Detecting primary HIV infection in Africa without viral load test is possible

US and African researchers have developed an algorithm, based on rapid test results, symptoms and risk behaviours, that makes it possible to accurately detect acute HIV infection without widespread use of HIV RNA assays. Identifying acute infection has significant implications for curbing the spread of HIV infection, particularly in resource-poor settings. The findings were published the October edition of AIDS.

Acute or primary HIV infection is the three- to four-week period after infection but before seroconversion. Standard HIV antibody tests yield negative results during this time; only more expensive and specialised HIV RNA and p24 antigen tests can identify people with primary infection.

Such tests are not routinely used even in developed countries, and are prohibitively expensive for resource-poor settings in the developing world. However, people are extremely infectious during this period due to very high HIV viral load, as well as more likely to transmit infection due to a lack of knowledge of their status.

Researchers from the University of North Carolina and Kamuzu Central Hospital, Malawi, therefore set about developing an algorithm that could be used to screen for people most likely to have primary HIV infection, for targeted HIV RNA screening. A cross-sectional study was conducted on 1,448 adults presenting at the outpatient STI clinic of Kamuzu Central Hospital, Lilongwe, Malawi, between February 2003 and October 2004. Other than age of 18 years or older, the only requirements were informed consent, willingness to return for follow-up, and no prior exposure to antiretrovirals.

Participants were screened using two rapid HIV antibody tests (Abbott's Determine and Trinity Biotech's Unigold); 588 (41%) were classified as having established (chronic) HIV infection and were excluded from the remainder of the study. HIV RNA assays were run on blood samples from the remaining 860. (To reduce the total number of tests needed, samples were pooled at first; further tests were run on individual samples making up any pools testing positive for HIV RNA. See also this earlier aidsmap.com news report. These 860 patients had a median age between 24 and 25 years and were 69% male

Participants were classified as HIV-negative if the rapid tests were both negative or were discordant (one positive and one negative) with no detectable HIV RNA. (See also this < href="/news/jan-2007/one-forty-negative-hiv-tests-malawi-may-mask-acute-infection-need-more-sensitive" > report on aidsmap.com)

A Western blot test (Bio-Rad Laboratories) was run on all individuals with discordant rapid tests and detectable HIV RNA. Participants were classified as having acute HIV infection if they had detectable HIV RNA and either both negative rapid tests, or discordant rapid tests and a negative, indeterminate, or weakly positive Western blot. Acute infection was confirmed by repeat rapid and Western blot tests at intervals up to 16 weeks after baseline.

Of the 860 participants without established infection, 21 (2.4%) were found to have acute infection, and 839 (97.6%) were HIV-negative. Variables from self-reported patient demographics, sexual and medical history, symptoms, and physical exam findings were treated as potential predictors of acute infection. Those with the highest unadjusted odds ratios were combined into several potential multivariate regression models, which were then compared to find the strongest overall predictive model.

Discordant rapid test results were extremely predictive of acute infection (odds ratio [OR] 29.5, 95% confidence interval [CI] 8.56 to 92.48), as were genital ulcer disease (OR 7.03, 95% CI 2.53 to 22.35) and diarrhea persisting at least a month (OR 5.57, 95% CI 1.29 to 18.31). More than one sexual partner in the past two months, fever, and body aches (symptoms persisting for at least one month), were the next strongest predictors; the resulting model combined these six into a single predictive formula.

The formula uses a weighting system which assigns a given individual a score of 4 for discordant rapid tests, 2 for each of genital ulcer disease and persistent diarrhea, and 1 for each of: fever, body ache, and more than one recent sexual partner. (The scores assigned to each variable were the logarithms of the adjusted odds ratios in the final combined model, rounded off to integer values.)

A resultant risk score of 2 or greater was found to be 92.5% sensitive and 90.5% specific in predicting acute HIV infection. The net result was that this predictive formula, or algorithm, could be used to identify 95.2% of all actual cases of acute HIV infection, while performing resource-intensive HIV RNA or p24 antigen tests on only 40.9% of presenting individuals. The performance of this algorithm "compares favourably" with similar algorithms developed in other settings for similar purposes.

The researchers concluded that "risk score algorithms for identifying [acute HIV infection] in high-risk, resource-poor settings could be powerful tools in HIV prevention and control", making acute-infection screening more feasible by guiding the efficient use of RNA or p24 tests.

This particular analysis was based on specific factors in both the population and the assays used; the report cautions that this specific algorithm "is intended for use with similar HIV tests in sub-Saharan … clinics, where [clinical factors] are likely to be comparable with those in the study setting," and that "generalisability to dissimilar settings is likely to be limited".