MS analysis is highly reliable if the instrument is of sufficient resolution and the technician's interpretation of the results is competent. While some factors rarely affect MS analysis, some factors are absolutely essential for the use of reliable MS evidence. In all cases a technician must process a standard sample containing a verified composition identical to the presumed contents of the collected specimen. This standard sample must be processed under identical conditions, both before and after processing the collected specimen . Any identification based on output from the collected specimen that does not match the standard sample is inconclusive. "

from "GC/MS Combination

The GC device is generally a reliable analytical instrument. The GC instrument is effective in separating compounds into their various components. However, the GC instrument can not be used for reliable identification of specific substances. The MS instrument provides specific results but produces uncertain qualitative results. When an analyst uses the GC instrument to separate compounds before analysis with an MS instrument, a complementary relationship exists. The technician has access to both the retention times and mass spectral data. Many scientists consider GC/MS analysis as a tool for conclusive proof of identity. "


Although many consider GC/MS to be the "gold standard" in scientific analysis, GC/MS does have some limitations. Because great faith is maintained in GC/MS analysis, erroneous results are not expected and hard to dispute. However, false positives and false negatives are possible.

Some problems with GC/MS originate in improper conditions in the GC portion of the analysis. If the GC instrument does not separate the specimen's compounds completely, the MS feed is impure. This usually results in background "noise" in the mass spectrum. If the carrier gas in the GC process is not correctly deflected from entering the MS instrument, similar contamination may occur.

Also, the MS portion suffers from the inexact practice of interpreting mass spectra. An analyst must correlate computer calculations with system conditions. The typical memory bank for MS identification contains about 5000 spectra for a particular group of compounds. Even if a competent analyst could find conclusive results pointing to one substance out of 5000 substances, this does not rule out the remaining over 200,000 known existing chemicals. For the 5000-spectra memory bank, the typical computer result is limited to as many as six possible identifications.

In one instance, erroneous GC/MS results may have been responsible for a criminal defendant receiving a death sentence. John Brown killed a police officer and wounded two bar patrons in a shoot-out on June 7, 1980 in Garden Grove, California. Mr. Brown's diminished capacity defense to capital murder relied on the assertion that Mr. Brown was under the influence of narcotics at the time of the shooting. The prosecution introduced GC/MS evidence that showed Mr. Brown's blood to be free of narcotics. The California Supreme Court overturned the jury's death sentence because the prosecution never introduced evidence from a radioactive immunoassay ("RIA") test that detected phencyclidine (PCP) in Mr. Brown's blood. Obviously, an example like this demonstrates that analytical evidence, including GC/MS, should always be confirmed with another reliable technique.

A more advanced analytical method is MS/MS, a tandem series of instruments, which has the advantage of increased sensitivity. One court states that MS/MS analysis has never produced a false positive in the FBI laboratory. However, MS/MS is not widely used yet as the instrument's cost is prohibitive.


GC and MS are useful tools for chemical analysis, especially when used together. An attorney can present an effective attack or defense of GC/MS evidence with a basic knowledge of the analysis process and an insistence on documentation of important indicators that may affect GC/MS results. At the minimum, a technician must process standard samples before and after analyzing a specimen in question. In litigation an adverse party should seek hard copy output, including system conditions. Finally, no analytical technique produces results that are completely without doubt. An effective advocate should always seek corroboration of GC/MS results. "

Russ said...

syi or m or Larry?
One of you asked about noise, notice it is partially illuminated in the limitations section for GC/MS above.

Ali said...


I reread your piece and I'm confused now.

At one point, with regard to the 5a peak shifting, you claim that...

"The probability of this happening is close to zero."

However, you end by saying that...

"I assign a probability of 95% to 1, and 5% to 2."

Where 2 is the peak shifting scenario.

It can't be exactly 5% and, at the same time, "close to zero".

I think I need to see how you worked this out, m. Otherwise I'm going to have to "close to zero" confidence that your numbers are correct.


Michael said...

m 11:22

Landis couldn't challenge the science, so how could he prove that something else was in his urine and not what LNDD found without using science?

Mike Solberg said...

m said: This is why Mike has made such a big to do about Landis taking Cortisol.

