Combining pooling and alternative algorithms in seroprevalence studies.

Data from two seroprevalence studies and one comparative study of confirmatory algorithms were used to compare the costs and sensitivities of six algorithms for determining seropositivity to human immunodeficiency virus (HIV). We evaluated confirmatory strategies by using the CBC Recombigen HIV enzyme immunoassay (EIA; Cambridge BioScience, Worcester, Mass.) and immunoblotting followed by radioimmunoprecipitation assay to confirm indeterminate immunoblotting results with and without pooling of samples during screening. The least expensive algorithm was that in which sera were pooled during screening and EIA was used to confirm positive test results. The cost savings associated with this confirmatory test were greater when the prevalence of HIV infection was higher. Savings from pooling of sera for screen testing diminished as HIV prevalence increased. The sensitivity and specificity of EIA with respect to immunoblotting and radioimmunoprecipitation assay were estimated to be 0.9992 and 0.9977, respectively. We found that the implementation of pooling during screening and the use of EIA as the confirmatory test do not affect the statistical reliability of estimates of seropositivity but do result in considerable cost savings.