Re: Re: The "HIV" antibody tests are not diagnostic 21 October 2004
Previous Rapid Response Next Rapid Response Top
Nicholas Bennett,
Infectious Disease Postdoc/Clinician
Department of Pediatrics, University Hospital, Syracuse NY

Send response to journal:
Re: Re: Re: The "HIV" antibody tests are not diagnostic

I'll try to answer Mark Bartlett's questions, but my experience is necessarily limited to that of the developed world. His clarifications based on my last response make sense, so I'll try to be more specific.

From what I can gather, his main point is that: if these SA infection rates are true, then (a) where did some of the people get HIV from and (b) why aren't more of them at the stage of AIDS.

I agree that the sex distribution is very different in the African cohorts that in the developed world. This is attributed largely to differences in mode of transmission (how many women do gay men have sex with, for example!).

I would add that there might well be a higher rate of false-positive ELISA in pregnant women, probably due to antibodies to endogenous retroviruses (which are produced, albeit at relatively low levels). The highest figure I've seen quoted in 1.6% false positive, which I believe is by ELISA alone. Far lower rates have been published for ELISA/WB testing. In non-pregnant women I think the false-positive rate is probably equal to men.

Without knowing the trial methodology I can't comment on the diagnosis stigma: blinded, anonymous tests would not have that problem (unless the participants didn't believe the investigators). Without knowing asymptomatic status it's impossible to comment at all: symptoms might well be relatively minor (candidiasis) or common (TB) and overlooked. As such, any question like "Why do they not have AIDS?" is extremely difficult to answer, since some might well have AIDS! Further studies are required in order to formulate an opinion on the situation, which I'll try to pull up now. A good place to start is the study in question [1], which I think I have a full-text copy of (all 133 pages!).

For starters, the ELISA used was a specific rapid test with 99% sensitivity and specificity called Orasure. This is one of the newer kits I sometimes refer to when talking about African AIDS, developed specifically because of the need of a rapid but accurate test. The tests were not performed in the field, but at three experienced laboratories at Universities, examined by the Medical Research Council (MRC) one of which is approved by the CDC in Atlanta. They also undertook an internal MRC- run quality control audit prior to the study. I would therefore put a fair bit of weight on the test results.

They state that:

"There is limited evidence that PLWAs (people living with AIDS) or those at high risk of HIV infection are less likely to participate in surveys involving HIV testing."

But I would pull out the caveat that absence of evidence is not evidence of absence. Overall though, it is at least encouraging.

The rates of 7% in the older age groups (55+) aren't a problem if one considers that people don't become celibate once they reach 40 years of age :o) However, one is assuming that they acquired the infection relatively recently, or as Mr Bartlett states are LTNPs. They in fact directly address the issue of cohort aging of people moving up from the 30 -40 year age groups, and state that as a cross-sectional study they cannot possibly unravel this concept.

I'm interested where the figures of 96000 infections versus 570 MTCT in the 2-14 white age group came from. The dirty needle explanation is one which is specifically mentioned, in relation to vaccination programs. Throwing some figures in the mix, the 43.2% response rate for that group is considered poor, and the results therefore should be interpreted with great caution. 11.3% of the 145 white children in the study would be 16.385 (makes you wonder about the data actually since neither 16 nor 17 is 11.3%), which is around 0.2% of the relevant study population of 8428. Extrapolated to the 41 million or so people in the weighted study population would be 80035 HIV+ white children. In 2003 96228 children were estimated to be infected by their mothers, of whom 5946 (6.18%) can be expected to be white, based purely on how many kids overall were white.

That's not too bad: 6000 HIV+ kids a year and the child age group runs from 2 up to 14 years old (that would account for the majority of the 80,000 HIV+ white kids even assuming deaths). I dunno where the 570 came from, but if Mark Bartlett has some math then it'd be worth seeing.

I don't think anyone has said that AIDS was in the Western world since the 1930's. I have heard estimates of the species crossover occuring in the 1950s with the upper limit of accuracy being in the 1930s, but that would have occured in Central Africa. Mr Bartlett is quite right that if HIV was introduced into the heterosexual population then prostitutes would have been a prime target for the era prior to AIDS being recognised. However, all the evidence suggests that it was introduced into the homosexual population, relatively recently in the 1970s.

I think there might well be evidence that bi-directional transmission is more likely in Africa, but right now (after spending the entire evening ploughing through 133 pages of the Mandela report!) I don't have the energy to hunt anything down. I do know that among the porn industry, HIV testing is frequent and routine: they are well aware of the risks of the industry.

As regards the questions about clinical suspicion of HIV infection, I honestly cannot easily say. I know that HIV infection is more common in SA among whites than in the developed world, and higher again in blacks. Additionally, the ELISA used in this situation seems remarkably good (high specificity) so I might ironically trust that more than a bog-standard ELISA in use as a screening tool in the developed world.

I think more important than the 30,000ft view of the epidemic based purely on infection rates, would be a proper analysis of the route of infection. Notoriously difficult in this kind of study I imagine, if not impossible. If say 6-7% of 75 year olds came back as HIV+ but zero percent reported a high-risk activity, I'd be very suspicious of those rates. Even assuming that _some_ would lie about risky activities, that kind of result would be hard to explain. I don't think any such thing has ever been done.

I'm interested in Mark Bartlett's views on the fact that HIV infection rates show a neat temporal relationship in South Africa [2]. If the ELISAs were inherently bad, and in fact they've got better (more specific, less false positives) over the years, why has the measured rate steadily increased in antenatal testing? By the dissident criteria one would expect the false positive rate to remain stable, since it would not behave like an infectious epidemic.

I appreciate that may not have answered all the questions Mr Bartlett had, but I think I went some way to explaining a few things.

Nick Bennett


1. Nelson Mandela/HSRC Study of HIV/AIDS 2002


Competing interests: None declared