[This post originally appeared at Lean Left, a general-issues blog KTK also contributes to. The founders of Lean Left graciously allowed me to re-post it here to bring all my health-related posts into one place. Original posting: 10/21/2004]

Phillip Johnson – self-appointed fundamentalist scourge of science everywhere – is off on another ignorant rampage, this time directed at the “sweet racket” of the worldwide AIDS conspiracy. He is not just wrongheaded and wrongly-informed; he has allied himself with a crank theory on the fringes of (what you can barely call) science, and is championing not only crackbrained conspiracy theories of his own, but the literally deadly fantasies of some of the world’s most irresponsible people involved in the global AIDS effort. His recent article is nominally about procedural problems in the estimation of the total prevalence of AIDS in Africa. But he uses this point to slip in some unexplained hints that the entire scientific underpinning of AIDS research is faulty, and to extol the ravings of near-lunatics who have almost destroyed AIDS work in Southern Africa. When Johnson was merely blathering about evolution, his ignorance was ignorable; when he tries to inject himself into medical treatment for a desperate population suffering lacerating resource shortages, he is unforgivably irresponsible.

Joe Carter, at Evangelical Outpost, has linked Johnson’s article, echoing his implications of a conspiracy to inflate AIDS statistics to take advantage of aid money. That in itself seems rather overwrought, but aside from that, Joe does not seem to have picked up on Johnson’s insinuation that the standard AIDS treatments themselves should not be used. If this idiotic article begings to get traction in the evangelical community (and especially if Bush wins re-election and they continue to dictate scientific policy in the US), the results could be disastrous.

(Read on.)

The Problem

There is, in fact, a problem in estimating rates of HIV and AIDS in Africa. The lack of precise diagnostic technology, and the lack of easy access to random population samples for statistically valid survey testing, make all AIDS prevalence estimates from that continent questionable. Johnson�s point about the importance of accurate epidemiological surveys is perfectly reasonable, and hardly new – accurate information is best in any circumstances. And if the epidemic-level estimates of AIDS in some African countries really are as far wrong as Johnson’s sources indicate, then there is likely a serious resource misallocation going on. But Johnson breathlessly implies that there is some sort of conspiracy to suppress accurate information about AIDS rates in Africa � and Joe imputes the conspiracy to some sort of shadowy AIDS underworld that is deliberately inflating disease rate estimates to scam foreign aid dollars. In fact the problems with AIDS rate estimates in Africa are well-known and have been widely discussed for years. They result from simple practical exigencies: it is difficult to do sophisticated laboratory testing in countries that do not have a medical-technology infrastructure, so diagnosis is based on direct observation of clinical symptoms rather than on laboratory tests; it is difficult to survey disease spread directly in a population with little contact with healthcare workers, so indirect estimates of spread rates are used. These methods are well-established and widely used, but they have significant limitations.

Johnson’s article is characteristically off-base. The issues he raises are serious, but they have been discussed by serious people all along. Johnson believes he has blown the lid off some dark secret, when in fact he has only confused himself with his lack of technical understanding and his bizarrely obtuse refusal to use simple common sense. To the extent that he is taken seriously (he never has been yet, which is a good thing, but we don’t want to set a precedent here), he could cause a backlash against a problem that doesn’t exist as he imagines it, and is already being dealt with as well as possible by dedicated and informed people.

The Confusion

Both Joe Carter and Phillip Johnson imply that there is some kind of conspiracy going on to inflate AIDS statistics and to suppress accurate prevalence estimates. Johnson thunders:

Every year, all over Africa, blood samples are taken from small numbers of women at pregnancy clinics and screened, not for the virus itself, but for proteins thought to be indicative of antibodies to HIV. From the premise that the presence of the antibody equals incurable infection, the Epimodel [computer] program calculates an estimate of the total number of African women infected by HIV. If so many women are infected, it follows that a like number of their husbands and lovers must be infected also, and, according to the underlying virus theory, all these will sicken and die at a predictable rate.

When these estimates are extrapolated to the general population, the computer modelers can arrive at seemingly precise tallies of the doomed, the dying, and the orphans left behind, with no need for anyone to verify the figures by counting bodies on the ground. Do the funded researchers regularly perform searches of mortality records to check if their estimates are accurate?

