Identifying human herpesvirus 8 infection: performance characteristics of serologic assays

Epidemiologic studies of infection with the oncogenic human herpesvirus 8 (HHV-8) depend on serologic methods to diagnose infection. However, optimal strategies for identifying HHV-8 infection remain undefined. We therefore evaluated four enzyme-linked immunoassays (EIAs) and one immunofluorescence assay (IFA) using sera from 87 individuals with the prototype HHV-8 disease, Kaposi's sarcoma (KS), and 210 participants in a hemophilia study (who were presumed not to be infected with HHV-8). Assays performed reasonably well in distinguishing between infected and uninfected persons, with receiver operator curve areas between 0.86 and 0.96. Nonetheless, IFA had only 86% sensitivity and 88% specificity, and no EIA simultaneously had sensitivity and specificity above 90% for any of the optical density (OD) cutpoints used to define seropositivity. Some assays were markedly less sensitive with diluted KS sera, suggesting that they poorly identify low-titer antibodies present in asymptomatic infection. We also developed a classification tree that categorized individuals as seropositive if they had OD > 2.00 on recombinant K8.1 protein EIA or if they had both K8.1 OD between 0.51 and 2.00 and positive IFA results; this strategy had between 80% and 90% sensitivity and 95% and 100% specificity. Overall, assays performed adequately for use in most epidemiologic investigations, but wider applications will require improved tests.