Henry Bauer’s harebrained disproofs II: “Frequency of positive tests does not correlate with incidence of AIDS”

HERE'S WHERE HENRY REALLY GOES TO TOWN inventing new methods and principles of epidemiology. Even a quick check of available statistics exposes the nonsense of this claim, which breaks down to four parts:

1. HIV and AIDS are not correlated geographically.
2. HIV and AIDS are not correlated chronologically.
3. HIV and AIDS are not correlated in their relative impact on women and men.
4. Nor are HIV and AIDS correlated in their relative impact on white and black people.
http://failingsofhivaidstheory.homestead.com/

CLAIM: “HIV and AIDS are not correlated geographically”

REALITY: Per capita rates of HIV and AIDS correlate closely for different geographical areas of the United States, allowing for differences in the proportion of prevalent HIV cases infected earlier in the epidemic.

In 2006, the national ratio of “HIV (not AIDS)” to AIDS was 143.7 to 178.6 (55.4% of people diagnosed with HIV in the US had progressed to AIDS). The numbers of AIDS cases in each state is commensurate with the numbers of HIV diagnoses.

Some reporting areas have higher relative rates of AIDS than this: in New York, for example, 62.6% of people with HIV have AIDS, as New York City was an early epicenter of HIV infection, and there are a relatively greater proportion of people with HIV who have had it longer. There are no HIV notification data for California, but the reporting areas with the highest ratios of AIDS to non AIDS HIV diagnoses are those with significant numbers of cases early in the epidemic: New York, Florida, Texas and Puerto Rico.

For 2006 US data on HIV and AIDS per capita prevalence by state, See: http://www.cdc.gov/hiv/topics/surveillance/resources/reports/2006report/map1.htm

CLAIM: “HIV and AIDS are not correlated chronologically”

REALITY: Pre 1996-7 incident AIDS rates closely follow the incident HIV infection rates of a decade or so earlier in affected groups. Since the introduction of HAART for a proportion of people with HIV pre AIDS, the median latency is extended somewhat, but the chronological relationship between incident HIV infection and incident AIDS in each affected group remains clear.

Henry bases his assertion on the fact that overall incident AIDS diagnoses in the US peaked around 1992 to 1993 (which is true), and a claim that HIV prevalence has remained steady at about 1 million ever since 1985 (which is not).

"What's more, not only has the distribution of HIV remained the same ever since testing started, so has the total number of HIV-positive Americans: [table inserted]. On the other hand and in stark contrast, the numbers for AIDS increased into the early 1990s and then decreased: [graph inserted]. So HIV and AIDS are not correlated chronologically either."
"Truth is stranger than fiction: HIV is not the cause of AIDS" p3-4

Bauer’s first error is to confuse prevalence with incidence: the correlation he is looking for is between incident (new) AIDS diagnoses with incident (new) HIV infections. Prevalence, on the other hand, is all incident infections ever, cumulated, and subtracting total deaths.

The sharp peak in AIDS incidence among predominantly white gay men in 1992 and among IDUs in 1993 corresponds with their peak rates of HIV infection a little under a decade earlier in the early to mid 1980s, and is a function of the median clinical latency of HIV of around 10 years – slightly longer among those using the treatments available at the time. AIDS incidence in other groups show less marked peaks, which occur later (typically around 1996-7) and similarly correspond to rates of new HIV infections in those groups around a decade earlier. Incident AIDS continues to rise in non-IDU heterosexuals of both sexes, particularly among black women , corresponding to rises in heterosexually transmitted incident HIV in the late 80s which did not begin to level out until the 90s and which continue to remain high. The overall US peak around 1993-4 and sharp drop thereafter reflects the fact that historically the predominantly white gay men and IDUs infected in the late 70s and early to mid 80s made up the overwhelming majority of early HIV and AIDS cases, however this has changed significantly over the past decade or so. The introduction of HAART made a contribution to reduced AIDS incidence from 1996 on by extending the median period of clinical latency, but its major effect has been on reducing incident deaths.

The early to mid 80s peak rate of HIV infections among gay men (and slightly later peak among IDUs) cannot be established from contempraneous HIV incident diagnoses because the tests did not come into common use until the second half of that decade. However, retrospective testing of stored samples from cohorts of gay men and IDUs confirm these early peaks in new infections.