No. The HIV-scientists have so much confidence in their model that they see no need for corroborating the figures it generates, so any verification is strictly pro forma. Continent-wide verification is impossible because no reliable mortality records exist in most of Africa.

[emphases added]

Sounds pretty shifty! They aren’t counting virus loads – they’re just counting antibodies and making . . . an assumption about infection rates. Then they . . . extrapolate! And they never check against mortality records! Who are these frauds? Where do they get off with such sloppy methods?!

Well, first, as anyone familiar with basic biology or immunology would recognize, but I suppose many self-half-educated law professors do not, antibodies are produced in response to specific foreign proteins (such as virus particles), so the presence of antibodies to a particular virus is almost certain proof that that person has been exposed to that virus. (This may not prove that the virus is currently active – the antibody remains after the virus is gone – but since the virus almost never spontaneously disappears in the case of HIV infection, positive HIV antibody titres have a very high correlation with active HIV infection.) Most immunological screening tests use antibody reactions, because they are much cheaper and easier than viral screening; even today, standard practice in the West is to use antibody tests for initial diagnosis of HIV infection, with viral-load assays used diagnostically only for confirmation in certain cases.

Plasma Viral Load Testing in the Management of HIV Infection

(Am Fam Physician 2001;63:483-90,495-6)

The combination of a screening enzyme-linked immunosorbent assay (ELISA) followed by a confirmatory Western blot test [i.e., two types of antibody screenings] has been more than 99 percent accurate in detecting HIV infection. However, this protocol may have negative or indeterminate results, especially during the first weeks of HIV infection.

. . . The occurrence of . . . high levels of [virus load] during primary HIV infection has led some physicians to use PVL [plasma viral load] assays as diagnostic tests for early HIV infection in high-risk patients with a negative ELISA or an indeterminate Western blot technique.

A drawback to PVL testing is the high cost of the assays. . . . To minimize the occurrence of false-positive results, only patients who have a high pretest probability of a positive result should be evaluated for HIV infection using PVL testing. Such patients include those with definite or probable recent exposure to HIV and a clinical syndrome suggestive of acute HIV infection. . . .

Even if PVL is detectable, repeat HIV antibody testing is indicated to rule out a false-positive PVL assay.

[references omitted]

So, antibody screening is not only the most reasonable technique for that region for reasons of cost and available technology, it’s actually the test used in other countries as well, where viral-load tests are used only as a backup in ambiguous or high-risk cases. (Viral load assays are widely used to monitor the effectiveness of treatment after a positive diagnosis has been made, but that is a different matter.)

Second, extrapolating from patient contacts and known transmission rates is a standard part of epidemiology – it’s done in every country for every disease. There’s no other way to estimate prevalence in cases in which you cannot take random samples from the entire population. In this case, the only source of test material from a wide spectrum of the population is women who come into health clinics, so they – perfectly reasonably – test that sample, then extrapolate to the larger female population and from there to the male population. What else could they do? Of course this introduces potential error – they know this perfectly well. But they are using a standard epidemiological technique to make the best estimates possible.

Third, there is no deliberate refusal to calibrate the accuracy of the estimates against official mortality records. As Johnson himself points out in the same paragraph:

[N]o reliable mortality records exist in most of Africa.

It makes no sense at all to say that “HIV-scientists have so much confidence in their model that they see no need for corroborating the figures”. They don’t check against mortality records because there aren’t any! It has nothing to do with “confidence” or “seeing a need” – as Johnson himself states in his very next sentence, it’s physically impossible to do what he suggests! Why he imagines that they should do, or that they are deliberately indifferent to doing, what they simply cannot do in any case, is incomprehensible.

This strangely paranoid attitude infects the rest of his prose. His remarks, in various places, include:

[T]he premise that the presence of the antibody equals incurable infection . . .

[A]ccording to the underlying virus theory, all [those estimated to be HIV-positive] will sicken and die at a predictable rate.

When these estimates are extrapolated to the general population, the computer modelers can arrive at seemingly precise tallies of the doomed, the dying, and the orphans left behind . . .

Any person with such common conditions as persistent fever, coughing, and weight loss can and will be diagnosed as a doomed AIDS sufferer.

(The last quote is from the full text of the letter, linked above.)