You kidder, you. Ha ha. He did not take Cortisol. Dexamethesone and methylprednisolone were injected into his hip.


Larry said...

Russ, the article you quoted is pretty good, but IMHO it's unclear in a few places, and I think it's wrong in a few places.

The stuff on wikipedia is really very good. The problem with wikipedia is that the information you need to understand chromatography is spread out across a number of articles, so it's not easy to piece together. But I haven't seen anything about chromatography on wikipedia that's been proven to be wrong.

The LandisCase wiki (see links of interest) is also very good, as far as it goes.

I've been trying to learn the science of chromatography in my spare time (!) for the last 3 months. It is a difficult challenge, made more difficult by the lack of simple, comprehensive material on line that explains it all to us. Understanding the distinction between the mass spectrum data and a TIC (which I think stands for "total ion chromatogram") took me a long time. All I can do is to assure you that to the best of my knowledge, I drew a correct distinction for you earlier today between the mass spectrum data and the TIC data.

bk said...

I'd like to offer possibility #3: That something else is under the 5A peak in both the GC/MS and GC/C/IRMS. There may be a reason the MS data was deleted. Wasn't the MS data deleted off 2 different machines?

What is the probability of fumbling incompetence v.s. systematic deletion of data?

Ali said...


Regarding noise, my understanding is that you have principally two types, chemical and electrical.

The chemical noise is unwanted "stuff" which is just floating about in the machine and will have it's own C13/C12 ratio. This ratio doesn't appear to be constant and can vary at different points along the spectrum.

The electrical noise will probably be either internally generated within the IRMS equipment electronics (the detectors and their amplifiers and ADCs), or as you suggested, induced from external sources. I'd imagine that the internal electronics are well screened from external interference, so internally generated noise may dominate the electrical aspect.

Both chemical and electrical sources of noise combine to produce a composite background noise level which is digitised and sent to the computer. I wouldn't think it is possible to determine what the relevant contributions of chemical and electrical noise are to the overall background.

The C13/C12 ratio for the background will be calculated from the ratio of the background level on the m45 plot and the background level on the m44 plot (at a particular point). I don't think it matters what the sources were of that figure (chemical or electrical). From a data processing point of view, it is what it is.


m said...


I took a look at the F2 peaks.

My general claim is that peak patterns should be the same in the GCMS and GC-IRMS chromatograms. If a peak is large in the GCMS, it contains a lot of the substance, and generally the peak should also be large in the GC-IRMS because that peak will likely contain a lot of carbon atoms, unless the substance for some reason contains no carbon.

In the B sample F2, the Internal Standard peak is small in the F2, but large in the GC-IRMS. I should note that in the F2 blank urine, the IS is large in both the GCMS and the GC-IRMS, and this appears to be the case in all the other samples and blank urines also. So what might have happened in that one F2?

One explanation is that during the sample preparation a different and smaller amount of the IS was injected into the GCMS sample F2, compared to the amount injected into the GC-IRMS sample F2. This problem did not occur with the IS in the F2 blank urine.

In re-reading the USADA pre-hearing brief which describes to some extent the sample preparation steps, it states at paragraph 69 that the IS is inserted into the IRMS samples, after it has been fractionated. It's not absolutely clear, that 2 separate sample preps are undertaken, but that appears to be what happens.

In any case, I believe the F3 shows no such problems, although I didn't look at it again. So I think my general proposition that a large peak in the GCMS should be a large peak in the GC-IRMS is probably valid for these analytes.

m said...


"In the B sample F2, the Internal Standard peak is small in the F2, but large in the GC-IRMS. "

Should read as follows:

In the B sample F2, the Internal Standard peak is small in the GCMS F2, but large in the GC-IRMS.

m said...

Re: Mass spectrography

I would point you to two sources:

lecture notes from U.Col.

And a great diagram which illustrates the difference between a mass spectra and a total ion chromatogram and selected ion profile.

It's not clear to me however that the chromatograms that we have been looking at are these TICs.

