Background:  The Centers for Disease Control and Prevention (CDC) recently recommended voluntary HIV testing of all adults in U.S. health care settings. While the availability of rapid, non-invasive HIV tests facilitates large scale screening, emerging data suggest that these tests may not perform as well in field settings as they did in pre-approval studies.

Methods:  In the context of a large clinical trial on HIV-testing in the emergency department setting, we evaluated the performance characteristics of the oral sampling method of the OraQuick ADVANCE HIV 1/2 Antibody Test (OraQuick, OraSure Technologies, Inc). Patients with a reactive screening test were confirmed with enzyme-linked immunoassay (EIA), Western blot, and HIV RNA. We determined the positive predictive value and specificity of the OraQuick test, using the confirmatory battery of tests as a standard.

Results:  From February 7 until August 15, 2007, 662 patients had valid test results. Of these, 30 patients had a reactive test (4.5%, 95%CI 3.0 to 6.1%) of whom 5 were confirmed as HIV⁺ (prevalence 0.76%; 95%CI 0.10 to 1.41%), 20 were HIV^–, and 5 declined antibody confirmation. Among those who confirmed, the positive predictive value of a reactive oral test was 5 of 25 (20.0%, 95%CI 4.3 to35.7%). Conservatively assuming there were no false negative results, the estimated test specificity was 96.9% (95%CI 95.6 to 98.2%), statistically significantly lower than the manufacturer-reported 99.8% (95%CI 99.6 to 99.9%). Western blot for confirmation, as suggested by the CDC, provided only 15 of 25 patients with a definitive HIV status at their follow-up visit (60.0%, 95%CI, 40.8 to 79.2%); 10 of 25 patients with a reactive OraQuick test had an indeterminate Western blot, but were all HIV^–. The addition of EIA and HIV RNA to the confirmation protocol increased the proportion of conclusive results at first follow-up to 24 of 25 (96.0%, 95%CI 88.3 to 100%).

Conclusions:  Rapid HIV screening with oral sampling in the emergency department may perform at lower positive predictive value than initially anticipated. Further rapid test confirmatory algorithms may require re-evaluation because patients with rapid HIV test reactivity frequently have indeterminate serum Western blot even though they are uninfected. While routine HIV screening remains an important public health goal, these results suggest that scientific and public health communities must recalibrate their expectations of HIV screening test performance. Information given to patients who undergo HIV screening should also reflect these findings.