The suspicion score: UCI rankings leaked, published and debated
I'm pretty sure that by now, most of you have read the
following news, in which L'Equipe has obtained a confidential document from the UCI in which all the riders in the 2010 Tour de France have been scored on a scale from 0 to 10 based on their biological passport data
(I've copied the full list from this article and pasted it at the end of this post)
On the scale, 0 means no suspicion of doping - a rider with an extremely clean passport record. Scores between 6 and 10 apparently imply a high degree of circumstantial evidence because of large fluctuations in the passport data.
Not surprisingly, the reaction from the cycling world has been angry, with the UCI condemning the leak and the riders reacting angrily to both the leak and the fact that they're being rated on some kind of suspicion scale.
But let's cut through the expected emotive reactions and look at some of the issues behind the story.
The first point is that the leak and the fact that riders are being rated on a suspicion scale are
two unrelated events. That is, the fact that there is a leak should be condemned, mostly because it's a symptom of an organization that doesn't have it's house in order. And a leak has some far-reaching implications, in that it undermines the trust that biological passport system relies on. So that is certainly a concern for those involved in the fight against doping.
But the fact that riders are being ranked on this scale is a separate issue. It may even be commended, depending on how it is done, but the leak has created confusion and speculation which shouldn't be confused with the actual concept of rating riders (a practice that certainly can be debated).
The rating concept is an issue that has people asking some very pertinent questions: "How is the score calculated? What are the UCI doing with the scores? How can a rider be suspicious and not simply positive or negative? Are they targeting testing based on the ratings? Is this fair?".
For the answers to these questions, let's go back to go forward.
The biological passport concept
We've said a fair amount on the biological passport in 2011, most of it very positive. If you missed those pieces, I'd encourage you to read them here:
- The legal, scientific and performance viewpoint of the biological passport
- Evidence that the biological passport IS proving effective in the fight against doping
Without wishing to rehash those (lengthy) posts, a summary is important because it sets the context for why a rating system is part of the control and policing of the passport system. First, the passport is a very stringently designed scientific tool. It has the backing of substantial research, some of which we discussed in that first post on the legal issues. It's not perfect, and of course more research is needed, but it's not a stab in the dark at catching dopers.
Now you may recall that in order for the biological passport to be legally acceptable, it
has to have a very low risk of producing "false positive" results. And false positives do happen, because you are looking for abnormal variations in a biological system that has natural variation to begin with.
And so what has been done is to set a limit of probability that allows 99.9% confidence that a a value (for reticulocytes, hemoglobin or off-score) lies outside the 'normal' range. All measurements are analysed using software to answer the question "what is the probability of finding these values in an undoped sample?" Only when a value lies outside this 99.9% confidence limit does it constitute a 'strike'.
That's the first line of defense against unfairly pursuing doping cases. The second is that a rider is not sanctioned on a once-off strike. A case is only opened if several different variables are beyond these boundaries on more than one occasion.
If this happens, then a third layer is that a team of experts evaluate and analyze the values. If they feel that the profile is typical for a certain doping intervention, the athlete is contacted and questioned about potential reasons for his values. His justifications are again evaluated by the experts. Only if they are still convinced that the profile is typical of doping and is not caused by the explanations put forward by the athlete (as has happened for Pellizotti and Valjevic), do they suggest the opening of a procedure against the athlete.
Legal clout, but expanding the impact of the passport through ranking the "near strikes"
So quite clearly, by the time a case is opened, there is already high certainty that doping is taking place - that's why
three out of three cases that have gone to the CAS have been won by cycling. This is a fantastic boost for the passport, because the stringency and strength of the system has been affirmed at the highest level.
This strength and "legal clout" comes from the high probability limit of 99.9%, which means that false positives will happen only once in a 1000 samples. This is essential, because it must protect riders against false accusation. The "problem" is that it also
sets the bar so high that a lot of sample that are doped will perhaps fall into a 99% confidence limit, or even a 95% confidence limit, and thus not counted as strikes. The legal requirement is a "conservative system", which doesn't necessarily ensure maximum impact on doping.
