MEASUREMENT UNCERTAINTY Alpha genotype Dependent controls Produced for a specific system
Same sources as calibrators – reduces sensitivity to subtle changes
2nd Party Semi-dependent controls Recommended by the manufacturer of the system
Produced according to the system specifications
3rd Party Detecting trends Not optimised for a single
closed analytical system, kit, method etc.
As long as it is commutable – it more closely reflects patient sample performance
Table 2. Contrast of different sources of IQC material and the impact on traceability, commutability and the application to the QC Component II approach.
n Short term data collection Prior to the 1920s, statistical analysis was largely by z- and t-tests. That was until the introduction of the analysis of variance (ANOVA) credited to Ronald Fisher. ANOVA assesses the variance between and within groups of a system, in the laboratory case IQC across multiple runs, days, analysers or laboratories. The origins of ANOVA, or at least the statistical underpinnings, can be traced back as far as Pierre Simon Laplace and Carl Freidrich Gauss, two giants of statistical method development in the late 1700s and early 1800s. Those early works were interested in combining measurements to obtain more accurate estimates in astronomy, and Fisher’s work applied the ANOVA method to crop variation. Despite these apparently diverse, somewhat unrelated applications, the ANOVA was the basis of determining imprecision in 5 x 5 studies and is still the basis of many CLSI documents (EP05, EP15, for example). CLSI EP29A, the first guideline document for MU in medical laboratories, uses ANOVA. Early applications highlighted its utility with short-term data leading to underestimation of MU. It can, however, represent what the minimum achievable MU in a local laboratory may be, and also allows confirmation of performance characteristics provided by the manufacturer for verification. Moving forward, it may be an option to handle MU in laboratory networks.
n Choice of material to represent metrological traceability The use of homemade quality controls has been replaced by provision of control material from commercial manufacturers. When aligned with a specific assay and/or analytical platform they are commonly referred to as first party QC. Such QC is optimised to verify the alignment of the method for batch acceptance and routine random access activities (Table 2). Alternatively, IVD manufacturers may choose to procure QC from a manufacturer of IQC that have optimised the controls for the
20
manufacturer’s system, based on analytic performance requirements. These are termed second party and provide a useful alternative to first party when it is not available. However, both first and second party QC are optimised for the specific measurement system. This may not reflect the true random variability of the analytical system if optimisation is such that changes remain undetected. The final classification is third party QC; this is not optimised for any individual analytical platform, so provides an independent assessment of the platforms performance, in both random and systematic error detection. The use of third party is defined as an “attribute to be considered” in the selection of IQC (ISO TS/20914:2019) and “should be considered, either as an alternative to, or in addition to, control material supplied by the reagent or instrument manufacturer” in ISO15189:2022 (7.3.7.2 [a.3]). Recent publications have suggested that this concept should be taken even further when specifically discussing MU.
n QC Component II – an additional QC activity?
To implement metrological traceability in IQC, and enable MU quantification, the approach to QC may need to consider a combination of two different types of QC. Firstly, one to test alignment of the measurement system, using first or second party QC. Secondly, an independent measure of the measurement procedure variability using a third party or patient pool solely
to calculate uRW . There are obviously
consequences including time, money and effort so it remains to be seen if it develops further. The irony of this is that we seem to have come full circle, back to the days where we used in-house QC and are now again looking at serum/plasma pools as QC material (Panteghini, 2023),5 albeit to provide a different function than we were all those years ago.
n Additional considerations on the subject of sample handling The nature of some measurands in clinical laboratories makes IQC material fundamentally different to patient samples (see Table 1). The extent to which this impacts MU should be assessed. Such examples may be lyophilisation of IQC material, a necessity for stability and transport in some disciplines. The processing of haemolysates in some tests (HbA1c and many others) may also not be represented in the IQC material, and stabilisers and preservatives within QC material again may not be reflected in patient samples. It must be shown that the variability in such material is equivalent to that seen in patient samples for them to be used for MU assessment to achieve the assumptions in Table 1.
n Testing for differences To investigate equivalence, the building block of the ANOVA method (the F-test) can be used to compare variances of the patient and control material. If there is no statistically significant difference in variance, IQC material is adequate for representing imprecision within the method as well as in a single patient result. Of course, statistical assumptions around what is the size of the difference to be detected, and the resultant sample number needed, should be considered.
n Data handling - Excluding of outliers All IQC data used for MU assessment must be obtained while the system is under steady state control. Data obtained during process deviations or when significant analytical problems are identified do not represent the routine random variability of the method. This
Prior to the 1920s, statistical analysis was largely by z- and t-tests. That was until the introduction of the analysis of variance (ANOVA) credited to Ronald Fisher. ANOVA assesses the variance between and within groups of a system, in the laboratory case IQC across multiple runs, days, analysers or laboratories
MAY 2024
WWW.PATHOLOGYINPRACTICE.COM
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46 |
Page 47 |
Page 48 |
Page 49 |
Page 50 |
Page 51 |
Page 52 |
Page 53 |
Page 54 |
Page 55 |
Page 56 |
Page 57 |
Page 58 |
Page 59 |
Page 60
