Thursday, March 22, 2007

Thu, Mar 22

This weekend, Floyd Landis will be in Ephrata, Lancaster County, PA, for the so far two last installments of his Tour de Innocence. The March 25 event sold out quickly, but Mike Farrington, owner of Green Mountain Cyclery and old friend of Floyd, added a second event on the night before. Details can be found at Floyd's Blog. Sources tell me that the Left Coast will send a delegation, and possibly there will be a report by Monday. No, not me; I will be taking another shot at Mt. Diablo.

At DPF, duckstrap posts a great summary about the GC/MS and GC/IRMS analysis, which seems to point to contamination of Floyd Landis' A and B sample. Since DPF tends to lose important posts once in a while (and who doesn't remember Floyd's rant regarding Greg Lemond), here's duckstrap's post:

You3, I think you may have found something important. Bottom line for those who don't feel the desire to wade through the following is that I believe there is evidence of significant chromatographic interferences in the GC-IRMS analyses that could lead to skewing of the CIR results.

First a review of analysis process. I will focus on the A analysis, but the B shows quite similar results. We have previously considered that there were 3 main analyses that the samples were subjected to: 1) Screening T/E, 2) Confirmatory T/E, and 3) IRMS. However, the IRMS analysis is actually two assay procedures. The first, in section 2.4.1, starting on p. USADA 0123 is another GC/MS analysis whose aim is to qualtitatively identify the peaks of interest. This should unequivocally show that the chromatographic separation is sufficiently complete that the subsequent IRMS results arise solely from the metabolites of interest. The second analysis is the actual GC-IRMS analysis (Section 2.4.2, starting on p USADA 0152), and is carried out on yet another instrument, with setup and GC conditions that are similar, but not identical to the analysis in 2.4.1.

[Click to continue reading]We first note that in the sample prep. three sub-samples were prepared, each containing an elution "fraction" (sorry, you3, it is not a "fragment") from the solid-liquid extraction (procedure is on p. 117). Of passing interest is that Fraction 1 should contain compounds that are more hydrophilic than those in Fraction 2, which in turn is more hydrophilic than fraction 3. As OMJ notes, this separation is good but far from perfect, as varying (usually minor, but …) amounts of substances found in one fraction can be found in the other fractions. So after the extraction procedure, we are left with these 3 fractions to be analyzed further.

1. Qualitative GC/MS analysis

The goal of the qualitative GC/MS analysis is to definitively identfy the chromatographic peaks whose carbon isotope ratios will be subsequently measured by IRMS. As we will see, similar to the confirmatory T/E test, LNDD did not do a complete job of this identification (or at least do not provide the complete dataset).

1.1 Standard mixture of metabolites (p. 130-131)

On page USADA 0130 are the results for a standard mixture of the metabolites of interest disolved in acetonitrile. This gives us the elution times for the peaks, and the abundance ratios for 3 diagnostic ions arising from each metabolite. The abundance ratio is simply the ratio of the response for the two secondary ions to the response from the most abundant ion arising from that metabolite. As shown on p. 131, the chromatographic peaks for these ions coelute exactly. What is not shown here is the entire mass spectrum for each of the chromatographic peaks. Therefore, we can see that each peak does contain the metabolite of interest, but the data shown do not verify that the peak contains only the metabolite of interest. As we have previously discussed, both things must be true for the IRMS results to be trustworthy. For this standard mixture, however, it is reasonable to expect no interferences.

Of special note is that the retention times for 5b-Androstanediol (5bA) and 5a-Androstanediol (5aA) are 15.17 min and 15.57 min, respectively.

1.2 Fraction 1 for the urine bland (UB) and FL samples (pp. 132-135)

Page 132 contains the chromatogram fraction 1 (F1) of the urine blank (UB). There are two peaks identified, one belonging to an internal standard, and the other belonging to 11ketoetiocholanone (11KE), as well as several other major peaks migrating later, but not identified or quantified. On the next page (p 133) are seven closeups of the chromatogram, one for each metabolite plus the internal standard (SI, which is added to each sample fraction). You can see that only the peaks corresponding to the SI and 11KE are clearly defined, while the others show hints of peaks but mostly noise. This is because these other 5 substances are not contained in F1, but will appear in the analysis of samples F2 and F3. Floyd's sample F1 follows, and is more or less similar to the urine blank, except that there are a number of very large peaks eluting before the SI and these are not identified. Note that the abundance scale on this chromatogram runs from 0-8,000,000 for the UB, and 0-2X10^7 for Floyd's F1 sample. Bottom line, though is that the F1 samples don't show much cause for concern, as the abundance ratios are similar those found in the earlier standard and urine blank samples.

1.3 Fraction 2 for the UB and FL samples (pp. 136-139)

OK, now look at F2. The UB shows only two major peaks besides the SI, etiocholanonolone (E) and androsterone (A), while the individual chromatograms show clear signals for these molecules, and mostly noise for the others. There are no peaks where 5bA or 5aA should be. However, let us look closely at Floyd's sample (p. 138; this is what you3 has noticed—very sharp!). The E and A peaks are quantified, and appear to elute appropriately. However, there is an enormous peak, the largest of all of them, that elutes where 5bA should be, and a smaller one where 5aA should be (note that the abundance scale runs from 0-1.25X10^7 on this page). So one significant difference between the UB and the FL sample is that, apparently, a very large portion of the 5bA and 5aA turn up in F2, rather than where they are suppose to be, in F3. But are these substance that elute at approximately the same time as 5bA and 5aA really those substances? If I look at the UB individual chromatograms on p. 137, I see mostly noise for the 5bA and 5aA traces, as expected. However in the FL chromatograms, p. 138 there appears to be a reasonably clear signal for 5bA, and a small bump for 5aA. However, although the peak corresponding to 5bA on p. 138 is larger than either the E or A peaks, the individual peak heights (p. 139) for 5bA and 5aA are much smaller than those for A and E. This would suggest that an unidentified substance is coeluting with 5bA in F2!.