This language is both weird and false. There is nothing about “underlying virus theory” that categorizes HIV-positive patients as “doomed”, “dying”, or �incurably infected�. It is certainly true that untreated HIV infection follows a roughly predictable clinical course, but the whole point to these screening efforts is to develop treatment strategies. Aggressively treated HIV infection offers an essentially unlimited lifespan; less-optimal therapy gives more limited, but still highly effective, results. The issue for Africa is that AIDS therapies are extremely expensive and difficult to implement in areas with limited health infrastructure. So the expected clinical outcome for HIV in Africa depends on many factors, including the patient’s access to healthcare, the country’s AIDS therapy budget and availability of foreign aid, which therapy regimen is offered, and so on. But “doomed” is not part of any clinical diagnosis in any part of the world. Why Johnson keeps speaking in these terms, and why he imputes such an attitude to people who are actively dedicated to providing treatment that will prolong patients’ lifespans, is hard to say, but it’s very strange at the least.

And, finally, his discussion of the “Bangui criteria” is ill-informed. As Johnson does not explain, and apparently does not understand, this is a clinical case definition (a diagnostic definition based on how the patient looks under examination, as opposed to a laboratory diagnostic definition based on the results of laboratory tests). Clinical case definitions are perfectly standard forms of medicine, and in fact are used in most ordinary medical diagnoses (if you’ve ever had a doctor say “Yep, you’ve got xxxxx” without waiting for the results of a lab test, that doctor was working off a clinical case definition). AIDS was originally defined in terms of clinical presentation, but the standard case definition in the West shifted to laboratory-based diagnostic criteria when antibody assays became widely available. Note something important: the case definition changed when the available technology changed. A “case definition” says nothing about the underlying disease mechanism – AIDS is caused by HIV no matter what diagnostic criteria are used – it just defines the best method of identifying the disease in practical terms. Clinical case definitions are widely used for AIDS in areas in which laboratory tests are not available, not affordable, or not practical (for instance, when patients from rural areas cannot be expected to return to clinics for test results, and must have a diagnosis made and treatment initiated on their first visit). The World Health Organization currently utilizes 6 different AIDS case definitions, three for areas with “sophisticated laboratory facilities” and three for areas with “limited laboratory facilities”. They explain that the differences are dependent on “population factors (children, adults, relative occurrence of opportunistic infections) and on the laboratory infrastructure and training available”.

Johnson is wrong both on the facts and in his comprehension of what a diagnosis means when he says that “AIDS in Africa has a definition . . . so completely different from the definitions of AIDS used in North America and Europe that it is altogether a different condition”. There is nothing about the WHO’s clinical case definition (i.e., it’s method of diagnosis) that implies that AIDS itself is “a different condition” or that the disease being diagnosed as AIDS in Africa is something other than the disease being diagnosed as AIDS in the West. The WHO uses different methods of diagnosis in the two regions in an attempt to identify the same disease using technology most available and most appropriate in each. Again, what else could they do? If they used a high-technology definition requiring an antibody test (note: precisely the technology Johnson had previously been jumping on WHO epidemiologists for using in screening surveys – none of the WHO case definitions requires a viral load titre) in a region in which such tests are not available to clinicians, what would the result be? There would be almost no WHO-recognized diagnoses of AIDS in Africa at all! It’s simple, really: if the criteria, by definition, require a laboratory procedure that costs too much to afford or is not available, and for those reasons that procedure is not performed, then by definition that patient did not meet the criteria and will not be diagnosed as having the condition. So it is totally counterproductive to insist on diagnostic criteria that it is actually impossible to apply, because that will hugely undercount the affected population, not because the test is bad but simply because nobody is actually taking the test as prescribed. Thus, the WHO uses a clinical – not a laboratory – case definition in regions where that is most appropriate.

Johnson is equally wrong in his description of what the test actually requires (a seemingly simple matter – he could have just looked it up). He claims that:

Any person with such common conditions as persistent fever, coughing, and weight loss can and will be diagnosed as a doomed AIDS sufferer. These symptoms are characteristic of both malaria and tuberculosis, which are very common throughout Africa, as well as other diseases . . .

This is idiotic. The actual WHO case definition employed in Africa is perfectly straightforward:

For the purposes of AIDS surveillance an adult or adolescent (>12 years of age) is considered to have AIDS if at least 2 of the following major signs are present in combination with at least 1 of the minor signs listed below, and if these signs are not known to be related to a condition unrelated to HIV infection.