"First the intensities of all the ions in each spectrum can be summed, and this sum plotted as a function of chromatographic retention time to give a total ion chromatogram (TIC) whose appearance is similar to the output of a conventional chromatographic detector. Second, as shown in the diagonal display in Figure 15, any of the spectra can be diplayed. Each peak in the TIC represents an eluding compound that can be identified by interpretation of the mass spectra recorded for the peak. Finally, as shown in the lower part of Figure 15, the intensity at a single mass-to-charge ratio over the course of a chromatographic run can be displayed to yield a selected ion current profile or mass chromatogram. This technique can be used to find components of interest in a complex mixture without having to examine each individual mass spectrum."

Mike Solberg said...

m, I wonder if you could explain something to me? I have mentioned before the report from the USADA symposium on anti-doping science in 2003. Here is the link:

On page 106, there is a summary of a discussion that the participants had about "harmonization of inter-laboratory reports" of IRMS results. In that summary it says:

The WADA ISL and Technical Document for LDPs apply, although there are specific reporting issues that should be addressed for GC-C-IRMS analysis. The amount of documentation should allow a third party expert to confirm the conclusions derived from the data. (first bullet) Due to the nature of the data, chromatograms should not be required...

I wonder, how can a key argument about peak identity be based on information that the USADA experts agree doesn't even need to be in the document package, and obviously isn't needed to "confirm the conclusions derived from the data"?

On the face of it, it seems like a pretty strong sign that the "peak matching" method can't be the basis of a strong argument.

Remember, I actually think you've proved something about peak id., although not specificity, with the argument. But this sure seems to imply that we shouldn't put too much weight on the argument.

It IS very odd that Brenna was part of that symposium, yet he argued (belatedly) at the hearing for the peak matching method. Those two things seem highly inconsistant.


m said...


That document is hell to read. I'll take a shot at it sometime.

But I wouldn't take a cryptic comment like that as a basis to throw out retention times and peak patterns (they are linked) as a method of identification even if it's applicability to IRMS, as opposed to GC-MS, isn't perfectly understood by you and me.

Mike Solberg said...

Yes, it is a pain. I can't even figure out how that pagination could come about in any copying process. And I agree that that cryptic comment shouldn't make you throw anything out too quickly. But it is surprising, and it must have been motivated by some thought that the 'grams have serious limitations.

Of course, that applies both to trying to use them to support the AAF (as you do wrt peak id) and challenge the AAF (as I do wrt specificity).

Let me know if you can make sense of what might have motivated that comment and how it might impact your argument.


blackmingo said...


The PDF should be printed out -the pages are folded in half and a staple would be place at the folded end as in a cheap booklet.

It is complicated by the fact that the copier copied every third page or so twice and did not bother to sort out the duplicates.

dan said...

Working link to the symposium document that is still hard to read and print, but at least you can get to it.


m said...


The preface notes that two approaches seemed to have developed wrt to IRMS target analytes, some use etio and andro, others use 5A and 5B.

With respect to 5A and 5B the preface notes:

"Several groups have reported a difference in delta values between the 5A and 5B, but the biochemical explanation of this phenomenon is not clear."

This may shoot a huge cannon hole in the Amory testimony claiming that 5A and 5B must always move together. I always thought Amory's testimony was anecdotal anyway.

Ali said...

Maybe there are systematic issues with resolving 5bA and 5aA, such that many groups suffer the same problems that LNDD demonstrated.

The observed degree of difference in delta values between these two metabolites may be proportional to how poor the chromatography is.

Push them too close together and the errors you get processing them trips you over the 3-delta unit threshold, resulting in a false positive.

I wonder whether labs which use these two markers have a higher positive rate than labs which don't ? It wouldn't surprise me at all.


blackmingo said...

My Favorite Quotes from the 2003 symposium:

Page 40:

The pattern where you see the 5a Adiol but not the 5b Adiol, couldn’t that be explained by someone taking a combination of DHT and T, and shouldn’t that be reported as a positive?

It could be explained that way and other ways (my emphasis). It is the “shouldn’t it be proven as positive” that is the problem. Because as soon as you report it positive you have to prove it. Not only do you have to prove it, you better well have a lot of clinical studies that demonstrate it or you will lose in CAS.”

