Nice Graph – No Data

Today, on a NIAID website entitled “The Relationship between the Human Immunodeficiency Virus and the Acquired Immunodeficiency Syndrome” you will find the following figure:

This figure (Figure 4) is described as showing the “Typical course of HIV infection” and references a paper identified as “Pantaleo, 1993a” which is described as: Pantaleo G, Graziosi C, Fauci AS. The immunopathogenesis of human immunodeficiency virus infection. N Engl J Med 1993a;328(5):327-35. In other words, NIAID did not collect the data themselves, but are relying on the data produced by other scientists. And, having provided readers with a reference this appears to be a legitimate method used by scholars in all fields of study. The reference is provided so that we can verify the source and see if we agree with their methodology and the interpretation of the anonymous NIAID authors. So that is where we must go…

Pantaleo’s 1993 Figure

If we go to the cited Pantaleo paper, we see a graph that is similar, but not identical:

Let’s play “Spot the Differences” (I love games). What I found different in the NIAID figure compared with this figure from Pantaleo, 1993 were:

• Addition of an HIV RNA axis (and data for this axis shown with triangles).
• Replacement of the word “Possible” with the symbols +/- in front of “acute HIV syndrome”.
• Replacement of “Plasma Viremia Titer” with “Culturable Plasma Virema (dilutional titer).

The similarities are more numerous, showing that these two figures do have a common ancestor, these include:

• Identical horizontal axis showing time in weeks up to 12, a break, and then time in years up to 11.
• The vertical axis labeled on the left has a very similar title and, more importantly, data with exactly the same values.
• The vertical axis labeled on the right has, apart from the different name, has the same values on the scale and data points at identical coordinates.

What appears to have happened is that NIAID has referenced the wrong paper, we have found a much more similar figure in another paper (we will get to this). But this raises a problem with the NIAID website document. Unlike a scientific publication, the NIAID website has a number of undesirable characteristics:

• It is anonymous, so we do not know who to contact (and there is no provision on the web page for any kind of contact). So it is very difficult for anyone to report an error.
• If an error was reported, there is no guarantee that it would be acted on.
• If there is an editor, their identity has not been made public.
• There is no “Letters” section where a complaint could be lodged.
• There is, in fact, no contact information on the NIAID page.
• If a change was made it might not be highlighted, so people coming back to the page might not realize that a correction had been made.
• This document is not listed on PubMed or other indexes where a correction or erratum could be listed along with the citation.

In summary, the NIAID paper is not on the scientific record where there is a moral obligation to ensure the integrity of published work, including correcting it if errors or fraud are discovered. Admittedly this does not always happen even with science, but a process is at least in place, and errata are often published. Corrections due to fraud are much less likely to occur, but that is a whole different topic. The point is that the scientific record does have processes in place to admit corrections. Web pages could have a process in place, but the NIAID web page does not. It is ironic that people who are horrified that Mbeki downloaded information on HIV and AIDS from the internet are sanguine about relying on the NIAID documents as the response to dissidents rather than the scientific record.

But where did Pantaleo get the data from this figure from? Well, from an earlier (1991) paper by one of his co-authors, Anthony Fauci, head of NIAID. We will get to that paper in time, but first we need to locate the paper with the correct figure. For that we need to go to another paper by Anthony Fauci of NIAID with Pantaleo as one of the co-authors (are you starting to see a pattern yet?).

Fauci’s 1996 Figure – The Right Stuff

Fauci’s 1996 paper (Fauci AS et al. Immunopathogenic mechanisms of HIV infection. Ann Intern Med. 1996 Apr 1; 124(7): 654–63.) has the identical figure as the NIAID website. You can play “Spot the Differences” again, but this time you will not find any. Even the shapes used to identify points on the graph are identical. The axes have the same units and titles and the annotations are the same. Clearly NIAID must have got their figure from Fauci’s paper which was not documenting scientific research, but the outcome of a 1994 conference of which Anthony Fauci was moderator and the other panelists (including Pantaleo) were the co-authors.

The question you must be asking is, since this is not a research paper, “Where did Fauci draw this graph from?”. Well, now we have a bit of a problem. Because Fauci references the Pantaleo 1993 paper which we know does not have the same figure!

Fauci does in fact note that he “adapted” the graph from Pantaleo, which must have meant adding the third data series and vertical axis, among other changes. But if this was based on real data, where did the data come from? Surely, if Fauci had access to additional data he would have referenced that source, but he didn’t. Perhaps a clue is that Fauci labelled the graph “Typical course of human immunodeficiency virus (HIV) infection”. But surely the data is not imaginary! Does Fauci have perfect recall of the viral load of thousands of AIDS patients in his head so that he could just do a quick average and throw them on the graph? This would still not be right. The data should have been reported from scientific studies, and those should have been reported, even if Fauci was an idiot savant and could keep thousands of data points safely in his head.

Perhaps it is worth considering what you would do if you were preparing an illustrative graph, one that illustrated approximate trends without being based on real data. Without asking how you can map out even approximate trends without some data (and if you have data, why not somehow average it to produce a graph?) would it be reasonable to display axes with precise quantities when the data were not precise? Would it be reasonable to graph through data points with small irregularities from a smooth curve, simulating what happens with real data, or would it be more honest to graph smooth curves without data points to emphasize the point that this is just an approximation?