Major signs (2 signs or more):

• Weight loss of at least 10% of body weight
• Chronic diarrhoea for >1 month
• Prolonged fever for >1 month (intermittent or constant)

Minor signs (1 or more):

• Persistent cough for >1 month
• Generalized pruritic dermatitis
• History of herpes zoster
• Oropharyngeal candidiasis
• Chronic progressive or disseminated herpes virus infection
• Generalized lymphadenopathy.

The presence of either generalized Kaposi sarcoma or cryptococcal meningitis is sufficient for the diagnosis of AIDS for surveillance purposes.

Note first that Johnson has combined “major” and “minor” symptoms from the list, and that he has not noted that the clinical diagnosis requires 2 of the 3 symptoms be present. The presence of greater-than-10% weight loss and diarrhea or fever for more than a month at a time (or any two of those) is not common in many diseases. Even then, confirmation requires the presence of at least one characteristic opportunistic infection, some of which (oral thrush, notably) are almost exclusive to AIDS. And, finally, the criteria explicitly specify that they only apply in cases in which the diagnostic symptoms are not due to some other condition. It’s not like anybody in Africa needs Phillip Johnson to tell them about malaria. Clinicans there know perfectly well what conditions can produce what symptoms. They just need a standard set of criteria which more or less reliably distinguish AIDS from non-AIDS in cases in which AIDS is a possible diagnosis.

For what it’s worth, clinical evaluations of the African WHO clinical case definition of AIDS that were performed – contrary to Johnson’s assertion – demonstrated that “The WHO/Bangui definition in adults has a sensitivity [true-positive rate] of 60%, a specificity of 90% [true-negative rate] , and a high predictive value especially in endemic areas.” These accuracy rates would be quite low for a laboratory-based diagnostic test, but, as noted above, that’s not an option in much of Africa. Note, too, that the relatively low sensitivity of the test would tend to under-count cases, not overcount them as Johnson asserts. Finally, the “positive-predictive value” of the test depends not only on sensitivity but on the actual rate of disease in the population, and, as the researcher notes, in Africa these criteria are working well.

So, Johnson is wrong is his characterization of the diagnostic criteria, badly confused in his apparent understanding of what a diagnosis actually means, and completely wrong in his direct, explicit statement of how the criteria are used clinically and under what conditions and AIDS diagnosis is produced. He also seems to think that African clinicians do not both to distinguish malaria or tuberculosis from AIDS in making their diagnoses. This is hack work.

The Implications

Johnson’s conclusion that AIDS is being grossly overestimated as a result of these issues is unfounded. The evidence of overestimation that is given � Joe�s example of the overestimates of AIDS in the Angolan military, and Johnson’s second-hand citations of wrong estimates in certain countries – are weak at best. As Joe himself points out, the Angolan problem was a case of faulty sampling – it has nothing to do with the case definition (clinical or laboratory-based); if they had used a clinical test for diagnosis but had sampled the same (wrong) population, the results would have been just as skewed. The best evidence, from Johnson’s original source (Rian Malan of the UK Spectator, article here, reprinted here), includes useful statistics suggesting that death rates from AIDS predicted by epidemiological computer models were too high � but also that the models were sequentially improved. (He also notes, but Johnson somehow overlooks, that the original program Johnson criticizes by name has not been used since 2000.) While, obviously, everyone would have liked the first-generation predictive model to have been perfectly accurate, revisions of this kind are par for the course in such cases. Again, the simple practical exigencies of technology use have been inflated into some grand conspiracy, and the fact that predictive uncertainties exist is taken as evidence that the predictions have been systematically skewed.

The “STATS” project of George Mason University (a project to provide a neutral analysis of figures used in news stories) did an article on precisely the issues Johnson discusses, in which they agree that the clinical case definition and limited-sample extrapolations are sources of concern. They conclude:

[W]hile estimates of the extent of HIV infection have been forthcoming (ideally based on blood analysis), the true scope of the crisis is simply unknown.

There could be vastly more cases lurking than have been dreamed of in the current nightmare – or there could be substantially less.

This seems much more reasonable than an assertion of a widespread conspiracy to systematically distort evidence in one particular direction.

