Friday, September 27, 2013

Analysis of Chris Horner's Biological Passport: One in a Million?

Following the post-Vuelta-win scrutiny, Chris Horner, to his credit, released his SRM data from stage 18.  Ghisalberti's estimate (437 Watts), as I argued, was inaccurate.  He'd inflated Horner's output by about 12 Watts. 

Horner has also released his biological passport data on his personal website this past week.  I've compiled that data in a spreadsheet here.

The data includes 39 tests from 2008 - 2013.  Most of the tests are out-of-competition, but there are concentrations of in-competition tests, especially during Grand Tours--particularly the 2013 Vuelta.

I'm no expert in reading blood values (there are probably less than 50 true experts in the field), but using the rules laid down by the experts, I'll take the side of the prosecution on this one.

Reasons to be suspicious

1.  Rapid increases in hematocrit.

According to an article in the British Journal of Pharmacology, a 10% increase in hematocrit "would be evidence of unapproved activity." Horner's data shows two relevant increases:
  • April 27-30, 2009: Horner's hematocrit increases 9.2% in 3 days. 
  • February 11 - March 7, 2010: Horner's hematocrit increases 10% in 25 days
The 2009 increase occurred during the Tour of the Gila, with Horner racing as part of the Mellow Johnny's 3-man team with Lance and Levi.  The increase is not quite at 10%, and the effect of altitude and the exertion of racing may also influence hematocrit.

The 2010 increase was not so sudden, but it was larger and occurred at the same time as other suspicious factors (discussed below).

According to a group of Australian researchers who conducted a large study on the effects of endurance sports on blood values of athletes, any movement upward in "values of hemoglobin and hematocrit in endurance athletes during the competitive season should be considered abnormal." (Haematologica 88/5/570, 2003)

This leads to our next marker...

2. Hemoglobin (Hg) does not decline by 10% from start to finish of stage race.

There are several instances of this in Horner's data:

  • 2013 Vuelta: Horner's Hg declines 3.9%, from 15.2 (8/22) to 14.6 (9/14).  
  • 2010 Tour de France (9th): Hg declines 1.3% in first ten days (7/1 - 7/11)
  • 2009 Tour of the Gila: Hg rises 10% over 3 days
While such increases (or unusually slight decreases) are irregular, they are not impossible.  A comprehensive study of the 2010 and 2012 GiroBio found the general trend of decline in Hg to be true for most riders, but not all.  While Horner's Hg pattern during stage races does not conform with the general pattern, it's merely slightly unusual.

That doesn't mean he's off the hook; other data are more worrisome.


3. Low % of new red blood cells (reticulocytes).


Scientists measure the ratio of new blood cells to old ones, a measurement that can detect both the infusion of blood and the use of EPO.

A low % of new red blood cells (ret) may indicate an infusion, the dumping of a bag of entirely old blood (i.e., it was withdrawn a month or more prior).  Horner's ret values twice fall below the normal levels (.5 - 1.5% of total red blood cells).
  • .39%: 8/29/2013 
  • .45%: 2/11/2010
Both of these values are abnormal, but not necessarily indicative of certain doping. As Science of Sport notes, merely having low ret means little, since genes and even environmental effects (e.g., a strenuous bike ride) can alter production of new blood cells.

Rapid changes in ret, however, suggest infusion.

4. High rate of change in new red blood cell (ret) %.

Horner's data reveals three instances or unusual ret rate of change:
  • 2/11/2010 rate of change - 31% decrease over 6 days.
  • Vuelta (8/22 - 8/29/2013) -  54% decrease in 7 days (from .85% to .39% in 7 days). 
  • 2010 Tour (7/1): 33% decrease in 14 days.
Of these three instances of ret decreasing over a short period, Horner's Vuelta data is most irregular.  In a week his body halved production new red blood cells (ret) but somehow he still had plenty of red blood cells (i.e., his new blood cell count fell in half but his total blood cell count feel only by around 5%).  The odds of this seem very slim to me, but perhaps that's because I'm not Michele Ferrari.

To maintain high ret levels following a transfusion, some athletes micro-dose EPO intravenously.  EPO pushes up ret, preventing levels from falling.  The athlete introduces both new (through EPO) and old (through transfusion) blood at a natural ratio.

Closer scrutiny of that ratio is the purpose of the next variable of interest: OFF-score.  OFF-score is the ratio of Hg to ret.  OFF-score brings together all the data we've looked at so far--HG and ret--and puts it in context. As with Hg and ret, there is a normal range for OFF-scores:  80-100.

