A comparison of the 12s rule and Bayesian approach for quality control: application to one-stage clotting factor VIII assay.

An ideal medical biology internal quality control (IQC) plan should both monitor the laboratory methods efficiently and implement the relevant clinical-biological specifications. However, many laboratories continue to use the 12s quality control rule without considering the high risk of false rejection and without considering the relationship of analytical performance to quality requirements. Alternatively, one can move to the Bayesian arena, enabling probabilistic quantification of the information coming in, on a daily basis from the laboratory's IQC tests, and taking into account the laboratory's medical and economic contexts. Using the example of one-stage clotting factor VIII assay, the present study compares frequentist (12s quality control rule) and Bayesian IQC management with respect to prescriber requirements, process start-up phase issues, and abnormal scenarios in IQC results. To achieve comparable confidence, the traditional 12s quality control rule requires more data than the Bayesian approach in order to detect an increase in the random or systematic error of the method. Moreover, the Bayesian IQC management approach explicitly implements respect of prescriber requirements in terms of calculating the probability that the variable in question lies in a given predefined interval: for example, the factor VIII concentration required after knee surgery in a hemophilia patient.

Impact of quality control matrix effect: application to the calculation of uncertainty of measurement in one-stage clotting factor VIII assay.

In Europe, the ISO 15 189 standard requires uncertainty of measurement to be calculated for all measurands. We calculated the analytical imprecision and bias of our factor VIII coagulometric assay method between 5 and 80 U/dl, using plasmas expected to be at 5, 30 and 80 U/dl of factor VIII. We implemented Meijer et al.'s [Clin Chem 2002; 48:1011-1015] long-term coefficient of variance, bias and also uncertainty of measurement calculations. Assessments used reference plasma diluted in severe haemophilic plasma, in immunodepleted factor VIII-deficient plasma and in bovine serum albumin. With plasmas diluted in severe haemophilic and immunodepleted factor VIII-deficient plasma, calculated uncertainty of measurement was 10% compared with 15% (i.e., 50% greater) for plasma diluted in albumin buffer or as calculated from European Concerted Action on Thrombosis consensus values. It is thus important to approximate the patient sample matrix to obtain as precise an estimation as possible of assay method uncertainty of measurement.