The Insanity

Johnson spends most of his time discussing � badly � the limitations on AIDS diagnostica and epidemiological techniques used in Africa. But he slips some rather strange-sounding remarks in along the way:

Almost exactly 20 years ago, American health authorities announced the discovery of the virus (HIV) they said was the probable cause of AIDS, and predicted that a vaccine would be available within two years. Neither the exposure of the long-suppressed doubts about the validity of the scary [epidemiological] statistics nor the failure of the vaccine trials has motivated the researchers to consider the possibility that there might be something wrong with their understanding of the epidemic.

As predictions fail and anomalies pile up, the AIDS experts cling to their theory as dogmatically as they have done since 1984. If anyone ever wonders, �Could we have made a mistake?� the unwelcome question never appears in the mass media or in the scientific journals. . . .

The impasse in AIDS research suggests two questions. One is . . . whether the statistics are wrong because the official understanding of the underlying syndrome is wrong.

The validity of the statistics is tied to the validity of the underlying virus theory . . .

[emphases added]

It�s not clear exactly what he�s referring to, until you get to this eye-opener:

There is one hope. South African President Thabo Mbeki has read the scientific literature, including articles by scientists who dispute the nature of the health crisis that threatens Africa, and he has become skeptical . . .

You have to be reasonably well up on AIDS politics to recognize this name, and how frightening it is. (Even a novice might wonder, however, why the personal interpretation of the scientific literature by the elected president of South Africa carries any weight in this debate. Note that Mbeki � like Johnson, for that matter – has no scientific background at all. Would you trust your AIDS diagnosis to him?) Mbeki holds a racially-inspired conspiracy theory to the effect that AIDS does not exist as a distinct disease, and that the connection between HIV (the virus) and AIDS (the disease) was a myth promulgated by drug companies to allow them to conduct racist experiments on blacks. When he was elected President of South Africa, he turned his party officially against standard treatments for AIDS, and blocked distribution of AZT and other retroviral drugs in South Africa for years. He also invited members of the AIDS conspiracy fringe to advise the South African government and sit on an AIDS study panel there.

As a digression, I note that this points up the dangers of pseudo-science. Johnson�s AIDS conspiracy theorizing is not conceptually related to his creation-science beliefs, but both obviously flow from a deep conceptual difficulty with science, and a tendency to imagine that mainstream science is riddled with organized conspiracies to promote falsehoods. I don�t understand how anyone pretends creationism or AIDS conspiracy theories make sense, but I can easily see how someone who believes in the former would believe in the latter � and claim the same �we must look at the data in a new way� pseudo-intellectual rationalization for both. This is just what Johnson, playing true to form, has done � only this time, other people�s lives hang in the balance.

At any rate, Johnson�s invocation of Mbeki makes his true aims clear � he is not just criticizing methodological difficulties in AIDS monitoring, he is laying the groundwork for a claim that HIV is not the AIDS virus at all (and presumably that AIDS efforts aimed at treating HIV replication � i.e., all of them � should be discontinued). This is not just ignorant (the �HIV does not cause AIDS� claim has been debunked many, many times � like all conspiracies, it refuses to die), it is dangerously insane. It is irrationality that literally threatens the lives of everyone it affects � just as Mbeki allowed AIDS to go untreated for thousands upon thousands of the poorest South Africans.

Summary

Johnson�s egregious shortcomings aside, it remains true that the actual prevalence of HIV and AIDS in Africa are not well known. The problems with the clinical case definition and the epidemiological predictive model (they are real) are well known, and they significantly complicate the matter of identifying and treating AIDS and of designing treatment programs and resource-allocation priorities for the region. (These issues are also not some hidden conspiracy – references to them can be found in professional literature going back almost to the beginning of the AIDS crisis.) But it is ridiculous to pretend that these problems are deliberately manufactured by vested interests trying to obscure the (minimal) severity of AIDS. These issues – every one of the points Johnson raises that are not just stupid on their face (like demanding viral-load tests as a diagnostic procedure) – are a direct result of the simple facts that there is no widespread access to clinical HIV antibody testing for African clinicians, antibody testing costs far too much for almost all African countries to implement on a broad scale even if they had the technology, and laboratory-based diagnostic procedures are not appropriate in some patient populations (particularly those who cannot make repeated trips to clinics for testing and followup, and those whose care is provided by overworked clinics that cannot budget two or more visits per patient). The clinical case criteria and limited-sample prevalence estimates used in managing AIDS in Africa are both longstanding medical/epidemiological methods that are, under the circumstances, about the only practical way of going about things in areas. There is nobody who would not like to do better, but both budgetary and simple logistical constraints put severe limits on what can actually be achieved.

