Current HIV incidence tests do not work in Africa

Current assays which aim to distinguish between recent HIV infections and chronic ones, and therefore establish incidence (the annual rate of new infections), do not work in a west African context, a study from the Côte d’Ivoire finds.

In nearly all scenarios, the researchers calculated, the tests would tend to grossly overestimate HIV incidence and underestimate declines in incidence.

The authors say that their study supports a recent warning issued by UNAIDS and the WHO that the BED assay, one of those tested, should not be used as a surveillance tool in Africa until further research has been done and incidence assays are adapted for African use.

Establishing true incidence is crucially important for the global response to AIDS for two reasons. Changes in incidence predict, in conjunction with changes in the death rate, how prevalence (the total number of people with HIV) will change and therefore the future human and economic costs of HIV. Changes in HIV incidence are also the ‘gold standard’ when it comes to measuring the efficacy of HIV prevention programmes.

Incidence, however, is challenging to measure in a chronic, slowly-developing disease like HIV/AIDS because of the difficulty of distinguishing between recent and chronic infections, since newly-diagnosed people may have been infected from anything between six weeks previous to their test to well over a decade ago.

The gold standard in measures of incidence is the prospective cohort study; researchers follow a large group of people in a high-risk population or area, test them for HIV regularly, and measure how many seroconvert. This has been done in Africa (in Uganda and Rwanda, for instance), but is expensive and labour-intensive.

When it has been done, however, it has yielded surprises, such as the finding (Wawer) that incidence in Uganda had not declined during the 1990s despite significant declines in prevalence. This underlines the importance of accurately establishing incidence.

A number of assays have been developed to get round this problem, and are used as research and epidemiological tools in the developed world. They all rely on the same idea; the body’s immune response to HIV develops quite slowly, and HIV antibody titres (concentrations) do not reach their final level till six months to a year after infection.

Incidence assays measure this rising antibody response in different ways and by comparing it with the titre typical of people in long-term infection, can establish with reasonable accuracy whether an infection is recent or not.

The first assay to do this was the so-called ‘detuned’ or STARHS (Serological Testing Algorithm for Recent HIV Seroconversion) assay which simply compared results from a standard HIV antibody test to one that had been deliberately made less sensitive so it only gave a positive result for infections typically more than six months old.

This now has successors in the shape of assays which directly measure the level and potency of different aspects of the HIV antibody response.

In the current study, four incidence assays were tested against a panel of 135 samples from the blood of people in a Côte d’Ivoire cohort who, as former blood donors, had quite precise dates for their HIV infection established. Of the 135, 26 had recent infections (that is, within the last six months), 15 were from people with clinical AIDS and the other 94 were from people with chronic HIV infection, 27 of whom had been infected from six months to a year ago.

These samples were tested against four incidence assays:

• The BED Assay, which measures a part of the antibody response called immunoglobulin G (IgG), and which has superseded the STARHS assay as the incidence assay used by the US Centers for Disease Cotnrol.
• The Vironostika assay, developed by BioMerieux, which measures the intensity of another part of the antibody response.
• The IDE-V3 assay, a similar ‘in house’ assay developed by the virology laboratory in Tours, France.
• The Avidity assay, developed by Abbott. This works in a different way by measuring the antibody response’s robustness against an agent (guanidine) that disrupts it.

The sensitivity and specificity of the four assays were measured against the individual Côte d’Ivoire samples.

In this context, the sensitivity of the assays was their ability to confirm that recent infections were not chronic ones, and the specificity of the assays was their ability to confirm that chronic infections were not recent ones.

This varied widely. While the Avidity assay, for instance, was 100% sensitive (i.e. it produced no ‘false negatives’, which means it did not class any recent infection as a chronic one) it was only 49.5% specific (i.e. half of its positive results were ‘false positives’, in that half of the infections it classed as recent were in fact chronic.) It would therefore, in these African samples, grossly overestimate the proportion that were recent infections.

Conversely the IDE-V3 assay was 96.3% specific (i.e. less than 4% of the infections it classed as recent ‘false positive’ were in fact chronic) but only 42.3% sensitive (i.e. more than half of the recent infections were misdiagnosed as chronic ones and were ‘false negatives’). It would therefore tend to grossly underestimate the proportion of infections that were recent.

The assay with the ‘best fit’ was the Vironostika assay, which was 100% sensitive and 80% specific, so 20% of its ‘recent infections’ were false-positive and were in fact chronic ones.

The majority of samples misdiagnosed as less than six months old were from people infected between six months and a year ago. When this was taken into account, the specificity of the Vironostika assay, for instance, went up to 89%.

In a situation of very high incidence, it might not matter if individual chronic infections were misdiagnosed as incident ones as this might be balanced by a similar number of incident infections being misdiagnosed as chronic. However in the real situation, where roughly 10% of the Côte d’Ivoire population has HIV and incidence is 1% a year, one ‘false positive’ result has considerably more weight, i.e. biases the true result more, than one ‘false negative’ result.

As a result, the researchers calculated, in the Côte d’Ivoire situation three of the four assays would overestimate true incidence five- to ten-fold. Only the IDE-V3 assay, because of its poor sensitivity, would yield anything like the true incidence, and even so would overestimate it by 20%. In a lower-prevalence and lower-incidence population it, and all the other assays, would overestimate true incidence even more.

Consequently, the researchers conclude, “none of the four methods could currently be used to estimate HIV-1 incidence routinely in Côte d’Ivoire.”

Exactly why the incidence assays performed so poorly isn’t clear. The researchers said that it may be because they were developed using HIV subtype B, and the predominant subtype in Côte d’Ivoire, accounting for all but one sample, is a recombinant, CRF02_AG, a virus consisting of a mixture of subtypes A and G. Subtle differences in the body’s antibody response to different subtypes may account for some of the difference.

Most of the difference, however, may be accounted for by a higher rate of immune activation and therefore of antibody titres in the African population. Total IgG levels, for instance, are known to be raised in the African population, and especially amongst people with malnutrition.

The authors say that changing the criterion for recent infection from six months to a year after infection would improve the specificity of the tests, but add that further research on a variety of HIV subtypes is needed, especially in light of the small sample size in the study.