Analytical quality control of INR measurements in primary care
MetadataShow full item record
In Norway, most patients on oral anticoagulation with warfarin are treated in primary care. The treatment is monitored with the laboratory method prothrombin time, expressed as International normalized Ratio (INR). It is important that the INR methods have good analytical quality because the treatment (medical dose) depends on the INR result. Overdosing can cause severe bleedings and under dosing can lead to thrombosis. The laboratories in primary care control the analytical quality of their INR methods by performing internal quality control (IQC) and external quality assessment (EQA). There are, however, some challenges regarding these quality control systems. The aim of this thesis was to evaluate and suggest improvements of the analytical quality control of INR methods used in primary care.
The primary care laboratories perform IQC mainly by two different approaches; 1) a commercial lyophilized control material is analyzed on the INR method and the result is compared with some control limits, 2) a fresh patient sample is analyzed both on the INR method and on a hospital method, and the difference between the methods is compared with some control limits. The latter approach is called split sample procedure. The primary care INR method is considered “in control” if the result is within the limits and “out of control” if the result is outside the limits (error alarm). The aim of paper I was to evaluate and compare these two IQC approaches in their ability to detect systematic errors. Power functions were created by computer simulations based on empirical data from 18 primary care laboratories using the INR methods Thrombotrack, CoaguChek S, or Hemochron Jr. Signature. The control rules 12S, 13S, exponential weighted moving average, and the deviation limits of ± 10% and ± 20% were evaluated by their probability of error detection and false alarms. The results showed that the probability of detecting systematic errors was higher when lyophilized control materials were used compared to patient split samples. The probability of false alarms was, however, the same. The conclusion in paper I was that IQC of INR methods in primary care should be performed by using control materials rather than the split sample procedure. The split sample procedure with native patient samples should be restricted to method bias estimation.
International guidelines recommend that primary care laboratories should participate in an EQA scheme whenever available. The aim of paper II was to investigate if and how the European countries provide this service for point-of-care (POC) INR methods. Thirty European countries were asked, and nineteen countries reported that they do not provide EQA schemes for POC INR methods, while 12 organizations from nine countries (Austria, Czech Republic, Denmark, Finland, Hungary, Netherlands, Norway, Switzerland and United Kingdom) reported that they offer this service. All 12 organizations answered a questionnaire regarding their schemes, and the results showed that there is a vide variation in how the schemes are organized. However, the most common is to use lyophilized control materials, establish peer group target values, use an acceptability limit of 15% and distribute four samples per year. Most of the countries organize educational activities with focus on quality improvement. The study in paper II demonstrates that most European countries do not provide EQA schemes for POC INR methods, and that the disadvantages in most of the provided schemes were the use of non-commutable control materials making comparison between different POC methods impossible.
An important objective in EQA is to evaluate systematic deviations (bias) between methods. This is, however, not possible when non-commutable control materials with peer group target values are used. The aim of paper III was to develop a new EQA model in which an evaluation of method bias was incorporated in EQA schemes that use non-commutable materials. The model was developed based on the concept that a selected group of primary care laboratories should establish an estimate of the systematic deviation of the POC method from a designated comparison method by using fresh patient samples, and this information should then be incorporated in the feedback to the participants in the EQA scheme using non-commutable control materials. As a consequence, the participants will get more information about the analytical quality of their method. The model was applied twice in POC INR surveys among 1341 and 1578 participants, respectively. To estimate bias for each POC INR method, about 100 native patient samples were analyzed both by a selected group of expert primary care laboratories (72 and 69 in the first and second survey, respectively) and on a designated comparison method. Both method bias and the deviation of a single-participant result in the EQA schemes were evaluated against separate analytical quality specifications. Two POC INR methods (CoaguChek XS Plus and Simple Simon) fulfilled the quality specification for bias, whereas one did not (Thrombotrack). More than 90% of the participants received results within the quality specification for a deviating EQA result. In conclusion, a new EQA model for POC methods was proposed in paper III. This model can be used in situations where commutable control materials are not available. An editorial in the journal Clinical Chemistry has recommended that EQA organizers should implement this proposed EQA model.