Road Bike, Cycling Forums banner
1 - 20 of 34 Posts

·
Bluebiker
Joined
·
88 Posts
Discussion Starter · #1 ·
Just read this

http://www.velonews.com/article/81414/scientist-anti-doping-tests-don-t-pass-statistical-muster

regarding a new study published in Nature. Talks about how anti-doping tests being based on faulty science and statistical methods. I think tests have a long way to go for 100% certainty one way or another. The Ricco tests at the TdF seem to indicate that, only 2 of his 13 tests showed positive. I'd think it can work the other way as well (false positives).

Anyway, just starting a thread.
 

·
Registered
Joined
·
11,561 Posts
blbike said:
Just read this

http://www.velonews.com/article/81414/scientist-anti-doping-tests-don-t-pass-statistical-muster

regarding a new study published in Nature. Talks about how anti-doping tests being based on faulty science and statistical methods. I think tests have a long way to go for 100% certainty one way or another. The Ricco tests at the TdF seem to indicate that, only 2 of his 13 tests showed positive. I'd think it can work the other way as well (false positives).

Anyway, just starting a thread.
It is unrealistic to expect tests to be 100% certain either way. The only real question is have there been any false positives as it is patently obvious there have been literally 1000s of false negatives.

The only false positive we know of is the triathlete Beke (sp?) for EPO and one has to wonder about that since the scientist who got him off the hook was prepared to go to bat for Jeanson. Who of course later admitted to her EPO use.

The author seems to think a doping positive just falls on a rider out of the blue, when there is often any number of other indicators a rider is doping diminishing the probability the test is a false positive. IOW, the dope test isn't a stand alone indicator of doping, but it is pretty much the only thing the ADAs can hang their case on.

The two biggest criers of "foul", Landis and Hamilton were both almost certainly not victims of false positives.
 

·
Registered
Joined
·
3,375 Posts
Dwayne Barry said:
The two biggest criers of "foul", Landis and Hamilton were both almost certainly not victims of false positives.
There is no objective evidence to support that statement and this pattern of targeted testing only reinforces that fallacy. If the only riders tested are those who have been deemed "suspicious" then of course all false positives will appear to be true and support the underlying prejudice.
 

·
Reality Star
Joined
·
930 Posts
I read the article, too. Very interesting points that are well stated and valid. Currently, the accused rider is guilty until proven innocent and given the current system, proving innocence is nearly impossible.

OTOH, recent spectacular displays of climbing prowess have largely been discredited by high profile busts of Landis, Vinokourov, Sella, Ricco, Piepoli, et al. No doubt there is a lot of doping being done under the radar in pro cycling today. The recent move by the ASO to "take over" the grand tours does not do anything to further the pursuit of a more fair and transparent doping control system. At some point the pro cycling community and the scientific community need to meet and develop a system that also protects the "innocent".
 

·
Registered
Joined
·
11,561 Posts
asgelle said:
There is no objective evidence to support that statement and this pattern of targeted testing only reinforces that fallacy. If the only riders tested are those who have been deemed "suspicious" then of course all false positives will appear to be true and support the underlying prejudice.
The sheer probability in the Landis case suggests it could not be false positives, he had 7 positive tests from 7 different samples. You would have to assume a very high probability of a false postive from any one test to get any sort of even remote chance that these 7 tests weren't detecting testosterone doping.

The biggest problem I have with believing false positives are any sort of problem is that every other indicator is that there was a culture of nearly ubiquitous doping, and the tests were largely ineffective at detecting it.

In addition to the numerous specfic reasons to believe that both of these guys were involved in doping during their careers.

Or another way to look at it is, whatever concerns the researcher raises, at least in this case, it is not like the sample population that is being tested has a low probability of doping use and therefore we should be really concerned about false positives. Just the opposite is true, every indicator is that this a population very likely to be engaged in doping.
 

·
Registered
Joined
·
3,375 Posts
Dwayne Barry said:
The biggest problem I have with believing false positives are any sort of problem is that every other indicator is that there was a culture of nearly ubiquitous doping, and the tests were largely ineffective at detecting it.
There is no correlation required between false positives and false negatives. A test can be high in one, the other or both. To argue a low probability of false positives based solely on a high rate of false negatives has no scientific basis.
 

