In recent years a variety of studies have been carried out to compare the accuracy (generally expressed in terms of sensitivity and specificity) of commercially available anti-HTLV tests. None of these studies were performed in Brazil or in any other South American country. During the characterization of our Brazilian reference panel we evaluated the sensitivities and specificities of the Abbott HTLV EIA (100%; 89.7%) and the Biochrom HTLV-1/-2 ELISA (100%; 42.4%). Our conclusion was that both assays may be problematic in terms of correctly identifying HTLV-negative sera. We therefore adjusted the cut-off values using receiver operating characteristics (ROC). ROC analysis, which involves calculating sensitivity and specificity for several cut-off values, can be used to ascertain the co-variation in the specificity and sensitivity of any assay giving quantitative results. The optimum cut-off value for the assay in a given study population is the point that gives highest possible sensitivity in conjunction with a small false-positive fraction. Using the HTLV-1/-2 Western blot as the "gold standard", we were able to improve the specificity of the Biochrom HTLV-1/-2 assay to 95% without affecting its sensitivity of 100%. However, it seems that when using the Biochrom HTLV-1/-2 ELISA, there may be problems in separating positive and negative sera. In the case of the Abbott HTLV EIA, our ROC analysis revealed that the cut-off value suggested by the manufacturer was nearly identical to the optimum cut-off value. Adjustment will affect neither sensitivity nor specificity. However, a slight adjustment of the cut-off value result in a clearer separation of the positive and negative populations. Furthermore, we assume that this adjustment will help to avoid false-positive results when larger serum panels are investigated. Further investigations will show whether or not this problem is linked to the geographical regions where the test is performed (e.g. polyclonal stimulation due to parasitic infections in tropical countries).