Henry Bauer’s harebrained disproofs III: “The age distribution of positive HIV tests superposes on the age distribution of deaths”

HERE HENRY IS CLAIMING that HIV diagnoses and HIV/AIDS deaths occur in the same age distribution, therefore there is no latent period between infection and death, therefore “HIV/AIDS theory” is wrong.

“Another shibboleth [sic] of HIV/AIDS theory is that infection by HIV is followed by a latent period averaging [sic] 10 years before symptoms of illness present themselves; and this pre-symptomatic period is supposed to have been lengthened by contemporary antiretroviral treatment. It follows that the ages at which people die from “HIV disease” should be much greater than the ages at which they become “infected”. Yet the ages at which people most often test “HIV-positive” are the same as the ages at which people are most likely to die of “HIV disease”, in the range of 40 ± 5 years. There is no indication of a latent period, nor that antiretroviral drugs have extended it.”
- "How 'AIDS deaths" and 'HIV infections' vary with age and why"

This claim of no HIV “latent period” is so self-evidently absurd that when I first saw it I thought it had to be a hoax, and that Henry’s real agenda was to demonstrate how easy it was to fool people into accepting a patently ridiculous claim if you baffle them with enough statistics and hand waving. But sadly, he doesn’t seem to be joking.

The claim is self-evidently absurd because when people die with HIV disease it is always after an HIV diagnosis (with the exception of rare HIV diagnoses first made at autopsy), and individuals who die usually do so years and sometimes decades after that diagnosis. Henry’s “explanation” of this apparent paradox is to wave his hands about and claim that the “hypothesized” period between infection and death does not exist, and therefore HIV/AIDS theory must be wrong.

This “explanation” makes no sense. Even if HIV diagnoses had nothing to do with the presence of an infection, and even if there were no causal relationship between HIV and subsequent deaths, that would make no difference to the existence of the intervening period. Furthermore, the existence of a period between HIV seroconversion and death is not “a hypothesis” – it is a fact established through numerous longitudinal studies following subjects with a known time of seroconversion, and is a fact that exists independently of any HIV/AIDS theory. It is also a fact which is the everyday experience of millions of people currently living with an HIV diagnosis years and sometimes decades after their first positive test.

The alternative resolution of this “paradox” – that HIV diagnoses and deaths supposedly occur with the same age distribution – is to recognize that it is simply wrong, and that Henry has either misunderstood his data, or has comprehensively botched his analysis, or is fudging and dissembling in his exposition. In keeping with his status as a crank, none of these possibilities seem to have occurred to Henry, but as we shall see, he has done all three.


TO ILLUSTRATE THE SUPPOSED “superposition” of HIV diagnoses by age over deaths by age, Henry has drawn us this graph, which he has reproduced in assorted variations throughout his blog, see for example "HIV/AIDS and age - HIV theory is wrong"

Now there is no y-axis scale provided here, so it’s difficult to know exactly what figures these curves are supposed to refer to, but even a quick glance leads us to the startling conclusion that between 1999 and 2004 people in their early sixties were diagnosed with HIV at the same rates as people in their early 30s (whatever those rates were)!

Another version of the graph can be found on slide 9 of a presentation he gave to the "Society for Scientific Exploration", from which we discover that young babies test HIV positive at 3-4%, a rate significantly higher than that of people in their 30s and 40s who seem to test positive at around 2.8%.

This startlingly high level of positive HIV tests among babies is also asserted in his seminar notes from the talk he gave at the Virginia School of Osteopathic Medicine:

Babies are infected at about the highest level found among adults who appear to be in good health. Infection rates drop sharply in the first year after birth, and begin to rise again in or after the teens. Males are always infected more than females, except in the low teens when females are more infected than males.

- “Truth is stranger than fiction: HIV is not the cause of AIDS” p.6

That is nuts. Between 2003 and 2006 the CDC received between 100 and 200 notifications annually of diagnoses of perinatally acquired HIV from the 33 reporting states, out of around 4 million births per year for the whole country. This works out to a rate of about 0.005%, not 3-4%, allowing for the notifications that weren’t received from the non-reporting states. This compares with over 5000 HIV diagnoses annually in each 5 year age group of 35-44 year olds.

Furthermore, from the scale provided on the version of the graph presented to the SSE Henry is claiming HIV positive rates for most age ranges of around 1-3%, which is odd since the total prevalence of HIV in the US is currently only around 0.3%. Worse, he seems to be claiming these rates as referring to incident diagnoses (new diagnoses made each given year) when the CDC estimates an annual infection rate of only 0.017% (around 50,000 new infections annually per 300 million Americans).

And are 62 year olds really diagnosed with HIV at the same rate as 32 year olds?

Well, no.

