236   Rates of False-Positive Results on Quantitative HIV RNA Assays.

D. Brambilla*¹, S. Granger¹, C. Jennings², and J. Bremer².
¹New England Res. Inst., Watertown, MA and²Rush Med. Coll., Chicago, IL.

Background:Quantitative HIV RNA assays are being used to diagnose HIV infection, but little is known about the specificity of the tests in practice. This study examined rates of false positives on the Standard (ST) and UltraSensitive (US) Roche HIV Monitor Test and the Bayer v3.0 bDNA (BD) assay.

Methods:Coded test panels of 20 samples, including 12—18 HIV-positive samples and 2—8 (median: 3) HIV- negative samples, were sent to participating laboratories (ST: 67 labs, median 7 panels/lab, range 1—16; US: 28 labs, median 5 panels, range 1—8; BD: 7 labs, median: 8 panels, range 1—13). Assays of 2,026 negative samples were examined (ST: median 23 negative samples/lab, range 2—57; US: median 14, range 3—22; BD: median 20, range 3—34).

Results:False positives were obtained from the ST kit in 9 (13%) labs, from the US kit in 3 (11%) labs and from the BD kit in 4 (57%) labs. The rate of false positives was higher on the BD assay than on the US and ST tests and slightly higher on the US than on the ST test [ST: 1.4% (22/1,554); US: 2.6% (9/351); BD: 5.8% (7/121)]. However, if one site with 6 false positives in 53 assays on the ST kit and 7 in 22 assays on the US kit were excluded, the rates on these two assays would be 1.1% and 0.6% respectively. Estimates of HIV RNA level in false positives were often well above the limit of detection (ST: median 924 copies/mL, range 95—379, 844; US: median 89, range 12—5, 969; BD: median 190, range 51—369). Potential causes of the errors were examined. It is highly unlikely that they were simple chance events, given the large distances, in standard deviations, between the average signals (e.g. optical densities on the ST and US assays) from negative samples and the positive/ negative cutoffs for the kits (e.g. >8 SD on the ST and US assays). Signals from true negatives in runs with and without false positives were similar, so background drift was an unlikely cause. One switch of a negative and a positive sample was identified.

Conclusions:We postulate that most of the false positives were caused by mistakes in sample handling. We recommend that if one of these tests is used for diagnosis, a positive result should be confirmed with a new specimen, even if the HIV RNA level in the first assay was very high.