MEASUREMENT UNCERTAINTY
Review of the clinical utility of the measurand under study
The measurand is central to the disgnosis and monitoring of a specific disease?
Yes
Reliable and appropriate clinical outcome data is available
Yes Model 1 (Clinical Outcome model)
What is the best achievable anaylytical performance based on the state of the art?
No
Model 1a (Direct clinical outcome data is available)
No Fig 3. The method recommended to assign a specific measurand to a defined analytical quality specification. Figure modified from Ceriotti, 2016.10
that to clinical practice. Studies should be designed, including sampling intervals, to reflect that utility. Ensuring appropriate APS from BV,
related directly to clinical utility, has been the vision for the last decade of a working group of the EFLM on biological variation. The wealth of studies (in their hundreds) have all been interrogated against a defined set of publication standards as the biological variation data critical appraisal checklist (BIVAC). It is widely recognised that the work done toward critiquing data within the EFLM database has significantly improved what was available prior to its implementation. However, there are still some limitations. The discussion centred around what to do with the data that is currently in existence in the database, that does not achieve the BIVAC standard. Should it be kept in, should it be flagged or should it be removed?
State of the Art
A good example of setting APS on biological variation inappropriately is shown with C-reactive protein (CRP). Initially it was assigned to Model 1, but clearly the clinical utility of this measurand is not central to the diagnosis and treatment of a single disorder. As an acute phase protein it is associated with inflammatory states, infection and many other clinical states. As an option, BV was considered but data derived for within and between subject biological variation showed very large values. These were above 50% to 70% respectively and therefore for setting specifications were not clinically useful.
It has been shown that a CV of >33% was indicative of a non-Gaussian distribution for the measurand and that in the case of CRP this was most likely due to subclinical raised CRP levels skewing the data and forming a log normal distribution. Accordingly, for CRP, Braga and Panteghini published a method for determining APS from the best available analytical performance in commonly used methods.1
In this study
in all other methods, if any do not achieve the stated goals identification of what contributor (IVD – calibrator or local laboratory imprecision) should be evaluated and reduced if possible.
four methods of CRP measurement were assessed and their MU calculated. The best performing method was set as the optimal performance and desirable performance for the rest was determined as 1.5 x the MU of the best method. It is against that APS that the other methods are compared. As such, and by reviewing uRW
and ucal
Selecting the Model Ideally, according to Figure 3, the first step in assigning APS to a measurand is to thoroughly consider the clinical utility of the measurand in question. Again, this may sound easier than it actually is as utility may vary for a single measurand. Based on the criteria above, a measurand is first classified as either being central to the diagnosis and management of a single disease, and related to clinical outcome of that disease or has a steady state under close control within the body in a state of health. Only at that point, once the decision is made, can the appropriate model, either 1 or 2, be
WWW.PATHOLOGYINPRACTICE.COM APRIL 2024
Using APS to budget your MU The models above tell us what the maximum acceptable MU may be, but the question of how that is divided into the separate parts of the traceability chain was not clear initially. Since the first discussion of MU budgeting in the context of traceability in 2012, three major contributors to the MU of patient results have been identified.11
These
begin with higher order reference measurement procedures and material provided by reference measurement laboratories. Secondly, IVD manufacturers assure traceability through the hierarchy through calibration processes and finally the measurement system local performance contributes (Fig 4). The requirements for each contributor can be summarised as below. As the MAU for patient results is
, by being added through the root sum of squares methodology (discussed in article 2) to the uncertainty of the calibrator assigned value. As such ucal
, in 31
derived from the APS this is taken as the starting point. From there the reference measurement procedure or material must not exceed 33% of the total MAU. This 33% then contributes of course to ucal
decided upon. If appropriate and well verified data is available, the APS can be set using that model. If not, until such a time as appropriate data are derived, Model 3, the current state of the art may be temporarily used. It is important to note that Model 3 is not recommended to be used as a rescue model for measurands where the APS are difficult to achieve.
Model 1b (Simulation data is available or can be derived)
No No
Reliable biological variation data is available and appropriate
Yes Model 2 (Biological Variation model)
The measurand has a steady, well- controlled level in the body, in health
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
