Some Welcome Perspective on Sports Doping

Here is a very interesting and, to my mind, useful letter to the editor of Sports Illustrated, written by Brandon Gaut of Irvine, Calif., whose home page is here:

As a scientist and a sports fan, I believe the current doping scandals compromise science as much as sports. The tests are performed by entities motivated by and funded to achieve the goal of detecting cheaters; their objectivity is suspect. Also, it is a scientific fact that there will be positive tests even when there are no cheaters. From my perspective, the puzzle is not the occasional prositive test, but why there aren’t a great many more. The system is broken, and I fear it is not always due to cheating athletes.”

This is yet another example of why I believe that letters to the editor are one of the most worthwhile parts of any publication.


Mack

It's good to see someone else who thinks that readers' letters are the best part of many magazines and papers. I've felt the same way for years.

I couldn't care less about doping 'scandals' though, or sports in general.

akatsuki

I guess it would just depend on how their incentives are structured. I personally believe (admittedly unfounded by any true evidence) that sports are rife with doping. One could also argue that the agencies are actually more interested in making a show of catching people and not actually doing it to try and preserve the integrity of the competition.

110phil

Are false positives based on characteristics of the sample? That is, if my test shows a false positive, if they repeat the test on the same urine sample, will it still show positive?

If the answer is no, that would explain why there are so few false positives -- they just test the same sample several times.

But if the answer is yes, then, indeed, this is a very insightful point. Why *aren't* there more false positives? What is the false positive rate for a steroid test?

blaisepascal

110phil:

It is my understanding that, yes, the false positives are characteristic of the sample. That is one reason why confirmatory tests are usually done using different techniques.

For instance, the recommended protocol for employment-based drug testing is to use an ELISA test (which is a relatively inexpensive antibody based test, and has a relatively high false-positive rate), followed by an expensive, but highly accurate, gas-chromatography confirmation test if there is a positive.

For sports, I understand that they often take multiple samples for retesting purposes.

Dr. Funk

Seems like if you ask the athlete being tested, almost all positive tests are false positives.

110phil

Thanks, blaisepascal, much appreciate the info.

This is yet another example of why I believe that comments are one of the most worthwhile parts of any blog. :)

steve

Gaut says "Also, it is a scientific fact that there will be positive tests even when there are no cheaters."

There is a counter-intuitive catch to testing here that I'd like to elaborate:

MESSAGE: For any test (doping, HIV, etc.), ironically, there will be MORE false positives, the less frequent the target is (doping, HIV, etc.) when holding the "quality" of the test (e.g. false positive rate) constant.

To see why, imagine the following scenario (the exact numbers don't matter too much to the argument, feel free to experiment with other numbers):

Assume that we have a doping test that correctly identifies a person as "doped" with a probability of 98%, or in other words, 98 out of 100 doped people are detected and only 2 people are not detected.

Further assume that with the probability of 5% the test gives us a false positive result, or in other words, 5 out of 100 people are diagnosed to be "doped" although they are in fact not. (One can of course argue about the exact value of this number because one can repeat the test. However, even with repetitions, there will always be some false positive rate. Let's take 5% for illlustration).

Assume now a world ("heavy doping world") in which, say, 500 out of 1000 people are doped. This means that of the 500 dopers, 490 (500 x 98%) are detected and 10 are not. Of the 500 people that don't dope, 25 (500 x 5%) are wrongely accused of doping and the others are OK. This means that we have 490 + 25 = 515 people with a positive test and 490 real dopers. Thus, the probability that somebody with a positive test actually really is doped is be 490 / (490 + 25) = 95%, which is not too bad.

Now assume another world ("moderate doping world") in which, say only 100 out of 1000 people are doped. This means that 98 (100 x 98%) people are detected and two are not. And of the 900 people that are not doped, nevertheless 45 (900 x 5%) will get a false positive result. This means, that in this world, the probability of person with a positive test to be a real doper is 98 / (98 + 45) = 69%. Thus, roughly one out of three people with a positive test is wrongly accused of doping. Please note that the exactly the same assumptions about the test are used here, only the frequency of dopers is lower in this world.

