I went back to the thread that I started after Contador lost his CAS case. Here is the link I posted at that time to a report by NZVT, the Netherlands Security System, which tests various supplements, and if it can’t detect any banned substances, gives them its seal of approval:
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2465258/?tool=pubmed
In addition to the problem of intentional spiking of contaminants, they note that contamination can occur accidentally:
The difficulty of finding possible contaminations of a nutritional supplement was shown in one of the very first studies addressing this problem. A group from an anti‐doping laboratory in Los Angeles, California, USA [Don Catlin] proved the existence of tablet‐to‐tablet variation in contaminations.24 This variation was confirmed later that year17 and is still likely to exist. The experiences from the laboratories show that contaminations can be present in the raw materials that are used, both in the active ingredients and in the substances used to make tablets or capsules. This type of contamination is often referred to as “cross‐contamination”. A second source of contamination might result from a lack of sufficient hygiene in the machinery that is being used during the production process.
The article goes on to list testosterone specifically as a contaminant found in some supplements, and notes that "contaminations in nutritional supplements are most likely to occur with substances that are part of the groups of anabolic agents or stimulants."
In November 2001, the athletes nominated to go to the Winter Olympics in Salt Lake City 2002 were given an opportunity to have their supplements tested for doping substances. They were asked to buy a supply of the nutritional supplements they were going to use during their preparation for the Olympics from a controlled sample of one batch. From this supply, a random selection of supplements was tested for several anabolic steroids, their precursors, and several stimulants.
The results of this preliminary study gave a clear insight of the seriousness, size, and scope of the problem. Of the 69 supplements that were submitted (mainly vitamins, minerals and creatine), 13 (19%) contained unlabelled doping substances.42 Most products showed traces of caffeine and/or ephedrine, one product contained a small amount of 3,4‐methylenedioxymethamphetamine (better known as MDMA or XTC), and five products contained anabolic steroids. By pure chance, two different batches of one single product were tested as well, yielding one positive and one negative finding.
Now this was quite a while ago, and as McR notes, athletes are well aware of the dangers of taking supplements now, or should be. But many still do, and supplements are not any cleaner, because they aren't required to be:
These results were reported to the relevant public authorities. The local authorities took appropriate steps to eradicate the amphetamine traces from the public food supply, but concluded that this particular issue is not a concern from a public health perspective but is first and foremost a sport and a doping problem, as the trace amounts found in supplements would not endanger general health.
The NZVT then notes its own criteria:
Based on the published facts that a precursor of an anabolic steroid in an amount between 1–10 μg can cause a positive doping test,18,24 and based on the fact that athletes easily use 50 g of supplements per day or more, a reporting threshold value of 10 ng/g or 10 ppb for all anabolic steroids is used in all tests. This value also allows for individual variations in metabolism. Excretion studies for stimulants are rare, but similar considerations led to the conclusion that for stimulants, a reporting threshold value of 100 ppb is opportune.
The 1-10 ug estimate is probably lower now. I think WADA uses MS to detect testosterone initially, and that should have a limit of < 1 ng/ml, or less than 200 ng in a urine sample, maybe ten times lower than that. There are some newer methods that can detect as little as 10 pg/ml. So while the 10 ng/g threshold sounds reasonable, it's not close to the detection limit, and if it's still in force, a sample that tested clean, if taken in large quantities by an athlete, might result in a positive test.