Tip:
Highlight text to annotate it
X
Now, let us focus on Analytical Variation. Analytical Variation is the variation that
occurs during the analysis of a sample. This variation can take two forms: random and systematic.
Let’s look more closely at each form and what contributes to their causes.
Random variation is also known as imprecision. Random variation is the measurement of on-going
randomness in a test system.
Each instrument or kit has a certain amount of “expected” randomness as described
by the manufacturer in the manual or product insert.
As instruments become more computerized and as methods become more stable and specific,
imprecision has slowly improved.
The tests most susceptible to imprecision are those that require manual steps or are
technique sensitive.
Several sources can contribute to imprecision in a testing system.
The age of the instrument, number of tests it has produced over its lifetime, and the
number of hardware and/or software changes can all have an adverse affect on the imprecision.
The computerized components of modern laboratory equipment can be sensitive to environmental
changes like temperature, electrical power surges and humidity.
It is important for laboratories to focus on training “consistent” technique among
operators so that a change in operator due to shifts, vacation, extended leave or termination
of employment has minimal impact on imprecision.
Fresh reagents can be slightly more reactive while still being within specifications. Therefore,
there is an expectation of performance change when reagents are refreshed or changed.
Now that you have an understanding of what can cause imprecision, we will look at methods
of measuring imprecision.
One way of measuring imprecision is using data collected from repetitive testing of
quality control materials. This method provides a good estimate of long-term random variation
in actual everyday practice.
The laboratory uses the data collected to calculate imprecision expressed as the Coefficient
of Variation (CV). To calculate for CV, divide the standard deviation by the mean and then
convert the result into a percent by multiplying the result by 100.
Another method is to perform more formal testing by following one of two Clinical & Laboratory
Standards Institute (CLSI) standards: EP5-A2 Evaluation of Precision Performance of Quantitative
Measurement Methods and EP15-A User Demonstration of Performance for Precision and Accuracy.
Both standards describe protocols for controlled experiments used primarily to verify that
the laboratory can meet the manufacturer’s claim for precision.
The second form of Analytical Variation is systematic variation. The terms systematic
variation and bias can be interchanged. Systematic variation, or bias, is a sustained variation
where measurement and direction are relatively persistent over a period.
Both shifts and trends can be an expression of bias. A shift is a sudden and persistent
increase or decrease in QC values. A trend is a gradual increase or decrease in QC values
that eventually stabilizes.
There are two types of bias: true bias and comparative or relative bias. We will now
look at these two types of bias in more detail.
True bias is the difference between an observed value and the value for the same analyte derived
by a reference method that is metrologically traceable to an international standard. Most
control materials do not have such traceability and those that do are usually very expensive.
Comparative or relative bias is the difference between an observed value and the calculated
value of a consensus group for the same analyte.
There are three types of consensus groups familiar to laboratories participating in
control vendor interlaboratory comparison or proficiency testing programs.
The first consensus group, peer, is the ideal and preferred group for statistical comparisons.
The peer group contains analyte data from laboratories using the same instrument, method,
reagent(s), temperature and unit of measure.
The second consensus group, method, contains analyte data from laboratories using the same
method, temperature and unit of measure. Use this comparison group when the peer group
is too small for reliable statistical comparison.
The last consensus group, all labs, contains analyte data from laboratories using the same
temperature and unit of measure. This is the least statistically valid consensus group.
Only use this group as a comparative when there are not sufficient data in either the
peer or the method groups.