So enough math ;-)

The point is, that even if a test seems to be quite "good" (a false positive rate of 5% is not ideal, but does not sound absolutely terrible), the meaning of a positive test result can only be really understood if one has an idea about how frequent the target event (e.g. doping) is. This is of course, not so easy, because this is why people do the testing in the first place.

If you are interested in this kind of thing, then you might check out the following two papers (one in "Science" and in "Scientific American").
http://www.psychologicalscience.org/pdf/pspi/sciam.pdf#search=%22%22Better%20decisions%20through%20science%22%20filetype%3Apdf%22
http://www.mpib-berlin.mpg.de/en/mitarbeiter/gigerenzer/pdfs/2000_%2520communicating_statistical_information.pdf

Read more...

pkimelma

To add to what blaisepascal said: Tour De France takes two samples (an "A" and a "B" sample) to avoid the risk of a contaminated sample or sample container. These are separately transported, etc.

They also use two testing protocols, although not necessarily based on cost difference, but time to measure.

zbicyclist

pkimelma is pretty much right, but for some reason the Tour de France uses one lab to do all test on both samples. This means the potential for either sloppy or fraudulent lab procedures -- which may create the temptation to repeat the sloppy procedure to defend the "integrity" of the lab.

That's a surprising thing to do. Why not test the A sample at the French lab, and ship the B samples to the Swiss lab? That would make the conspiracy argument almost impossible to take seriously. (It's not very credible, anyway.)

That said, it's pretty clear there is a lot of doping in high level $port$. There are many temptation$ and considerable folklore about how to dope undetectably.

dbrower

A conspiracy isn't necessary it the work is done incorrectly, or rarely enough (IRMS) that errors are hard to detect.

In the T/E test, it's common for the E value to be so low as to be in the range of measurement error. This results in a value of denominator that is at best an estimate -- think divide by zero. Depending on what guess you want to use, you can come up with any ratio you like. In these cases, the absolute value of T is more likely to be interesting, along with the "longitudinal" study to see if similar patterns occurred over time. We don't have the actual numbers to know with Landis yet. (French summer vacation).

The IRMS test isn't done very frequently, which raises lots of technical correctness questions that the lab -ought- to be very open about answering. Whether they are or not will be an interesting test of good intentions.

Those interested will find a roundup of Landis news at http://trustbut.blogspot.com.

TBV

Read more...

landaa

The fact that athletes are enhancing illegal substances is due to the lack of parenting. I know this may sound odd, but think of every criminal in the world. All criminals were either brought up in a lower class neighborhood or were too spoiled by their rich family. Because most professional athletes grew up in a poor neighborhood, many children believe that by having a more “ghetto” name such as these professional athletes, this will make them a better athlete or at least make them appear to be good at sports. This however cannot help a person become a better athlete. The only thing that a parent can do is to simply act as a good parent. Once in a blue moon, a child who was abandoned at thirteen years of age will become a success. Most of the time though, any child who is abandoned, regardless of age, will either be put into jail or in a coffin. According to Freakonomics, children from lower class neighborhoods are given different names from the upper class. Most of the professional athletes today grew up in poor neighborhoods. For instance, Mugsy Bogues, the NBA's smallest professional basketball player, told reporters that he would practice on his basketball everyday at the park because there was nothing else to do where he grew up. Mugsy had great parents who told him the consequences of drugs and guns, and has made him fear these illegal items throughout his life. Surely, Mugsy received much of his talent from his athletic family, but he in no way appeared fit for the NBA. Although Mugsy had much of his inherited talent from his parents, he also was forced by his parents to maintain a certain grade point average in order to play in elementary, middle, and high school. His parents also made him practice the SATs in high school, so he could play at a prestigious college basketball university. As a result, if Mugsy would have been abandoned like many sixteen year olds today, he would have never played basketball, and would most likely have gotten into drugs or gang-related violence. I am not trying to editorialize, but rather merely state the facts outline in Freakonomics.
Respectfully,
Adam L.

Read more...

Ralph

Well the way I see it it so easy to get a drug test and results in minutes. I tested my teen with a drug test strip fom www.medicaldisposables.us and pay $1.20 for a quick result. Testing politicians, police orgoverment employee is as simple as that.

Mack

It's good to see someone else who thinks that readers' letters are the best part of many magazines and papers. I've felt the same way for years.

