Summary of the Centers for Disease Control Human Immunodeficiency Virus (HIV) Performance: Evaluation Surveys for 1985 and 1986

Abstract

During 1985 and 1986, the Centers for Disease Control conducted eight surveys to evaluate laboratory performance in testing for antibody to human immunodeficiency virus (HIV). (Other names for this virus include “human T-lymphotropic virus type III” and “lymphadenopathy associated virus.”) The first survey was conducted with ten samples and 50 laboratories; the remaining surveys used six samples, but enrollment increased to as many as 475 laboratories. The purpose of these surveys was to measure test performance by the laboratories under the actual conditions of use. The surveys contained duplicate samples to permit measurement of within- and between-survey reproducibility. One survey included positive and negative reference materials as evaluation samples. Results of the Western blot (WB) test were not always positive on samples that were positive by enzyme immunoassay (EIA), and sometimes they were positive on samples that had negative results by EIA. The percentage of positive results reported by laboratories using Electronucleonics tests was lower than that reported by users of Abbott and of Litton tests. Positive results on the negative reference material were reported by 13.7% of the Abbott EIA^® test users. The tests do not seem to be calibrated against an equivalent standard. Within-survey reproducibility was usually above 95% for EIA and WB methods and for the major manufacturers of EIA tests. Between-survey reproducibility was usually above 90%. Reproducibility was below 75% for some combinations of samples, method, and manufacturer. The test performance observed in these surveys may be lower than is actually achieved on patient samples because the samples were selected to measure technical competence and contain a higher frequency of samples with low reactivity than would be encountered with clinical samples. Statification and weighting of the results in these surveys to estimate performance parameters for a blood bank population indicate that the tests are performing well in this application. If single EIA tests at the performance levels achieved in these surveys were used (no repeat and no confirmation), about 1 sample with positive results in every 50,000 samples tested would be missed and about 2.5% would be false positive results. Because tests are duplicated and confirmed in blood banks, better performance would be expected. Testing for antibody to HIV virus continues to undergo rapid change because of the introduction of new tests, changes in technical aspects of existing tests, and application of testing to changing populations. Such dynamic factors necessitate an ongoing, comprehensive monitoring of test performance.