Johnson bungles the technical issues that underly such limits, makes grandiose and ignorant demands for standards that make no sense (viral-load testing as a primary epidemiological screen – something nobody does; antibody-based diagnosis in countries that do not have antibody testing labs) and would be positively counterproductive if employed, and then attributes the well-known issues he thinks he has discovered to some sort of nebulous conspiracy when those problems are obviously the direct result of the bare facts of logistics and finances in the countries in question. He also holds that there is something wrong with employing the best and most applicable standards and procedures in a country if those standards are not equivalent to those employed in other countries with completely different technological infrastructure. His confusion would be comical if he weren’t so self-righteous about it. But whatever his problems may be, it doesn’t help to inflate well-known issues – that are widely agreed to be serious and frustrating – into conspiracies. The people Johnson ignorantly accuses of causing the problems he points out are the people actually working to solve those problems. They deserve better than to be asked to respond to irresponsible and confused carping of this kind.

And those are merely the least of his conceptual and logical difficulties. Far worse is his open support for anti-science, fringe claimants with no supportive data and at times no scientific background at all, and AIDS conspiracy theories that already have killed thousands of victims of ignorance and stupidity. Whatever it is that drives his perfervid ravings about scientific conspiracies that only he and a few like-minded visionaries can see, when he brings them into the realm of medical diagnosis and treatment he is not just being intellectually irresponsible, but morally irresponsible and quite possibly murderous.

* * * * *

[NB: Aside from his many conceptual errors, it’s worth nothing that some of Johnson’s article is also plagiarized. Compare Johnson’s text to the text of his main source, Rian Malan’s article in the Spectator:

JOHNSON

Every year, all over Africa, blood samples are taken from small numbers of women at pregnancy clinics and screened, not for the virus itself, but for proteins thought to be indicative of antibodies to HIV. From the premise that the presence of the antibody equals incurable infection, the Epimodel program calculates an estimate of the total number of African women infected by HIV. If so many women are infected, it follows that a like number of their husbands and lovers must be infected also, and, according to the underlying virus theory, all these will sicken and die at a predictable rate.

When these estimates are extrapolated to the general population, the computer modelers can arrive at seemingly precise tallies of the doomed, the dying, and the orphans left behind, with no need for anyone to verify the figures by counting bodies on the ground. Do the funded researchers regularly perform searches of mortality records to check if their estimates are accurate?

MALAN

Every year, all over Africa, blood would be taken from a small sample of pregnant women and screened for signs of HIV infection. The results would be programmed into Epimodel, which transmuted them into estimates. If so many women were infected, it followed that a similar proportion of their husbands and lovers must be infected, too. These numbers would be extrapolated out into the general population, enabling the computer modellers to arrive at seemingly precise tallies of the doomed, the dying and the orphans left behind.

(It is also fair to note that Malan�s article is much better than Johnson�s and does provide significant support for his claims. He is also not an AIDS-virus conspiracy theorist. But he is something of an AIDS-community conspiracy theorist:

I think it is time to start questioning some of the claims made by the Aids lobby. Their certainties are so fanatical, the powers they claim so far-reaching. Their authority is ultimately derived from computer-generated estimates, which they wield like weapons, overwhelming any resistance with dumbfounding atom bombs of hypothetical human misery. Give them their head, and they will commandeer all resources to fight just one disease.

"Fanatical"? "The powers they claim"? Their "authority"? Their "weapons"? "Atom bombs"? "Overwhelming any resistance"?

I'd like to see the day when the medical-aid community in non-industrial Africa had that much power. And I�ve never heard anyone argue that all resources should go toward AIDS.

Like Johnson, Malan congratulates himself for discovering a deep cabal engaged in deceptive practices. In fact, he has done useful service pointing up the (already known) deficiencies in some computer models for predicting the spread of AIDS (and undercut his own conspiracy theory by also pointing out that those models are getting increasingly accurate over time), but what he has to say is hardly new, hardly earth-shaking, and evidence of nothing more than the difficulties of doing a difficult job.)]