
This article is dedicated to the memory of Prof. S.W.P. Mhlongo, without whom nothing reported below would be available to report, and in the hope that its publication might in some manner encourage the rapid return to excellent health of the very Hon. Manto Tshabalala-Msimang, Minister of Health of South Africa.
Seven years ago this spring, The President of South Africa, Mr. Thabo Mbeki, startled the world by very publicly announcing that he for one was unconvinced that sex, not poverty, was at the root of his nation's woes, and convened a panel of expert advisers to help him sort through the various inconsistencies concerning the "epidemic of HIV/AIDS" that was said to be ravaging his Beloved Country, just when it could finally be called his own, and that were so deeply troubling him.
It is perhaps the most personally satisfying distinction ever accorded me to have been invited to become a member of that Panel, which I serve on to this day at the pleasure of The President since it has never been officially dissolved. The fact that seven years later Mbeki is still unconvinced says all one really needs to know regarding the determinations he made after viewing (from several angles) recordings of all the proceedings of the meetings, and reading most, if not all, of the thousands of pages of internet discussion that went on between May and July of 2000, and for several years thereafter.
At the time, the clearest indication that the dissident advisers (out-numbered 2 to 1) had made chopped liver of the establishment representatives (some of whom were caught on tape napping during afternoon sessions) was the sudden appearance of the Durban Declaration, in the once august pages of Nature, in the week between the close of the formal deliberations and that year's international AIDS fiasco, held in the city of the same name. This "declaration", which perhaps marks the low point of all orthodox attempts to silence legitimate debate over two decades, has been throughly rebutted in this exhaustive, and extensively referenced, interlinear.
When the final panel discussion concluded, there was a large press conference at which Helene Gayle (then the director of the Afrika Korps of the CDC) introduced me to her "good friend", Charlene Hunter-Galt from CNN. I remember shaking her hand and saying, "Gee, you look just like you", before Helene and I were whisked to the dais to jointly announce the single, substantive accomplishment of the Panel.
It had been agreed upon by a unanimous vote of the working group assigned the task, that there were several, straightforward experimental ways to test critically the cornerstones of HIV/AIDS theory. First among these was the agreement to devise and implement a protocol to determine, accurately and for the first time, to what degree the well documented, cross reactions of HIV antibody tests with proteins derived from a wide range of microbial parasites common to South Africa, affected tests results reported by the laboratories authorized to carry them out.
But between the cup and the lip, is sometimes a long trip, and it took almost three years before the pilot study of a protocol, that itself took more than two years "to negotiate", was finally carried out by myself and Roberto P. Stock, a senior investigator at the Institute of Biotechnology of the Autonomous National University of Mexico, where I was, until very recently, a resident scholar.
We performed our experiments, in May of 2003, on sera drawn from the five, major clinics serving Shoshanguve, the Soweto of Pretoria, in the laboratory responsible for all HIV testing in the township, at the Medical University of South Africa (MEDUNSA) where Prof. Mhlongo was the Head of Family Practice. Our first surprises were discovering how few patients of any kind were to be found, how difficult it was for the nursing sisters to draw blood from those who presented on the few days we had to collect our samples, and how heavily lipídized many of the samples appeared.
The second set of surprises came when we discovered that it was impossible, given the way the antibody test that was used at MEDUNSA performed, to obtain an answer to our original question of whether preadsorption of sera to immobilized, TB antigens, could remove sufficient "reactive" antibody to convert a "positive" test result to a "negative".
I am tempted to explain exactly how badly these imported test kits perform, when used according to manufacturer's directions in an African setting, in language much less temperate than Roberto and I used in our report to the President. I will refrain from such temptation, however, and present below only those graphs showing assay variability, and some very deliberately, 'unemotional' text that goes along with them. The complete report as it was prepared for President Mbeki and a few others, and a set of supplemental material that fills in most of the lacunae in this brief condensation of almost four years effort, are located at the bottom of the article.

"B. Assay variability.
As the untreated samples were not tested in duplicate, we decided to evaluate whether the source of variation was in the preadsorption treatments or whether it was due to an inherent variation in the "test itself". To do this, a subset of untreated samples with readings below the positive control value for the assay (which, depending on plate-to-plate variation, is around 1) were re-tested in triplicate. The scatter plot on the left shows the outcome of the three tests for each individual serum sample. The bar graph on the right shows the average of the three determinations with the standard deviation.
