Several enzyme immunoassay (EIA) tests for the detection of HIV-1 and HIV-2 serum antibodies have been developed and are available for the diagnosis of HIV infection. In areas with HIV seroprevalence <10%, the algorithm of choice is represented by an initial screening with a sensitive assay, followed by the confirmation of positive specimens with a more specific assay. Final confirmation in discordant cases is usually obtained with a positive Western Blot. Differences, however, exist in the overall performance of distinct algorithms, depending on several factors, including not only HIV seroprevalence but also economic resources, availability of laboratory equipments and the composition of prevailing circulating HIV-1 strains. Our study compared a very specific test vs. a sensitive test in the second diagnostic step. Using an oversensitive test in the confirmation step is actually inappropriate, but this is what was being done in the Republic of Congo, according to a survey conducted in the few local laboratories. However, soon after starting the project, we realized that some patients, identified as HIV positive with the diagnostic approach that put in sequence Genscreen and Determine, were probably cases of false-positivity, at least based on clinical evaluations. Biomolecular studies confirmed the clinical impression, because 5.2% (11/211) of positive patients were actually not infected when tested with WB and/or biomolecular techniques. The same individuals would have been correctly diagnosed as HIV negative if, since the beginning, Vironostika were used as second-line test. Although the difference in specificity was low (99.68%vs. 100%), this diagnostic mistake would have led to the unnecessary treatment of mother and child, with obvious consequences in terms of toxicity and cost, without considering the psychological and social impact of misdiagnosing a dreadful infection like HIV. Both diagnostic approaches were in agreement with WHO guidelines, but the choice of each test was crucial in influencing the overall performance of the diagnostic approach. Indeed, false positive testing appeared to be related to the performance of each assay: a highly specific EIA assay as second confirmatory step allows to correctly identify these samples when compared with less specific ones. Previous experiences in low-income countries suggest the usefulness of rapid field testing (Brattegaard et al. 1993; Stetler et al. 1997; Nkengasong et al. 1999; Koblavi-Deme et al. 2001; Rouet et al. 2004). However, a recent study in Uganda evidences the risk of over diagnosis of HIV infection using an algorithm based on three rapid tests: Gray et al. (2007) show a significant increase in PPV when weak bands were not classified as positives. Moreover, it has to be noted that these strategies were applied to WHO II areas with >10% seroprevalence, while the present study was performed in an area with a 5% prevalence. Similar data reported from the Indian region are in agreement with our findings (Iqbal et al. 2005). It is possible that in regions with median/low seroprevalence as in the Republic of Congo, intervention efforts may favour the use of rapid field testing in order to increase the screening efficiency especially in rural areas (Giuliano et al. 2007). Our work shows that rapid testing could be maintained, but, as suggested by Rouet et al. (2004), a serial-testing strategy using at least one assay with high specificity has to be used to avoid unnecessary treatment of uninfected mother–child pairs.