If we go back to the Panteleo paper (with Fauci as a co-author) we find a reference to a 1991 paper by Fauci (with Pantaleo as a co-author).

Fauci’s 1991 Figure – Almost Pantaleo

Fauci’s 1991 paper (Fauci AS et al. NIH conference. Immunopathogenic mechanisms in human immunodeficiency virus (HIV) infection. Ann Intern Med. 1991 Apr 15; 114(8): 678–93) turns out to be an earlier version of his 1996 paper. It is also derived from a National Institute of Health conference with Fauci as moderator and Pantaleo as one of the panelists and co-authors. The graph, upon first glance, does look very much like the Pantaleo 1993 graph.

First appearances can be deceiving. Although the curves on this graph are similar to Pantaleo’s 1993 paper, there are a number of significant differences:

• The left vertical axis has a slightly different name with identical numerical values, but the right axis has changed from “p24 Antigenemia (pg/ml)” in 1991 with a scale from 0 to 1500 changed to “Plasma Viremia Titer” ranging from 0 to 1:512.
• The data points for both vertical axes are in slightly different positions.
• Most notably while p24 Antigenemia levels off after opportunistic infections start in the 1991 graph, Plasma Viremia Titer continues to rise on the 1993 graph.
• The first peak of p24 Antigenemia in the 1991 graph is only about half the level reached by the time of death, but in the 1993 graph the first peak is equal to the level reached at the time of death.
• The initial CD4 drop in both graphs is to around 500 cells, but in the 1991 graph it is shown rebounding to about 800 by week 9, while in the 1993 graph it is only about 650 by week 12.
• The horizontal (time) axis has changed from 0-9 Weeks before the cut point in 1991 to 0-12 weeks in 1993. And, after the cut point, the axis runs from 1 to 10 years in 1991 but 1 to 11 years in 1993. AIDS victims were given one more year of life.
• The annotations are considerably different as well. From left to right, “Initial Infection” in 1991 is changed to “Primary Infection” in 1993. An annotation at the initial peak in the level of virus stating “Possible acute HIV syndrome. Wide dissemination of virus. Seeding of lymphoid organs.” only appears in the 1993 version. Similarly, the annotation “Clinical Latency” that extends from 9 weeks to almost 8 years in 1993 does not appear in the 1991 graph. Finally, “Early Symptoms” on the 1991 graph at year 5 is replaced by “Constitutional symptoms” at year 8; “PCP” at year 8 is replaced by “Opportunistic diseases” at year 9; and “Death” occurs at year 10 in 1991 and 11 in 1993 (due to the miracles of AZT and other early nucleoside analogs?
• Fauci is author in 1991, with Pantaleo as co-author. In 1993, Pantaleo is author, with Fauci as co-author.

This paper leads us to a bit of a dead end because there simply is no reference for the source of this data. There are two 1989 New England Journal of Medicine papers referenced in the vicinity of references to the figure, but neither of them provide either a similar graph or raw material that could be made into this graph, at least not without a great deal of imagination. Neither of them provide a time series that would be necessary to construct any of the graphs discussed so far.

Haseltine, 1988

The trail might be undocumented, but it hasn’t necessarily gone cold. Just because there are no bank account transactions in a fraud or bribery case doesn’t rule out the possibility that cash was changing hands in parking garages in plain paper bags. Similarly there is a graph from 1988 that is of great interest (redrawn from a photograph in John Lauritsen’s classic AZT: Poison by Prescription, p. 160):

John Lauritsen describes a scene at a 1988 AmFar sponsored conference (designed to discredit Dr. Peter Duesberg in the opinion of many people) where William Haseltine (a man with strong financial connections to a HIV testing corporation, Cambridge Biosciences) put a slide with this graph up. He claimed that it showed that Duesberg was wrong, that there was evidence that virus titer (the solid black line) peaks early, and then after a latency period rises again, this time causing fatal illness. The presence of the virus, he implied, is proof that it is what is doing the killing. This chart was immediately criticized by Peter Duesberg upon presentation for having no units. Haseltine’s colleague Robert Redfield confided afterwards that the chart was just illustrative, but Haseltine had introduced it as if it represented real data, proof that free virus was present later in infection.

One of the other audience members present was Anthony Fauci. Perhaps he took Duesberg’s criticism seriously. Perhaps he said to himself “If Duesberg wants a chart with units, he will get a chart with units!”. And, lo and behold, Fauci did produce a series of graphs which, it appears, are not based on data at all – but do have axes with units. One would think that such powerful and convincing graphics, if representing reality, should be easy to generate from real data, after all enough time has passed. But perhaps that is the problem, that none of these graphs represent reality. The chart was so popular and powerful and so convincing to people who want so badly to believe, that few asked questions about their validity so it didn’t matter that it was based on the imagination of a couple of people.

And, not more than a couple of people. The chart evolved solely within the control of two people – Pantaleo and Fauci – and then ended up on the website of an organization controlled by Fauci. By NIAID citing Pantaleo (when it should have cited Fauci) and Pantaleo citing Fauci and Fauci not citing Haseltine, outsiders might not realize the incestuous nature of the development of this graph.

© Copyright 2007 – Alberta Reappraising AIDS Society