Page 81 (end of comments / questions section following Brenna’s presentation)

I want to make two points. One, better chromatography solves a host of problems, no question about that. Two, peaks that are buried under your peak are not going to be resolved by curve fitting or any other method whatsoever.”

Ali –what do make of Figure 2 and Brenna’s estimates of uncertainty with the different algorithms?

Mike Solberg said...

Dan, that Matthews quote is my favorite. Seems like he and Dr. WMA might see more eye to eye than some others.

m wrote: This may shoot a huge cannon hole in the Amory testimony claiming that 5A and 5B must always move together. I always thought Amory's testimony was anecdotal anyway.

I'd have to know more about that difference btn the 5bA and 5aA. If I remember right Amory's testimony was based on years of his own research and clinical practice with testosterone and a review of the available literature.

Ali wrote: I wonder whether labs which use these two markers have a higher positive rate than labs which don't ? It wouldn't surprise me at all.

Yes, that would be interesting. I think UCLA uses etio and andro right, not the 'diols? I wonder why they went that route? It could be that the close eluting 5bA and 5aA make things unnecessarily difficult.


Michael said...

Even better quote from Catln:

It could be explained that way and other ways (my emphasis). It is the “shouldn’t it be proven as positive” that is the problem. Because as soon as you report it positive you have to prove it. Not only do you have to prove it, you better well have a lot of clinical studies that demonstrate it or you will lose in CAS.”

A lot of clinical studies?? What's CAS going to say about the deletion of data critical to the case.

DId USADA know that if Landis took his case to CAS, that it would be overturned? Is this a possibility?

Could this be a case where CAS will let Landis off, but of course it doesn't matter because Landis has been fighting this for two years. And now he can't race in France until 2009 regardless due to the AFLD decision.

Landis has been railroaded by the system. Landis has essentially served a two year suspension. It's just more proof that the powers that be don't give a sh** about the athletes.


I hope Landis is acquited by CAS and sues the heck out of the USADA.

Ali said...


Figure 2's quite alarming.

They take one set of data and process that data using two different commercial IRMS analysis packages

What did they find ? ... Systematic differences as big as 4 delta units. Not a random spread, but consistent repeatable error (to within 0.4 delta units). That's nice.

It's a pity Brenna never investigated this and determined the cause. That would have been interesting. Still, what's 4 delta between friends.

I think what was missing from his study was the magic touch of a LNDD employee. That would have sorted his results out (one way or another)


Ali said...


"...It could be that the close eluting 5bA and 5aA make things unnecessarily difficult."

In fact, so difficult that their blank urine tests positive. They have to intervene and keep applying "qualty control" to the results, until they didn't test positive.

It's just as well they knew which were the blank samples, that's all I can say. Otherwise Mr Blank would have been sanctioned as well and I know for a fact that he didn't dope either.


Ali said...


We keep coming accross examples of how data processing techniques can significantly effect the calculated value for a peak. In retrospect, I know for a fact that if I found myself in the position of being accused of doping, the first thing I'd do is look at the raw data and make sure that it had been processed in the best possible way for the chromatographic conditions present.

The fact that this data has been destroyed or made unavailable means that the only cast-iron means of establishing the validity of the analysis has been removed.

You'd think that would mean that both USADA and Floyd were screwed. However, somehow, the burden of proof fell on the athlete, even though the evidence had been destroyed.

Well, that's a fine position to be in, isn't it ... prove that the evidence doesn't make you guilty, whilst being denied access to the evidence.



Michael said...

Yeah Ali. I agree.

My gut feeling all along was that the USADA went through with this whole thing against Landis because they were/are playing the game.

I was hoping that USADA would stand up to WADA and let Landis off based on the test packet, but alas, Tygart didn't have the guts. Heck, he didn't have the guts - as a lawyer- to prosecute Landis himself so he hired a high priced law firm. That just about says it all.

Too bad the major media outlets didn't follow up on the Landis case like the Mitchell report. Crazy thing is that everyone is assuming Clemens is lying about Steriod use and there isn't anything close to a positive test.

Everything about doping control sucks right now and peoples reputations and careers are being flushed down the toilet. Maybe we should do some research on Mr. Pound and Mitchell to see what kind of skeletons they have in the closet...