Another minister and doping story (see my April 23 blog), but this time the minister is of the religious persuasion.
Anti-doping agencies have a “strict liability” rule. If a drug is in your body you are guilty. The authorities don’t have to prove how it got there. Therefore “false positive” tests are a huge problem in developing a new test. The example I give in my book is of a test for gene doping that is accurate 999 out of 1,000 times. This is much better than the usual statistical tests that scientists use. If the 1 in 1,000 chance is of missing a cheat (false negative) this might be acceptable. However, it is unacceptable if the 1 in 1,000 chance is of catching an innocent (false positive). This is because doping tests are not used to confirm whether someone is cheating; in this case a 1 in 1,000 error would probably be considered “beyond reasonable doubt” by a jury. Instead they are the primary way of catching someone in the absence of any other evidence. A statistical calculation shows that for a 1 in a 1,000 chance of a false positive result, it would only take on average 693 tests to convict an innocent person. To put this in context, London 2012 will conduct over 6,000 drug tests.
We can relate this idea to our own lives by thinking of what might happen when we see a doctor. Say she gives us a test for a fatal disease that affects 1% of the population. The test is 99 percent reliable; so 99% of people who have the disease test positive and 99 percent of those are healthy test negative. The doctor gives you a test and it comes back positive. How worried should you be? How likely are you to die? The immediate thought is that you are 99% likely to die. Start writing your will and putting your affairs in order now. In fact the answer is that your doom is only 50% likely. You still have a chance.
The statistical tests that are used to make these calculations are based on Bayes's theorem. Thomas Bayes was an 18th century English mathematician and Presbyterian Minister based in Tunbridge Wells in Kent. In an essay published posthumously he devised a theory that used prior knowledge of a distribution to determine the likelihood of a subsequent event being correct.
What has this got to do with doping? Well Bayesian probability underpins the test for blood doping used in the athlete’s blood passport1. Upon being enrolled in the passport program, the athlete gives blood regularly. At first it is assumed that his blood parameters are the same as the general population. As more tests are taken a statistic is calculated that is biased towards the individual parameters of the specific athlete. The athlete becomes his own control. If his blood deviates from his own normal readings, such as might happen following EPO doping or a blood transfusion, he can fail a doping test even if no specific doping product is found in the blood.
This is not an academic question. Last week saw the very first successful prosecution of a runner for blood doping via the use of a biological passport. A four-year ban2 was given to the Portuguese marathon runner, Helder Ornelas, solely for anomalous readings in his blood passport during the 2010 season.
We may be entering a new era in anti-doping (although see my April 3 blog for a different view).
1 you want to find out more about the biological passport and Bayesian theorem in layman's terms, check out these links: Biological Passports; Bayesian probability.
2 At 38 years old any ban effectively ends Ornelas’s career. The unusually long ban for a first offence in this case suggests that the authorities may be trying for a deterrent effect; they want the dopers running scared of the passport system.