Accuracy

Comparison tests between different rapid tests often show a wide range in performance, with some tests being much less sensitive and specific than others.^{1 2 3 4 5}

Rapid tests that use oral fluid (saliva) rather than blood may be subject to more sampling variation, influencing the sensitivity of the test.⁶

One comparison of six rapid tests found that HIV-positive samples which had false-negative results were more likely to be from individuals with an HIV viral load of less than 200 copies/ml.⁵

As with any other test, a reactive ('positive') result must be confirmed with a confirmatory test. This approach greatly increases the accuracy of the testing process.

A 1998 evaluation of Determine HIV-1/2 by the Health Protection Agency found that it was the second most sensitive rapid test (of 13 tested) and was almost as sensitive as some laboratory-based methods. Specificity was 100%.⁴

An evaluation of Determine for the World Health Organization found that, against 250 samples, it had a sensitivity of 100% (95% confidence limit 95.5%-100%) and a specificity of 99.4% (96.7%-100%). It detected HIV about three days later than the most sensitive third-generation test. It scored highly for ease of use.²

A validation study by the Chelsea and Westminster Hospital, enrolling 1623 people, found that the sensitivity of Determine – the ability to detect all true positive results – was 100% (95% CI: 95.3-100%) and the test’s specificity – the number of negative tests correctly identified as negative – was 99.9% (CI: 99.5-100%). There were no false negative tests, but there were two false positive results, which were picked up on standard confirmatory testing.⁷

A French study tested Determine against 200 HIV-positive samples, and found that it had a sensitivity of 94.9%.⁵

Evaluation of the use of Determine at Terrence Higgins Trust outreach sites found that of 50 people who tested positive, there was one false positive (probably due to another viral infection at the time of the test).^{8 9}

However, a study from Uganda highlighted problems with the specificity of Determine, in other words a high rate of false positives. In this setting, reactive results from Determine were confirmed with two other rapid tests (neither of which are available in the UK), and a result was considered positive if all three tests agreed. The samples were then re-tested with third-generation ELISAs and Western Blots to confirm that the diagnosis was true.

Based on 1517 samples, the sensitivity of this sequence of tests was 97.7% and the specificity was 90.4% (in other words, of 1000 HIV-negative people, 96 were incorrectly diagnosed as positive). The researchers identified that part of the problem stemmed from ‘weakly positive’ results, in other words when the testing device gave a weak signal that the manufacturer had nonetheless said should be interpreted as positive. When the ‘weakly positive’ results were excluded, the sensitivity and specificity improved to 99.6% and 97.7% respectively. The researchers concluded that the tests need to be interpreted differently, and reactive (positive) results confirmed by traditional methods.¹⁰

The manufacturer claims a sensitivity of 100% and a specificity of 99.75%.

The first ever fourth-generation test in a rapid format was introduced in 2009. This assay looks for both antibodies and p24 antigen, thus reducing the window period.

The manufacturer reports that the assay was evaluated on 1179 positive and 2343 negative samples. Sensitivity was 100% (95% confidence interval 98.04-100%), and specificity was 99.2% (98.9-99.6%).

The ability to detect p24 allowed for the identification of people in primary infection an average of 5 days (range: two to 20 days) before the previous Determine test.

A study from San Francisco assessed the performance of the combo test against samples from gay men testing during primary infection. Sensitivity in this population was 96%, which was superior to the performance of third-generation tests used in the study, and a little lower than a laboratory fourth-generation test (98%).¹¹

However a French study tested the kit against 200 HIV-positive samples, and found that it had a sensitivity of 95.8%. Two samples of people in primary infection were not detected. In addition, in 33 cases, the first test performed could not be read (results are considered invalid if an internal control bar does not appear).⁵

A French study tested the INSTI against 200 HIV-positive samples, and found that it had a sensitivity of 99%.⁵

On fingerprick blood samples, the manufacturer claims a sensitivity of 99.6% (95% confidence limit 98.9-99.9%) and a specificity of 99.3% (95% confidence limit 98.9-99.5%).

An HPA evaluation of the VIKIA HIV1/2 test, using a limited number of samples, found that it had a sensitivity of 100% and specificity of 100%. VIKIA was compared to the Determine antibody test in its ability to identify infection in blood samples of people seroconverting. Vikia detected seroconversion a mean of 2.8 days later, but performed significantly better than two non-CE marked rapid tests that were tested at the same time.¹² 

A French study tested VIKIA against 200 HIV-positive samples, and found that it had a sensitivity of 98.5%.⁵

When used with whole blood, the manufacturer claims a sensitivity of 99.95% and a specificity of 99.86%.

