Limits of maximum allowable analytical imprecision have been defined o
n the basis of normal range (Tonks), composite biological variation (C
otlove), intraindividual biological variation (Aspen Conference, Colle
ge of American Pathologists), medical significance (Bar-nett, Skendzel
), and a combination of medical significance and biological variation
(Fritsche, Klee). Because error limits based on medical significance a
re less stringent than those based on biological variation, performanc
e goals based on medical significance are more likely to be attained b
y laboratories than those defined by biological variation. Most clinic
al scientists realize that the goal to limit the analytical C.V. to on
e-half or less of the biological C.V. is arbitrary, and for the large
majority of laboratory tests of no benefit to the patient, for the maj
or component of total variation is not the analytical imprecision; but
the intraindividual variation itself Furthermore, the purpose of labo
ratory testing is to help physicians confirm or exclude potential diag
noses and monitoring therapy rather than detecting small deviations fr
om normal values. Small changes in test values are quite often uninter
pretable or lead to costly albeit fruitless investigations. In view of
diminishing resources for health care we must establish the accuracy
and precision required for each test. While improving the accuracy of
some tests would be desirable, for most of them further improvement wo
uld be irrelevant to patient care because few tests are pathognomonic
by themselves and quite often diagnosis is not made on the basis of a
single laboratory result. If more accuracy is desirable, it must also
be affordable and benefits should outweigh costs. (C) 1997 Elsevier Sc
ience B.V.