34 |
Klinisk Biokemi i Norden · 1 2013
They were shipped on dry ice to the logistic office of
NKK, where they were stored again until shipment
in the frozen form to the study participants. Each of
them received 2 aliquots per sample. The two EQA
materials were normal and spiked human serum
pools, filtered (0.22 µm) and frozen (Labquality,
Helsinki, Finland).
Laboratories
All Norwegian laboratories participating in the
Labquality’s clinical chemistry EQA schemes were
invited to take part in the study and the 56 labora-
tories accepted.
Measurement design
The participating laboratories measured S-Ca, S-Mg,
S-Alb and S-Prot in the survey samples. They were
required to perform duplicate measurements per sam-
ple, but each replicate in a different run. In the first
run, the samples were measured upwards (#1 to #20)
with the 2 EQA samples positioned after sample #7 and
#14, in the second run, in reverse order.
Assays
Peer groups (n ≥ 6, except Abbott Mg: n = 4) were
identified for laboratories that used the same test sys-
tem combination in terms of instrument, reagent and
calibrator (further referred to as homogeneous systems),
i.e., the Abbott Architect, Ortho Vitros, Roche Modu-
lar, Roche Cobas, and Siemens Advia peer group. The
measurement principles used in the different systems
were: S-Ca arsenazo III (Architect, Vitros) and o-cre-
solftaleine complexone (Modular, Cobas, Advia); S-Mg
arsenazo (Architect), formazan (Vitros), xylidyl blue
(Modular, Advia), and chlorophosphonazo III (Cobas);
S-Alb bromocresol purple (Architect) and bromocresol
green (Vitros, Modular, Cobas, Advia); S-Prot biuret
(all systems).
Statistical approach
Target values.
Peer group targets were calculated for
each sample (native and EQA-) as mean of the mean
of duplicates. Before inclusion, the input data were
investigated for outliers by means of the Grubbs test.
Overall target values (“ManuMean”) were calculated
as average of the peer group means in order to give the
same weight to each manufacturer.
Commutability assessment.
To assess commutability of
the EQA samples, first ordinary linear regression ana-
lysis of the peer group data versus the ManuMean (for
Þingvellir, Island. Foto: Henrik Alfthan.