In other words, at say 95%, a sample would be suspicious and indicative of doping, but not legally enforceable, because you could get a false positive once every 20 measurements. But that cyclist may still be doping and if they
repeatedly produce these kinds of measurements, then these "near-strikes" also need to be accounted for. And this, I suspect, is how the rating system works.
A justified process of control and one way of implementing the passport
Now, this is speculative, because we simply don't know that this is the basis for those scores. However, for the sake of illustration, this is how I would use a rating system:
Consider a rider whose measurements would produce a 'strike' at a 99% confidence level, but not 99.9%. The first such occurrence may be worth a score of 1. If that same rider repeatedly 'strikes' at this lower probability level, he accumulates more points. These points could be scaled afterwards by some correction factor to produce the 11-point scale (0 to 10) that was used, and there's the value you see in the lists below.
Our rider never exceeds a 99.9% confidence level, and so quite rightly, a case can't be opened, but he's been identified as a suspicious rider based on a long sequence of unusual values that lie on the border of being strikes. The more often he is flagged with a near strike, the higher his score, which is reasonable. And I believe it is entirely justified to watch this rider more closely. Why? Because we know from testimony that riders are able to manipulate their blood values through masking with EPO and by micro-dosing, and so it's quite conceivable that the precision of doping allows riders to stay beneath a 99.9% "radar" but not the 99%.
"Too normal" - sometimes variation is good
The other interesting possibility, mentioned in the L'Equipe report, is that some riders are suspicious precisely because they are "too normal". That is, variation in blood markers is supposed to happen, and when it does not, then that too is a flag, and so is part of the designation of what constitutes suspicious. I'm not sure how this is incorporated into a score, but it's hopefully understandable that if a rider presents with "perfect" values time after time, that's as much a problem as wild fluctuations outside the limits.
The fact that he is then rated on a scale, receiving a value of say 4 or 5, is simply part of the management of such a system. In my opinion, this is entirely justified and acceptable. It is part of the effective implementation of a passport system, which would lose much of its "bite" if there was not some avenue to monitor and target testing of riders who did not hit the 99.9% level.
The fact that these
scores are now public knowledge, without necessarily explaining where the number comes from, that's the problem, for the following reasons:
- The possibility of subjective scoring and generalization
Honestly, in the immediate aftermath of seeing that list, did you look at some cyclists and say "Ha, I knew it! Doper!", or did you look at some names and scores and say "No way is Rider X more suspicious than Rider Y!"? Because if you did, then you were, at least partly, comparing riders based on subjective perceptions of guilt and innocence!
The problem is that it's quite possible that whoever developed those rankings did so objectively, but with some subjective moderation, and that would not be a good thing. If the rating is developed solely on the basis of biological passport measurements (as we are led to believe by the L'Equipe article), then I'd be satisfied. But if it is subjectively moderated, then that may be an issue.
An extension of this is that once public, then the scores can further re-inforce stereotypes about certain teams or riders. For example, L'Equipe took the step of using the UCI scores to calculate which teams and nationalities were most suspicious. In what some will describe as 'vindication', it turns out that Astana and Radioshack are the most suspicious teams, and Russia, Ukraine and Kazhakstahn are most suspicious nations. I'm reminded that sometimes, stereotypes exist because they're true - sometimes, reputations are earned and deserved. But one can sympathize with the rider from these teams/countries who is NOT a doper, but is now implicated by association and generalization, which would be unfair.
For the public, the subjective perception will always feature, but what a leak does is allow confusion, and confusion allows more subjectivity and generalization.
