Here’s an article on cancer-sniffing dogs.

The samples come to the dogs — the dogs never go to the patient. At the moment, our dogs would be screening about between a .5- to 1-ml drop of urine [or 1/5 to 1/10 teaspoon], so a very small amount. In the early days, of course, we know whether the samples have come from a patient with cancer or if the patient has another disease or condition, or is in fact healthy.

They come to the dogs at our training facility. They’re put into a carousel, and the dogs go around smelling samples. If they come across a sample that has a cancer smell, they’ll stop and stare at the sample and wait. They won’t move on.

One thing this reminds me of is that the organic compounds in our bodies, our food, and the rest of nature are just incredibly complex. When we try to measure and recreate them, we tend to miss the mark. A vitamin pill is not as good as a salad, baby formula is not as good at breast milk, and food grown with synthetic fertilizers is probably not as nutritious as food grown in healthy soil (although the evidence on this is not entirely conclusive). So it makes sense that when we try to devises a test for a particular compound, we may only be testing for some of what is actually there.