I will also note that there is a second injection of Floyd's F2 sample on pp. 140-141. This appears to be a severely overloaded sample, with badly misshapen peaks. This appears to be the first sample analyzed (at 13:47 on 7/23) with the other one at 14:33 on the same day. Presumably, LNDD noticed that the column was overloaded and repeated the analysis with more dilute sample (this is not explained, however).

1.4 Fraction 3 for the UB and FL samples (pp. 142-145)

For the UB F3 sample, there are three peaks of interest: 5bA, 5aA, and 5b Pregnanediol (5bP, eluting at 19.14 min). Note that the 5bA and 5aA peaks elute at 15.17 and 15.51 min, very close to where these substances eluted in the standard mixture—nothing remarkable here. On the idividual chromatograms (p. 143), it is noteworthy that small amounts of E and A are found (especially A), suggesting that some of the A and E do not come out in the F2 eluate, but are retained in F3. Note that these are minor amounts, with the abundance scale running to 52K, compared to 360K for 5aA and 5bA.

Now let's compare the peaks for 5bA and 5aA in Floyd's F2 and F3 samples. In F2, the sample corresponding to 5bA is the largest peak in the chromatogram, with a peak height of ~1.25X10^7. The peak corresponding to 5aA is about 10-fold smaller, with peak height around 1.5X10^6. However, in the F3 sample (p. 144), we see that the 5bA peak height is about 7-fold smaller (peak height ~1.7X10^6) than in the F2 sample, although on the individual chromatograms, the abundance is similar between F2 and F3. Again, this clearly points to something extra in the F2 peak corresponding to 5bA. If there is a contaminating substance in F2 interfering with 5bA, how can it be known that there is nothing in the F3 sample without the mass spectra?


2. GC-IRMS Analysis

This is the portion of the analysis where the actual CIR's are measured, and its description and results start on p. 0153. Broadly, the sample is injected onto a GC with the GC conditions close to those used for the qualitative GC/MS analysis described above. The GC eluate goes into the IRMS instrument where the sample is burned, and the isotope ratio of resulting CO2 gas measured. Note that two substances eluting at the same time from the GC will have their IRMS results mixed, thus it is important that the GC peak be pure—hence the issues with the F2 5bA peak above. One of the first things to notice is the disconnect between the elution times observed in the GC/MS analysis, described above, and the elution times noted here. In the IRMS analysis, the chromatograms start after a delay of 760 sec (12.7 min), with the SI eluting in the neighborhood of 870 sec (14.5 min). For comparison, the SI elutes at 10.7 min in the prior GC-MS analysis, with the elution times differing for the metabolites as well (first number is from GC/MS, second from IMRS): E (14.35 vs. 20.5 in F2), A (14.6 vs. 20.95 in F2), 5bA (15.2 vs. 21.7 in F3), 5aA (15.6 vs. 22.25 in F3), 11KE (17.1 vs. 24.6 min in F1), and 5bP (19.2 vs. 27.5 in F3).

For each sample in the LNDD presentation of the IMRS analysis, a reference chromatogram from the GC/MS analysis is given first presumably to show relative peak order ,since the elution times don't match. This page is followed by a table of CIR results (containing the elution time of the quantified peaks, peak height and dC13Pk or the CIR value, among other numbers). Finally, the IRMS chromatogram for the sample (see for example pp. 0156-0158 for BU F1 sample) is shown, with each quantified peak being labeled with its CIR. It is also worth noting that the integration limits for each peak appear to be marked, with a typical large peak covering approximately 30 s. Thus any peaks eluting within approximately 30 s are likely to have some overlap. Turning to the F2 analysis, for UB we note only three peaks quantified, the SI, E and A, while in Floyd's sample, there are four peaks quantified, one of them corresponding to the F2 contaminant. Of note is that its CIR is -32.3 (again you3 observed this), very close to the value observed for 5aA. Further, the minor peaks that are clearly visible in the reference chromatogram are nowhere to be found or greatly diminished in the IRMS trace. Clearly the chromatographic conditions are not that similar—hence the questionable quality and utility of the qualitative GC/MS above.

In summary, there is clear evidence of a major contaminating peak in Floyd's sample that would coelute with 5bA, and this contaminant has a strongly negative CIR, comparable to that observed in Floyd's single positive metabolite, 5aA. We see evidence that these substances are not confined to just one elution fraction, and that the chromatography conditions vary considerably from the qualitative GC/MS where the metabolite identification should occur. LNDD do not provide definitive evidence to rule out the coelution issue in the form of complete mass spectra, which again, they did collect. Therefore, it is critical that the FL team get their hands on the computer files for all analyses. The complete mass spectra for the metabolites of interest are key.

No comments: