In the US, when it comes to quality requirements, we have CLIA and clinical decision criteria and a whole host of other choices (many of them bad). But, if you can believe it, the situation in Europe is even more complicated! Dr. Per Hyltoft-Petersen discusses external quality assurance schemes, ISO, analytical goal-setting, and finally biological goal-setting.
Since Tonks [1], a number of different approaches to analytical goal-setting have been introduced based on the state of the art’, biology, and clinical usefulness. In the US, the introduction of the CLIA concept has focused on total error criteria for acceptable performance in proficiency testing surveys. These CLIA criteria are available on this website and have been discussed by Westgard in Quality goals, requirements, and specifications.
In Europe, Germany has chosen a concept close to the US, whereas the other European countries prefer external quality assessment with the so called educational schemes’ in order to establish a continuous improvement of quality. Two approaches to goal-setting -- clinical and biological -- have been evaluated and two European working groups (EGE-Lab and EQA-Organizers) have outlined recommendations for analytical quality specifications based on biology. Further, a new group under ECLM is working on quality specifications, also for quantities measured on ordinal and nominal scales. Internationally, ISO has a working group on analytical quality specifications and in the US there have been attempts to evaluate analytical quality specifications based on 'clinical outcomes’.
The Nordic approaches (NORDKEM-projects)[2-3] are based on clinical situations/strategies where specific clinical decisions using a single analytical quantity are described under optimum analytical conditions. Then, increasing analytical bias and imprecision are assumed or simulated resulting in poorer clinical/diagnostical outcome and the minimum analytical quality requirements are defined from the worst, but still acceptable situation. A special evaluation of analytical quality specifications for S-Cholesterol has been performed by Westgard et al. [4] and Klee has now taken up the Nordic concept for S-Calcium [5]. These approaches, all give separate specifications for allowable analytical bias and imprecision.
A comparable approach for evaluation of imprecision specifications, in clinical decisions for pre-decided changes during patient monitoring has been evaluated by Fraser, Lytken Larsen, and Hyltoft Petersen [6-8] and the Barnett approach by questionnaires sent to clinicians [9] has been used by Elion-Gerritzen [10] and Skendzel et al. [11]. A special approach for quality specifications in drug monitoring has been elaborated by Fraser [12] (for imprecision only).
Two main strategies for analytical quality specifications based on biology have been evaluated for imprecision and bias (in combination with imprecision), respectively.
The Cotlove/Harris concept [13] defines the maximum allowable imprecision, CVA, as the maximum imprecision, that when added to the within-subject biological variation, CVI, will maximimally increase the total CV by 12%, which is achieved when:
Gowans et al. [14] based their evaluation on the concept of sharing common reference intervals for homogeneous populations. IFCC’s recommendation for establishing reference intervals has a minimum of 120 reference individuals in order to reduce the confidence intervals about the reference limits [15]. By increasing the sample size considerably (more than 800 individuals), this uncertainty will become negligible. Then, the inevitable uncertainty of the recommendations can be used as the acceptable analytical uncertainty, if a laboratory wants to use these common reference intervals, and thus the analytical quality specifications are defined. This concept assumes reliable method standardization, but when established, no further creations of reference intervals are needed in the laboratories. Thus, the reference intervals are well documented for all laboratories and no extra work on reference intervals need to be done in the single laboratory, where such work has often been difficult to perform. This model has been used for establishment of specifications for nine plasma proteins in the Nordic countries using traceability to the CRM 470 preparation [16]. The specification for analytical bias (for using the common reference intervals) is:
when analytical imprecision is negligible, and where CVG is the between-subject biological variation [14]. Lists of biological data on CVI and CVG are available from references 17 and 18.
A working group under EGE-Lab (European Group for the Evaluation of Reagents and Analytical Systems in Laboratory Medicine) has adapted the two biological concepts, mentioned above, as the basis for recommendations for analytical quality specifications for bias and imprecision [19, 20]. Illustrative examples are shown in the table below; the full list is available in the Table of European biologic goals. For quantities where the biologically derived specifications are too demanding for the current methodology, interim specifications are given in parentheses.
