Pharmacogenetic tests are part of a broader class of predictive medical tools.

Suppose a viral infection made patients' noses glow in the dark. A doctor would immediately know, without doing a diagnostic test, whether his or her patient had caught the virus. The glowing nose would be an example of a 'marker' - something that can be conveniently measured and is a reliable way of seeing the presence or absence of a biological trait.

A pharmacogenetic test is a form of marker for disease. Typically, a test would involve detecting a particular gene variant that is associated with an adverse drug reaction. You might possibly need to test for several variants that act together. Tests might be based on the detection of protein product of a gene (as is the case with Herceptin).

In the case of the glowing-nose disease, a doctor would need to be confident that (a) everyone who caught the new bug developed a luminous nose and (b) that no uninfected people had glowing noses. Otherwise people who needed to be treated wouldn't be, and healthy people would be given drugs they didn't need and which might cause them harm.

The same is true of tests developed based on pharmacogenomic research: they need to have good predictive value (a high standard of actually reflecting whether people have or will get the disease). Unfortunately, this is often difficult to achieve.

For example, there will rarely be an absolute one-to-one link between gene and effect. A particular genetic variant will often increase the likelihood of an adverse reaction, but not guarantee it. And some people without that variant may experience an adverse reaction to the drug anyway.

In some cases, however, there are strong links between a variation and a particular response and simple yes/no tests may be helpful. This is the case with the anti-HIV drug abacavir.

If several genetic factors are involved in response to a drug, however, things become more complicated. Clinical decisions may then need more careful thought - taking into account risks, possible alternative drugs, the severity of potential side-effects, and a patient's views.

Again, this emphasises why pharmacogenetics is likely to be slow to enter clinical practice. The discovery of a genetic association is only the beginning of the process. A test has to be developed and assessed (validated), and doctors need to see how it would benefit their clinical decision-making before they will use it.