Since 2003 the CDC has published annual data for incident HIV/AIDS diagnoses broken down by 5 year age groups for 33 states, and between 1999 and 2002 provided similar data for 30 states by 10 year age groups. AIDS diagnoses and deaths have been recorded for the whole country since the epidemic was first observed. The age distribution of HIV diagnoses in 2006 mapped to AIDS deaths in 2006 look like this:

Source: Cases of HIV Infection and AIDS in the United States and Dependent Areas, 2006, Table 1 and Table 7

A few points to note:

1.The 33 states that reported new HIV/AIDS diagnoses in 2006 accounted for only 62% of all people living with AIDS in the entire 50 states plus D.C. Therefore the amplitude of the HIV diagnoses curve is likely to underestimate all diagnoses of HIV in the 50 states plus D.C. by a factor of about two thirds. However, this is unlikely to significantly affect the shape of the age distribution.

2. HIV diagnoses are not the same as HIV infections and seroconversions: diagnosis can occur at any stage from seroconversion until presentation with an AIDS defining illness (and occasionally later). The CDC estimated that 38% of AIDS diagnoses occurred within 12 months of first HIV diagnosis: this tendency for late diagnosis of HIV infection was more marked in older age groups than younger: more than half of HIV diagnoses in the over 55s were followed by an AIDS diagnosis within 12 months, while less than 20% were among 15-24 year olds.

3. The 2006 HIV diagnosis and 2006 AIDS deaths curves refer to different populations. AIDS deaths in 2006 occurred in people diagnosed with HIV any time during the previous two decades. Current annual mortality for people living with HIV in the US is only one or two per cent (15,000 out of about a million), which means that deaths in people diagnosed in 2006 will be distributed over a large number of years into the future. The median age of death with HIV/AIDS has increased by around 0.67 years per year since the availability of HAART: on those trends the median age of death for people diagnosed with HIV in 2006 will be significantly greater than the median age of those who died in that year.

(A side note: commenting on a previous post, Chris Noble remarked on the fact that the age distribution of incident syphilis has developed a bimodal distribution since 2006 with peaks in the 20s and 40s. It appears that a similar bimodal pattern is starting to emerge with new HIV diagnoses clearly visible on the above graph – yet another refutation to Bauer’s claim that the demographics of HIV are unlike any other STI.)


EVEN WITHOUT CONSIDERING the three points above, it is obvious that there is a marked difference in the age distributions of HIV infections and HIV/AIDS deaths, corresponding to the “latent” period. So how did Henry manage to make such a hash of his data and end up with the ludicrous graph he keeps hawking round the internet and elsewhere?

Here’s how:

To compare the actual years of that peak on “HIV” tests with the peak years of "HIV” deaths, I wanted “HIV”-test data for the population as a whole, since the death-data in Table A are also for the population as a whole. The most appropriate data-sets are those, totaling nearly 10,000,000 tests, published in 1995-8 by CDC for all public testing-sites (clinics for TB, HIV, STD, drugs, family planning, prenatal care, and more, as well as prisons and colleges and some reports from private medical practices). Pooling the actual numbers for each of those four years and making the appropriate calculations delivers the following results...

In other words Henry is assuming that if, for example, the CDC funded 4,511 HIV tests for 0-4 year olds in 1997 of which 149 (3.3%) were positive, then that percentage can be extrapolated to the 15 million or so 0-4 year olds in the population as a whole, ignoring the fact that these kids were selected for testing from the few thousand in the country actually at risk of infection because they had been born to HIV positive mothers.

Similarly, each other age group in the public test site data is not a representative sample of that age group in the population as a whole: each group consisted of people with identified risks for HIV infection, and who chose to undergo testing using the CDC funded services. The percentage of positive tests in each age group, therefore, is not just a function of the overall incidence in each age range, but also of the percentage in each group who (a) fit the CDC criteria for funding on the basis of HIV risk, and (b) choose to use public sites for testing rather than private, and (c) chose to test at all in that year. Some groups of people test regularly, others rarely, and the frequency of testing does not necessarily reflect the probability of having acquired HIV since the last test. Actual risk, perceived risk, and the options available for testing change in different age groups. Because of this, comparing rates of percentage positive tests between different age ranges does not give you relative rates of incidence for the population as a whole.

Unfortunately,this elementary error of extrapolating data from highly and differently selected groups to the population as a whole is a recurring theme throughout Henry’s thesis.

ONE FINAL COUPLE OF POINTS for those who have bothered to read down this far. When Henry says, “the ages at which people most often test HIV-positive are the same as the ages at which people are most likely to die of HIV disease, in the range of 40 ± 5 years” he is being... well vaguely correct (or he was in describing the figures for 2004), but this is of very limited value in describing the overall age distributions of diagnoses and deaths. The “the ages at which people most often test “HIV-positive” (the mode) is not the average (mean) age, nor is it the median (the midway point with half above and half below): it is substantially older than either of these values, and is even further removed from the mean or median ages of seroconversion because diagnoses at older ages tend to be later in the course of HIV disease than at younger ages.

The “40 ± 5 years” is a fudge, too, spanning as it does an entire decade. Median age at diagnosis and age at death have both been increasing over the course of the epidemic, the latter rising more quickly than the former. And in fact both the median and modal age of incident deaths - not predicted lifespan - is now in the 45 to 50 range, the median increasing by around 0.67 years per year.