12 |
Klinisk Biokemi i Norden · 1 2017
The standard deviation for each observation is only
known up to some common scaling factor. The stan-
dard errors for the parameters have been corrected
by an estimate of this factor, the square root of the
reduced chi-square.
Hypothesis Testing:
Test for slope = 0
F
= 5295,9198
P
= <0,0001
Test for slope = 1
F
= 2,8469
P
= 0,1088
Table 2
: The parameters of the ortogonal/Deming regression
as calculated by SigmaPlot 12.5.
Recalculating results
For the data above (tables 1 and 2) the slope of the
Deming regression line is 0,9683 and the intercept on
the Y-axis is 0,2688. This means that if you are mea-
suring samples for a study which extends in time over
the change of methods you will multiply the results
performed with the new method by 0,9683 and add
0,2688 in order to transform all results to the same
measurement level as the one which existed during
the time period of the old method.
Provided you wish to transform the results of the
old method to the results performed with the new
method you subtract 0,2688 from the results of the
old method and divide by 0,9683.
Commutability
Commutability is basically a characteristic of samples
which expresses their ability to result in very similar
measurement results using different measurement
systems. More formally it is a property of a material/
sample demonstrated by “the closeness of agreement
between the relation among the measurement results
for a stated quantity in this material, obtained accor-
ding to two given measurement procedures, and the
relation obtained among the measurement results
for other specified materials”. (31). Natural patient
samples are by definition commutable.
When comparing measurement methods during
verification and validation, it is crucial to include
natural patient samples as commutable materials in
a very substantial part of the the procedures in order
that the results ultimately measured in the patient
samples are comparable.
Medical laboratories process very substantial num-
ber of patient samples and have usually material to
spare that can be used for maintaining and increa-
sing the quality of the measurement methods used
in the laboratory. Medical laboratories therefore have
through the availability of patient samples a very
substantial advantage compared to the manufactu-
rers of measurement systems and methods when it
comes to the availability of patient samples in their
quality processes.
Method validation
Method validation (13, 16, 17) is commonly thought
of as a highly standardized single linear process, when
it actually consists of several dimensions of repetitive
sub-processes as depicted in figure 5.
In addition to the repetitive sub-processes for opti-
mation, there are a number of principally different
validation processes appropriate for the context that
the method(s) will be used in as follows.
Single laboratory method validation
Single laboratory method validation
is appropriate
when one method is used for a specific purpose in one
laboratory (13, 16, 17). This is the type of validation
provided by manufacturers to end-users.
Full method validation
Full method validation
in a conglomerate of labo-
ratories includes, in addition to the procedures of
single laboratory validation, a study of the fitness
for purpose of measurement systems in a number of
locations, several operators etc. including a study of
the performance characteristics of the measurement
systems over extended periods of time including the
effects of lot-to-lot variations etc.
Coefficient
Std. Error
95% Conf-L
95% Conf-U
Intercept
0,2688
0,0473
0,1693
0,3682
Slope
0,9683
0,0186
0,9293
1,0072