- Uncertainty over the relative rankings and scaling
How is a rating of 8 different from a rating of 4? How close to a doping sanction is a guy who scores 9? Is a score of 3 supposed to be interpreted as possible doping, likely innocence or neither? Does the score actually have any doping relevance? For one thing, it doesn't mean that the rider who scores 8 is doping twice as much as the rider scoring 4. It doesn't mean he is twice as likely to be doping either. The truth could well be the other way around, but it's just that the one is smarter at avoiding larger changes in bio-passport data. But the problem is that by the time you get to 9 or 10, then that rider is so strongly suspected that a lot of people would be wondering why he isn't just sanctioned. Or why the rider who scores 5 (Contador) is caught whereas riders who score 9 or 10 (Menchov, Popovych) are not caught? Hopefully, in my short summary above, and in this previous post, I've explained why bringing a legal case has a more stringent requirement than being suspected.
I've also given my illustration of how the bio-passport MIGHT be used to create these scores, by combining 'near-strikes' according to an objective criteria with the number of times they occur. The cyclist scoring 8 or 9 might have produced 9 suspicious values at a probability limit of 99%, but not one at 99.9%. Of course, this is speculative, which is why a leak undermines the credibility of the process, because it invites speculation.
- The witch-hunt phenomenon
The third point, very relevant, is that this kind of scoring (independent of it being leaked) will lead to a witch-hunt, where some riders will be targeted for testing. I actually don't have a problem with this, provided that the classification of riders is done objectively using solid science (see point 1). I will concede that some cyclists may be targeted unfairly, perhaps based on their reputation or allegations. But honestly, I don't even have a problem with this, as long as it's not prejudicial to the rider's participation (that is, cases are opened unfairly).
But that doesn't seem to be the case - every case opened so far has been won. The expertise behind the passport is world class (even if the management and administration by the UCI is not - see the leak as Exhibit A), and so I'm confident that targeted testing is not a witch-hunt, but a more cost-effective and intelligent way to weed dopers out. It's much like the sniper approach compared to the machine gun approach - if finances are limited, then rather spend the money where it may be most effective, and that is part of the value of the passport.
Incidentally, targeted testing has helped catch dope cheats before - Ricco was suspected for a long time and targeted for testing before eventually being caught, and so there is definitely merit. I guess the key question here is what do you actually do with a rider who is targeted? By definition, the fact that he is suspicious and not positive means that you'll never be able to bring a case against him unless he slips up. And perhaps more frequent testing for the higher rated cyclists will improve the chances of catching these slip-ups. That's one way of focusing testing. This kind of focus saves money, and as long as the process is objective and fair, and not played out unnecessarily in public, then I think they're positive. But I accept that some may disagree.
The other thing that is really interesting to consider is that some cyclists may be better than others as disguising doping. Now, because of this leak, cheaters who are better at manipulating their values look squeaky clean, and have this rating as their defense, when in fact, all they are is more effective at doping! That's part of the reason why linking names to scores may not be ideal, and certainly why a leak doesn't help because we lack the context to fully understand how the score is derived.
As an aside, it is interesting to note that last year, during the Tour, a lot of the riders who were rated 8 or higher were NOT tested (as raised by an independent WADA Report) - this point is raised by Festinagirl in the comments section below, and it again opens the UCI to alleged favouritism as a result of lax testing. If the point of this rating system is to target testing, then testing had better reflect it. If it doesn't, well, what's the point of the scores? Too much unanswered.
Fascinating insight into the process, but have we moved backwards?
I'd be lying if I said the scores were not fascinating to see. I've often argued for full transparency in anti-doping efforts, that the bio-passport values should be made available (as some promised to do but never did). Seeing the rankings of riders is about as transparent as it gets, but the confusion and uncertainty over exactly what the scores mean and how they're used means one step forward for transparency, one step backward for improved knowledge! And perhaps
most worryingly of all, two steps back for trust and credibility of the whole anti-doping effort.
So perhaps the scores might have been made available without the associated names. Revealing that riders are rated, and telling us that say 50% of the riders have a score of 4 or higher would send the same message to the professional peloton that they are being watched as publishing names.
