Sensitivity (detection of true positives) and specificity (detection of true negatives) of a test are important properties. In an ideal case each should be equal to 100% or the value of 1.0: all true positives should be identified by a test and all true negatives rejected, or fall below a cut-off value. In practice, this is rarely the case. A trade-off usually has to be sought between a good enough sensitivity and an acceptable specificity.

The name ‘receiver operating characteristic’ (ROC) curves was conceived during the Second World War to name the ability of radar operators to discriminate between friend or enemy blips on the screen, whilst also discounting the background noise (false positives). When tested, correct or false answers could be plotted and the result was the ROC curve. This was later taken up by the medical profession to describe the accuracy of laboratory tests.

To visualize how false positives and false negatives are related to the true values and to each other, consider Figure 42. It shows the distribution of fasting blood sugar in the normal (non-diabetic) population and in the diabetic population. Note that the distribution curve for diabetics is skewed to the right. Some overlap exists between the two populations, as not all diabetics will always have a raised fasting blood sugar, whilst a raised fasting blood sugar may on occasions be seen in a non-diabetic person.