This evaluation was based on the testing of whole blood samples on the five selected rapid HIV assays. Determine, SD Bioline and Uni-Gold™assays had initial sensitivity of 100% (95% CI; 99.1–100) while the initial specificity of the Uni-Gold™ assay was 100% (95% CI; 99.6–100). Overall the performance characteristics of First Response and HIV 1/2 Stat-Pak assays were lower compared to those of Determine, SD Bioline and Uni-Gold™ assays.
The findings of sensitivity of 100% and specificity of 99.4% for SD Bioline in the present evaluation are in agreement with results reported elsewhere [12
]. The findings of 100% sensitivity for Determine and Uni-Gold™ assays found in this evaluation were consistent with findings documented in other studies [10
]. The 100% specificity for Uni-Gold™ assay is in contrast with findings from other studies [13
]. In a previous study that was conducted in Uganda [14
], blood donor samples from low HIV prevalence population were used [14
] compared to blood samples from different groups including high prevalence population such as hospital patients and pregnant women. In the present evaluation the sensitivity of First Response was lower [(99.5%) 95% CI; 98.2–99.9] than that reported by the manufacturer (100%) while the specificity in the present study [(99.6%) 95% CI 99–99.9%] was higher that that reported by the manufacturer (99.18%). The sensitivity of 97.7% for Stat-Pak assay is lower compared to 100% sensitivity found though the specificity is comparable with that reported from a study done in another setting [14
]. Variations in performance of rapid HIV assays in different settings especially with different levels of antigenic challenge are well known. In addition, differences in the performance may relate in part to their inability to detect low antibody titres in people taking antiretroviral therapy with low viral loads. However, in the present study, no specific information was solicited regarding the use of antiretroviral drugs making it difficult to explain the lowest performance of some of the rapid HIV assays evaluated.
Of the five rapid HIV assays evaluated, Uni-Gold™, Determine and SD Bioline were shown to have the required sensitivity and specificity for inclusion into the National rapid HIV testing algorithm. It is possible to have different algorithms based on different combinations of Determine, SD Bioline and Uni-Gold™ assays. From the site specific performance characteristics of the tests, it would appear that a combination comprising of Uni-Gold™ as a first test and Determine or SD Bioline as a second test would be suitable in all settings. On the other hand, a combination comprising of SD Bioline followed by Uni-Gold™ would possibly be useful, in a descending order of priority in: PMTCT and stand alone VCT (equally); facility-based VCT; and blood bank. A combination comprising of Determine followed by Uni-Gold™ would possibly be useful, in a descending order of priority in: PMTCT, stand alone VCT, blood bank and facility based VCT while a combination involving use of SD Bioline followed by Determine would be useful in facility-based VCT, PMTCT setting and stand alone VCT settings. Since all assays have sensitivity of 100%, in the event of shortage of any of the three assays, any available assay (among the three) can be used as the first assay. Determine is one of the assays in the currently used algorithm while SD Bioline is licensed for use in the country. In view of the present results showing acceptable performance characteristics of these three assays and taking into account the multiple interventions for treatment, control and prevention by different partners, it is prudent to consider introducing more than one algorithm for use in the country. In the event that one of the tests is withdrawn from the market, the other algorithm(s) can continue to be used without disrupting service provision in various HIV testing programs.
It is apparent that strategies involving the use of Uni-Gold™ as first or second assay are much more expensive compared to those which use combinations of the other assays. The prices for these tests were obtained from the WHO bulk procurement scheme [15
]. Taking into account the criteria for inclusion into the National algorithm and the element of cost, the possible rapid HIV serial testing algorithms are: screening by SD Bioline followed by confirmatory testing of reactive samples by Determine with Uni-Gold™ assay as a tiebreaker for discordant results. The sensitivity and specificity of this strategy is 100%. A strategy based on screening by Determine followed by confirmatory testing of reactive samples by SD Bioline with Uni-Gold™ assay as tiebreaker for discordant results also has a sensitivity of 100% and a specificity of 100%. It would be cheaper to use the former (involving initial screening with SD Bioline) than the later algorithm when serial HIV testing is performed because SD Bioline is relatively cheaper than Determine assay. However, if testing is done in parallel, the cost would be similar for both algorithms. An algorithm based on screening by SD Bioline and confirmatory testing of reactive samples by Stat-Pak Dipstick assay would have a similar cost to that of the above named two algorithms but the initial sensitivity of Stat-Pak Dipstick was 97.7%, which, according to the protocol did not qualify it for inclusion into the National algorithm. The cheapest assay combinations are those involving the use of SD Bioline and First Response assays, with SD Bioline as the first assay followed by First Response as the second assay or vice versa. This algorithm, however, is associated with a concordant false positive reactivity on the two assays as well as the fact that First Response had initial sensitivity of 99.5%. It should be noted that the cost of rapid HIV testing algorithm can change to due to changes in prices of the test kits. A rapid HIV testing strategy that is consistent with recommendations by WHO [16
] is essential for promoting HIV care, treatment and prevention programs.
Results from the current evaluation indicate that a number of assays have good performance characteristics suggesting that considering additional operational characteristics other than sensitivity and specificity would be useful in developing a national rapid HIV testing algorithm for Tanzania. This aspect was taken into account in the current evaluation during selection of the tests for inclusion into the evaluation as well as assessment of additional characteristics.