As mentioned at the outset, linking the names with the scores only undermines credibility of the system, and the system is so reliant on the trust of anonymous analysis of blood values. It also provides a convenient public defense to some riders who may not necessarily be innocent, and possibly disadvantages riders who may be innocent but now have high index scores.
The reality is that putting names to the biological passport should be done only in the very final step of the process - the leak has given us an insight into the process, but perhaps the same effect might have been achieved without the risk to the credibility and trust within the between rider and tester.
Intriguing insights into the UCI, the leak, and conflicts of interest
But the UCI's reaction, combined with other UCI anti-doping actions only provides more fodder for critical of the UCI. For one thing, the leak undermines the credibility of an organization that can't seem to sort out its own internal processes. One of the key requirements of a successful anti-doping programme is trust - every party must trust the others to be fair and accurate in their actions and a leak like this undermines that trust. The UCI already has this clandestine reputation, and this doesn't help.
It also doesn't help when the UCI can't reach agreements over testing at races and ends up marginalizing independent organizations, as happened for the upcoming Tour of California. Rather than having independent and comprehensive testing from USADA,
Bonnie Ford reports that the UCI will now handle the testing, which further increases the perception that the UCI are hiding something, giving preferential treatment to certain teams and riders, and generally having an enormous conflict of interest.
When the UCI deplore the leak and publication of their suspicion scores, and follow this up a day later by marginalizing independent testing authorities as a result of unwillingness to make available test results or allow targeted testing, it only reinforces the perception of a clandestine approach. Perhaps full transparency is the way to go - but not for the first time, it's the media who have forced it onto the UCI, who continue to send mixed signals about anti-doping efforts.
Difficult times for cycling. And all the while, a bike race goes ahead in Italy, but it's anti-doping on the back pages.
Ross
And finally, here is the complete list:
0 Mario Aerts, Yukiya Arashiro, Stephane Augé, Michael Barry, Francesco Bellotti, Jose Alberto Benitez, Edvald Boasson Hagen, Lars Boom, Maxime Bouet, Pavel Brutt, Fabian Cancellara, Manuel Cardoso, Dries Devenyns, Samuel Dumoulin, Julien El Farès, Simon Gerrans, Anthony Geslin, Bert Grabsch, Christopher Horner, Robert Hunter, Kristjan Koren, Burt Lancaster, David Le Lay, Christophe Le Mével, Adriano Malori, Koes Moerenhout, Amael Moinard, Lloyd Mondory, Damien Monier, Juan Jose Oroz, Remi Pauriol, Mathieu Perget, Gregory Rast, Mark Renshaw, Nicolas Roche, Jurgen Roelandts, Pierre Rolland, Anthony Roux, Jeremy Roy, Mathieu Sprick, Rein Taaramae, Sebastien Turgot, Niki Terpstra, Brian Vandborg, Kristof Vandewalle, Ivan Velasco, Thomas Voeckler, Fabian Wegmann, David Zabriskie
