We thank Brown and colleagues for their comments. Their analysis highlights an important problem, both for PCR and for other diagnostic tests for HIV: the effect of the genetic diversity of HIV on sensitivity and specificity. It is possible to classify HIV-1 into two main groups, M and O, on the basis of nucleotide sequencing [1]. Strains of HIV in group M are responsible for the vast majority of infections. Nine subtypes of group M (A through I) have been described, and subtype B causes most infections in North America. The sensitivity and specificity of current enzyme immunoassays and PCR have been evaluated primarily on the basis of detection of subtype B strains. The evaluation by Brown and colleagues indicates that the sensitivity of the widely used SK462/SK431 primer varies from 0% to 80% for the detection of non-B-subtype viruses. Detection of group O viruses is also problematic with current tests. Of 10 HIV antibody tests licensed by the U.S. Food and Drug Administration, 6 failed to detect group O HIV infection in at least one of nine specimens from Africa [2]. The first case of group O HIV infection in North America was recently identified [3] in a woman in whom the results of standard enzyme immunoassays for HIV-1 and HIV type 2, DNA PCR for HIV-1, and reverse transcription PCR were negative. We did not comment on the sensitivity and specificity of PCR for non-B-subtypes because the studies included in our meta-analysis did not perform these evaluations.

These developments serve as a reminder that the sensitivity and specificity of diagnostic tests are dynamic and require ongoing assessment. Changes in PCR technology [4] and the changing molecular epidemiology of HIV have affected, and will continue to affect, the accuracy of PCR. We agree that inclusion of a universal primer in PCR would be useful for surveillance purposes. We caution, however, that such changes should be carefully evaluated with attention to the principles of study design for diagnostic tests. Although rare exceptions exist, an increase in the sensitivity of a diagnostic test is usually associated with a decrease in the test's specificity. We should evaluate PCR against HIV subtypes that reflect the clinical population in which the test will be used to ensure that increased sensitivity for rare strains of HIV does not degrade the sensitivity or specificity of PCR for the predominant strains.