·
Registered
Joined
·
11,561 Posts
Don Duende said:
At some point the pro cycling community and the scientific community need to meet and develop a system that also protects the "innocent".
Don't you think a big problem here is that the pro cycling community doesn't even believe that there are really "victims"?

Especially in light of the fact that there does seem to be a perceptual shift amongst many teams and riders towards anti-doping, are they really going to want to weaken a system that finally seems somewhat capable of detecting and detering doping?
 

·
Banned
Joined
·
608 Posts
Don Duende said:
The recent move by the ASO to "take over" the grand tours does not do anything to further the pursuit of a more fair and transparent doping control system.
Yes, it does. The ASO put the AFLD in charge of anti-doping this year, and the AFLD is independent of cycling. They have no vested interest in pretending that the sport is clean like the UCI does. The AFLD has also been aggressive and smart about how they went about testing. Compare that with the UCI, which refused to allow the Giro to test riders at night. The AFLD has also done its testing and given out the results in a timely manner. Results from the UCI sometimes spend months in some sort of mystery limbo that has never been explained.

From the ASO's statements during this year's Tour, it is clear that they think independent testing is the way forward. That is a good thing. The UCI, which paid money for a bogus investigation to cast doubt on Armstrong's use of EPO, cannot be trusted.
 

·
Registered
Joined
·
11,561 Posts
asgelle said:
There is no correlation required between false positives and false negatives. A test can be high in one, the other or both. To argue a low probability of false positives based solely on a high rate of false negatives has no scientific basis.
While that may be true. My understanding is that setting the criteria for a test affects it's sensitivity and specificity. In the face of ubiquitous doping yet few positives that suggests to me that the criteria were selected for low sensitvity and high specificity as one would expect for something like a dope test. But perhaps I have this wrong?

For example, the T/E ratio for postive use to be much higher than the 4/1 simple logic would suggest that it previously had a much higher rate of false negatives and lower rate of false positives given the natural variation in T/E, no?
 

·
Registered
Joined
·
3,375 Posts
Dwayne Barry said:
For example, the T/E ratio for postive use to be much higher than the 4/1 simple logic would suggest that it previously had a much higher rate of false negatives and lower rate of false positives given the natural variation in T/E, no?
That would be a case where false positives and negatives would be related. However one example is not a proof. There could be a random error in the test that causes T:E ratio to read over 20:1. No fiddling with the threshold value around 4,6 or 10 would affect the rate of false positives from this type of error.
 

·
Back from the dead
Joined
·
20,800 Posts
Dwayne Barry said:
are they really going to want to weaken a system that finally seems somewhat capable of detecting and detering doping?
Yes, there will be a backlash, if and when the demands of dope testing finally intrude too much on rider's private lives. I'm surprised with what they put up with already, with all the reporting they have to do about their whereabouts, three months in advance, and having to be available at all times for testing. What about the recent story of Kevin Van Impe, who was forced to submit to a test as he was preparing the funeral of his dead son? That's crossing the line. Doesn't Lance tell the story about how he was going to the hospital with his pregnant wife when the testers showed up, and wouldn't let him go until he peed in a cup? That crosses the line. With teams now doing their own additional dope testing, I'm guessing that most riders have reached their limit. The system may rack up a few successes, but it does so by assuming that all riders are guilty, and the burden is on the riders to regularly and repeatedly prove their innocence.
 

·
Registered
Joined
·
5,671 Posts
Why do people keep insisting Landis had multiple positives during the 2006 Tour? As far as I can tell, from both looking at the Malibu transcripts, WIKI defense etc., Landis was tested 8 times and only the sample after stage 17 was positive. I've read Berry's commentary in Nature (not a study, but a commentary, with some statistical analysis). Based on this, the possibility of a false positive is not as small as posts here suggest:

Quoting Berry's commentary: "Therefore, Landis's false-positive rate for the race as a whole would be about 34%. Even a very high specificity of 99% would mean a false-positive rate of about 8%."

