An international quality control programme for PRISM chemiluminescent immunoassays in blood service and blood product laboratories

Background Laboratories screening for blood‐borne virus infections in blood and blood products are required by international standards and guidelines to ensure that their testing processes remain within control. An effective means of ensuring this aim is through participation in a quality control programme. Analyses of results from a quality control (QC) programme conducted for the Abbott PRISM (PRISM) assays are reported. Materials and methods Laboratories participating in the National Serology Reference Laboratory, Australia's PRISM QC programme were provided with aliquots of a multimarker QC sample which were tested regularly in each PRISM subchannel. Test results were submitted to a single database using an Internet‐based QC monitoring system, EDCNet. The QC test results submitted between 15 October 2001 and 5 March 2006 for each PRISM instrument and each lot of PRISM reagent were analysed to determine the imprecision and bias in each test system. Results A total of 157 404 test results from approximately 47 000 test runs submitted into the EDCNet database were analysed. Six batches of the multimarker QC samples were tested in 454 PRISM reagent lots. The coefficient of variation of QC sample test results ranged from 9·17 to 15·83%, 8·29 to 9·44%, 10·50 to 15·38% and 7·05 to 10·32% when tested in the PRISM anti‐hepatitis C virus, anti‐human immunodeficiency virus, anti‐human T‐cell lymphotrophic virus and hepatitis B surface antigen assays, respectively. Analysis of QC test results reported from testing in the anti‐HTLV assay detected one lot of reagent (10572HN00) which was identified to be an outlier using Tukey's filter. Discussion Analysis of test results of an external QC sample can be used as a statistical process control through ongoing measurement of imprecision. When laboratories test the same QC sample in the same assay and submit test results to a single database, the results can be compared and a measure of bias can be calculated. The resulting QC programme can offer detection of unexpected variation in the testing processes and the source of variation investigated.