Saturday, 19 February 2011

AIDS Denialism: Deadly Ignorance Part III: Fuzzy Math and Distorted Reality

In the previous two part of this series (see Part I and Part II), I tackled the forerunners of the AIDS "skeptics" community, and addressed how Koch's Postulates support HIV as the cause of AIDS. In this part, I will assess how some of the deniers distort studies and misinterpret math to support their dangerous ideas.

If you spend any time on AIDS denialist websites, there are likely a few statistics that will show up repeatedly. This is because it is a common practice for AIDS denial websites to copy and paste the same arguments verbatim from one website to the other, no matter how old the information might be, or if the arguments have been debunked already. One such statistic that the deniers constantly toss around is that half – or more - of Africans who qualify as having AIDS test as HIV-negative1. Specifically, they routinely cite three specific studies2:

122 patients with "AIDS": 69% test HIV-negative (Brindle, 1993)
227 patients with "AIDS": 59% test HIV-negative (Hishida, 1992)
913 patients with "AIDS": 71% test HIV-negative (Songok, 1994)

But if one were to take a critical look at each of these studies, they do not support the claims of the AIDS "skeptics" in the least. Let's take a look at each of them.

The first paper by Brindle et al. is entitled "Quantative bacillary response to treatment in HIV-associated pulmonary tuberculosis"3. In this study, the authors looked at 122 patients with "culture-proven pulmonary tuberculosis". These patients were divided into two groups, and each group had their tuberculosis treated with one of two different treatments. They then looked at how the HIV-positive individuals in each group took to their treatments compared to the HIV-negative individuals. The purpose of this was to determine if patients with HIV would respond to chemotherapeutic treatments differently than those without HIV (and they found that there was very little difference).

It should be noted that the 122 patients in the study were NOT patients with AIDS. They were patients with pulmonary tuberculosis. This is a symptom of AIDS, but of course, pulmonary tuberculosis can arise in humans in a variety of ways; it is not exclusive to AIDS patients. Of the 122 patients, only some of them had AIDS. So where did the denialists get their numbers? Of the patients that were given the first treatment, 17 were HIV-positive and 57 were HIV-negative. Of those that were given then second treatment, 20 were HIV-positive and 35 were HIV-negative. In total, 37 of the 122 patients were HIV-positive, so 85 out of 122, or 69%, were HIV-negative. But this is not the number of AIDS patients who were HIV-negative; rather it is the number of tuberculosis patients who were HIV-negative. This is a different thing altogether. The AIDS "skeptics" misrepresent this number to support their case, when in fact, it does not.

The second paper is "Clinically diagnosed AIDS cases without evident association with HIV-1 and 2 infections in Ghana" but Hishida et al4.The authors' aim in this paper was simple: to test patients suspected of having AIDS for HIV. It is important to remember that the patients they looked at were suspected cases of AIDS – the authors state "CD4 cells were not counted [in these patients] because of insufficient facilities". They looked at 227 cases and found the following results : 48 were positive for HIV-1, 17 were positive for HIV-2, 11 were positive for both stains, and 16 were of indeterminate HIV status. The remaining 135 patients were determined to be HIV seronegative. Those 135 of 227 patients amounts to the 59% the AIDS denialists toss about. So at a superficial glance at the numbers might be seen as giving support to the denialist arguments.

But the denialists apparently didn't read any further than the raw numbers, because the authors continue: the 135 seronegative samples were from patients that had "weight loss, prolonged diarrhea, chronic fever" – so is this AIDS? The authors didn't think so. They state "We believe that many patients of this group were perhaps improperly diagnosed or had other unidentified diseases…[t]he existence of other agents causing AIDS-like syndromes might be possible for these so-called HIV-negative cases." In other words, it's likely the HIV-negative cases were from patients who did not have AIDS to start with – remember that these were samples from suspected AIDS cases, and not confirmed cases. The denialists take a quick look at the numbers, point out the 59% HIV-negative cases and ask "How could this be?" without bothering to read further and realize that the authors have answered that very question!

I also feel that I should point out that the authors used serological techniques to detect the presence of HIV antibodies in the samples, rather than using more sensitive PCR techniques. I am confident that if this study was repeated today, using PCR and improved diagnostic techniques, the suspected AIDS cases that test HIV-negative would be quite fewer.

