Reference methods are a key element to the objective of traceability in lab
oratory medicine. However, to serve this purpose adequately, minimum analyt
ical quality specifications are required. Here, possible strategies for der
iving such specifications are presented, being based on concepts developed
before by a European Working Group. Distinction is made between "genuine re
quirements" for reference methods (direct calibration with primary referenc
e materials: absence of sample-related effects) and "performance specificat
ions" (limits for random, systematic and total error, the latter in associa
tion with the number of measurements). While the former requirements are co
nsidered as conditio sine-qua-non, the latter specifications should be vari
able, which means that they should be tailored to the specific application
of the methods. In general, it is advocated to derive performance specifica
tions for reference methods from desirable specifications of routine method
s (analyte-related), although other models should not be ruled out beforeha
nd. Further, it is recommended that reference laboratories make special eff
orts to demonstrate and maintain a uniform level of quality of reference me
thods.