Rapid HIV test performance poorer with oral fluid than blood, especially when prevalence is low

The performance of OraQuick, a widely used rapid point-of-care test for HIV diagnosis, is slightly poorer when testing oral fluid samples than when testing blood samples, according to a meta-analysis published in the online edition of Lancet Infectious Diseases this week.The authors also found that if the test is used in low-prevalence settings with oral fluid samples, over one-in-ten reactive results may turn out to be false positives.

Rapid diagnostic tests, which can be operated at the point of care by doctors or nurses, have a number of attractive features in comparison with conventional laboratory tests. Such devices are essential in resource-limited settings where laboratories are inaccessible or unavailable. As patients do not need to come back for their results on another day, far more people receive their results.

Rapid tests also have the advantage of not requiring a blood sample taken with a needle. Most require a tiny sample of whole blood, taken by pricking a fingertip with the sharp point of a lancet. Some tests can be used with oral fluid, obtained by swabbing an absorbent pad around the outer gums, adjacent to the teeth. 

The OraQuick Advance Rapid HIV-1/2 is, by far, the most widely used test which can test oral fluid samples. It is also one of the rapid tests which has been most extensively evaluated by independent researchers. Whilst accuracy has generally been reported to be very good, a few reports have identified problems including false positive results and limited sensitivity in people with recent infection. Such problems seem to have been associated in part with the use of oral fluid samples.

Dr Nitika Pant Pai of McGill University and colleagues conducted a systematic review and meta-analysis in order to identify and pull together all the studies that have been conducted on the use of OraQuick in clinical practice.

Twenty-four publications were identified, approximately half from the United States and half from African or Asian settings. Some of these studies were very large, and just under 40,000 samples were tested in total.

The key measures of accuracy in such studies are sensitivity (the percentage of results that are correctly positive when HIV is actually present) and specificity (the percentage of results that are correctly negative when HIV is not present).

The meta-analysis confirmed the very high sensitivity and specificity of OraQuick when testing blood samples.

The primary objective of the meta-analysis was to compare the test’s performance with oral fluid and blood samples. In seven studies which made a direct comparison of performance on these two types of sample, the pooled sensitivity with oral fluid was lower than with blood – a difference of around 2%. In contrast, the difference in pooled specificity was not so marked.

While the differences may be relatively small and potentially compensated by the high acceptability of testing with oral fluid, it is important to note the 95% confidence intervals when reading the following figures. Confidence intervals represent plausible estimates within which the ‘true’ figure may lie – for example, sensitivity could be as low as the lower figure given.

The researchers also pooled data from six studies which tested oral fluid only and eleven studies which tested one form or another of blood sample. As these involve different study populations, the comparison is somewhat less reliable, but similar results were found.

The differences in performance will have a particular impact when the test is used in populations in which HIV prevalence is relatively low. As with other tests, a few people who test will have a false positive result. In a setting where very few people have HIV, the majority of apparent positive results will in fact be incorrect. However, as the proportion of people with HIV being tested increases, the true positives start to outnumber false positives.

The authors found that when testing oral fluid in a population where prevalence is below 1%, the positive predictive value (the proportion of individuals with positive test results who are correctly diagnosed) was 88.6% (95% confidence interval 77.3 - 95.9%). On the other hand, in populations with HIV prevalence above 1%, positive predictive value with oral fluid was 98.6% (95% confidence interval 85.7 - 99.9%) and with blood 98.5% (93.1 - 99.8%).

These data derive from evaluations of OraQuick as used in clinical practice and are lower than the manufacturer’s claims. However almost all data from manufacturers are derived from assessments in carefully controlled laboratory settings.

The authors examined reports of false positive results in order to identify factors that could have caused them: errors in test performance and conduct of test (i.e. inaccurate specimen collection, gum swabbing more than once), and errors in the interpretation of results (interpreting weakly reactive lines). These were generally linked to inadequate training of staff or lapses in quality assurance.

However Nitika Pant Pai says that the key reason for poorer performance with oral fluid is that there are lower quantities of HIV antibodies in oral fluid than in whole blood. The quantities of antibodies in oral fluid are especially low after recent infection.

The authors suggest that people who think they may have recently been exposed to HIV should treat negative oral fluid results with caution. Similarly, people who are at higher risk of HIV infection who get a negative result should seek further testing.

On the other hand, if the test is used in low-prevalence settings, there will be a higher number of false positive results than with blood-based testing. As with any HIV test, an apparently positive result must be confirmed with supplementary testing.

Pant Pai N et al. Head-to-head comparison of accuracy of a rapid point-of-care HIV test with oral versus whole-blood specimens: a systematic review and meta-analysis. Lancet Infectious Diseases 12: 373-380, 2012. (Full text freely available).