HPA evaluations of Core Diagnostic’s Immunoflow and Core devices against seroconversion panels found that they both had similar performance to Determine, but were superior to some other devices tested. Sensitivity and specificity were not evaluated.^{13 14}

For Core, the manufacturer claims a sensitivity of 100% and a specificity of 99.75%. For Immunoflow, they claim a sensitivity of 100% and a specificity of 100%.

OraQuick is the only rapid device that is both approved for use in the United States and currently marketed in the UK. Because of its position in the American market, its effectiveness has been more widely examined than that of some other rapid devices. Some problems, including false positive results and limited sensitivity in people with recent infection, have been identified.

More problems have occured when the test is used with oral fluid ('saliva') than when used with fingerprick blood, venepuncture blood, plasma or serum.

The CDC reported on their experience of using OraQuick in two separate reports in 2006. The ‘four studies’ pooled data from over 12,000 people.¹⁵ Post-marketing surveillance tracked the use of OraQuick with over 160,000 people.¹⁶ Both papers suggested that their experience of using OraQuick was broadly consistent with the manufacturer’s trial data. Additional data comes from a smaller WHO study.² 

Evaluations of OraQuick

Source: See above for source details

However, there have also been unexplained clusters of false positive results (i.e. lower specificity) at specific locations. These problems were reported as part of the CDC studies cited above, and in a separate report:

• During a five-month period in Minnesota, specificity dropped to 95.9%. ¹⁵
• During a nine-month period at one location in San Francisco, specificity dropped to 98.7%.¹⁶
• In New York, the specificity each month ranged from 98.88% to 99.98%.¹⁷ 

In each case, the problems arose with the oral fluid samples, and not the blood samples. Investigation into manufacturing problems and compliance with operating procedures failed to identify causes for the increase in false positives, though unknown operator errors are a possibility.

The authors of these reports stress the need to confirm initial positive results with confirmatory testing; pre-test discussion so that people who are testing are aware of the need for confirmation; quality-assurance procedures to monitor compliance with operating procedures and test performance; and the importance of alerting the manufacturer to increases in false positives.

In London, evaluation by Barts and The London NHS Trust on the effectiveness of OraQuick with sexual health patients considered ‘high-risk’ (HIV prevalence 5.7%), found that the test’s sensitivity was 93.6% (95% CI 82.5-98.7%), its specificity 99.9% (99.3-100%). The three false negatives were considered to be due to observer error; further training increased test sensitivity to 100%.¹⁸

A French study tested the kit against 200 HIV-positive blood samples, and found that it had a sensitivity of 94.5%. However, when testing saliva samples from the same individuals, sensitivity dropped further to 86.5%.⁵

A San Francisco study assessing the performance of OraQuick Advance against samples from gay men testing during primary infection found that the test had poor sensitivity (86%) when using saliva samples, but did better when blood samples were tested (92%).¹¹ 

Another study, conducted among gay men in Seattle, including high numbers of men during primary infection, found that the sensitivity of OraQuick was only 80%. A mix of saliva and blood samples were taken.¹⁹ Other studies have found that OraQuick was only able to identify one of 43 samples from people with recent infection¹ and that OraQuick was unable to detect seroconverters any sooner than a Western Blot, which has a window period of approximately five weeks.²⁰  Because OraQuick relies on detecting antibody to gp41 any delay in producing this antibody could reduce the sensitivity of the test in early infection.²¹

The manufacturer claims that OraQuick detects people with recent infection an average of 2.5 days (95% confidence limit: 1.2 to 3.8 days) after a third-generation test.

There has also been a case report of a situation in which a man with late-stage AIDS was given repeated false negative results with OraQuick. The 28-year-old man had oral candidiasis and PCP and was tested using oral fluids several times over a nine-month period. Each test was negative, but standard tests later confirmed HIV infection. The man had a CD4 count of 4 cells/mm³ and a viral load >100,000 copies/ml.

Examination of the the Western blot results showed a strong gp160 band, but other bands, including gp41, were very weak. Because the OraQuick test contains only a single glycoprotein antigen (anti-gp41), it will not identify HIV-infected individuals who fail to mount a substantial response to that particular antigen. These individuals may include the newly infected as well as people with such late-stage infection that their antibody responses to HIV have been compromised by the infection itself.²²

Similar problems could occur with other rapid tests; if symptoms suggest HIV infection, but results of a rapid test are non-reactive, confirmatory laboratory tests should be used.

More positively, evaluation of OraQuick in southern Africa (where subtype C is most common) found that it had a sensitivity of 100% (95% confidence interval 94.9%-100%) and a specificity of 100% (98.2%-100%).²³

A WHO report rated OraQuick highly for ease of use.²