The most important thing to note is that 7 of 17 samples (marked with asterisks) tested yield inconsistent results that in practice means supplying presumably quantitative and accurate information when in fact that is not possible [italics added]. This situation is worsened because in actual HIV antibody testing samples are never tested even in duplicate.
The second thing to notice is that although not giving inconsistent results (in terms of positivity or negativity, or crossing the cut-off), another 2 samples exhibit very high standard deviations.
These results, completely unexpected given the guidelines supplied by the manufacturer, imply that in order to quantitatively assay the cross-reactivity between presumed anti-HIV antibodies and antigens from other sources (M. tuberculosis in this case), we must first optimize the performance of the test to obtain results that are reproducible within acceptable bounds, ideally less than 5% variation from replicate to replicate -- see Recommendations (below) for details on how to address these problems experimentally.
Conclusions and Comment.
The variability of the HIV antibody test. The variability observed warrants
no extensive commentary save that, considering the importance for the life of
the patient, the outcome of the test is of unacceptable precision, at least when
done following the manufacturer's recommendations. The excessive variability in
replicate testing of the samples can be addressed experimentally by adding
additional steps in the procedure, beyond those claimed sufficient by the
manufacturer. These might include clarification of the sera (by centrifugation)
and possibly de-lipidization. As most of the samples gathered for this study
fall out of the linear range of the spectrophotometer (where significant drops
in antibody levels, if occurring, can be meaningfully measured), it may
be
necessary to normalize the samples by dilution prior to preadsorption to
values in the vicinity of the positive control supplied by the manufacturer of
the test kits." [Read
the entire report.]
Supplemental material:
Final Protocol for Validating HIV Antibody Testing in South Africa
Part of the discussions that led to the approval of the protocol above in November 2001.
Appendices to the Pilot Study: I, II, III, IV, V (Raw Data. 7.5 MB), VI.
A Powerpoint presentation of the background to, and results of the Pilot Study prepared by Dr. Stock.
A "lab rat" and his rusty assistant at the bench.
Harvey Bialy is the founding scientific editor of Nature Biotechnology, and the editor of "You Bet Your Life".
Let me first thank you for making this compilation available as you have. It provoked an almost all day discussion here at the Univ. of Fort Hare among myself and a number of my colleagues and our students.
We all agree that the title, however, is a bit disingenuous, as the text, which is an insider, historical introduction to the larger question of "AIDS and Africa" and its political ramifications, belies the extravagant headline in so far as it presents data that call into the most serious question only one antibody test, which may not even be used any longer at MEDUNSA.
We do understand that a title that read, "GIGO - or Why HIV Antibody Tests in South Africa May Be Worthless" is not nearly as catchy, but it started us to thinking of a larger study that I will do my best to implement in the next months (not years).
It seems fairly certain that the reason the test that failed so miserably did so was because it was not calibrated or tested for performance against typical, impoverised African blood, which is biochemically quite different from the blood likely to be drawn from an average European.
It is disturbing to say the least that no one has ever systematically addressed this question, which is so obviously important.
What we therefore are proposing to do is gather all of the 7 or 10 antibody test kits that are currently in use in SA, and test them, as you did, for internal quality control against blood from (a) reasonably well nourished and healthy folks like us, and (b) blood drawn on different hospital wards, and in the villages where poverty, lack of clean water and crowded living conditions are perennial problems.
We are not interested in "HIV" status per se, but rather in seeing how reliably the tests perform on these different samples.
We will of course, since we are not racially biased, include as many of our colleagues and countrymen of European extraction as we can, realizing that not too many will be found in the infectious disease hospital wards and in the villages.
If we do find the kinds of horrendous problems that you did in the Pilot Study apply to most or all of the tests, we will then proceed in a systematic way, as you described, to see if we can improve any of them to the point of biochemical reproducibility.
But until such time, it is perfectly clear from your own results, that any African given a positive antibody diagnosis should demand that the test used to give him or her a death sentence is as internally reliable as the one used to tell them their blood sugar levels.
Professor Anthony I. Okoh
Microbiology Unit
University of Fort Hare
Alice, South Africa
Posted by: Anthony I. Okoh | February 27, 2007 at 10:13 AM