Rapid HIV tests: false positive risk without confirmatory test highlighted in Uganda

Patients who have a `weakly positive` result on a rapid HIV antibody test should not be told they are HIV-positive without a confirmatory ELISA or Western Blot test, since 86% of these weakly positive results turned out to be negative when checked with a confirmatory test, say researchers from the United States and Uganda in the July 28th edition of the British Medical Journal.

Rapid diagnostic tests for HIV diagnosis have made it possible to provide point of care results without having to invest in costly laboratory facilities. Rapid diagnostic tests have been used for voluntary HIV counselling and testing, antenatal surveillance, and population screening.

With increased access to antiretroviral therapy (ART) being the target for many resource-poor countries lacking good laboratory facilities, the use of rapid diagnostic tests will be scaled up to an even greater level.

An ideal rapid point of care diagnostic test must have several important attributes. It must be cheap and affordable, it must provide rapid results, it must be simple such that technicians can be trained to use it within a short time, and, significantly, it must be as reliable as the enzyme immunoassay and western blot found in sophisticated facilities.

This is important whether they are being employed in the context of clinical decision-making or for routine screens.

Many national HIV testing protocols in Africa now recommend the use of two different rapid tests rather than a rapid test followed by a confirmatory ELISA test in order to increase the volume of tests that can be carried out.

However, the reliability of the rapid diagnostic test is crucial, and researchers in Uganda spotted a potential for a high false-positive diagnosis rate during recruitment to two large studies of circumcision during 2003-4.

The study population were HIV-negative and HIV-positive uncircumcised males aged 14-49 years old from Rakai District in rural south-western Uganda.

The rapid diagnostic tests Determine HIV-1/2/O, HIV1/2 Stat-Pak Ultra Fast, and Uni-Gold Recombinant HIV-1/2 were used to screen the men for enrolment into the two clinical trials.

The algorithm consisted of a first screen with Determine HIV-1/2/O; if the test result was negative a negative HIV diagnosis was made; if it was positive, the sample was re-tested with HIV1/2 Stat-Pak Ultra Fast. If both Determine and Stat-Pak Ultra Fast were positive, a positive HIV diagnosis was made with no further testing. However, if the results did not agree, the sample was re-tested with Uni-Gold Recombinant.

When a sample was tested by all three tests, two positive test results were interpreted as a positive diagnosis and if two of the three tests gave a negative result, a negative diagnosis was made. For quality control (QC) purposes, all samples were batch re-tested using two enzyme immunoassays; discordant results were confirmed using western blotting.

In all, 1,517 samples were tested using the algorithm, and 261 were found to be positive on the Determine assay but negative on the Stat-Pak assay; 121 of these indeterminate results were subsequently declared to be HIV-negative after testing with the Uni-Gold assay.

Quality control analysis of 878 samples further suggested that there were false positive results. Subsequently, to address this issue, technicians recorded the intensity of positive bands in 639 samples. One hundred and twenty-five out of the 639 samples tested positive for HIV and 37 (5.8 %) of these had weak positive bands.

The algorithm was evaluated for sensitivity, specificity, negative predictive and positive predictive values when compared with the conventional enzyme immunoassay and western blotting. There were three evaluations: for the all the 1,517 samples, for the 639 samples for which the intensity of bands were recorded, and for 602 samples without the weak positive bands.

For the total 1,517 samples initially screened without taking into account band intensity, the sensitivity, specificity, negative predictive and positive predictive values were 97.7, 90.4, 99.7, and 56.3 %, respectively. The false positive rate was very high at 43.7 %. This explained the low positive predictive values.

For the 639 tests in which weak bands were coded, the specificity and positive predictive values were still low at 94 and 74%, respectively, and the false positive rate was high at 26 %. In the 602 subsamples with no weak bands, the specificity and positive predictive values increased significantly to an impressive 99.6 and 97.7%, respectively, and the false positive rate fell to 2.3 %.

The authors considered various explanations for the high false positive rates, including observer error, cross-reactions with other co-endemic pathogens in Uganda, and HIV-1 subtypes. Samples sent for QC re-testing in the USA ruled out observer error. The authors recommend the implementation of a QC system in which samples are re-tested by enzyme immunoassay and western blotting in every endemic setting using rapid diagnostic tests.

The findings demonstrate that weakly positive bands in rapid diagnostic tests severely compromise the diagnostic accuracy of HIV rapid diagnostic tests due to high false positive rates. The policy implication is that a rigorous QC system for validating rapid test results must be implemented under research and normal clinic settings, and that patients must never be tested using only one rapid antibody test.