The third study the denialists refer to is Songok et al's paper "Low Prevalence of Human T-Lymphotrophic Virus Type 1 (HTLV-1) in HIV patients in Kenya"5. They claim that this study showed a huge 71% of AIDS patients were HIV-negative; such a bold claim better have a wealth of evidence supporting it. The first thing one might notice when reading it, however, is that it isn't a peer reviewed research study; it's just a letter to the editor. Nevertheless, the authors report on some original research they were doing: they wanted to check to see the prevalence of HTLV-1 infections in HIV infected individuals. They did this by collecting 913 samples from suspected AIDS cases in Kenya. Again – remember that these are not confirmed cases of AIDS. Using these samples, they tested for HIV-1 using both ELISA and western blots. They reported that 265 (or 29%) of these tested positive for HIV-1 on both tests. This is where the denialists get their statistic. However, this number represents the samples that were positive on both the ELISA and the western blot – it excludes those samples that tested positive on one or the other tests. They also tested for HIV-1 alone; they did not test for HIV-2, and this undoubtedly reduced the number of positive results. And again, the researchers did not use more sensitive PCR techniques, so it is entirely plausible that many positive cases were missed.

What all three papers have in common is that the estimates of HIV-positive individuals are underestimated. All three papers were written before sensitive PCR-based methods for detecting HIV were developed and put into common use, and before doctors were able to accurately diagnose AIDS and differentiate it from other immunodeficiency syndromes. And what deniers need to realize that it is impossible to obtain a 100% detection rate. Multiple factors exist that prevent the presence of HIV from being detected 100% of the time. If the AIDS "skeptics" really wish to show that HIV does not cause AIDS, then perhaps they should find a study to cite that isn't two decades old.

So as usual, critical examination of denialist claims shows that they don't have a leg to stand on. Better luck next time, guys. Come back when you have recent research that you can cite without twisting the statistics into misleading conclusions.


  1. See for an example.
  2. These three statistics were also stated by the Youtube user who prompted this series on debunking AIDS denial nonsense.
  3. Brindle et al. Quantative bacillary response to treatment in HIV-associated pulmonary tuberculosis. 1993. American Review of Respiratory Disease
  4. Hishida et al. Clinically diagnosed AIDS cases without evident association with HIV-1 and 2 infections in Ghana. 1992. The Lancet
  5. Songok et al. Low Prevalence of Human T-Lymphotrophic Virus Type 1 (HTLV-1) in HIV patients in Kenya. 1994. Journal of Acquired Immune Deficiency Syndromes


Snout said...

Okay, a few points:

1. I haven't been able to access the study, but is completely unsurprising that there should be a "Low Prevalence of Human T-Lymphotrophic Virus Type 1 (HTLV-1) in HIV patients in Kenya". HTLV-1 has nothing to do with AIDS. It's a completely different retrovirus from the two that cause AIDS in humans (HIV-1 and HIV-2), which were briefly known during the mid 1980s as HTLV-3.

2. The conclusions of Brindle et al (and I'm only going on the abstract here) seem to be that HIV positive TB patients did as well in the first month as HIV negative TB patients in the first month of TB treatment in their study. This has nothing to do with the question of whether African AIDS patients have HIV or not.

3. The Hishida et al paper is miscited. The correct title is "Clinically diagnosed AIDS cases without evident association with HIV type 1 and 2 infections in Ghana." The miscitation "with HIV-1 and 2" is a mutation which occurred on a denialist website probably many years ago (such mutations commonly originate on the virusmyth site), and has been extensively copy-pasted on the net by subsequent denialist websites.

Hishida et al, in my understanding, applied HIV testing to examine the specificity of the 1985 World Health Organization AIDS surveillance case definition, also called the "Bangui definition", used for a couple of years in the mid to late 80s in areas so resource poor that basic HIV testing and diagnostic methods for most AIDS defining opportunistic diseases were impractical. The purpose of the Bangui epidemiological surveillance definition was to try to get some preliminary estimates of AIDS prevalence and incidence in such areas, not to inform treatment of AIDS.

Hishida et al and two other studies found that the specificity of the Bangui definition was fairly poor in determining late stage HIV/AIDS when compared with HIV testing (and more rigorous diagnosis of AIDS defining opportunistic diseases). It was also fairly insensitive - many people with late stage HIV/AIDS by more rigorous criteria did not meet the "Bangui" surveillance definition. However given the resource limitations, "Bangui" was the best epidemiologists could do for the times and places. Any examination of epidemiological data based on this surveillance definition should bear this in mind.

C.W.G.K said...

Thanks for the comments, Snout (sorry it took a while for your comment to be posted; it got caught in my spam filter for some reason).

I was also not surprised that the Songok paper showed a low prevalence of HTLV-1 infections in HIV patients. The paper had nothing to do with AIDS, really, but denialists will cling to whatever papers they can, regardless to the purpose of the study.

The same can be said for the Brindle paper. The authors were uninterested in the HIV/AIDS link and were only interested in treating TB. But again, the deniers distort studies to serve their purpose.

As for the Hishida paper, I have seen it miscited multiple times, though I have cited it correctly in my post. However it is cited, though, it still does not support the denialist claims.

What all three papers do show is that denialists rarely make an attempt to understand primary research, and will distort whatever papers they can to support their cause.