[Improvement of erythropoietin bioassay.].

The relevance of bioassay standardization results from the lack of consistent national regulatory requirements for evaluation of recombinant human erythropoietin quality and the need to harmonize these requirements with international ones. Precision studies were carried out in 6 experiments on Balb/C mice. The factors that can influence the accuracy of the method were altered during the experiments. Each experiment included three levels: 20, 40 and 80 IU/ml, and 8 replicates for the reference and test samples. The trueness was estimated by bias relative to the reference value at 5 levels: 10, 20, 40, 80 and 160 IU/ml, and 4 replicates for the reference and test samples at each level. The test samples were prepared by a series of independent dilutions of the reference standard. Reticulocyte count was performed using a flow cytometer. 5 μmol acridine orange solution was used as a dye. Experimental study of accuracy and optimization of erythropoietin bioassay procedure helped to obtain two validation characteristics (trueness and precision). It was shown that logarithms of erythropoiesis registered values could reasonably be used in statistical calculations of erythropoietin specific activity and evaluation of the method's validation parameters. The theoretically and experimentally justified test procedure includes three levels of doses: 20, 40 and 80 IU/ml, and 8 animals for each level, which is consistent with the international requirements for accuracy. According to the results of experimental studies, the trueness is characterized by a bias of no more than 9 % and does not exceed the range of the calculated activity (80-125 %). Statistical processing of the test results by the parallel-line method makes it possible to check the assumption of equivalence of the test and reference samples and to calculate the test sample activity. The confidence limit of the calculated activity for intra-laboratory precision of 5.6 % is equal to 76-131 % which complies with the proposed range (64-156 %, P=0.95).