5. High OFF-score.

OFF-scores above 100 are irregular, but OFF-scores over 110, according to a team of Italian researchers, indicates a "99.9% probability that blood doping has occurred," even weeks prior to the test.

Horner has 15 OFF-scores over 100 and 2 scores over 110:
  • 5/19/2008: 114
  • 3/7/2010: 112.3
The odds of Horner twice achieving, without manipulation, OFF-scores above 110 are, if I've done my math right*, 1 in 1,000,000.

Keep in mind that Horner's values, although suspicious, are well below 133--the Off-score level experts consider to be "irrefutable" evidence of doping.  This possibly says more about the weakness of the UCI's passport than it says about the likelihood of Horner's having doped.


When Dr. Christopher Gore put the odds of Lance Armstrong's blood values occurring naturally at "less than one in a million," those odds weren't good enough to persuade anyone at the time.  We thought Lance was special; if anyone had blood that would do the miraculous, it was Lance.

Or so we thought.

That doesn't mean cyclists, including Lance, are not one-in-a-million.  What they do and how their bodies respond to training load--that's probably rarer than one-in-a-million.  And winners of Grand Tours are even rarer than that.

There are three possibilities here:
(1) I've applied expert theory correctly, and they are right and the odds of Horner being clean are somewhere around one in a million;
(2) I've applied expert theory incorrectly and I'm wrong about the odds;
(3) I've applied expert theory correctly...and they are wrong.

I welcome corrections to my own work here, or to the work of the experts.

*Two events of 1/1000 likelihood: 1/1,000 * 1/1,000 = 1/1,000,000


Anonymous said...

"The odds of this seem very slim to me, but perhaps that's because I'm not Michele Ferrari"

Exactly. If you hang around the pro peleton enough, you'll see these kinds of values are, in fact, quite common.

Anonymous said...

You have assumed independence of two 1 in a thousand events, in order to deduce that the occurrence of two such events is one in a million. Given the commonality of his one body between the two occurrences, that is likely an extremely erroneous assumption, thereby totally negating the validity of your one in a million conclusion, even if the 1 in a thousand is correct for a given occurrence.

Also, how big are the error margins in these measurements, and in particular, for the difference in readings on different days - note that the difference in readings may be larger (or potentially smaller) than the error margin for any one reading. It seems that this should be taken into account.

That being said, I am by no means stating that Mr. Horner is as pure as the driven snow.

Please note that I am a different "anonymous" than the previous poster. Perhaps I can be referred to as Anonymous_not_one_in_a_million.

djconnel said...

I'm also no expert, but I made this plot of Lance's data from 2009, which seemed to me to be fairly damning (blog post here). I played around with Horner's data and failed to produce a simular result. Other than the "1 in a million" analysis, which as Anonymous said assumes uncorrelated deviations (which isn't the case here), you do point out some nice features from the chaos of numbers, but it's hardly conclusive.

UncleTbag said...

I don't know what was going on in February of 2010, but would not the decline in new red blood cell production be consistent with going pretty much directly from the Tour of Utah, where he had raced at high altitude for a week to the Vuelta, where he was racing at sea level or thereabouts?

Not that reassuring said...

"If you hang around the pro peleton enough, you'll see these kinds of values are, in fact, quite common."

This is the same pro peloton that for much of the last 2 decades was juiced to the eyeballs?

Robert Burgholzer said...

I have seen the range of OFF-scores listed in several places as 80-110 for normal values, in several places, one of which is here:

Robert Burgholzer said...

One other thing that I suggest note to provide some context on your analysis, would be to include an analysis of the 2010 data when Horner was deemed a "0" risk of doping by the UCI Passport program. That year saw a pre-tour high of 0.81 reticulocytes (June), Horners highest OFF-score of 112.3 (March), and his 2nd lowest reticulocyte percentage of 0.46 (February). It would be interesting to see what your impression of those numbers was. (see below) From my acdmittedly untrained view that seems like a substantial amount of variation.

Date Hemoglobin Hematocrit OFF Score % Reticulocytes
2/5/2010 15.4 45.6 104.9 0.67
2/11/2010 14.5 43.2 104.3 0.46
3/7/2010 15.8 47.5 112.3 0.58
4/5/2010 15 45.4 103.52 0.6
4/21/2010 14.7 44.5 100.9 0.59
6/16/2010 13.6 42.9 82 0.81
7/1/2010 14.6 44.6 101.9 0.54
7/11/2010 14.4 43.7 96.4 0.63
10/23/2010 15.4 44.5 99 0.84

Music Cases said...

I’m glad it was useful to me. Thanks for your work.