Berry's commentary had a number of 'between the lines' charges that weren't picked up on. 1) the need to have the tests done blind. In addition to describing 'prosecutor's fallacy' about prior probabilities, he should have discussed the more obvious problem of 'confirmation bias.' Namely, Frelat knew she was testing Landis' B sample, even described her job as "confirming the A sample" and ran complex tests (which Berry does note) not only without the expertise (Frelat's training in these tests was dubious) but re-ran critical steps because of various problems/errors - which involved over-writing files - until the 'right' result was produced. This is a classic scenario for confirmation bias - which doesn't have to be a conscious manipulation - to influence the results.

The bottom-line is that the LNDD's practices were shoddy, uninformative in the deeper statistical sense Berry has in mind and violated basic scientific principles (blindness, independence in replication).
 
G

·
asgelle said:
That would be a case where false positives and negatives would be related. However one example is not a proof. There could be a random error in the test that causes T:E ratio to read over 20:1. No fiddling with the threshold value around 4,6 or 10 would affect the rate of false positives from this type of error.
From my experience with a frequently run tests is that large random errors aren't really random at all - they were typically operator error (ie me!). So long as lab techs are ensuring that their equipment is properly calibrated and that any reagents are kept in a proper environment and kept free of contamination, large random variations are typically very minimal. If they are not, the everything in the testing protocol needs to be examined to ensure that there isn't a systemic flaw.

As I understand it, all of WADA's tests are biased towards false negatives - in order to avoid the obvious immediate detrimental effects of a postive (true or false). Typically, false negatives and false postives are directly related since it comes down to where you draw the line in the output curve .

Statiscally, there is always a problem with running a test only once or twice on a given sample (assuming the sample is divisible in a non-destructive fashion). I've seen this is production QA runs. In R&D, where possible, I would run at least 20 tests on a sample where possible (and I had to fight for that since time is money).

There are no perfect systems and often we are forced to settle for far less than perfect. And it does sound like that's where we are on PED testing - which is why tests must be biased to false negatives.
 
G

·
asgelle said:
It's hard to accept that after looking at Fig. 1B in the Berry paper.
I just printed out the article now that it's free. I'll go over it and comment. Wouldn't be the first time I've taken a trip down the primrose path over WADA and the UCI's comments vs reality - someday I'll learn.
 

·
Banned
Joined
·
2,241 Posts
Berry has a history of defending dopers and admits that he got his information from the Landis defense team. That he bases his theory on a single test, yet ignores the mountains of additional evidence, including addition positive tests. This completely negates the "Prosecutors fallacy"

It appears the Mr. Berry is auditioning for some more of that lucrative expert witness work, I would not be suprised if he is calling Suh and Jacobs right now offering his services for the post Olympic work that will sure come their way.
 

·
Banned
Joined
·
2,241 Posts
asgelle said:
It's hard to accept that after looking at Fig. 1B in the Berry paper.
It is if you take into account the 5 other postives samples that Berry conveniently ignored. You also do not know the threshold that the LNDD used as positive, it is possible that many of those negatives were actually postives that did not pass an artificially high threshold
 

·
Banned
Joined
·
2,241 Posts
stevesbike said:
Why do people keep insisting Landis had multiple positives during the 2006 Tour? ."
Because he did. They retested all of his samples for synthetic testosterone using the IRMS test.

Stage 11, 15, 17, 19, 20 showed the presence of synthetic testosterone. That Berry ignores this and bases his theory on one test shows he is either clueless or did not intend to write an unbiased article.
 

·
Back from the dead
Joined
·
20,800 Posts
bigpinkt said:
Berry has a history of defending dopers and admits that he got his information from the Landis defense team. That he bases his theory on a single test, yet ignores the mountains of additional evidence, including addition positive tests. This completely negates the "Prosecutors fallacy"

It appears the Mr. Berry is auditioning for some more of that lucrative expert witness work, I would not be suprised if he is calling Suh and Jacobs right now offering his services for the post Olympic work that will sure come their way.
If you can't attack the message, attack the messenger.
 
1 - 20 of 34 Posts
This is an older thread, you may not receive a response, and could be reviving an old thread. Please consider creating a new thread.
Top