I couldn't care less about doping 'scandals' though, or sports in general.

akatsuki

I guess it would just depend on how their incentives are structured. I personally believe (admittedly unfounded by any true evidence) that sports are rife with doping. One could also argue that the agencies are actually more interested in making a show of catching people and not actually doing it to try and preserve the integrity of the competition.

110phil

Are false positives based on characteristics of the sample? That is, if my test shows a false positive, if they repeat the test on the same urine sample, will it still show positive?

If the answer is no, that would explain why there are so few false positives -- they just test the same sample several times.

But if the answer is yes, then, indeed, this is a very insightful point. Why *aren't* there more false positives? What is the false positive rate for a steroid test?

blaisepascal

110phil:

It is my understanding that, yes, the false positives are characteristic of the sample. That is one reason why confirmatory tests are usually done using different techniques.

For instance, the recommended protocol for employment-based drug testing is to use an ELISA test (which is a relatively inexpensive antibody based test, and has a relatively high false-positive rate), followed by an expensive, but highly accurate, gas-chromatography confirmation test if there is a positive.

For sports, I understand that they often take multiple samples for retesting purposes.

Dr. Funk

Seems like if you ask the athlete being tested, almost all positive tests are false positives.

110phil

Thanks, blaisepascal, much appreciate the info.

This is yet another example of why I believe that comments are one of the most worthwhile parts of any blog. :)

steve

Gaut says "Also, it is a scientific fact that there will be positive tests even when there are no cheaters."

There is a counter-intuitive catch to testing here that I'd like to elaborate:

MESSAGE: For any test (doping, HIV, etc.), ironically, there will be MORE false positives, the less frequent the target is (doping, HIV, etc.) when holding the "quality" of the test (e.g. false positive rate) constant.

To see why, imagine the following scenario (the exact numbers don't matter too much to the argument, feel free to experiment with other numbers):

Assume that we have a doping test that correctly identifies a person as "doped" with a probability of 98%, or in other words, 98 out of 100 doped people are detected and only 2 people are not detected.

Further assume that with the probability of 5% the test gives us a false positive result, or in other words, 5 out of 100 people are diagnosed to be "doped" although they are in fact not. (One can of course argue about the exact value of this number because one can repeat the test. However, even with repetitions, there will always be some false positive rate. Let's take 5% for illlustration).

Assume now a world ("heavy doping world") in which, say, 500 out of 1000 people are doped. This means that of the 500 dopers, 490 (500 x 98%) are detected and 10 are not. Of the 500 people that don't dope, 25 (500 x 5%) are wrongely accused of doping and the others are OK. This means that we have 490 + 25 = 515 people with a positive test and 490 real dopers. Thus, the probability that somebody with a positive test actually really is doped is be 490 / (490 + 25) = 95%, which is not too bad.

Now assume another world ("moderate doping world") in which, say only 100 out of 1000 people are doped. This means that 98 (100 x 98%) people are detected and two are not. And of the 900 people that are not doped, nevertheless 45 (900 x 5%) will get a false positive result. This means, that in this world, the probability of person with a positive test to be a real doper is 98 / (98 + 45) = 69%. Thus, roughly one out of three people with a positive test is wrongly accused of doping. Please note that the exactly the same assumptions about the test are used here, only the frequency of dopers is lower in this world.

So enough math ;-)

The point is, that even if a test seems to be quite "good" (a false positive rate of 5% is not ideal, but does not sound absolutely terrible), the meaning of a positive test result can only be really understood if one has an idea about how frequent the target event (e.g. doping) is. This is of course, not so easy, because this is why people do the testing in the first place.

If you are interested in this kind of thing, then you might check out the following two papers (one in "Science" and in "Scientific American").
http://www.psychologicalscience.org/pdf/pspi/sciam.pdf#search=%22%22Better%20decisions%20through%20science%22%20filetype%3Apdf%22
http://www.mpib-berlin.mpg.de/en/mitarbeiter/gigerenzer/pdfs/2000_%2520communicating_statistical_information.pdf

Read more...

pkimelma

To add to what blaisepascal said: Tour De France takes two samples (an "A" and a "B" sample) to avoid the risk of a contaminated sample or sample container. These are separately transported, etc.

They also use two testing protocols, although not necessarily based on cost difference, but time to measure.