The “flat 1 million prevalence” argument originates with Duesberg and Rasnick and is a staple HIV/AIDS denialist canard that has been repeatedly debunked. It is a result of cherry picking prevalence estimates, including the upper limits of very broad ranges, which were postulated tentatively or with misplaced confidence. Prevalence estimates from the mid to late 80s in particular were often based on very scanty data and had large ranges of uncertainty.

The best available evidence is that HIV prevalence rose sharply from near zero in the late 1970s till the mid 80s, before rising more slowly and even leveling out toward the mid 90s as deaths began to approximate new infections. Since 1996-7 following the introduction of substantially improved treatments, HIV prevalence began to increase again as death rates fell relative to those of new infections.

For US data on prevalent and incident AIDS 1985-2006 by various demographic indicators, see http://www.cdc.gov/hiv/topics/surveillance/resources/slides/trends/


CLAIM: “HIV and AIDS are not correlated in their relative impact on women and men”

REALITY: Yes they are.

In 2006 males accounted for 72.95% of people with HIV infection in the 33 states and 5 U.S. dependent areas with confidential name-based HIV infection reporting.

Males also accounted for 73.36% of new AIDS diagnoses, 72.91% of AIDS deaths and 76.98% of adults and adolescents living with AIDS in the US.

Rates of HIV and AIDS are, in fact, correlated in their relative impact on men and women.


CLAIM: “Nor are HIV and AIDS correlated in their relative impact on white and black people”

REALITY: Yes they are.

In 2006, blacks accounted for 47.17% of people living with HIV in the 33 states with confidential name-based HIV infection reporting.

They also accounted for 43.87% of people living with AIDS, 48.77% of new cases of AIDS and 52.98% of deaths from AIDS.

In 2006, whites accounted for 33.76% of people living with HIV in the 33 states with confidential name based HIV infection reporting.

Whites were correspondingly 35.38% of people living with AIDS, 29.68% of new AIDS cases, and accounted for 27.54% of deaths from AIDS.

Rates of HIV and AIDS are, in fact, correlated in their relative impact on white and black people

For 2006 HIV and AIDS data by sex and by race see: http://www.cdc.gov/hiv/topics/surveillance/resources/reports/2006report/default.htm


GIVEN THAT BAUER’S CLAIMS are so clearly and evidently at odds with reality, the obvious question is this: how did he manage to botch his analysis so comprehensively?

Part of the answer is that he has made several kinds of fairly elementary epidemiological errors, and made them repeatedly. For example, he often fails to distinguish incidence from prevalence (see for example the chronological argument above). He fails to understand that a series of cross sectional studies – for example HIV rates among blood donors or military recruits from year to year – don’t amount to a longitudinal study of any population, particularly since a prior HIV diagnosis excludes anyone from selection for future cross-sectional studies in these settings. One of his most common and serious errors, though, is trying to derive population-wide prevalence or incidence statistics from the ratios of positive and negative tests (what he calls “F(HIV)”) within quite specific subgroups and settings, without considering the criteria used to select the population under study, including the reasons for testing in the first place.

Much of his argument and analysis combines all three of these methodological flaws, often combined with basic faulty assumptions such as that HIV is diagnosed on the basis of a transient antibody response or that people diagnosed with HIV rarely progress to AIDS. He is hopelessly confused about the semantic relationship between AIDS (an immune system disease) and its indicator illnesses, and has no insight into why such illnesses are medically significant. He uncritically regurgitates misconceptions from other denialists about the sensitivity and specificity of diagnostic testing algorithms, and either does not understand or deliberately misrepresents the scientific literature on treatments. Occasionally he simply fails to read his data properly.

What makes him a crank, though, is that he holds on to his assertions despite the fact they are obviously contrary to evidence, and that he continues to make the same methodological errors even when they have been pointed out to him. Encouraged by his tiny fan-base he persists in his sad fantasy that one day his “insights” and “analysis” will be discovered by the scientific mainstream and he will be lauded for the iconoclastic genius he believes he is.

Unfortunately, his “insights” are flawed and ignorant, and his “analysis” is inept.