Examples of EGE-Lab biologic goals
Quantity | Allowable Bias (BiasA) | Allowable Imprecision (sA) |
Alanine aminotransferase | 13.6% | 13.6% |
Cholesterol | 4.1 % | 2.7 % |
Glucose | 1.9 % (4.4 %) | 2.2 % |
Urate | 4.2% | 4.0 % (8.4 %) |
A Working group under the auspices of European External Quality Assessment Organizers (EQA-Organizers) has confirmed the EGE-Lab recommendations with a minor difference [21]. Whereas the EGE-Lab recommends separate criteria for bias and imprecision that contribute to a total error, the EQA-Organizer group pays attention to the original work of Gowans et al. [14] and recommends the same figures, but the value for bias is only valid when imprecision is negligible.
This working group further recommends a maximum allowable difference in bias between two instruments in a laboratory measuring the same quantity based on the work of Hyltoft Petersen et al. [8]:
The EQA-Organizer working group has also described how the Validator program can be used for design of internal control systems according to the biological criteria [22]. The group pays attention to the importance of keeping the Pfr very low, when effective trouble- shooting procedures are maintained and fail-safe procedures are established. When Pfr is low, e.g. 0.1 %, then the probability of false reject is so small, that it is worth to start an intensive trouble-shooting procedure each time the run is rejected. This means there should be no immediate re-run, but a thorough investigation of the error in order to find the source. When this has been found, then fail-safe procedures are established in order to prevent future errors of the same kind - and consequently the frequency of errors are reduced. Therefore, a Ped of less than 90 % may be acceptable as the frequency of reported errors is reduced as well, and without frustrations from re-running falsely rejected runs. This concept requires a lot of discipline in the beginning until the technicians (and others) really understand the system.
The European goals for imprecion and inaccuracy can be transformed to a biologic allowable total error (TEBA), as described by EQA-Organizers [22], according to the following formula:
|TEBA| = |BiasA| + 1.65 * sA (or CVA)
where the absolute value sign is used to be sure that bias and imprecision add together to give a maximum total error. This calculated total error can then be used as the total error parameter in the Validator program and the QC planning tools and planning process applied as demonstrated on this website.
The EGE-Lab specifications are related to analytical bias and imprecision separately, so an alternate QC planning approach is to calculate the critical increases in random and systematic that are important to detect by QC, as shown below:
REcrit = sA/sobserved
SEcrit = (BiasA - Biasobserved )/sobserved
where the division by sobserved is necessary to present the critical errors as multiples of the observed method imprecision; note that sobserved is called smeas and biasobserved is called biasmeas in the terminology used on this website. These calculated critical-sized errors can then be drawn on power curves to prepare critical-error graphs for estimating the probabilities for error detection and false rejection. The EZ Rules program can be used as a source of power curves for most commonly used QC procedures.
The EQA-Organizers group included applications for alanine aminotransferase and urate in its description of how to use the EZ Rules/Validator program [22].
Additional applications of the European biologic goals are provided on this website for cholesterol and glucose to permit comparison with earlier examples that are used analytical and clinical quality requirements. See QC Planning Applications using European Biologic Goals.
Per Hyltoft Petersen is a biochemist (MS) from University of Copenhagen and has been in the field of clinical chemistry since 1963, first at The Finsen Institute and from 1971 at Odense University Hospital, Denmark. He spent a sabbatical semester as a Visiting Scientist at the University of Wisconsin in 1983. He has been involved in analytical goal-setting (or analytical quality specifications) since the late seventies, when the Nordic approach was initiated in a Nordic project (NORDKEM-project) on analytical quality specifications based on clinical situations/strategies. He has elaborated on this concept for a number of clinical strategies, but also realized that only few clinical strategies are accepted internationally, and that a more general concept is needed as well, e.g., for external quality assessment schemes. So together with Elizabeth M. S. Gowans he evaluated the quality specifications for sharing common reference intervals. These specifications have been applied in a Nordic project on establishing common reference intervals for nine specific plasma proteins, traceable to the certified reference material CRM 470. As a member of European working groups on analytical quality specifications he has been involved in the recommendations from EGE-Lab and EQA- Organizers on quality specifications for external quality assessment schemes based on biology, and he is now a member of an other European working group under ECLM, where the goal is to establish quality specifications also for quantities measured on ordinal and nominal scales. He has further been involved in external as well as internal quality control for which the EQA-Organizers working group has published recommendations as well.