Skip to content

Detecting Autologous Transfusions: Time-Series Metrics & Bio-Passport

November 15, 2013

A recent blog by @VeloClinic (What Does 0 Vuelta Positives Mean?) pointed out some research that bio-passport experts (some of them members of the UCI biopassport panel) have been doing to improve the certainty of their detection methods (Morkeberg et al, http://www.ncbi.nlm.nih.gov/pubmed/22119492).  The paper in question evaluated 3 algorithms for automated detection/screening of an athletes blood values to find evidence of re-transfusion of stored blood – the subjects had either 1 or 3 bags of blood re-transfused (stored either refrigerated or frozen).  There are two very important facets to what they are doing (in my opinion):

  1. Time Series – These methods are “time series” methods in that they evaluate two successive readings, taken 7 days apart, for a single athlete to detect a re-transfusion event.
  2. Unsupervised – The methods were “unsupervised”, that is, they ran without human intervention to give a “positive” or “negative” reading.

About the “unsupervised” algorithm: An unsupervised algorithm is not a replacement for a panel of experts, but rather a supplement to one.  Thus, the results of a particular metric or set of metrics should not be considered as the state of science, but rather as an interesting glimpse into the ins and outs of the analytical tools that supplement the minds of the bio-passport panel.  Also, this approach is interesting because they take the basic tools Hg and OFF, and add the factor of time, comparing successive measurements.

The results of this study showed that two of the three methods had low percentages of catching doping athletes (some with only a 10-15% chance of catching dopers), but I think that this paper shows some very encouraging signs, particularly in the area of the “intuition” of the passport panelists.  The most successful method was one that was based on the “Best Professional Judgement” of the panelists, and saw very high rates of positives (so if you’re a doper and you think that you have a 90% chance of evading detection – think again).

Hits & Misses
In the Morkeburg study, the three methods evaluated were called the “3G”, “AP” and “Absolutist” methods — the “Absolutist” method was based on the Best Professional Judgement of the scientists involved – the scientists set thresholds for changes between two successive values of OFF-score, Hg, or other markers that they thought indicated re-transfusion.   This “Absolutist” algorithm was the most sensitive method, since it had an overall detection rate of 80% based on the Hg score, and 67% on the OFF-score, meaning that there was only a 1 in 5 chance that an athlete who WAS transfusing would avoid being flagged by this test.  The other metrics (3G and AP) had a very very low level of false positives (0% for for all tests except the Hg Mass test which had a single false positive, < 1%), but similarly had low levels of success – detecting between 11% and 32% of actual transfusions.  Figures 1 & 2 show the results of the Absolutist method applied to 5 bio-passport values taken in a single month for a single athlete.  In these charts, if the lines cross, the reading is flagged for further review.

absolutist_hb

Figure 1: Plot of actual [Hb] (blue) and the [Hb] that would be indicative of possible blood re-transfusion based on the previously measured [Hb] value. The profile in this example shows a single positive screening values for [Hb] change at the end of the time series. This athlete would be referred for further examination under a “screen” and “detect” methodology.

absolutist_off

Figure  2: Plot of actual OFF-score (blue) and the OFF-score that would be indicative of possible blood re-transfusion based on the previously measured OFF-score value.  The profile in this example shows no positive screening values for OFF-score change.

Screening vs. Detecting
The different algorithms proposed had varying detection and false positive rates – with the overall detection rates ranging between 10-80% and the false-positive rates varying between 0-20%.  Given the wide range of detection and false-positives, the most likely outcome of this research would be use each of these different tests according to their strengths and weaknesses.  Some tests would be used for “screening”, that is, to flag a set of results for further review, while other tests would be used for “detecting” or accusing an individual of doping – these two activities would require different tests.  For a screening test, you would want a test that is hyper-sensitive, that is, one with a very small likelihood of missing a doping act, while accepting that a substantial number of false positives might result that would later be reviewed further and discarded.  For a “detecting” test, you would want one that was very conservative, that is, unlikely to produce a false positive, since as we have seen in the Jonathan Tiernen-Locke case, once you are elevated to a formal bio-passport review, the likelihood of a leak and taint of scandal makes a permanent mark on an athletes career in the public eye.

Screening With the Absolutist Method
Since the Absolutist Hg (AbHg) and Absolutist OFF-Score metric (Ab-OFF) detected 80% of transfusions & 67% of transfusions respectively (100% and 43% for the set of subjects who had only 1 bag of refrigerated blood re-transfused) it would be a good screening method.  The Absolutist metrics can be considered as very conservative: they added 20% and 5% of false positives to the 80% and 67% actual positives that they caught.  Figures 1 & 2 show the Absolutist method applied to 5 athlete bio-passport values taken in a single month.  These values would be flagged as positive for [Hb] (Figure 1), but not flagged as positive for OFF-Score (Figure 2).  Based on these as a screening method, this athletes profile would be elevated for further examination.

One logical outcome of this type of research would be to have the results during a grand tour screened with this Absolutist method, then reviewed further and subjected to the other methods.

Advertisements
Leave a Comment

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

Digital Geography

We blog about GIS, geodata, webtechnology and have jobs too!

The 512 Project

Learning to eat healthy on a food stamp budget.

wordsfromtherollonation

ramblings from a former distance swimmer who enjoys hours on the trail

Rees Vineyard

Wine Growing From Ground Zero

V.K.Rees

Exploring the Body & Mind Through Sport

Athletic Algorithms

Exploring the Body & Mind Through Sport

%d bloggers like this: