The ideal situation with regard to quality assurance in clinical patho
logy would be that quality goals in numerical format were available fo
r all the practicability and reliability characteristics of laboratory
tests, particularly the latter. Many strategies have been used for th
e setting of quality goals for the most important reliability characte
ristics, namely, imprecision and bias. These have included fractions o
f the reference interval, opinions of clinicians, the state of the art
, views of expert individuals and groups, the influence of analytical
error on clinical utility and biological variation. All of these have
advantages and disadvantages. In spite of some interesting recent prop
osals, quality goals based on biological variation appear to be the be
st currently available and seem to have led to a consensus among Europ
eans that, in simple terms: 1. desirable maximum imprecision < one-hal
f the within-subject biological variation. 2. desirable maximum bias <
one-quarter the group [within- plus between subject] variation and 3.
maximum difference between methods < one-third the within-subject bio
logical variation. There appears a need for further work in setting qu
ality goals for laboratories which deal with samples from animals; int
ernational collaboration has proved productive in human clinical bioch
emistry and may be worthy of emulation.