1 Marcus Burghardt, Sandy Casar, Anthony Charteau, Sylvain Chavanel, Julian Dean, Mickael Delage, Martin Elmiger, Johannes Fröhlinger, Jakob Fuglsang, Robert Gesink, Xavier Florencio, Adam Hansen, Ryder Hesjedal, George Hincapie, Andreas Klier, Roger Kluge, Alexander Kuchinsky, Daniel Lloyd, Mirco Lorenzetto, Martijn Maaskant, Aitor Pérez, Alan Pérez, Jerome Pineau, Ruben Plaza, Alexandre Pliuschin, MaartenTjallingii, Rafael Valls, Maarten Wynants
2 Eros Capecchi, Mark Cavendish, Stephen Cummings, Remy Di Gregorio, Arkaitz Duran, Mathias Frank, Oscar Freire, John Gadret, Francesco Gavazzi, Volodymir Gustov, Thor Hushovd, Christophe Kern, Thomas Löfkvist, Sebastien Minard, Daniel Navarro, Grischa Niermann, Stuart O'Grady, Rubén Pérez, Christophe Riblon, Thomas Rohregger, L. L. Sanchez, Carlos Sastre, Fränk Schleck, Simon Spilak, Bram Tankink, Stijn Vandenbergh, Benoit Vaugrenard, Jens Voigt, Eduard Vorganov
3 Ivan Basso, Grega Bole, Brent Bookwalter, Dimitri Champion, Gerald Ciolek, Rui Costa, Damiano Cunego, Mauro Da Dalto, Francis De Greef, Markus Eibegger, Imanol Erviti, Tyler Farrar, Fabio Felline, Juan Antonio Flecha, Maxim Iglinskiy, Vasil Kiryienka, Roman Kreuziger, Matthieu Ladagnous, Robbie McEwen, Maxime Monfort, Sergio Paulinho, Joaquin Rodriguez, Andy Schleck, Chris Anker Sörensen, Sylvester Szmyd, Paolo Tiralongo, Amets Txurruka, Johan Van Summeren, Gorka Verdugo, Charles Wegelius
4 Lance Armstrong, Janez Brajkovic, Bernhard Eisel, Cadel Evans, Pierrick Fédrigo, Juan Manuel Garate, Andriy Grivko, Jesus Hernandez, Ignatas Konovalovas, Sebastian Lang, Levi Leipheimer, David Millar, Daniel Moreno, Serge Pauwels, Manuel Quinziato, Luke Roberts, Samuel Sanchez, Christian Vande Velde, Nicolas Vogondy
5 Alessandro Ballan, Matti Breschel, Alberto Contador, Cyril Gautier, Inaki Isasi, Sergei Ivanov, Vladimir Karpets, Alexandr Kolobnev, Karsten Kroon, Steve Morabito, Benjamin Noval, Jose Rojas, Nicki Sörensen, Alexander Vinokourov, Bradley Wiggins
6 Linus Gerdemann, Christian Knees, Egoi Martínez, Alessandro Petacchi, Francesco Reda, Mauro Santambrogio, Geraint Thomas
7 Jeremy Hunt, Andreas Klöden, Tony Martin, Christophe Moreau, Michael Rogers, Wesley Sulzberger
8 David De la Fuente, Ivan Gutiérrez, Danilo Hondo, Matthew Lloyd, Iban Mayoz, Dmitriy Muravyev, Rinaldo Nocentini, Daniel Oss, Kevin Seeldraeyers, Kanstantsin Siutsou, Jurgen Van Den Broeck
9 Denis Menchov
10 Carlos Barredo,Yaroslav Popovych
And here, courtesy L'Equipe, are the rankings of teams and nations:
The ranking of the least suspicious teams:
1.Cofidis 4
2. Bbox Buoygues Telecom 14
3. FDJ 15
4. AG2R-La Mondiale 16
5. Garmin-Transitions 17
6. Cervelo 20
Footon-Servetto 20
8. Rabobank 21
9. Liquigas 22
Team Sky 22
11. Milram 23
Saxo Bank 23
13. Euskaltel-Euskadi 24
14. Katusha 26
15. Lampre 28
16. Quick Step 30
17. Omega Pharma-Lotto 31
18. HTC-Columbia 32
BMC 32
Caisse d’Epargne 32
21. Astana 39
22. RadioShack 40
The ranking of the least suspicious nations:
1. France 1.23 (based on the average of 35 riders)
2. Netherlands 1.25 (8)
3. Switzerland 1.60 (5)
4. Portugal 2.0 (3)
5. Slovenia 2.25 (4)
6. USA 2.37 (8)
7. Belgium 2.69 (13)
8. Denmark 2.80 (5)
9. Austria 3.0 (3)
10. Germany 3.27 (15)
Australia 3.27 (11)
12. Spain 3.27 (32)
Great Britain 3.27 (8)
14. Italy 3.70 (17)
15. Belarus 4.0 (3)
16. Russia 4.33 (6)
17. Kazakhstan 5.33 (3)
Ukraine 5.33 (3)
Your thoughts welcome!